n

.avrnmtw...u_ .44
sfiwnaufivm. .
. .
...n..£.. yam

. ,.zs.:nm.whuv¢d

.. .1
. : u
I

311‘

11‘

firmmfiw§

«E

5.01
mm-

as .32 , . A .
: . . ‘. . A
.. my“... . , U ‘
. x .
.r .m .
. v ‘
,. . 45%:

a.“
. .F’

3% nmr

‘

2
02.5 .
. c

.Rfim:
r.- .. .

3. a}

w .I‘ 7
kg. Lw.
rum:

3: at.
. 31.:

 

 

infirm. a. 5512 p

 

 

 

THESlS

K)
n
(3*

 

LIBRARY

Michigan State
University

 

 

 

This is to certify that the

thesis entitled

An Epidemiological Critique of Two
Childhood Leukemia Cluster Studies:
Woburn, Massachusetts and West Central Phoenix, Arizona

presented by

Allen William Stout

has been accepted towards fulfillment
of the requirements for

 

Master's degree in Epidemiology

 

 

 

Date December 7, 2000

 

0.7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

11/00 chlRC/DateOuopfiS—pJA

 

AN EPIDEMIOLOGICAL CRITIQUE OF TWO CHILDHOOD LEUKEMIA
CLUSTER STUDIES: WOBURN, MASSACHUSETTS AND WEST CENTRAL
PHOENIX, ARIZONA

By

Allen William Stout

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Epidemiology

2000

ABSTRACT

AN EPIDEMIOLOGICAL CRITIQUE OF TWO CHILDHOOD LEUKEMIA
CLUSTER STUDIES: WOBURN, MASSACHUSETTS AND WEST CENTRAL
PHOENIX, ARIZONA

BY

Allen William Stout

This paper was undertaken to critically review and analyze the childhood
leukemia studies conducted in West Central Phoenix, AZ and Wobum, MA in
light of Leo Kinlen and Melvin Greaves’ infectious etiology hypotheses, and to
provide a general epidemiologic critique of the studies. The focus is on Acute
Lymphocytic Leukemia (ALL) (ICD-9 code 204.00 and 204.01), as the infectious
etiology hypothesis pertains to this type of leukemia. The studies are critiqued
according to an epidemiologic investigation checklist (e.g. potential biases,
confounders, descriptive and analytic methods, validity of statistical
interpretation, etc.) to determine if the ALL cases fit a model that is consistent
with an infectious disease etiology. The findings of this paper indicate that there
is some evidence to support the infectious etiology hypothesis for ALL and study

methodology recommendations are made for future research in this area.

Copyright by
Allen William Stout
2000

Dedicated to my mother who has always encouraged me to excel.

ACKNOWLEDGMENTS

Michael Rip, Ph.D.
Michael Collins, MD, MS.
Alka lndurkhya, PhD.

Nigel Paneth, MD, MPH

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................. vii
LIST OF FIGURE ................................................................................................ viii
INTRODUCTION .................................................................................................. 1
CHAPTER 1

BIOLOGICAL ASPECTS OF LEUKEMIA ........................................................... 10
CHAPTER 2

ETIOLOGY OF ACUTE LYMPHOCYTIC LEUKEMIA ......................................... 15
CHAPTER 3

WOBURN, MASSACHUSETTS CHILDHOOD LEUKEMIA STUDIES ................ 44
CHAPTER 4

WEST CENTRAL PHOENIX, ARIZONA CHILDHOOD LEUKEMIA STUDIES...65
CHAPTER 5

EPIDEMIOLOGIC CRITIQUE OF THE WOBURN AND WEST CENTRAL
PHOENIX STUDIES ........................................................................................... 76
CHAPTER 6

CONCLUSIONS .................................................................................................. 88

vi

LIST OF TABLES

TABLE 1: Annual Incidence of ALL in the US. based on 9 Standard SEER
Registries ............................................................................................................ 18

TABLE 2: Average Annual Incidence of ALL in the United States from 1991 to
1995 .................................................................................................................... 18

TABLE 3: Annual Incidence of AML in the US. based on 9 Standard SEER
Registries ............................................................................................................ 19

TABLE 4: Average Annual Incidence of AML in the United States from 1991 to
1995 .................................................................................................................... 19

TABLE 5: SEER Incidence Trends of ALL (Age 0-14) in the United States: %
Change ............................................................................................................... 20

TABLE 6: SEER Incidence Trends of AML (Age 0-14) in the United States: %
Change ............................................................................................................... 20

TABLE 7: SEER Mortality Trends of ALL (Age 0-14) in the United States: %
Change ............................................................................................................... 21

TABLE 8: SEER Mortality Trends of AML (Age 0-14) in the United States: %
Change ............................................................................................................... 21

TABLE 9: SEER Relative Survival Rates of ALL (Age 0-14) in the US:
Percentage ......................................................................................................... 22

TABLE 10: SEER Relative Survival Rates of AML (Age 0-14) in the US:
Percentage ......................................................................................................... 22

TABLE 11: Comparison of Reports on Childhood Leukemia in Europe, NA &
Australia .............................................................................................................. 23

TABLE 12: Childhood Leukemia Cases in Woburn from 1/69 to 8/89 by Cell
Type .................................................................................................................... 78

TABLE 13: Childhood Leukemia Cases Sub-Sample in WCP 1965 to 1990 by
Cell Type ............................................................................................................. 78

TABLE 14: Childhood Leukemia Cases in Woburn from 1/69 to 8/89 by Age
Group .................................................................................................................. 99

TABLE 15: Childhood Leukemia Cases Born in WCP 1965 to 1990 by Age 100

vii

LIST OF FIGURES

FIGURE 1: Location of Woburn, MA and West Central Phoenix, AZ on

US Map ........................................................................................................... 106
FIGURE 2: Location of Woburn, MA on Massachusetts Map ........................... 107
FIGURE 3: Woburn, Massachusetts Map ......................................................... 107
FIGURE 4: Location of West Central Phoenix, AZ on Arizona Map .................. 108
FIGURE 5: West Central Phoenix, Arizona Map ............................................... 108

viii

LIST OF ABBREVIATIONS

Acute Lymphocytic Leukemia .......................................................................... ALL
Acute Myelogenous Leukemia ......................................................................... AML
Aberrant Response Model .............................................................................. ARM
Arizona Department of Health Services ........................................................ ADHS
Common Acute Lymphocytic Leukemia .......................................................... cALL
Centers for Disease Control ............................................................................ CDC
Chronic Lymphocytic Leukemia ....................................................................... CLL
Chronic Myelogenous Leukemia ..................................................................... CML
Department of Environmental Quality Engineering ....................................... DEQE
Electro Magnetic Field .................................................................................... EMF
Massachusetts Department of Public Health ............................................... MDPH
Massachusetts Health Research Institute ...................................................... MHRI
National Institute for Occupational Safety and Health .................................. NIOSH
Surveillance, Epidemiology and End Results ................................................ SEER
Trichlorethylene ............................................................................................... TCE
West Central Phoenix ..................................................................................... WCP
Woburn Environment and Birth Study .......................................................... WEBS

INTRODUCTION

Cancer is a major concern in the United States and many other regions of
the world. This concern is heightened when cancer occurs among children. A
notable form of childhood cancer that has received much attention in the past is
childhood leukemia. Childhood leukemia is still a fairly rare occurrence. Acute
lymphocytic leukemia (ALL) is the most common leukemia in young people (0-
19), and it has an incidence rate of only 3.1 per 100,000. The next most
common childhood leukemia is acute myelogenous leukemia (AML), and it only
has an incidence rate of 0.6 per 100,000 (Wang & Haines, 1995). In spite of the
small numbers of childhood leukemia cases and the improvements in treatment,
childhood leukemia is still viewed as a serious problem. Concerns about
childhood leukemia are heightened even more when a cluster of cases is
observed.

Cancer clusters are the occurrence of a greater than expected number of
cases of a particular type of cancer within a group of people, a geographic area,
or a period of time. Cancer clusters capture public attention and therefore must
be investigated, however they are extremely difficult to evaluate and to
conclusively link to a putative agent. In fact, from 1961 to 1982, the Centers for
Disease Control (CDC) investigated 108 reported cancer clusters in 29 states
and 5 foreign countries and all had negative results. The types of cancer clusters
studied were leukemia (38%), leukemia and lymphoma (30%), leukemia and
other cancer combinations (13%), and all other cancer or combinations (19%).

Although 14 different categories of associations were reported, no clear cause

was found for any cluster. (Caldwell, 1990) Also, Freda Alexander recently noted
that "no causal factor has been identified which can explain a single cluster of
childhood leukemia" (Alexander, 1999).

Childhood leukemia clusters in Woburn, Massachusetts and West Central
Phoenix (WCP), Arizona are prime examples of the difficult nature of cluster
analysis. After careful consideration, the studies of these two clusters were
chosen for critique in this paper. The clusters at Woburn and WCP were
identified as key clusters in the United States, and they also demonstrated stark
contrasts in their features and the way in which they were studied. Woburn, a
small community in existence for over 300 years, is located in the Northeastern
United States and has a long industrial history. West Central Phoenix is a much
newer Southwestern United States community inside the larger City of Phoenix.
These two communities also differ widely environmentally, climatically, and on
residential, commuting, and migration patterns. Additionally, Woburn
experienced its most significant population growth between 1940 and 1960, while
WCP grew more substantially in the 1970's and 1980's.

Woburn is the largest childhood leukemia cluster that has been
investigated by the CDC. The leukemia cluster led to thirteen environmental and
environmental health studies being conducted in Woburn from 1979 to 1997 by
the CDC, the Massachusetts Department of Public Health (MDPH) and
Department of Environmental Quality Engineering (DEQE). After nearly two
decades of cluster investigation, the Woburn researchers were only able to

provide a weak association between the excess of leukemia and the suspected

putative agent, contaminated well water exposure. The childhood leukemia
cluster in WCP, one of the largest in the western United States, was also
exhaustively studied over the period 1983 to 1997. Unlike Woburn, the most
significant excess of childhood leukemia was of the acute myelogenous type
rather than the acute Iymphocytic type. The lengthy and costly effort by the
Arizona Department of Health Services (ADHS), in investigating this excess of
leukemia, yielded only weak associations between childhood leukemia and
several potential putative agents out of the dozens that were evaluated.
However the final conclusion of the ADHS was that no specific putative agent
was ever identified that accounted for the cluster of childhood leukemia.

ln Woburn, researchers quickly focused on environmental exposures from
contaminated water supplies, largely because of a publicly perceived connection
between the water and leukemia. Conversely, investigators in WCP explored
virtually every possible risk factor for childhood leukemia, largely because of the
public expectation for the identification of a putative agent. Both the Woburn and
WCP clusters were highly publicized, with the Woburn story being made into a
book and feature film, "A Civil Action", and the WCP story being made into a
televised documentary. Thus, these studies demonstrate both the sociological
factors and political pressures that can influence the investigation of a cluster of
childhood leukemia and drive methodological decision making.

Due to the significance and contrasts of these clusters, the numerous
studies of these two large and notable childhood leukemia clusters were chosen

for careful review in this paper. My intention is to demonstrate the flaws in the

science and epidemiology of the cluster investigation efforts in Woburn and
WCP, and to advance potential solutions to the inherent problems of childhood
leukemia cluster analysis. I will also evaluate the studies for evidence that
supports an infectious etiology. And in this context, I will make recommendations
on how to advance research in the elucidation of an infectious etiology of
childhood leukemia.

Before examining the nature of the infectious etiology hypothesis one
must first understand the basic background of childhood leukemia. The
identification of the malignant hematopoietic disorders, collectively referred to as
leukemia, occurred in the early nineteenth century. The earliest observations of
patients who had marked elevation of their white blood cells was by European
pathologists who assigned the term “Weisses blut” or “white blood” as a
designation for the disorder. The pathologist Velpeau was likely the first to
clinically describe leukemia in 1827, but Virchow was the individual who assigned
the disease name in 1847. The term “leukemia,” which is derived from the Greek
words “leukos” meaning “white” and “haima” meaning “blood,” was used to
indicate the disease. Leukemia was differentiated into two subgroups, acute and
chronic, as early as 1889. Naegeli subdivided acute leukemia into the myelocytic
and Iymphocytic categories, subsequent to his description of the myeloblast cell
in 1900. The chronic subgroup was later differentiated in the same fashion as
acute leukemia. (Linet, 1985)

The modern classification for leukemia has changed little since the initial

discoveries. The major forms of leukemia are divided into four categories. The

terms myelogenous or Iymphocytic denote the cell type involved. Myelogenous
(also called nonlymphocytic, granuleocytic, myelomonocytic, myelosclerotic,
myelosis, myeloid, myelocytic, or myeloblastic) and Iymphocytic (also called
lymphoid, lymphatic, lymphogenous or lymphoblastic) leukemia have an acute or
chronic form. Thus, the four major types of leukemia are as follows: acute
lymphocytic leukemia (ALL), lCD-9 codes 204.00 and 204.01; chronic
lymphocytic leukemia (CLL), lCD-9 codes 204.10 and 204.11; acute
myelogenous leukemia (AML), lCD-9 codes 205.00 and 205.01; and chronic
myelogenous leukemia (CML) lCD-9 codes 205.10 and 205.11. (Note: lCD-9
code extensions .00 & .01 indicate no mention of remission and .01 & .11
indicate leukemia in remission) The differentiation of the types of leukemia has
been a problem in the past with regards to epidemiological research. Many
epidemiologic studies have failed in this differentiation either because of lack of
recognition of epidemiologic differences or because of the rarity of leukemia.
The fact that leukemia is not very prevalent has made it difficult to obtain
adequate sample sizes for studies. For this reason, sampling is problematic
when all types of leukemia are combined and even more so when separated.
(Linet, 1985)

In spite of the difficulties in obtaining samples for leukemia research, it is
critical that studies focus on individual types. Each of these main types of
leukemia has unique features, and more importantly, unique etiologies. The
basic pathophysiology and epidemiology of the four categories of leukemia will

be further discussed, but the emphasis will be placed upon those types most

commonly experienced by children, and specifically on ALL. The term childhood
leukemia is a general term encompassing the leukemias which afflict children,
and the acute leukemias are those most likely to be found in child populations,
while the chronic forms are rarely, if ever, observed in children. Therefore,
studies on childhood leukemia typically involve ALL and AML, but ALL is more
common and is responsible for the majority of childhood leukemia morbidity.
Additionally, due to potentially age—dependent causal pathways, cases of ALL in
people age 0 to 19 likely have a distinct etiological pathway as compared to adult
cases of ALL (Gilman, McNally and Cartwright, 1999). Therefore childhood ALL
(age 0 to 19) must be examined separately from other leukemias.

It is childhood leukemia of the ALL type to which the infectious etiology
hypothesis applies. The infectious etiology hypothesis has taken more than one
form, and the first was advanced by Leo Kinlen in 1988. Kinlen stated that in the
United Kingdom massive migration into previously remote and unpopulated
areas led to clusters of childhood leukemia. Kinlen asserted that:

1) The influx of individuals into an isolated community produces conditions in
which leukemia is more likely to occur.

2) Isolated communities (susceptible individuals) newly exposed to migrants
(infected individuals) from elsewhere could experience an unusual exposure to
some hypothetical infectious agent or to otherwise common infections for which
no immunity happens to exist.

3) This exposure, in turn, increases the risk of childhood leukemia.

4) Consequently, isolated communities may experience “mini—epidemics”
because they are too small to maintain common infections in endemic form,
which could lead to a cluster of ALL (Kinlen, 1988).

Melvin Greaves’ infectious etiology hypothesis takes a slightly different
form. Greaves hypothesized that ALL, which is almost exclusively responsible
for the childhood peak (3-5 years of age), may be due to two separate genetic
events:

1) spontaneous mutation of a pre-B-cell in utero

2) proliferation of the mutated pre-B-cell when exposed to a later antigenic
challenge (i.e. an infectious agent)

Greaves stated that the early in utero and later childhood (age 2-3 being the
crucial period of cell promotion) mutagenic events lead to acute lymphocytic
leukemia. Older children who have delayed exposures to specific agents may
experience more vigorous B-cell proliferation, resulting in the second genetic
event leading to leukemia. (Greaves, 1988)

Freda Alexander provides an Aberrant Response Model (ARM) that
unifies the Kinlen and Greaves models. The ARM states that a substantial
portion of ALL cases arise as a rare host response to certain patterns of
exposure to common infectious agents. This is at the core of the various
infectious etiology hypotheses. (Alexander, 1993) Therefore the infectious
etiology hypothesis could possibly follow either the Kinlen or Greaves model or
some combination thereof. There also is the possibility that Kinlen's model may

occur in the reverse order with susceptible individuals moving from rural areas

and mixing with infected individuals in more urban areas. This could very well be
the case in most industrialized countries, such as the United States. Also
Greaves' model could possibly be varied to state that an in utero antigenic
challenge is the event which causes mutated pre-B cell proliferation. Thus, this
paper will explore the evidence in the literature and the studies at Woburn and
WCP to identify evidence that would support some facet of the infectious etiology
hypotheses.

Objectives:

1) To critically review and analyze the childhood leukemia studies conducted in
West Central Phoenix, Arizona and Woburn, Massachusetts and identify the
flaws in the scientific and epidemiological methods and study results. The
studies will be critiqued according to an epidemiologic investigation checklist
(e.g. potential biases, confounders, descriptive and analytic methods, validity of
statistical interpretation, etc.).

2) To examine the literature and the studies in West Central Phoenix, Arizona
and Woburn, Massachusetts to elucidate evidence for the Leo Kinlen and Melvin
Greaves infectious etiology hypotheses (Kinlen, 1988) (Greaves, 1988). This will
be done to determine if the studies' childhood leukemia cases fit a model (see
objectives below) that is consistent with an infectious disease etiology.

2.1) Evaluate if Woburn and West Central Phoenix were relatively isolated from
newcomers and did not experience a great deal of population mixing prior to the

ALL epidemic.

2.2) Evaluate if Woburn and West Central Phoenix experienced (a) an influx of
newcomers and (b) increased incidence of ALL in children (age 0 to 19) following
an influx of newcomers.

2.3) Evaluate if Woburn and West Central Phoenix experienced significant
mixing of newcomer and indigenous populations prior to and during the elevated
incidence of ALL in children (age 0 to 19).

2.4) Evaluate if the cases of ALL occurred primarily in children (age 0 to 19) born
in Woburn and West Central Phoenix and whose parents were indigenous to the
community (susceptibles) or vice versa in the case of susceptibles moving into
an infected population (urban to rural or rural to urban).

2.5) Evaluate if the increased incidence of ALL in children (age 0 to 19) Woburn
and West Central Phoenix was a result of an aberrant response to an infectious
agent(s) (in utero or in vivo) introduced by the newcomers.

2.6) Evaluate if susceptible indigenous children exposed to the infectious
agent(s) at the ages of 3-4 were more likely to develop ALL according to the
typical peak in ALL incidence at that age.

3) To advance recommendations on how to better evaluate childhood leukemia

clusters and how to better study the childhood ALL infectious etiology hypothesis.

Chapter 1

BIOLOGICAL ASPECTS OF LEUKEMIA

Human leukemia is a heterogeneous group of disorders characterized by
disruption of normal blood cell production. The normal process of blood cell
development, called hematopoiesis, occurs in the spongy bone marrow. A small
group of cells, the pluripotent bone marrow stem cells, are responsible for
making all the blood cells in the marrow. The stem cells eventually transform into
the specific blood cells by a process of differentiation and proliferation. The
process begins with the differentiation of the self-renewable pluripotent stem cells
into multipotent stem cells, myeloid or lymphoid, followed by differentiation into
the committed stem cells. The proliferating and differentiating cell process then
continues until functioning blood cells are produced. The cells produced include:
erythrocytes, platelets, neutrophils, eosinophils, basophils, monocytes, and T and
B lymphocytes. When these cells are fully formed and able to function, they
leave the marrow and enter the bloodstream.

The leukemia disorders cause blood cell dysfunction through malignant
expansion of aberrant blood cells. The malignant transformation can occur at
any level of the proliferating stem cell process. The subsequent representation
of transformed cells depends entirely on which level of precursor cell is involved
in the malignant transformation. Thus, depending on which stem cell is involved,
different blood cells will be effected and rendered dysfunctional. This variability

accounts for the different categories of leukemia, which are identifiable by the

10

presence of specific transformed blood cells. If the pluripotent stem cells are the
source of the malignancy, a mixed lineage form of leukemia will be observed.
The rapid production of the abnormal blood cells causes the pluripotent stem
cells to limit their expression of differentiation to the particular cell lineage. Thus,
abnormal cells proliferate and few healthy cells are produced. This reduction of
normal circulating blood cells precipitates the clinical consequences of the
various types of leukemia. Acute leukemia is a rapidly progressing disease that
affects mostly cells that are not yet fully developed or differentiated. These
immature cells cannot carry out their normal functions, which readily debilitates
an individual. Chronic leukemia progresses slowly and permits the growth of
greater numbers of more developed cells than does acute leukemia. These
more mature cells can carry out some of their normal functions, lessening the
impact on the leukemia patient (Mauer, 1990).

The acute leukemias are characterized by the presence of malignant blast
cells. The term “blasts” refers to the earliest marrow cells identified by the light
microscope. Blasts represent approximately 1 percent of normally developing
bone marrow cells. The majority of blast cells are myeloblasts, which are cells
that develop into neutrophils. The remaining blast cells are lymphoblasts, which
are cells that are part of lymphocyte development. In the acute leukemias,
malignant blast cells, similar in appearance to normal blast cells, accumulate in
huge quantities of as much as 80 percent of all marrow cells. Abnormal
myeloblasts accumulate in AML, and abnormal lymphoblasts accumulate in ALL.

The leukemic blast cells fail to function as normal blood cells and act as a

11

blockade to the production of normal marrow cells, leading to a deficiency of
erythrocytes (anemia), platelets (thrombocytopenia), and normal white cells
(especially neutrophils, i.e., neutropenia) in the blood.

Within the individual acute leukemia categories, there exist several
subtypes. The size and appearance of the leukemia cells are the criteria utilized
in the subtype differentiation. Additionally, the leukemic cells are classified by
their morphological similarity to normal B or T lymphocytes. There are three
principal subtypes of ALL which can be distinguished microscopically, and they
have been labeled L1 (small), L2 (large), or L3 (large with other phenotypic
features). Approximately 85% of the ALL cases in children present small
malignant cells that are related to the B lymphocytes and are of the L1 subtype.
In adult cases of ALL the B lymphocyte cells are typically of the L2 subtype. The
L3 type is very rare and accounts for only 3% or less of ALL cases in both
children and adults. ALL cases of T cell lineage are more commonly found in
adolescent and adult populations. There are seven principal sub-classifications
of AML, which include the following: M1 (myeloblastic, without maturation), M2
(myeloblastic with maturation), M3 (promyelocytic), M4 (myelomonocytic), M5
(monocytic), M6 (erythroleukemic), and M7 (megakaryocytic). This sub-
classification system is based upon the dominant leukemic cells in the marrow at
the time of diagnosis of the leukemia. These sub-classifications of ALL and AML
are not necessarily a major concern from an epidemiologic perspective, but they

do bear noting as they have unique diagnostic and prognostic implications.

12

Distinguishing between ALL and AML is quite obviously a necessity. Not
only do ALL and AML have distinctive etiologies, but they also present different
clinical features, as do their subtypes. ALL is frequently associated with central
nervous system degradation and dysfunction while AML is not. Also, ALL has a
higher recovery, and consequently lower mortality, than that of AML. The
individual sub-classifications are significant as well. The L1 morphologic type of
ALL has a better outcome than the cases involving the L2 subtype. The
subtypes of AML also demonstrate different prognostic and clinical features. The
M3 cell type has been associated with the longest duration of remission among
the AML subtypes. The M4 morphologic type is associated with a higher
incidence of central nervous system leukemia, presenting with isolated tumors.
Hyperplasia and skin nodules are more common in the M5 subtype of AML, and
simultaneous or subsequent development of meningeal leukemia is more likely to
occur with this cell type. The white blood cell counts in AML patients have been
found to be higher in both the M4 and M5 cell types than in M2 or M3. There is
also evidence to indicate that there is a higher incidence of the M6 subtype in
secondary leukemias associated with radiation or chemotherapy exposure. The
M7 cell type appears to have the poorest prognosis as it has a low rate of
remission followed by rapid relapse and typically a fatal outcome (Mauer, 1990).
The variations in clinical presentation and prognosis among the various sub-
classifications of ALL and AML may provide clues to subtle etiological
differences, but achieving any significant results from epidemiological research in

this area is very difficult. The rarity of ALL and AML collectively makes research

13

problematic, but the successful study of subtype etiologies can be affected
through large population-based case-control studies that look at large regions or

nations rather than focusing on small clusters of disease.

14

Chapter 2

ETIOLOGY OF ACUTE LYMPHOCYTIC LEUKEMIA

A. Descriptive Epidemiology of Acute Childhood Leukemia

The acute leukemias, ALL and AML, are not only pathophysiologically
distinctive diseases, but they are also distinctive in their epidemiology. This fact
is demonstrated by a population study conducted in the Province of
Saskatchewan in Canada. The objective was to describe and analyze the
incidence and survival data, on the childhood and adolescent leukemias,
available from the Cancer Registry of the Saskatchewan Cancer Foundation.
The provincial population is small, but the findings of the study are among the
most reliable estimates of childhood leukemia time, cohort, and survival trends
that were available at the time. The Cancer Registry utilized is believed to be the
oldest, continuously collected cancer data source for a geographically defined
population. The data utilized included all incident leukemia cases, ages 0-19,
which occurred over a sixty year period from 1932-1991. Through review of
incidence trends and univariate and multivariate survival analysis, it was
observed that age-adjusted incidence of ALL is 3.1 per 100,000 while for AML
(ANLL) it is only 0.6 per 100,000. The findings of the study also revealed that the
incidence of ALL had been steadily rising over the sixty-year period, but it was
acknowledged that previous studies did not necessarily support this conclusion.

Hence a certain amount of controversy exists on the issue.

15

The demographic features of the Saskatchewan leukemia study
population are congruent with previous research. The occurrence of ALL in
these children was very significant with 46% of the cases occurring in children
before age five and 74% occurring before age nine. Additionally, incidence of
ALL has remained consistently higher in male children, and this is evidenced
among varied racial and geographic boundaries. The results of the AML data
analysis identified two key differences between ALL and AML. First of all, the
AML data did not demonstrate any statistically significant increasing time trends,
and the risk by age was considerably different. Compared with the ALL figures,
AML risk was as high or higher for ages 10-19 as for ages 04. Also, ages 5-9
had the lowest risk of all the age groups. Considering that the risk rates for ALL
decrease from age 0-19, AML and ALL are certainly unique. Another
observation in the data seems to indicate that while AML incidence in male
children is still higher than in females, the slightly increasing incidence in females
appears to be converging with the steadily decreasing incidence in males. A
dramatic increase in survival rates was also observed, especially in ALL. Period
effects likely account for the higher mortality in the earlier portion of the sixty year
interval, due to less efficacious therapeutic interventions being available. It was
also determined that, in ALL cases, children under five have a much better
survival outlook than do older children and adolescents. The inverse of this is
true for AML. (Wang and Haines, 1995)

The incidence, incidence trends, mortality and survival rates provided in

the Saskatchewan study are further supported by the statistics provided from the

16

Surveillance, Epidemiology, and End Results (SEER) nine standard cancer
registries in the United States. The data collected from SEER is representative
of the child population of the United States (ages 0-14), and is inclusive from
1973 to 1995. The SEER incidence information on ALL (see Table 1) indicates
that the annual incidence rate of males has been consistently higher, with few
exceptions, than females in each of the three age brackets, over the 23 year
period. The average annual incidence from 1991 to 1995 (see Table 2) also
shows the higher burden of ALL in male children versus female children. The
stark gender differences with ALL are not nearly so pronounced in AML. The
SEER data (see Table 3-4) indicate a quite different trend in AML. It appears
that AML annual incidence is actually higher in females age 0-4 and slightly lower
in the age 5-9 and age 10-14 categories. Overall, it appears that gender

differences are gradually disappearing in AML incidence rates.

17

Table 1:
Annual Incidence of ALL in the U.S. based on 9 Standard SEER Registries

Male Female

 

Year Age Age Age Age Age Age
0-4 5-9 10-14 0-4 5-9 10-14
1995 6.79 3.59 1.76 5.01 2.17 1.39
1994 5.98 2.22 1.44 4.84 2.09 1.40
1993 4.49 2.25 1.79 5.15 3.65 1.17
1992 5.79 3.17 2.05 4.19 1.66 1.92
1991 7.34 2.94 1.63 5.46 2.73 1.47
1990 5.06 4.09 1.44 5.53 2.98 1.51
1989 6.28 3.54 2.97 5.54 3.95 1.43
1988 6.52 4.15 1.01 5.89 1.93 0.53
1987 5.79 2.57 2.44 5.35 2.33 2.02
1986 6.50 3.81 1.54 4.90 1.87 1.21
1985 6.07 3.29 2.11 5.39 2.42 1.69
1984 4.14 3.49 2.54 6.87 2.35 1.77
1983 4.74 3.06 2.35 4.96 2.14 0.61
1982 5.64 2.07 2.31 3.81 2.70 1.69
1981 4.54 2.45 1.70 5.75 3.90 1.30
1980 6.51 2.25 1.59 4.63 2.75 0.83
1979 4.68 2.69 1.68 4.63 1.79 1.17
1978 4.79 2.99 1.86 5.70 2.24 0.80
1977 5.79 2.59 1.80 4.95 2.21 1.21
1976 5.53 3.25 0.82 4.26 2.54 1.60
1975 5.11 1.98 0.99 3.34 1.33 0.72
1974 6.08 3.23 2.33 3.95 2.32 1.43
1973 5.05 2.33 1.43 4.02 1.28 0.74

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Table 2:

AveragAnnual Incidence of ALL in the United States from 1991 to 1995

| Sex Age 0-4 Age 5-9 Age 10-14
I Male 6.07 2.84 1.73

| Female 4.93 2.46 1.47 W

 

 

 

 

 

 

 

Note: The crude/age-specific incidence rates are based on a standard
population of 100,000.

18

Table 3:
Annual Incidence of AML in the U.S. based on 9 Standard SEER Registries

Table 4:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Male Female
Year Age Age Age Age Age Age
0-4 5-9 10-14 0-4 5-9 10-14
1995 0.64 0.00 0.66 0.78 0.11 0.46
1994 0.52 0.22 0.33 0.44 0.23 0.35
1993 0.73 0.67 0.45 1.21 0.35 0.12
1992 0.21 0.57 0.80 0.66 0.36 0.36
1991 0.32 0.68 0.23 1.34 0.47 0.61
1990 0.75 0.45 0.12 0.68 0.36 0.63
1989 0.44 0.23 0.37 0.69 0.60 0.39
1988 0.22 0.23 0.38 0.59 0.36 0.13
1987 0.45 0.23 0.39 0.36 0.73 0.27
1986 0.68 0.24 0.51 0.24 0.25 0.27
1985 0.57 0.85 0.87 0.36 0.26 0.78
1984 0.23 0.12 0.12 0.60 0.13 0.63
1983 0.58 0.13 0.35 0.73 0.67 0.12
1982 0.12 0.00 0.81 0.49 0.27 0.48
1981 0.36 0.26 0.57 0.37 0.00 0.36
1980 0.49 0.25 0.91 0.64 0.39 0.47
1979 0.38 0.61 0.56 0.53 0.77 0.70
1978 0.65 0.36 0.33 0.81 0.12 0.23
1977 0.39 0.35 0.21 0.41 0.49 0.33
1976 0.13 0.46 0.51 0.55 0.36 0.53
1975 0.89 0.35 0.40 0.67 0.24 0.52
1974 0.81 0.75 0.64 0.56 0.90 0.55
1973 0.59 0.55 0.48 1.39 0.28 0.87

 

 

 

 

 

 

 

 

Avmge Annual Incidence of AML in the United States from 1991 to 1995

 

 

 

| Sex Age 04 Age 5-9 Age 10-14
| Male 0.48 0.42 0.50
1 Female 0.88 0.30 0.38 I

 

 

 

 

Note: The crude/age-specific incidence rates are based on a standard
population of 100,000.

19

The SEER incidence rates show the gender differences in both ALL and
AML from 1973 to 1995, and the data also demonstrate time trends in incidence
similar to those observed in the 1995 Wang and Haines childhood leukemia
study. The SEER ALL incidence trends (see Table 5) show an increase in males
and a significant increase in females from 1973 to 1995, but the incidence for
both genders has declined somewhat since 1991. The SEER numbers for AML
are much more encouraging (see Table 6). From 1973 to 1995 a remarkable
decline in overall AML incidence was observed in all children. The bulk of the
decline appears to have occurred in the past decade.
Table 5:

SEER Incidence Trends of ALL (Age 0-14) in the United States: % Charge

 

 

 

I F rom-To Male Female I
L 1 973-1995 +5.66 +22.78 I
I 1991-1995 -5.21 -3.59 I

 

 

 

Table 6:
SEER Incidence Trends of AML (Age 0-14) in the United States: % Change

 

 

 

| F rom-To Male Female |
I 1973-1995 -38.40 48.49 J
I 1991-1995 -1 8.75 -37.09 I

 

 

 

Note: The incidence trends could be slightly misleading as the leukemia
immunophenotyping methods have improved since 1973, thereby reducing
misclassification errors.

The SEER mortality trend data provide an encouraging prognosis for the
newly incident ALL and AML cases. ALL has experienced greater than a 60%

decline in mortality from 1973 to 1995, and the percent change from 1991 to

1995 indicates a continuation of this trend (see Table 7). The percent change in

20

mortality for AML is not quite so profound, but AML mortality has declined
approximately 35% over the 23-year span (see Table 8). The concern with AML
is with the recent change in mortality in female children from 1991 to 1995.
Mortality has increased nearly 2% (though this slight change could be due to
improved diagnostic methods rather than an actual increase) in this time period,
while male children’s mortality declined nearly 4%.
Table 7:

SEER MortalitiLTrends of ALL (Age 0-14) in the United States: % Chaege

 

 

 

| From-To Male Female |
I 1973-1 995 -64.62 -60.39 I
I 1991-1995 -7.96 -10.73 I

 

 

 

Table 8:
SEER Mortality Trends of AML (Age 0-14) in the United States: % Change

 

 

 

| From-To Male Female |
I 1973-1995 -35.31 -34.69 I
| 1991-1995 -3.80 +1.84 |

 

 

 

Note: The mortality trends could be slightly misleading as the leukemia
immunophenotyping methods have improved since 1973, thereby reducing
misclassification errors.

The SEER relative survival rate information echoes the observations from
the mortality trends for the acute leukemias (see Table 9-10). ALL 1-year and 5-
year survival rates are greatly improved in the 1989-1994 time frame as opposed
to survival rates in 1975. The 1975 survival rates were not available for AML, but
considering the decline in AML mortality over the respective time period, it is

likely that survival rates for AML have followed the same trend as that seen in

ALL. Comparison of the most recent survival rate data for ALL and AML

21

indicates that AML is significantly more lethal than ALL, as the survival rates for
ALL are nearly 30% greater than for AML. This could be a result of the greater

prevalence of ALL in children and the subsequent increased focus on treatment

and prevention for ALL, from a public health standpoint.

Table 9:

SEER Relative Survival Rates of ALL (Age 0-14) in the U.S.: Percentage

 

 

 

 

 

 

 

 

 

 

 

Year Male Male Male Female Female Female
1 Year 5 Year 20 Year 1 Year 5 Year 20 Year
1989-1994 95.58 79.93 59 95.59 80.48 64
1 975 88.18 40.43 28.53 88.17 59.64 45.36
Table 10:

SEER Relative Survival Rates of AML (Age 0-14) in the U.S.: Percentage

 

 

Year Male Male Male Female Female Female

1 Year 5 Year 20 Year 1 Year 5 Year 20 Year
1989-1994 69.67 50.44 NIA 68.43 37.99 N/A
1975 NIA N/A NIA N/A N/A NIA

 

 

 

 

 

 

 

 

 

 

 

The incidence information regarding ALL and AML, seen in the data from
the Saskatchewan study and the SEER, is fairly consistent globally. The
following table (see Table 11) was excerpted from the British Journal of Cancer
and demonstrates the global similarities in childhood leukemia incidence. The
incidence rates are not broken down by acute leukemia sub-type, but they show
how the leukemic disorders afflict children nearly identically in the different
geographical regions. The ALUAML ratio provided in the table also reinforces

the much greater prevalence of ALL throughout the world.

22

Table 11:

Comparison of Reports on Childhood Leukemia in Europe, NA 8. Australia

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Country Period Incidence Incidence Incidence Adjusted ALU

Age 04 Age 5-9 Age 10-14 Age 0-14 AML

Ratio
Sask,Can 1976-90 6.9 3.9 2.3 4.4 6.7
US/SEER 1973-82 6.8 3.5 2.3 4.4 5.4
Fla,US 1981-86 7.0 3.5 2.6 4.4 5.2
Australia 1977-82 7.4 3.6 2.2 4.6 5.3
Torino,lt 1967-86 7.7 4.2 2.6 4.8 5.1
Finland 1971-82 5.4 3.0 2.7 3.9 5.0

GB(male) 1974-76 7.0 3.7 2.3 - -

W.Germ 1980-86 6.5 3.5 2.2 4.4 6.6
Denmark 1980-84 5.7 2.7 2.8 3.9 5.3
Manch,UK 1971-83 5.7 3.2 2.2 3.8 5.7
LRF-UK 1984-86 6.4 3.0 1.7 3.9 5.1
The Neth. 1973-86 6.1 3.1 1.9 3.9 8.3

 

 

 

 

 

 

Note: The incidence rates are based on a standard population of 100,000.

Source: British Journal of Cancer 1989: 59: 100-105

While macro geography demonstrates tremendous similarities in

childhood leukemia incidence across the globe, there are indications that the

place where childhood leukemia occurs may actually contribute to the

leukemogenesis. There has been substantial evidence accumulated that

clustering of childhood leukemia cases occurs within geographical areas, and

many hypotheses have been advanced to explain this phenomena.

Potential Causal Factors of Acute Childhood Leukemia

1. Environmental Factors

Numerous physical and chemical agents have been implicated in the

 

epidemiological origin of childhood leukemia. Benzene has been identified as a

potential causal factor in childhood AML but not ALL (Alexander, Leon and

23

Cartwright, 1996). An ecological study was conducted in the municipalities of
Sweden, over an 11-year period from 1975-1985, to determine if environmental
exposure to gasoline was associated with leukemia incidence in children and
young adults (ages 0-24). Leukemia incidence rates were obtained from the
National Swedish Cancer Register and compared to the number of cars per area.
As expected, an association between car volume and AML incidence was found,
but no association was found for ALL. The incidence of AML in municipalities
with more than 20 cars per km2 was 5.5 (95% CI 4.4-6.8) per 1 million person-
years as compared to 3.4 (95% CI 1.9-5.7) in municipalities with less than 5 cars
per kmz. This association between childhood AML and car density is likely
attributable to benzene exposure from gasoline vapors and exhaust gases
(Nordlinder and Jarvholm, 1997). Another potential environmental contributor to
the onset of childhood leukemia is radon. Radon exposure was significantly
correlated with AML in a study conducted in Great Britain. The study analysis
concluded that 6-12% of the AML incident cases in the UK could be attributed to
radon and between 13-25% of world cases may be caused by radon (Henshaw,
Eatough and Richardson, 1990). Radium-226 contaminated drinking water in the
region of Ellweiler, Rheinland-Pfalz, Germany has also been implicated in
leukemogenesis. It is believed that prenatal exposure and dose accumulation of
the Radium-226 was the cause of the excess childhood leukemia cases in the
Ellweiler region (Hoffman, Kranefeld and Schmitz-Feuerhake, 1993).

Another environmental exposure, that may put children at risk for

leukemia, is the presence of pesticides in the home and at large. A case-control

24

study of children, under the age of 15, was undertaken in the Denver area
between 1976 and 1983 to determine if an association existed between home
pesticide use and childhood cancer. 252 cases (70.8% of eligible cases) and
222 control subjects (79.9% of eligible controls) were obtained for the study in
which the cases had home pesticide use and the controls had no home pesticide
use. Regarding leukemia, the study results indicated a significant risk in the
homes which used pest strips (OR from 1.7 to 3.0) and elevated risk in homes
which had home extemiinations or yard treatments (OR from 0.3 to 1.1). The
varying odds ratios arise from consideration of differing exposure periods
(prenatal, less than 2 years, and greater than 2 years of age) but are inclusive of
all of the data collected. Although the exposure measures were rather crude, this
study does indicate an association between home pesticide use and
leukemogenesis (Leiss and Savitz, 1995). A critical review of 31 previous
residential or occupational pesticide exposure epidemiologic studies published
between 1970 and 1996 reported that relative risks to pesticide exposure and
development of childhood cancers were modest. The review found that childhood
leukemia was more strongly associated with home pesticide use than with
professional pesticide use or yard treatments (Daniels, Olshan and Savitz, 1997).
A recent large case-control study in Germany evaluated 1,184 cases of
childhood leukemia and 234 with non-Hodgkin's lymphoma, and the data
provided some evidence for an increased leukemia risk for children living on
farms who are exposed to agricultural pesticides (OR 1.5, 95% CI: 1.0, 2.2).

Also a significant trend (p = 0.02) was identified between frequency of parental

25

use of household insecticides and risk of childhood leukemia. (Meinert et al.,
2000)

The exposure to pesticides prenatally and early in childhood seems to be
associated with onset of leukemia, and studies indicate a similar association in
children who were exposed to alcohol in utero or cigarette smoke in vivo. A
case-control study of childhood AML was formulated, from data obtained from
187 matched pairs, and was examined for evidence of associations between
parental cigarette smoking and alcohol consumption and subsequent onset of
AML. Alcohol consumption during pregnancy appears to associated with
increased risk of AML in children less than 2 years of age (OR 3.00; 95% Cl 1.23
to 8.35), but little evidence was found to associate cigarette smoking and AML
(Severson et al., 1993). A similar study was conducted utilizing 302 cases
matched to 558 controls, and parental alcohol consumption and cigarette
smoking were again the subject of interest. The study found slight associations
between maternal alcohol consumption during pregnancy and elevated risk of
ALL and AML, but maternal prenatal cigarette smoking was actually found to be
negatively associated with infant leukemia risk. However, a significant
association was found between maternal prenatal alcohol consumption and AML
of the M1 or M2 morphology (OR 7.62; 95% Cl 203-2864). Thus, in utero
alcohol exposure appears to contribute to the leukemogenesis of certain AML
subtypes (Shu et al., 1996). Prenatal exposure to Phenobarbital and other drugs
is also suspected of contributing to leukemogenesis, but there is little evidence to

support a strong association or causal relationship. Association between

26

childhood leukemia and administration of intramuscular vitamin K to newborns, to
prevent hemorrhaging, is also controversial, but some evidence from the UK
supports the hypothesis. Two studies on the subject found no association;
however, one study found “borderline significance” while another in northern
England revealed a 2-fold risk of ALL in children age 1-6 who received the
intramuscular vitamin K at birth (Voelker, 1998).

One more environmental exposure that has generated controversy is the
association between childhood leukemia and exposure to electromagnetic fields
(EMF). EMF exposure can occur as a result proximity to power lines, certain
household wiring codes, and other magnetic field producing electrical items
within the home or in a child’s overall environment. Numerous studies have been
undertaken on the subject, but the recent studies appear to be the most relevant.
The landmark case-control study conducted by Linet et al. (1997), in the United
States, enrolled 638 children with ALL and 620 controls. Data were amassed
through exhaustive measurement of magnetic fields in the previous and current
homes of the children, by blinded collectors, and a computer algorithm was
utilized to assign wire-code categories. The findings of the study revealed an
odds ratio for ALL of 1.24 (95% CI 0.86-1.79). Therefore, the results provide little
if any evidence that ALL is associated with EMF exposure (Linet et al., 1997). A
study of proximity to power lines in relation to childhood leukemia was
undertaken in Greece in 1993-1994, and its findings were in concordance with
Linet et al. (1997) (Petridou et al., 1997). Many have seen the Linet et al. (1997)

study as the conclusive debunking of the EMF hypothesis, but the hypothesis

27

should not be ruled out altogether. Less elaborate but more recent studies have
indicated that an association may still exist. The risk of childhood leukemia and
proximity to high-voltage transmission lines in three urban districts of Taiwan was
examined. Data from 1987 to 1992 showed a standardized incidence ratio of
2.69 (95% CI 1.08-5.55) of childhood leukemia in children living within 100
meters of the power lines compared to the total child population of Taiwan (Li,
Lee and Lin, 1998). EMF measurements were taken from 1992 to 1996 in two
population-based case-control studies on childhood leukemia in the northwestern
part of Germany and in Berlin. A total of 176 case subjects were utilized and
matched to 414 controls to determine if an association existed between EMF
exposure and childhood leukemia. The results were not statistically significant,
but they did reveal a trend that seemed to support the EMF hypothesis of
association to childhood leukemia (Michaelis et al., 1998). Two detailed EMF
studies conducted in 2000 revealed that distance from power lines, exposure
index, peak exposure, threshold values, measures of short-term variability, and
spot measurements demonstrated little or no association with risk of childhood
ALL (Kleinennan et al., 2000) (Auvinen et al., 2000).
2. Genetics

There is little evidence to show an association between heredity and
childhood leukemia, but heredity remains an interest in leukemogenesis
epidemiology. The impetus for researching genetic factors arises from the
observed examples of familial aggregation. A twin study undertaken in the UK

examined the various determinants of concordant childhood leukemia in

28

monozygotic twins. The results of the analysis indicated that the genetic
determinants of childhood leukemia operate before the time of cleavage, but the
study cautioned that these determinants alone were insufficient for childhood
leukemia to ensue (Knox, Marshall and Barling, 1984). An earlier twin study
found that the childhood leukemia concordance rate among monozygotic twins
was approximately 25% (Jackson, Norris and Klauber, 1969). Certain genetic
defects that cause some cancers are inherited and therefore are associated with
familial predisposition to cancer. However, research indicates that the genetic
defects, such as observed chromosomal abnormalities, that increase the
likelihood of leukemia are actually acquired rather than inherited (Cline, 1994).

A recent population-based cohort study was undertaken in Denmark to
assess the role of birth characteristics and sibling patterns in the development of
childhood leukemia. The study established a cohort of children whose mothers
had been born between April 1935 and March 1978, and children who developed
ALL or AML between April 1968 and December 1992. The cohort included
approximately 2.0 million children from which the childhood leukemia cases were
identified. The results of the birth order analysis were not significant for ALL and
only slightly significant for AML, but the birth weight findings were substantial.
The overall risk of ALL increased by a factor of 1.46 (95% CI 1.18-1.81; P=.0005)
for each kilogram of increase in birth weight, and the overall risk of AML
increased by a factor of 2.14 (95% CI 1.19-3.85; P=.009) for each kilogram of
increase in birth weight. The possible explanation for this effect was that with

increasing birth weight there is greater cell proliferation, which places a larger

29

number of precursors at risk for malignant transformation (Westergaard et al.,
1997). The evidence in the literature regarding heredity and familial aggregation,
as well as the confounding effects of siblings sharing common environments,
makes it is difficult to implicate heredity as a contributing factor in childhood
leukemia etiology. However, the possibility of hereditary involvement should not
be dismissed, as it could play even a small role in the epidemiology of the
disease state of childhood leukemia.
3. Radiation

There has been substantial evidence accumulated that clustering of
childhood leukemia cases occurs within geographical areas (Laurier and Bard,
1999), and many hypotheses have been advanced to explain this phenomena.
Clustering of childhood leukemia has been observed in various areas, but the
clustering around nuclear facilities has probably generated the most research. A
metanalysis of investigations into the occurrence of childhood leukemia near
nuclear facilities in the United Kingdom identified space-time clustering in the
vicinity. Following an initial report of childhood leukemia clustering near one
nuclear plant in northern England, numerous subsequent investigations were
undertaken. The investigations revealed that there is a consistent pattern of a
slightly elevated risk of leukemia for children living near nuclear establishments.
However, the author acknowledges that these findings do not imply any causal
relationship, and the presence of the nuclear plants may be nothing more than a
convenient etiological scapegoat (Beral, 1990). A study conducted in northern

Germany identified a cluster of six cases of childhood leukemia, between

30

February 1990 and December 1995, among the residents of the small rural
community of Elbmarsch. Five of the cases occurred in only a 16 month period
between February 1990 and May 1991. All of the cases lived within 4.5 km of
Germany’s largest capacity nuclear boiling-water reactor. The study was only
able to identify the cluster and did not attempt to rule out other potential risk
factors (Hoffmann et al., 1997). Clustering of childhood leukemia, within a ten-
mile radius of the U.S. nuclear plant at Three Mile Island, was observed over the
period of 1975 to 1985. The cases identified had been or were suspected to
have been exposed to varying levels of background gamma radiation from the
nuclear facility. The study found an elevated odds ratio (OR=2.4) among incident
cases of leukemia for children under age 15 who had exposure, but the
association was not statistically significant. Also, the data on childhood leukemia
mortality showed little or no association to background gamma radiation
exposure (Hatch and Susser, 1990).

Another area of concern in childhood leukemogenesis lies in paternal
preconceptional exposure to radiation. Kinlen undertook a matched case-control
study in Scotland to determine if paternal preconceptional radiation exposure in
the nuclear industry was at all related to childhood leukemia and non-Hodgkin’s
lymphoma. The findings of the study found no statistically significant excess of
either disease at any level of paternal radiation exposure (Kinlen, Clarke and
Balkwill, 1993). A more recent case-control study, of cancer incidence in the

offspring of radiation workers in Great Britain, confirmed the findings of Kinlen et

31

al. (1993). The results are not conclusive though, as the numbers were too small
to generate reliable estimates of elevated risk (Draper et al., 1997).

Between 1951 and 1958, twenty-six nuclear tests were conducted in
southwestern Utah. A retrospective cohort study was undertaken twenty years
later to determine if radiation from nuclear fallout was associated with increased
incidence of childhood leukemia in the effected areas. Mortality records for
children under 15 years of age were evaluated and the subjects were assigned to
either a low exposure or high exposure cohort. For some unknown reason, the
low exposure cohort in the high-fallout areas actually had a leukemia mortality
rate that was half that of the rest of Utah and the United States. The mortality in
the high exposure cohort increases by 2.44 times (95% Cl 1.18 to 5.02) to just
slightly higher than that of the United States (Lyon et al., 1979). A more recent
case-control study evaluated the same Utah population utilizing estimates of
dose to bone marrow computed from fallout deposition rates and the residence
locations of the subjects. Using 1,170 leukemia deaths as cases and 5,330
controls who died from other causes, only a weak non-significant association was
found between bone marrow dose and all types of leukemia. However, trend
analysis revealed a statistically significant (p<0.05) excess risk in the individuals
of the high-dose group, with ALL and AML, who were under 20 years of age and
died before 1964 (Stevens et al., 1990). While these Utah studies do not provide
strong evidence for the role of radiation as a risk factor in childhood leukemia, an
analysis of the data collected through the Leukaemia Registries in Hiroshima and

Nagasaki Japan provides stronger evidence. Following the dropping of atomic

32

bombs on these two cities, tremendous increases of both the acute and chronic
leukemias occurred in the surviving population. The acute leukemias witnessed
the greatest increase, especially in children less than age 15 at the time of the
bombing who experienced at least a 100 rad total air dose. From 1950 to 1955
the annual incidence rate of acute leukemia in this group soared to over 90 per
100,000. From 1955 to 1960 the rate fell off to approximately 50 per 100,000.
Similar trends were observed in the other age groups and in the chronic types
(Greaves and Chan, 1984). The gross excess of childhood leukemia and other
leukemias, as a result of the radiation exposure, certainly implicates radiation as
a potential inductive agent even if it is not at the etiological core.
4. Infectious Etiology

The presence of these various clusters of childhood leukemia around
nuclear sites would seem to implicate the nuclear facilities in the etiology, but this
is likely not the case. There could quite possibly be other underlying factors that
lead to the presence of the clusters as childhood leukemia clusters are observed
in numerous locations that are completely isolated from nuclear radiation
exposure. This fact led to Kinlen’s population mixing (infective) hypothesis,
which proposes that “the bringing together of rural and urban groups, or different
socioeconomic groups, can increase the frequency of childhood leukemia
through increased contact between susceptible and infected individuals.
Geographical isolation, high social class, and frequent population turnover all
promote susceptibility” (Kinlen, 1997). Kinlen’s hypothesis was first brought forth

in 1988 and 1990 in two separate studies. In Kinlen’s 1988 paper he challenged

33

the unfounded accusations that the childhood leukemia clusters around Sellafield
and Dounreay were a result of proximity to the only British nuclear power stations
that also reprocessed spent nuclear fuel. Kinlen hypothesized that the clusters
near Dounreay and Sellafield had nothing to do with radiation and were a result
of the same causes that explained clusters elsewhere. The three key elements
of Kinlen’s hypothesis in this study were as follows: (1) that influxes of population
into rural and isolated areas are conducive to epidemics of certain infections, (2)
that Sellafield and Dounreay are extreme examples of isolation and population
influx, and (3) that childhood leukemia is an aberrant response to some infectious
agent(s). Kinlen tested his hypothesis by examining the mortality records of the
New Town of Glenrothes, Scotland, which was identified as the only other rural
area that received a large influx at the same time as Sellafield and Dounreay.
The results of the study revealed that a significant increase of leukemia below
age 25 was found, with 10 cases observed compared to the expected 3.6 cases.
The greatest excess was found in children below the age of 5, with 7 cases
observed compared to the expected 1.5 cases (Kinlen, 1988). Kinlen’s 1990
study found further evidence to support his hypothesis. Kinlen et al. examined
the mortality data for childhood leukemia in rural and urban overspill areas,
where immigration into the towns was rapid and on a large scale, in Great Britain
from 1946 to 1985. The data indicated that the hypothesis was not limited to a
particular cell type of leukemia, and that the distribution seemed to follow the
same increased social contact pattern as in feline leukemia in house cats (which

is the result of a Ieukemogenic virus). It was also noted that due to the lack of

34

appreciable space—time clustering, if childhood leukemia is of an infective origin
then the spread of infection is primarily done by trivially infected subjects rather
than by leukemia subjects (Kinlen, Clarke and Hudson, 1990).

A study of leukemia clusters in Great Britain supported the Kinlen
hypothesis in its findings. The study focused on the space-time interactions and
geographical concentrations of childhood leukemia in England, Wales, and
Scotland. The data examined included all registered cases of leukemia and
lymphoma between 1966 and 1983 in children aged 0-14. Statistically significant
excesses of case pairs occurred jointly within 0.5 km of each other, and these
clusters were observed mainly in the more rural northern England and Scotland.
The study concluded that the occurrence of childhood leukemia in the recognized
clusters followed a common epidemic process, and that the distributions almost
certainly reflect some sort of biological process that is likely communicable (Knox
and Gilman, 1992).

The infectious etiology hypothesis is multifaceted. In addition to Kinlen,
others have advanced slightly different yet epidemiologically coherent
hypotheses. For comparison let us first consider the essential nature of the
Kinlen hypothesis: "Massive migration into previously remote and unpopulated
areas led to clusters of childhood leukemia. The influx of individuals into an
isolated community produces conditions in which leukemia is more likely to
occur. Isolated communities newly exposed to migrants from elsewhere could
experience an unusual exposure to some hypothetical infectious agent or to

othenivise common infections for which no immunity happens to exist. Perhaps

35

this exposure, in turn, increases the risk of childhood leukemia. Isolated
communities may experience “mini-epidemics” because they are too small to
maintain common infections in endemic form (Kinlen, 1988)." The Greaves
hypothesis, in contrast, takes a somewhat different approach: "cALL (common
form of ALL), which is almost exclusively responsible for the childhood peak (3-5
years of age), may be due to two separate genetic events: 1) spontaneous
mutation of a pre-B-cell in utero 2) proliferation of the mutated pre-B-cell when
exposed to a later antigenic challenge (i.e. an infectious agent). The early in
utero and later childhood (age 2-3 being the crucial period of cell promotion)
mutagenic events lead to leukemia. Older children who have delayed exposures
to specific agents may experience more vigorous B-cell proliferation, resulting in
a second genetic event leading to leukemia (Greaves, 1988)." Freda Alexander
provides an Aberrant Response Model (ARM) that unifies the Kinlen and
Greaves models. The ARM states that a substantial portion of ALL cases arise
as a rare host response to certain patterns of exposure to common infectious
agents. This is at the core of the various infectious etiology hypotheses
(Alexander, 1993).

Support for an infectious etiology hypothesis in the occurrence of
childhood leukemia has been provided by numerous studies which have
demonstrated the patterns of population mixing and clustering in the
epidemiology of the disease. The first investigation of an association between
population mixing and childhood leukemia in Asia was conducted in Hong Kong,

where the study analyzed the childhood leukemia incidence data from 1984 to

36

1990 for evidence of variation between small areas. The results showed that
ALL incidence was similar to that in Caucasian populations, but the peak age
was somewhat older and broader in Hong Kong. Also the typical male
predominance in ALL was observed with an overall male-female ratio of 1.4:1.0.
More importantly though, the Potthofi-Whittinghill test was implemented and
found evidence of overall spatial clustering for ALL in the 0-4 age group which
includes the peak age group. Less statistically significant but similar spatial
clustering results were obtained for total childhood leukemia in all age groups.
The results yield strong evidence of ALL clustering concentrated in the areas
where population mixing was most likely, and the data on T-cell ALL supports the
hypothesis linking common ALL to exposures to common infection
microepidemics (Alexander et al., 1997).

Kinlen led another investigation in 1995 in which the WHO database was
utilized to evaluate the childhood leukemia mortality records of 33 countries from
1958 to 1987 with an emphasis on the late 19505 and early 1960s. The
countries chosen were primarily rural and experienced major rural-urban
migrations; and, consequently, population mixing occurred. The study revealed
that the two countries with the greatest amount of rural population movement,
Greece and Italy, also experienced abnormally high mortality rates from
childhood leukemia. In fact, Greece, which was more rural than Italy and
therefore more greatly impacted by the mixing, had the highest mortality from
childhood leukemia from 1958 to 1972. The high childhood leukemia mortality in

Greece, where extensive rural migration took place, during the 1960s is

37

apparently the highest ever recorded in any country. Vifith the relative rarity of
childhood leukemia, it is very doubtful that the disease would have been over
diagnosed in Greece, and it would actually be more likely overlooked (Kinlen and
Petridou, 1995).

Kinlen and Hudson conducted an ecological mortality study of the impact
of servicemen being stationed in rural areas on the burden of childhood leukemia
from 1950 to 1963. The study aggregated rural counties and urban counties and
determined proportions of servicemen to indigenous populations. The rural
counties with the highest proportion of servicemen had the greatest excess of
childhood leukemia, and the distribution of leukemia reflected the spread of
poliomyelitis at the time (Kinlen and Hudson, 1991). Another study on the effect
of population mixing identified a significantly increased risk of childhood leukemia
mortality in children (age 0-14) in areas which experienced more than a 50%
increase in population over a ten year period (Langford, 1991). Three additional
studies by Freda Alexander support these findings. The studies implemented
cluster analysis of childhood leukemia clusters in England, and they indicated
that either persistent infection in utero or horizontal transmission models could
account for the density and proximity of cases (Alexander et al., 1990)
(Alexander et al., 1992) (Alexander, 1992).

Another interesting observation about the role of population mixing was
made by Kinlen. Kinlen identified increased childhood leukemia trends in cities
within the UK that had greater commuting levels. From this data he was able to

advance the plausibility that children in families with high levels of personal

38

contacts (of which the age and background may be significant factors) may be at
greater risk of manifesting childhood leukemia (Kinlen et al., 1991) (Kinlen,
1993). Several other investigations of possible leukemia clusters have found that
cases have occurred predominantly in Catholic families and members of the
same parish or church (Heath, 1974). The isolation of populations and
subsequent clustering of childhood leukemia, due to an infective etiology, was
again supported by a study of the relationship between car ownership and small
area variation in ALL incidence in children. The study found that ALL and car
ownership were inversely associated, which indicates that isolated groups may
be more susceptible to the hypothesized infectious agent(s) (Alexander, Leon
and Cartwright, 1996). Alexander also identified that an excess risk of childhood
ALL was associated with ecologically defined areas farthest from large urban
centers (Alexander et al., 1990). Similar findings were observed in Sweden
(Hjalmars and Gustafsson, 1998).

Recent evidence has also supported the infectious etiology through
examination of space-time clustering and population density. In 1999, a
European population density study was conducted that included 13,551 cases of
childhood leukemia diagnosed between 1980 and 1989 in 17 different countries
or regions. The investigators found that there was statistically significant
evidence of extra-Poisson variation of case counts in intermediate density areas
(specifically, those with a density of 250-499 persons/km2, p < 0.001 for
childhood leukemia). The evidence suggests that childhood leukemia epidemics

arise regularly in moderately densely populated areas and also sporadically in

39

areas which are somewhat less densely populated. This provides strong support
for an infectious putative agent in childhood ALL, in that susceptibles would be
most likely to be exposed in intermediate areas as they move into them and
occasionally in the more rural areas when a unique infection is introduced
(Alexander et al., 1999). Also, in a more recent study in the United Kingdom
similar infectious etiology supporting evidence was found. Utilizing Knox tests for
analysis of space-time interactions between cases of childhood leukemia,
researchers revealed that there was highly significant evidence of space-time
clustering based on place of birth and time of diagnosis, paiticularly for ALL
cases age 0-4 years. The researchers concluded that their findings were
consistent with an infectious hypothesis (Birch et al., 2000). Both of these
studies demonstrate the potential role of a putative infectious agent through
population mixing, as evidenced by density, and susceptibility, as evidenced by
temporal and spatial characteristics at birth.

The potential infectious etiology of childhood leukemia is additionally
supported by observations of birth order. Regarding childhood leukemia there
appears to be a birth order effect in which the first-born children are at greater
risk of leukemia than their siblings. This could be the result of the older sib
bringing home infections to the younger sibs and thereby conferring immunity
and reduced risk of leukemia to younger sibs (James, 1990). While the older
sibling risk has been identified, no association with breastfeeding and onset of
ALL has been identified (Shu et al., 1995). The lack of protective effect from

breastfeeding could be viewed as negative evidence for the infectious

4O

hypothesis, but it is not necessarily so and further research in this area is
indicated. Another interesting finding, from a large Danish cohort study, that may
support the role of an infectious agent in childhood ALL is the weak association
found between parental autoimmune disease and hematopoietic malignancies in
their offspring during childhood. Results were not statistically significant but did
show an elevation in malignancies among children with parents having
autoimmune disease (Meelemkjaer, Alexander and Olsen, 2000).

The role of viral agents in the development of other human cancers also
supports the notion that childhood leukemia may have an infectious origin. The
Human T-cell leukemia virus (HTLV-1) has been identified by serological and
molecular analyses as the causal factor of adult T-cell leukemia. The chronic
immunosuppression caused by the human immunodeficiency virus (HIV) has
been linked to the increased incidence of some tumors in infected individuals.
HIV infected individuals have also manifested Epstein-Barr virus (EBV) related
immunoblastic lymphomas, Kaposi’s sarcoma, and human papillomavirus (HPV)
related anogenital cancers such as cervical and anal cancer. At least 40% of
Hodgkin’s disease tumors carry genetic traces of EBV. Also, hepatitis B virus
(HBV) has been found to increase the lifetime risk of hepatocellular carcinoma by
a factor of 30-100 (Morris, Eddleston and Crook, 1995) (Toren et al., 1996).
Research has indicated that as many as 15 percent of all cancers worldwide may
be caused by viruses (McCool, 1996).

Malcolm Smith proposes that the JC virus is possibly the infective agent

that leads to ALL (Smith et al., 1997). The JC virus is a polyoma virus with a B-

41

cell predilection. Approximately 50% of all pregnant women are seropositive for
the virus, and infection with the virus tends to occur earlier in developing
countries and in the lower socioeconomic stratum. Primary infection with the JC
virus is generally subclinical. Smith et al. (1997) did not identify the
seroprevalence in ALL cases, but previous studies indicate that this prevalent JC
virus is oncogenic in animal models (Davies and Ross, 1997). The hypothesis
that the JC virus is Ieukemogenic in ALL cases will require substantial further
research, but it does add to the growing body of information implicating an
infectious origin for childhood leukemia.

A growing body of evidence is also developing around the possibility of in
utero infectious origin. Greaves and colleagues (1997) have recently shown that
three children diagnosed with ALL at 5 months, 6 months, and 24 months each
had an abnormal "fusion gene" at birth. This MLL-AF4 gene is the product of a
translocation between two chromosomes and is commonly found in very young
children with ALL. They also found a pair of identical twins diagnosed with ALL
at age 3 and 4 years (within peak age range) who had a translocation between
two genes known as TEL and AML1. Both genetic abnormalities are not
hereditary and could have arisen as a result of an insult from an infectious agent
and/or the subsequent promotional event, necessary for development of clinical
leukemia, could be in response to an antigenic challenge (Greaves et al., 1997)
(Felix, 1999) (Wiemels et al., 1999). It was also recently concluded that the
frequency, age, distribution and clinical features of the TEL-AML1 fusion gene-

positive ALL cases in Brazil were similar in the diverse ethnic backgrounds of the

42

Brazilian children to those in other countries with predominantly white Caucasian
or oriental ethnicity (Magalhaes et al., 2000). Fetal infection models have also
been elucidated and are plausible (Knox et al., 1983). General horizontal
infectious models have also been substantiated by the sharp peak in incidence of
ALL at ages 2-5. Due to the potential for in utero antigenic challenge followed by
the 2 to 5 year latent period that precedes onset of ALL, the sharp peak in
incidence strongly suggests the effect of an infectious agent (MacMahon, 1992)
(Greaves and Alexander, 1993).

Animal models also provide coherence for the infective origin hypothesis.
Retroviruses have been determined to be the causative agents of
leukemogenesis in numerous animal species including: chickens, rats, mice,
guinea pigs, cats, cattle, and gibbon apes. Given that the morphology and
clinical features of animal leukemias and lymphomas closely resemble those in
man, the possibility that such viruses are involved in the induction of human
leukemias and lymphomas is quite high (Goldman and Jarrett, 1984).

The following critical analysis will elucidate some of the evidence that
lends support to the infectious etiology hypothesis through application of

hypothetical modeling of the study results.

43

Chapter 3

WOBURN, MASSACHUSETTS CHILDHOOD LEUKEMIA STUDIES

Introduction

Woburn, Massachusetts, is a community of approximately 35,000 people,
located 13 miles northwest of Boston. It has an extensive industrial history
spanning over 130 years, which included greenhouse operations, leather
manufacturing and chemical manufacturing. Products manufactured included
arsenic compounds used in pesticides, textiles, paper, TNT, and animal glues.
The deposition of hazardous material and waste products from these industries
was a serious point of environmental concern for citizens and government
officials, and led to the execution of several studies.

Thirteen environmental and environmental health studies have been
conducted in Woburn, Massachusetts by the Massachusetts Department of
Public Health (MDPH) and Department of Environmental Quality Engineering
(DEQE) since 1979. In 1979, environmental concerns were brought to the
forefront of public attention when excavation of a former industrial site unearthed
significant amounts of industrial waste that were laden with high levels of lead,
arsenic, and heavy metals. It was subsequently learned that two municipal
drinking water wells (Wells G & H), which had been installed near this site, were
contaminated with trichlorethylene (TCE), perchloroethylene, chloroform and
other organic compounds. These wells supplied public water primarily to the

eastern portion of Woburn between 1964 and 1979.

44

Woburn residents were concerned regarding health effects that may have
resulted from consumption of the contaminated water from wells G & H.
Between January 1969 and December 1979, twelve cases of childhood leukemia
had been diagnosed in Woburn. Six of these cases resided in a six-block area,
which was served directly by the contaminated wells. Identification of these
cases, and the potential link to the wells, prompted the Massachusetts
Department of Public Health (MDPH) and the Centers for Disease Control (CDC)
to begin a formal investigation of the health status of Woburn residents.

In January 1981 a case-control investigation was conducted by MDPH to
evaluate the situation in Woburn. The investigation concluded: (1) the incidence
of childhood leukemia was significantly elevated in Woburn (12 observed cases
vs. 5.3 expected cases between 1969 and 1979); (2) the majority of the excess
cases were males; and (3) six of the cases were diagnosed while residing in a
single census tract in east Woburn. At the time, the cases in this census tract
represented half the identified childhood leukemias, a number disproportionate to
the geographic distribution of the population in the community. In 1984, an
additional seven cases were identified that had been diagnosed through 1983.
This demonstrated a continued excess in the incidence of childhood leukemia
(19 observed cases vs. 6.1 expected).

By the middle of 1986, a total of 21 childhood leukemia cases had been
diagnosed in Woburn. Cases diagnosed after 1979 had birth residences which
were more evenly distributed throughout Woburn than the original 12 cases. As

a result of the continued elevated incidence and the change in the geographic

45

distribution of the cases, MDPH conducted the Childhood Leukemia Follow-Up
Study, the results of which are detailed later in this chapter.

The following describes each of the Woburn studies and summarizes their
findings: (note: each bold heading represents an individual publication by the
same name) (See Appendix A & B for maps indicating the location of Woburn
and Wells G & H)

Woburn Health Data Analysis 1969 - 1978 (Kotelchuck and Parker, 1979)

The first study conducted in Woburn was the Woburn Health Data
Analysis 1969 - 1978, which was presented December 21, 1979. The MDPH
ascertained that Woburn's cancer death rate was higher than what should be
expected considering the age and sex characteristics of its residents. The
MDPH also determined that although the mortality rates were elevated, they did
not represent an epidemic situation.

Based upon the statewide experience with cancer mortality from 1969 -
1978 for all causes (International Classification of Diseases categories 140 - 199,
Malignant neoplasms), the expected cancer mortality for Woburn was 503.5
deaths but 569 cancer deaths were observed. Woburn's cancer mortality was
13.7% greater than expected. Cancer mortality in Woburn was found to be on
the increase, but only during the last five-year period was the observed number
of deaths significantly higher than expected. During 1974 - 1978, 307 cancer
deaths were observed in Woburn with 248 expected. In the earlier 5-year period,
1969 - 1974, 262 cancer deaths were observed with 255.5 expected. From 1974

- 1978, Woburn had the highest rate of cancer out of the 70 largest

46

Massachusetts communities (defined as having a population greater than
20,000). Woburn had significantly elevated cancer mortality rates for only a
limited number of cancer types. In the period 1974 -1978, Woburn had elevated
mortality rates for bronchus/lung, breast, prostate, kidney, and liver cancers.
Childhood and young adult leukemia mortality was elevated but not significantly
elevated in Woburn during the 10 year study period. During the 1969 - 1973
period there was a statistically significant elevation of leukemias in Woburn (18
leukemia deaths observed with 10.9 expected), but there was no statistically
significant elevation for 1974 - 1978. The study also examined three specific
types of cancers in Woburn for clustering: liver, kidney, and childhood leukemia.
No strong pattern of clustering of deaths was found. Additionally, no increase in
infant mortality, Iow-birthweight infants, or congenital malformation was found.
(Kotelchuck and Parker, 1979)

Woburn: Cancer Incidence and Environmental Hazards 1969 - 1978 (Parker

and Rosen, 1981)

 

This study dealt primarily with questions concerning childhood leukemia,
renal cancer, and liver cancer. Cases with these cancers were investigated by
means of an interview study. An investigation into the incidence of bladder
cancer was also made. A case was defined as a person with confirmed
childhood leukemia, renal cancer, liver cancer, or bladder cancer, who was
diagnosed with the disease between 1969 and 1978 (1969-1979 for the leukemia
cases) and who was a resident of Woburn at the time of diagnosis. The study

questionnaire was developed, pretested, and revised by the MDPH and CDC.

47

The topics investigated included demographic information, disease processes,
past medical history, smoking, residence, schooling, occupational histories, and
environmental exposures. Trained interviewers from the CDC and NIOSH
conducted the majority of the interviews. Two age— and sex-matched controls
were interviewed for each of the twelve leukemia cases. One control was a
person of the same age and sex who lived close to the case. The other age and
sex-matched control lived in a distant part of Woburn. Interviews were
conducted with one or both parents of the leukemia cases and controls.

The investigation confirmed that there was a significantly elevated
incidence of childhood leukemia in Woburn for the period 1969-1979, with 12
cases observed and 5.3 expected. The excess observed was largely among
males, with 9 cases observed and 3.1 cases expected. Analysis of residence at
the time of diagnosis revealed a significant concentration of cases in the eastern
part of Woburn, where the incidence of the disease was at least seven times
greater than expected. For males, the incidence of childhood leukemia in this
area was over 12 times that expected. The incidence of childhood leukemia for
the rest of Woburn was not significantly elevated compared to national rates.

The investigation established that the overall incidence of childhood
leukemia was significantly elevated in Woburn and that there was a significant
concentration of cases in one particular area of east Woburn. No evidence was
ascertained that indicated that the increase in leukemia incidence was
associated with environmental hazards in Woburn, and specifically with the

contamination of drinking water supplies. (Parker and Rosen, 1981)

48

Retrospective StudiLof Woburn 1949 - 1968 (Rowe, 1981)

This study was undertaken to analyze cancer mortality in Woburn from
1949-1968. The main impetus was to evaluate the pattern of childhood
leukemia. Since the survival from childhood leukemia was 5% in 1950, mortality
rates closely reflected the actual incidence figures for that time. Thus, this study
relied on mortality records as a proxy for childhood leukemia incidence. Every
death certificate that mentioned cancer was abstracted. Childhood leukemia
cases were included if it was noted as an immediate cause of death and if the
individual was 19 years of age or under at death. Analysis of the SMRs for
leukemia showed a gradual increase from 76 to 123 in the period 1949-1968.
The increase in childhood leukemia mortality was greater than adult leukemia
mortality. Investigators concluded that "If one suspects an environmental toxin
such as the contaminated drinking water to have been a factor in the occurrence
of childhood leukemias, then there must have been a sufficient latent period such
that they didn't show up until the late 1970's. At the present time there is no
evidence to support an environmental toxin in the etiology of the cases of
leukemia in Woburn.” (Rowe, 1981)

CDC: Cancer in Woburn QDCJ 981)

 

In 1980 the Massachusetts Department of Public Health and the Centers
for Disease Control undertook a case-control study to investigate possible
associations between environmental hazards and the incidence of childhood
leukemia plus three other types of cancer (kidney, urinary bladder, and liver) in

Woburn residents for the period 1969-1978. This investigation confirmed that the

49

increase in incidence of childhood leukemia, concentrated in eastern Woburn,
was statistically significant when compared with nationwide rates. Because of
limited environmental data, the investigation did not establish an association
between known environmental hazards in Woburn and the elevated childhood
leukemia incidence. Also, kidney cancer incidence was slightly significantly
elevated, and rates for liver and bladder cancer were not significantly elevated.

Childhood leukemia cases were defined as Woburn residents under 19
years of age, whose cancer was diagnosed between 1969 and 1979. Trained
interviewers from CDC and NIOSH conducted interviews with one or both
parents of the 12 childhood leukemia cases. For each leukemia patient, two age-
and sex-matched controls were interviewed (one control resided near the case).

Twelve cases of childhood leukemia occurred in Woburn from 1969-1979.
Nine (75%) of the children had acute lymphocytic leukemia, and all cases were
diagnosed before the patients reached 15 years of age. Childhood leukemia
incidence in Woburn in this period was significantly higher than expected with 12
observed and 5.3 expected (p=0.0058). Boys had an elevated incidence rate
(p=0.00380), and girls did not. However, all of the girls' cases were all diagnosed
when they were between ages 10 and 14, which was a significant elevation for
that age group (p=0.008).

At the time of diagnosis, the leukemia cases were concentrated in the
eastern part of Woburn, just north of Walker Pond. Six patients lived within or on
the border of census tract 3334, in an area approximately 1/2 mile in radius.

There was a significant concentration of cases in this census tract. The binomial

50

distribution probability that 6 or more of the 12 cases would occur in this area,
which contains only 17% of the town's population in the 1 to 14 year age group,
was less than 0.01. The six cases in this census tract were 7.5 times
(p=0.00018) higher than the expected 0.8 cases. Childhood leukemia for the rest
of Woburn was not significantly elevated (p=0.297).

There were no significant differences between case and control groups
with respect to medical histories, parents' occupations, or environmental
exposures. Interestingly, four of the ALL cases were born in a 6-month period,
indicating a potential in utero exposure. (CDC, 1981)

Cancer Mortality in Woburn: A Three Decade Study 1949-1978 LTelles, 1981)

 

This study utilized a systematic search through all death certificates at the
Woburn City Hall for the period 1949-1968. Every certificate that mentioned
cancer was abstracted and the data were analyzed according to the site of
cancer, year of death, age of death and sex. These observed deaths were
compared to the expected number of deaths, by computing standardized
mortality ratios. During the period 1949-1978 there were fourteen deaths due to
childhood leukemia (cases 19 years and younger).

One objective of this study was to test further whether there was an
association between the use of water from Wells G and H and the occurrence of
leukemia. These additional data weakened the association, in that the number of
childhood leukemia deaths began to rise from 1959-1963 (SMR = 120), before
the wells were drilled, and continued to rise in the period in which the wells were

drilled 1964-1968 (SMR = 169). Given an average latent period of two to five

51

years, childhood leukemia associated with Wells G and H should not have
started to increase until 1969-1973, when in fact, the rate was lower than
expected (SMR = 50). However, from 1974-1978 the rate increased substantially
(SMR = 357).

Additionally, childhood leukemia mortality was not elevated during the
period 1949-1958, but childhood leukemia mortality began to rise in 1959-1963.
No unusual geographic distribution of childhood leukemia mortality occurred
during the period 1949-1968. The investigators concluded that these data were
consistent with the possibility that, excluding a chance occurrence, the elevated
incidence of leukemia and the concentration of cases in eastern Woburn during
1969-1978 must have been due to some newly introduced factor which became
present and active during the late 1950's. (T elles, 1981)

Childhood Leukemia, Woburn, MA: An Update (Cutler, 1984)

This study evaluated the 28 cases to date in Woburn. The important
findings were for the period 1980 - 1983, in which there were 7 new incident
cases of childhood leukemia. The cases occurred in a scattered geographic
pattern, although 1 of the cases was in the Walker Pond area. All of the cases
white; 5 of 7 were male, the age range was 22 months to 14 years; all had ALL,
with unambiguous clinical and histological diagnoses.

Investigators also calculated expected and observed leukemia cases for
Woburn for the period 1969-1983. Although all the leukemias were ALL,
because 3 of 12 cases before 1980 were other leukemias, expected cases were

calculated for leukemia, all types, age 0-14. Male - 3.99 expected and 14

52

observed (ratio 3.00, 95% Cl 1.55, 5.25); Female — 2.02 expected and 5
observed (ratio 1.98, 95% CI 0.54, 5.07); Both - 6.01 expected and 19 observed
(ratio 3.16, 95% Cl 1.90, 4.93).

Geographical clustering was observed in east Woburn around Walker
Pond (census tract 3334) with 7 cases observed and 0.92 expected. The
number of cases was significantly elevated in tracts 3331 (4 observed/ 1.08
expected) and 3335 (6 observed/ 1.46 expected) as well.

When soil samples were analyzed from the vicinity of Walker Pond, no
single compound or group of compounds was found to be associated with the
cases of leukemia. In addition, air samples were found to be normal. The
investigators concluded that '"l'he problem of excess leukemia in Woburn
children continues, and there is no explanation." (Cutler, 1984)

Final Report of the Woburn Advisory Panel to the Massachusetts
Department of Public Health (MDPH, 1985)

This paper did not report any study findings, rather it was a list of
recommendations on how to proceed with future studies in Woburn. The
advisory panel advanced recommendations on epidemiological studies of the
cluster of childhood leukemia, etiological studies, environmental studies, as well
as health surveillance in Woburn. (MDPH, 1985)

Health Assessment Wells 68H, Woburn (MDPH, 1989)

The suspected Ieukemogenic contaminated wells in Woburn, Wells G & H,

were located about 10 miles north of Boston, Massachusetts. Woburn Municipal

Wells G and H were used from 1964 to 1979 to supplement Woburn drinking

53

water supplies. The ground water extracted from Wells G and H was found to be
contaminated with volatile organic compounds in 1979. The predominant ground
water contaminants were trichloroethylene, tetrachloroethylene, trans-1,2-
dichloroethylene and 1,1,1-trichloroethane. Five contaminated properties located
near the wells were known contributors to the ground water contamination.
Ground water and both surface and subsurface soil contamination on-site at the
facilities was extensive and resulted from liquid-waste spills on the ground,
sludge disposal, buried and surface disposal of 55-gallon drums, and leaking
tanks. Based on a 30-day aquifer test, Municipal Wells G and H were found to
intercept, either entirely or partially, ground water from beneath these five
contaminated properties. Wells G and H have not been used since 1979. As
residences and industries were located both on-site and nearby, there was a
potential for human exposure to contaminants by: (1) inhalation of dusts and
vapors from surface soils, industrial use of contaminated well water, and
migration of vapors from contaminated shallow ground water to the inside of
buildings; 2) skin contact with contaminated surface soils at Wildwood
Conservation Corporation, New England Plastics Corporation, and Olympia
Nominee Trust Corporation, sludge at Wildwood Conservation Corporation, and
contaminated industrial well water at New England Plastics Corporation and John
J. Riley Tannery; and 3) ingestion of dusts, surface soils, and, possibly,
contaminated fish and contaminated industrial well water. (MDPH, 1989)

Exposure to Wells G&H in Woburn Massachusetts (Murphy, 1990)

54

This report attempted to quantify the exposure of Woburn citizens to
drinking water from Wells G and H. The calculation of exposure was a function
of both the approximately 50 hydraulically distinct neighborhoods within Woburn
and of the 114 months of Wells' G and H operation.

The method used to calculate this exposure to water from the wells begins
with a computer model of the water distribution system that was developed by
the investigators under a previous contract with the Massachusetts Department
of Environmental Quality Engineering (DEQE). This Woburn water distribution
model was applied to the various pumping and water use configurations that
occurred during each month that Wells G and H were in operation. The results of
these calculations were then individually analyzed with a hydraulic mixing model
to calculate the mixture of water supplied to each neighborhood, the monthly
average exposure index for each neighborhood, and the cumulative exposure.

The levels of exposure to water from Wells G and H were found to vary
widely. Typically, the neighborhoods south and west of the center of the city
received no or very little water from Wells G and H. The neighborhoods of east
Woburn received water mostly from Wells G and H whenever those wells were
pumping. The mixture zone between the two water sources ran along, or just to
the east of Main Street (the center of the city). (Murphy, 1990)

Final Report of the Woburn Environment and Birth Study (MDPH, CDC and

 

MHRLI 994)

 

The Woburn Environment and Birth Study (WEBS) was implemented to

determine if adverse reproductive outcome rates in Woburn (such as specific

55

birth defects and stillbirths) differed from rates in other communities, and also to
determine if the opportunity for exposure to contaminated drinking water from
Woburn municipal Wells G & H was related to a higher rate of any of these
outcomes. The study population consisted of all live-births and stillbirths that
occurred among Woburn residents between January 1969 and March 1991
(more than 11,500). Rates of adverse reproductive outcomes occurring in
Woburn live-births and stillbirths were compared to corresponding rates in other
local and regional populations to determine if the Woburn rates were elevated.
WEBS included a 20-year (1969 - 1988) retrospective cohort study (N =
10,383) to evaluate the overall reproductive health of Woburn residents relative
to several referent populations and relative to exposure to the contaminated
water and a 27-month (1/lll989 - 3/31l1991) prospective surveillance study (N =
1,227) to evaluate the more recent reproductive health of Woburn residents
relative to the referent communities. Maternal exposure to contaminated water
was assessed on the basis of an estimate of the proportion of contaminated
water that reached each birth residence location during pregnancy.
Comparison with the referent populations indicated no differences in the
prevalence of most outcomes for Woburn births. Over both the retrospective and
prospective surveillance study periods, there was a trend over time for a greater
prevalence of male fetal deaths in Woburn, but this trend was significant only
during the prospective surveillance period (after closure of the contaminated

water wells). During the retrospective study period, there were both significantly

56

higher and lower rates of particular organ groupings and specific birth defects
relative to two referent birth defects registries.

Analysis of the effects of exposure to contaminated water, revealed a
trend for a lower prevalence of the birthweight, pre-term and infant death
outcomes for exposed groups than for unexposed groups. Conversely, the trend
was for a somewhat higher prevalence of the small-for—gestationaI-age and fetal
death outcomes for exposed groups, but in most instances the differences were
not significant. There were too few cases of birth defects by diagnostic category
to make reliable comparisons between the exposure categories, although in the
few instances in which there were sufficient numbers, no significant association
with exposure to contaminated water was observed.

Investigators concluded that the WEBS results provided little support for
the hypothesis that environmental contaminants in the public water supply had
an adverse effect on the reproductive health of exposed subgroups of Woburn
residents or of Woburn residents as a whole. (MDPH, CDC and MHRI, 1994)
Health Assessment - lndustriplex Site in Woburn (MDPH, 1995)

This study evaluated the 244-acre lndustriplex site located in Woburn,
Massachusetts. This was previously the location of various chemical
manufacturing plants, and most recently a hide glue manufacturing plant. The
U.S. Environmental Protection Agency designated the site in the Woburn study
area as a Superfund site. Various chemicals were detected on-site and off-site in
the surrounding study area. Arsenic, chromium, lead, benzene, and toluene are

the primary contaminants on-site and were generally detected at elevated

57

concentrations in soils and sediments. Volatile organic compounds (VOCs),
semi-volatile organic compounds (semi-VOCs), polynuclear aromatic
hydrocarbons (PAHs), polychlorinated biphenyls (PCBs), and other metals were
also detected on-site and in the study area.

Populations with the greatest potential for historical exposures to
contaminants originating from the site were on-site workers, unauthorized
individuals who accessed the site, and workers and residents in the site vicinity.
The only chemical present at concentrations at which adverse human health
effects are known to occur was arsenic. Individuals who may have accessed the
site in the past would have been at the greatest risk of exposure due to
contamination of the soil (0-12 inch depth), subsurface soil, and sediment
through ingestion of these contaminated media.

The health effects which are expected to occur as a result of exposure to
site-related contaminants were primarily associated with exposure to
contaminants in on and off-site soil 0-12" in depth. Ingestion of arsenic
contaminated soil may have resulted in abdominal pain, diarrhea, and sore
throat. Skin exposure to chromium in this media could have enhanced already
existing dermatitis, and ingestion of lead contaminated soil may have caused
mild hematological effects by disrupting enzyme activity. All these effects are
reversible and would have been effectively halted with the termination of
exposure. No linkage to childhood leukemia has been identified for any of these
exposures. (MDPH, 1995)

Health Assessment Addendum - Wells G8l-LWobum (MDPH, 1995)

58

The Wells G and H site encompassed two municipal wells located in
Woburn, Massachusetts that were used as a municipal water supply for the city
of Woburn from 1964 to 1979. The initial health assessment was completed in
1989, and this addendum was published to address the indoor air monitoring
studies conducted at the Wells G and H site following the initial assessment.
Three nearby residences and a local day care center were sampled to determine
the extent of indoor air contamination due to volatilization of contaminants from
the groundwater plume into the basements of the residences and the day care
center. The primary contaminants detected in indoor air were carbon
tetrachloride, benzene, 1,2-dichloroethane, trichloroethylene, 1,1-
dichloroethylene, methylene chloride, and chloroform.

Populations potentially exposed to the inhalation of indoor air
contaminants from the study area included residents and workers in the general
area and workers and children at the day care center. Adverse health effects
from exposure were determined to be unlikely due to the low concentrations
detected.

Investigators determined that the chemicals detected during these
sampling investigations were not likely related to the Well G & H sites. The
concentrations measured were typical of concentrations seen in many studies of
ambient or indoor air. The majority of the samples taken at three Woburn
residences and the day care center resulted in either non-detectable levels of
contaminants or low levels of contaminants. The highest levels were detected

consistently at one of the residences, which stored a considerable amount of

59

household products containing a variety of the compounds of concern associated
with this investigation (suggesting another potential source of contamination
other than groundwater). The proximity of various industrial companies,
including a printing press, may have been a source of contamination for the
indoor air of the day care center. Thus, it was concluded that identifying the
contaminated groundwater as the only source contributing to air concentrations
in the basements, could not be made with any certainty; and the indoor air in the
site vicinity represented no apparent public health hazard. (MDPH, 1995)
Woburn Childhood Leukemia Follow-Up Study (MDPH, 1997)

This was the final study conducted by MDPH and they do not wish to
analyze the situation at Woburn any further. Consequently, they refused to give
the author of this critique access to any of the original data that was utilized for
this or any other childhood leukemia study conducted in Woburn.

Study Methods:

The study was a matched case-control design with two controls selected
for each case. All cases were residents of Woburn who were 0 to 19 years of
age at time of diagnosis and were diagnosed with leukemia between 1969 and
1989. Controls were selected randomly from Woburn school records and
matched to cases based on date of birth (plus or minus 3 months), sex and race.
Controls must have been Woburn residents at the time of diagnosis of the
matched case. Residential, occupational and health history data were collected
during the interview for the etiologic period for each case and its matched

controls. The etiologic period was defined as the period of time from two years

60

before conception to the case’s leukemia diagnosis and was sub-divided for
analysis into three time segments: two years before conception to conception,
during pregnancy, and from birth to case diagnosis.

Study Results:

Univariate Analyses - Detailed analyses of data collected at interview
revealed five variables for which 10 or more total positive responses were
identified and that demonstrated odds ratios greater than or equal to 1.50 in
relation to the childhood leukemia incidence. Maternal alcohol consumption
during pregnancy (OR = 1.50, CI. = 0.54, 4.20), diagnosis of paternal
grandfather with cancer (OR = 2.01, CI. = 0.73, 5.58), having a father who
worked for industries considered high risk for occupational exposures (OR =
2.50, CI. = 0.78, 8.30), and the subject's consumption of public water as their
primary beverage (OR = 3.03, CI. = 0.82; 11.28) were all variables which
showed non-significant but positive associations with childhood leukemia
incidence. A statistically significant association was identified between
developing childhood leukemia and being breast-fed as a child (OR = 10.17, CI.
= 1.22, 84.58).

Multivariate Analyses and Exposure to Wells G and H - Multivariate
analyses of the relationship between childhood leukemia and exposure to water
from Wells G and H revealed that although five variables discussed above
showed elevated odds ratios as univariates in relation to the leukemia, they did
not significantly affect odds ratios specific to water exposure. Adjusted odds

ratios were calculated controlling for socioeconomic status, maternal smoking

61

during pregnancy, maternal age at birth of the child, and maternal alcohol
consumption during pregnancy. Of these variables, only maternal alcohol
consumption during pregnancy demonstrated a slightly elevated odds ratio in
univariate analysis.

Adjusted odds ratios describing the effects of Wells G and H water on
leukemia incidence showed a non-significant elevation for the overall etiologic
period (OR = 2.39, CI. = 0.54, 10.59) and each time period subcategory. The
strongest relationship between exposure and leukemia among time period
subcategories was during pregnancy (OR = 8.33, CI. = 0.73, 94.67), the second
was in the two years before conception (OR = 2.61, C.l. = 0.47, 14.37) and the
time period between the birth of the case and the diagnosis of leukemia (OR =
1.18, CI. = 0.28, 5.05).

Sub-stratification of the pregnancy time period to assess specific effects of
water exposure by trimester revealed high correlation coefficients between
trimester exposure values. Thus, independent effects of water exposure by
trimester on leukemia incidence could not be distinguished.

Analyses to assess potential dose response relationships were completed
using a trichotomous parameterization of the actual study subject exposure
values by time period. Results demonstrated elevated odds ratios between dose
categories for the preconception and pregnancy periods. A significant trend
across exposure categories was also identified for the period during pregnancy
(P < 0.05) suggesting a dose-response relationship for subjects whose mothers

drank Wells G and H water during pregnancy. Tests for trend for the etiologic

62

period overall and for each of the other time period subcategories were not
significant (P>0.05).
Study Discussion and Conclusions:

The investigators concluded that the results suggested the relative risk of
developing childhood leukemia was greater for those children whose mothers
were likely to have consumed water from Wells G and H during pregnancy. This
conclusion was based on the association of the relationship to the amount of
water households received (dose-response). However, there appeared to be no
associations between the development of childhood leukemia and consumption
of water from Wells G and H by the children prior to their diagnosis.

Relatively few positive associations were identified between childhood
leukemia incidence and residential parental occupation, and medical history
related risk factors, and none were statistically significant with the exception of
the breast-feeding finding.

Investigators noted that TCE, one chemical detected in the well water, is
known to have weak hematologic effects in mammals. No studies have shown
this effect in humans, although effects on the developing human fetus are
unclear. They also noted that the nature and extent of historical contamination of
Wells G and H is not known.

Investigators also acknowledged that the small number of study subjects
leads to imprecise estimates of risk. As a result, the exact magnitude of the
association between exposure to water from Wells G and H and risk of childhood

leukemia could not be stated. However, results suggested a dose-response

63

relationship and demonstrated a decrease in effect after the elimination of the
potential for exposure. The study authors concluded that the incidence of
childhood leukemia in Woburn between 1969 and 1989 was associated with
mothers' potential for exposure to contaminated water from Wells G and H,

particularly for exposure during pregnancy. (MDPH, 1997)

64

Chapter 4

WEST CENTRAL PHOENIX, ARIZONA CHILDHOOD LEUKEMIA STUDIES

Introduction

In 1982 a parent first identified an elevated number of children with
leukemia attending a school in West Central Phoenix. Subsequently, the
principal of St. \fincent de Paul School, a parochial school in West Central
Phoenix (WCP), reported to the Arizona Department of Health Services (ADHS)
that seven students had died of leukemia between 1961 and 1982. The ADHS
computed the leukemia mortality rate in the city of Phoenix and in the school
catchment area for children age 5 through 14 for the years 1970 through 1981.
The number of leukemia deaths in the catchment area of school during the
twelve-year period was thirteen, about seven more cases than would have been
expected in comparison to the leukemia rate in the remainder of the city. A 1983
evaluation by the ADHS did not reveal any unusual environmental factors. Public
concerns resurfaced in 1987, leading to two ADHS studies that addressed overall
death rates and childhood cancer incidence rates in Maricopa county and West
Central Phoenix (WCP), defined here as the 50 square mile area from 27th
Avenue to 83rd Avenue, and Camelback Road to Southern Avenue. The ADHS
reported the results of those two studies in 1988 and 1990. The incidence study
showed an elevation of the rate of leukemia in WCP among children 0-19 years
of age (29 cases expected between 1965 and 1986, with 49 observed in that

time period). The only leukemia subtype that was significantly elevated in WCP

65

was acute myeloid leukemia (AML), an uncommon subtype in children. These
studies were followed by a comprehensive case-control study and an updated
incidence study.

The following describes each of the West Central Phoenix studies and
summarizes their findings: (note: each bold heading represents an individual
publication by the same name) (See Appendix A & C for maps indicating the
location of West Central Phoenix)

Report on Mortality in Maricopa County 1966-1986 (Flood and Chapin, 1988)

This study was conducted to investigate the cluster of childhood leukemia
in WCP. This was a retrospective study of mortality in Maricopa county for the
years 1966 through 1986 that aimed to quantitatively confirm the 1982
observations, determine leukemia mortality rates in previous and subsequent
time rates, and to examine total cancer mortality within four suspect areas and
the remainder of Maricopa county.

This study confirmed the previous finding of a higher rate of childhood
leukemia deaths in West Central Phoenix and in east central Phoenix between
1970 and 1981. During 1966-1969, the leukemia rates were not significantly
elevated in these areas. During 1982-1986, leukemia deaths only in WCP were
significantly elevated, with eight deaths observed and 3.8 expected. The finding
of an elevated leukemia rate in WCP during 1982-1986 was found to be accurate
and not biased by a sampling problem that occurred during 1970-1981. These

findings corroborated by an another group that assessed the spatial occurrence

66

of childhood leukemia mortality in WCP using standardized rate ratios with a
simple linear Poisson model (Aickin et al., 1992).

When compared to U.S. rates, total cancer rates in east central Phoenix
were significantly elevated in children between 1970-1981 and in middle-aged
adults (45-64) between both 1970-1981 and 1982-1986. During this latter period,
the total cancer rate among elderly (65+) was found to be significantly elevated in
West Central Phoenix. Total cancer mortality rates in all adult age groups and
during all three time periods in the non-suspect area of Maricopa county areas
were all significantly lower than U.S. rates. (Flood and Chapin, 1988)

Incidence Study of Childhood Cancer in Maricopa Conny: 1965-1986

(Flood et al., 1990)

 

This incidence study was conducted to determine whether there was an
elevated incidence of childhood leukemia among children residing in WCP, to
explore the data for factors that could guide the search for causes of the
apparent cluster of leukemia, and to examine the rates of other childhood
cancers within Maricopa county.

For this study, a case was defined as a malignancy diagnosed from 1965
to 1986 in a person age 0-19, who was a resident of Maricopa county at the time
of diagnosis. All leukemias were aggregated together.

The analysis showed an elevated childhood leukemia incidence rate in
WCP in both the overall period, 1965-1986, and in the latter period, 1982-1986;
but there was no elevation in the other 3 suspect census tract areas where cases

had been identified. The magnitude of the elevation, as reflected in the

67

standardized incidence ratio, was relatively small, 1.67 and 1.91 respectively,
and not statistically significant. The analysis also showed that there was no
elevation in WCP in the rate of all childhood cancers combined, only the
leukemia rate was elevated. Rates of cancer and leukemia in the entire county
were comparable to the rates reported by other geographic areas of the national
SEER program.

Researchers reported that the analysis of the leukemia rates within WCP
revealed no obvious geographic clustering of cases within the area, nor was
there an elevation in the rates that was specific to gender, age-group, or time-
period. Analysis of leukemia rates failed to demonstrate an association with
three potential risk factors: proximity to wells which pumped TOE-contaminated
water, the catchment area of St. Vincent de Paul school, and proximity to the
gasoline tank farms. (Flood et al., 1990)

Follow-Up of Childhood Leukemia Incidence Rate in Maricopa County
Standardized Incidence Ratio of Area F to Area R for 1987-1990(ADHS,
@311

The purpose of this study was to calculate new incidence rates for
childhood leukemia (ages 0-19 years) in a previously defined combined census
tract area that comprises WCP (Area F) in Maricopa county. The comparison
standard area (Area R) was comprised of all of Maricopa County, excluding the 4
suspect areas (including WCP). This study compared childhood leukemia in

area F with area R for the years 1987-1990.

68

During the period of 1987-1990, childhood leukemia incidence was found
to be no higher in area F than the standard area R. The standardized rate ratio
(SRR) for area F standardized to area R for 1987-1990 was calculated to be
0.8547 (95% Cl 0.4758, 1.5354). The observed number of cases for area F was
9 during this time period and 10.53 were expected. The data indicated that the
incidence of childhood leukemia in WCP from 1987-1990 was much improved
with an inclination toward lower rates. (ADHS, 1993)

Case-Referent Study of Childhood Leukemia in Maricopa County, Arizona,

1965-1990 (Flood et al., 1997)

 

This final study, a case-referent (case-control) study, was undertaken to
evaluate children diagnosed with leukemia between 1965 and 1990 in Maricopa
County, Arizona. The Arizona Department of Health Services administered the
study and selected participants from West Central Phoenix and all of Maricopa
County.

The purpose of this case-control study was to investigate possible risk
factors for childhood leukemia in Maricopa County with a primary focus on the
factors that might have produced the elevated childhood leukemia rate in WCP.
The lack of known causes of leukemia presented fundamental difficulties for the
study. The investigators looked broadly for possible risk factors because of the
numerous competing etiologic hypotheses for childhood leukemia. The primary
of objectives of this study were to (1) characterize the association of residence in
WCP with the occurrence of childhood leukemia; (2) to assess whether the

residence of cases tended to cluster around local sources of environmental

69

exposure; and (3) to determine if the risk for childhood leukemia was associated
with exposure to pesticides, solvents, and petroleum products. The secondary
objective was to determine if there was an association of childhood leukemia with
other known or suspected risk factors, such as ionizing radiation, genetic and
familial factors, traffic volume, cigarette smoke, drugs, electromagnetic fields
(EMF), and migration patterns.

Study Methods:

Investigators collected information from questionnaires administered over
the telephone to the parents of cases and referents. In addition, parents were
asked to provide complete residential histories, and to document the
characteristics of the homes in which they had lived. Also, in a limited number of
homes, investigators collected samples of indoor air, soil, and household dust,
and measured the strength of electromagnetic fields.

A case was defined as a child, age 0-19, who resided in Maricopa county
when diagnosed with leukemia during the period 1965-1990, and was identified
from the Arizona Cancer Registry.

Of 413 eligible cases of leukemia there were 274 for whom investigators
located at least one parent and requested their participation. Of these, the
parents of 222 cases (81% of the 274 locatable; 54% of the 413 eligible) actually
participated in the study. These 222 participating cases were compared to 219
age and gender matched referents, that were selected by using random digit

dialing of telephone numbers in Maricopa county.

70

The participation rate of case families living in WCP when diagnosed with
leukemia was lower than the rate of participation for all cases. Of 58 eligible
cases registered from WCP, there were 36 families located. Of these, the
parents of 23 cases (64% of the 36 locatable; 40% of the 58 eligible) participated.

The general characteristics of the cases and referents (more precisely,
their parents) were very similar with respect to race and ethnicity, income,
education level, and the participation by a father and mother. However, there
was a major difference in the vital status of their children. None of the 219 child-
referents were known to have died, whereas 102 of the 222 child-cases had died
prior to the parental interview.

Study Results:

Living in WCP: There was no difference between cases and referents in the
proportion of subjects who were born in WCP. There was no difference in the
proportion of time the subjects had lived in WCP or in the length of time spent in
WCP prior to being diagnosed. There was no difference between the cases and
referents in the proportion of children diagnosed by 5 years of age who were
born in WCP. There also was no difference in the proportion of participants age
0-19 who were born in WCP. There was statistically significant evidence that
children diagnosed before 60 months of age and born in WCP were diagnosed at
an older age than those born outside WCP. Investigators interpreted this finding
as evidence that there may have been a prenatal risk factor in WCP.

The proportion of cases and referents who shared various attributes

related to living in WCP was statistically similar. There was no clear evidence

71

that being born in, nor living in, WCP was a risk factor for developing childhood
leukemia.

Emissions and pollution of the general environment: The study results indicated
that residence within 0.5, 1.0, or 2.0 miles of municipal wells contaminated with
TCE at levels above the allowable maximum contaminant level (MCL), within the
boundaries of one or more federal EPA Superfund areas, within 0.5, 1.0, or 2.0
miles of any Arizona Department of Environmental Quality 'Water Quality
Assurance Revolving Fund" sites, and within 0.5, 1.0, or 2.0 miles of any of one
of 162 different sources of airborne volatile organic chemicals in Maricopa county
were not risk factors for leukemia. Also, proximity to gas stations was analyzed
and showed no risk of living within 200 feet or 200 yards of a station.

Having lived within 3 miles of the gasoline tank storage facility was a
significant risk factor (OR = 2.2 95% CI = 1.04, ). However, no participants had
lived within one mile of the tank farm; those who had lived within two miles had
an OR of 1.8; and living within 5 miles, but not 4 miles, was a risk factor. Thus
the results appear to be inconsistent.

Solvent Exposure: Parental occupational exposure to 11 different solvents and
all the solvents combined was not associated with childhood leukemia.
Household exposure through hobbies was assessed by evaluating 10 specific
activities. Two findings were significant, with model building as a risk factor (OR
1.9; 95% CI 1.1 to 3.4; p=0.02) and use of power tools as a protective factor (OR
0.33; 95% Cl 0.17 to 0.66; p=0.002), both of which could be markers for some

other factor. The "father only" analysis indicated that automobile/truck repair was

72

a risk factor (OR 1.6; 95% CI 1.03 to 2.50; p=0.033). The elevated risks were not
linked to residence in WCP or to cases of AML.

When inferring exposure from residential characteristics and activities, two
significant associations were found. Parking the car in a lot 100 feet from the
residence was protective (OR 0.85; p=0.047), and children who spent less than
one hour per day outdoors seemed to have elevated risk (OR 1.23; p=0.044).
Pesticides: Parental occupational exposure to pesticides was not associated
with childhood leukemia. Analysis of household pesticide use only revealed an
elevated risk associated with use of liquid concentrate ant and cockroach
pesticides (OR 3.3; p=0.036).

Other: Other exposures that were associated with childhood leukemia included:
(OR between 1.8 and 5.0) living near various industries (based on potentially
biased parental recall); cigar/pipe smoking by father; child's consumption of ham,
bacon or sausage, hot dogs, hamburgers, coffee, and colas; (OR between 1.0
and 1.8) cigarette smoking by father or any parent; marijuana smoking by non-
parents; black and white TV use by child; and child's consumption of grapefruit or
grapefruit juice and charcoal broiled meats.

Other exposures that were assessed but no association found: skin
application of mosquito repellants; living near industries that emit chemicals into
the air; ionizing radiation to the mother or child; family history of cancer or
leukemia; vehicular traffic exposure (time commuting to school); incense burning
at home; cigarette smoking by mother; various household appliances; moving or

number of residences; schooling in WCP or near contaminated wells; source of

73

drinking water; playing in irrigation water; plastic water pipes in home; public or
private swimming pool use; c-section or other birth-related operations; perinatal
complications; herbicide exposures of parent while in military; parent's
occupation or industry; child's consumption of oranges and orange juice, apples
and apple juice, bologna etc., and milk.

Study Conclusions:

Investigators noted that there were some weak associations, as indicated
above, but their finally conclusion was as follows: "The reason or reasons for the
elevated number of leukemia cases from WCP observed between 1965 and
1986 was not specifically identified in this study". (Flood et al., 1997)

Follow-Up of Childhood Cancer Studyin Maricopa County Standardized

 

Incidence Ratio of Area F to Area R Maricopa County for 1991-1995 (ADHS,

1997)

This study was undertaken to compare incidence rates for childhood

 

leukemia (ages 0-19 years) in a previously defined combined census tract area
that comprises WCP (Area F) in Maricopa County. The comparison standard
area (Area R) was comprised of all of Maricopa County excluding the 4 suspect
areas (including WCP). This study compared childhood leukemia in area F with
area R for the years 1991-1995.

During the period of 1991-1995, childhood leukemia incidence was found
to be no higher in area F than the comparison standard area R. The
standardized rate ratio (SRR) for area F standardized to area R for 1991-1995

was calculated to be 1.04 (95% CI 0.597, 1.812). The observed number of cases

74

for area F was 15 during this time period and 14.43 were expected. The data
indicated that the incidence of childhood leukemia in WCP from 1991-1995 was

not significantly different than area R. (ADHS, 1997).

75

Chapter 5

EPIDEMIOLOGIC CRITIQUE OF THE WOBURN AND WEST CENTRAL
PHOENIX STUDIES

Hypotheses

The multiple studies in Woburn, Massachusetts and West Central Phoenix
(WCP), Arizona advanced various a priori hypotheses to account for the
apparent excess of childhood leukemia in these areas. However, the two regions
differed starkly in the focus of hypotheses. In Woburn, hypothesis generation
quickly- focused on environmental exposures related to the contaminated wells
and the lndustriplex. Conversely, WCP studies took a much broader approach,
in that hypotheses were advanced that addressed virtually all the potential risk
factors for childhood leukemia. However, like the Woburn studies, WCP studies
did tend to focus on environmental exposures.

Initial studies in both Woburn and WCP explored the possibilities of an
epidemic of childhood leukemia through descriptive analysis of mortality and
incidence data. As a related hypothesis, the possibility of residential clustering
was explored by both groups. Woburn and WCP researchers also collected a
variety of demographic and family medical history data for post hoc analysis.

The Woburn investigators then turned to exploring the possible connection
between exposure to contaminated drinking water, soil, and air and the onset of
childhood leukemia as well as other various health outcomes. In WCP, the focus
was then directed at testing as many hypotheses as possible, as evidenced by

the long list of risk factors examined.

76

In Woburn and WCP, none of the studies conducted advanced a single a
priori hypothesis to test the potential for an infectious origin of the excess
leukemia cases. Investigators did make reference to the possibility of an
infectious origin in their reviews and in some anecdotal findings, but they did not
set out to evaluate this important potential causal pathway.

Case Definition

 

The studies conducted in Woburn and in WCP were relatively consistent
in their definition of a childhood leukemia case. Cases were typically defined as
individuals age 0-19 at the time of onset of leukemia and in most cases with a
histiologically confirmed leukemia type. Also, residence at the time of diagnosis
within the community of interest was a criterion. In the case-control studies as
well as some of the descriptive studies, the specific subtypes were noted but not
stratified by type. While the Woburn cases were predominantly ALL and the
WCP more of a mix ALL and AML, neither group of studies treated the subtypes
separately in the study, with one exception. In the final case-control study in
WCP, investigators conducted their proximity to the fuel tank farm analysis by
stratifying on AML and non—AML. This was the only identifiable instance in which
the subtype of leukemia was considered in analysis. In all other instances the
ALL and AML cases were aggregated together into the "childhOod leukemia"
category.

This aggregation of ALL and AML cases was likely done to increase
sample sizes, but is a serious problem in the design of the studies. Both studies

defined cases as children with any type of leukemia, which even included one

77

case of CML in Woburn, and did not stratify their analyses on subtype. As
previously stated, even though many previous childhood leukemia studies have
lumped ALL and AML together, ALL and AML are distinctively different in their
epidemiological breakdown and thus presumably in their etiologies. For this
reason, the results of the Woburn and WCP studies could have been biased
toward the acceptance of null hypotheses.

The Woburn data presented the actual final breakdown of childhood
leukemia cases by subtype over the study period (Table 12), while the WCP data
presented the final breakdown for only a sub-sample of WCP-related participants
that were utilized for the fuel tank farm analysis and sub-typed by only AML and
non-AML (Table 13). However, given the population statistics for childhood
leukemia, it is fair to assume that most, if not all, of the non-AML cases were in
fact ALL cases, and that this sub-sample was representative of other WCP
cases.

Table 12:

Childhood Leukemia Cases in Woburn from 1I69 to 8I89 by Cell Type

 

 

 

 

 

 

Cell Type Cases
Acute Lymphocytic Leukemia (ALL) 17
Acute Myelogenous Leukemia (AML) 3
Chronic Myeldgenous Leukemia (CML) 1
T Total 21

 

 

 

Table 13:

Childhood Leukemia Cases Sub-Sample in WCP 1965 to 1990 by Cell Type

 

 

 

 

Cell Type Cases
Acute Myelogenous Leukemia (AML) 4
Non-AML (likely all ALL) 23
Total 27

 

 

 

 

78

Both of the tables above clearly demonstrate that ALL comprised the
majority of the cases. Thus, it would have been prudent to analyze ALL
separately from the other subtypes.

Case Ascertainment

 

Following the previous case definition, studies in Woburn and WCP
selected cases on certain spatial and temporal criteria. Both study groups
selected cases from specific geographic locations, typically census tracts, and
categorized them by the time period in which incidence or mortality occurred.
Other than the preconceived notions of clustering that influenced the selection of .
certain temporal-spatial criteria, there is nothing clearly wrong with the way in
which cases were ascertained. Because clustering of childhood leukemia cases,
or any cancer for that matter, can occur randomly, the possibility exists that
investigators in both areas could have missed important cases through inclusion
of only certain specific areas. The case ascertainment in Woburn was more
likely to be susceptible to this bias, since researchers focused on the residents
only within the small community of Woburn, while in WCP the entire population of
Maricopa County was taken into consideration.

Control Selection 8. Matching

Controls for both study areas appeared to be selected fairly consistently.
Controls were randomly selected and community-based, and were typically
matched on age and sex and occasionally on other demographic variables.
Some of the controls were also matched on residential status or proximity to the

case, but not all were, as this was one of the main proxies for exposure to

79

various environmental agents. Matching on an exposure variable is clearly ill
advised, as it strongly biases the results towards the null hypothesis. Putative
factors cannot be teased out if they are matched on and therefore masked in the
analysis.

Ascertainment of Exposure to Risk Factors

In Woburn, the main childhood leukemia risk factors considered were
related to the presence of environmental contaminants. The consumption of
water from two contaminated municipal wells, Wells G and H, was the key
exposure evaluated by investigators. The main contaminant of these wells that
was evaluated in the studies was TCE, but other environmental toxins were also
considered. Other potential exposures evaluated included such factors as family
medical history, parental smoking, parental occupation, and breast-feeding.

In contrast to Woburn, the WCP studies did not concentrate on a single
exposure like municipal drinking water. Rather, the WCP investigators examined
virtually every potential childhood leukemia risk factor known. The investigators
evaluated such factors as water quality, air quality, parental smoking, parental
occupation, place of residence, solvent exposure, pesticide exposure, family
medical history, ionizing radiation, types of food eaten, perinatal complications,
and many more. Thus, exposure assessment was much broader in WCP than in
Woburn.

The case-control studies conducted in Woburn and WCP relied heavily on
proxy measures for exposure and on parental recall. In Woburn, parents of

cases and controls were interviewed by trained interviewers, and in most cases

80

by the same interviewer, both by telephone and in person. The ideal was to
conduct an interview with both parents simultaneously in-person. The
interviewers were not blinded to the status of cases and controls and were likely
not blinded to the study hypotheses, as they were quite narrow. Also, in the final
study there was only one main hypothesis to test - exposure to Wells G and H.
In WCP, trained interviewers were not blinded to the status of the cases and
referents, but they were blinded to the numerous study hypotheses. WCP
interviews were conducted primarily by telephone with the parents (one or more
parent or a surrogate) of the study participants. In both Woburn and WCP,
structured interview questionnaires were utilized to collect exposure data on the
cases and controls.

The methods used by investigators in Woburn and WCP were not grossly
biased, but an unblinded interviewer can introduce bias to the results. While it
would be extremely difficult to blind an interviewer to the status of the cases or
control, efforts should have been made in Woburn to ensure that the interviewer
was blind to the hypotheses. Additionally, parental recall of exposure information
was fraught with bias, both because of the time elapsed since the incidence of
the leukemia and the strong preconceived notions of the parents about cause of
their child's leukemia. The latter is most striking in Woburn, where many in the
community had already drawn a causal link in their minds between exposure to
contaminants from the well water and the onset of childhood leukemia. This bias
is evidenced by the intense media coverage at the time of the investigation which

subsequently led to the writing of a book on the Woburn childhood leukemia story

81

which was made into a 1999 feature film "A Civil Action". The media induced
public bias would have made it nearly impossible to have blinded the
interviewers.

In addition to interview data collected from parents, investigators also
collected environmental samples and used proxy measures of environmental
exposures, such as proximity to fuel tank farms or high levels of vehicle
commuting in WCP and proximity to the lndustriplex in Woburn, to assess
exposure to potential putative agents. The collection of environmental samples
was most vigorously conducted in WCP, where investigators collected such
things as carpet dust samples, air and soil samples, and direct magnetic field
measurements within residences. The evaluation of proxy measures and
environmental sampling in both Woburn and WCP was comprehensive, but little
data was collected that would be useful in assessing the infectious etiology
hypothesis.

Ascertainment of Exgosure to ConfoundinLVariables

In WCP and in Woburn, investigators controlled for basic demographic
confounders as well as certain potential environmental confounders.
Demographic variables such as age, sex, race, and socioeconomic status were
matched in most cases so as to control for any covariation. Also, depending on
the study hypotheses, certain environmental exposures were controlled for to
determine if a specific putative risk factor was significantly associated with the
outcome. The final study in Woburn controlled for numerous environmental

factors in order to test the Wells G and H exposure hypothesis.

82

Analytic Methods and Study Conclusions

The descriptive studies in Woburn revealed that there was in fact an
excess of childhood leukemia in the period 1969-1989, which was largely
comprised of ALL cases. In West Central Phoenix, descriptive studies identified
an excess of childhood leukemia, but stated that only AML was statistically
significantly elevated. The case-control studies conducted in Woburn and WCP
utilized a variety of univariate and multivariate statistical methods to analyze the
data collected, but they both relied on the calculation of odds ratios based on
contingency tables to present their findings. These odds ratios are potentially
biased by differential misclassification through recall bias of parents of cases
versus parents of controls. The parents of cases would be more likely to recall
exposures, as their child's illness had caused them to closely focus on what
exposures may have occurred and to leok for reasons to explain the illness.
Conversely, parents of controls would not have necessarily focused on particular
exposures their child may have had, and they would likely be more objective in
recalling exposures. For example, in Woburn parents of cases would have been
strongly biased on recall of their historical exposure to water from the
contaminated wells, and parents of cases in WCP would have likely succumbed
to recall bias on the myriad of potential exposures they were questioned on. This
differential misclassification can lead to an increase or decrease in magnitude of
association and even a reversal in the direction of association which may have
been the case in Woburn and WCP. This differential misclassification may

account some of the seemingly spurious findings in WCP such as the association

83

found between grapefruit consumption and childhood leukemia. The use of
these contingency tables could have also produced results biased towards the
null as a result of the potential for non-differential and independent
misclassification in exposure assessment. Potential for independent
misclassification existed in both Woburn and WCP. In addition, the final Woburn
study implemented a dose-response trend analysis to make their final
conclusions, which is described below. Due to non-availability of the original
data from the Woburn and WCP studies, the analytic methods utilized by the
researchers and the subsequent results cannot be validated for accuracy (i.e.
data cannot be re-analyzed to confirm results). Therefore conclusions made by
the investigators must be viewed with caution for this as well as other reasons.
In Woburn, the final and "conclusive" study conducted to
investigate the excess of childhood leukemia relied on very weak results on
which to base their causal conclusions. The results were for the most part
statistically insignificant, and those that were modestly significant were based on
a very small case sample size of 19 and were not entirely consistent with each
other. The results lacked consistency in that both an association and lack of
association was found between the childhood leukemia and contaminated water.
The study concluded that the incidence of childhood leukemia in Woburn from
1969 to 1989 was associated with the mother's "potential for exposure" to
contaminated water (with TCE as the suspected agent) from Wells G and H
during pregnancy. There are several problems with accepting this conclusion.

First, the chemical trichloroethylene (TCE) has been identified as the most

84

common contaminant of public wells (ADHS, 1997), and the chemical is not
known to be a leukemogen in mammals. Second, earlier studies in Woburn
noted that the number of childhood leukemia deaths began to rise from 1959-
1963 (SMR = 120), before the wells were drilled in 1964, and continued to rise in
the period in which the wells were drilled 1964-1968 (SMR = 169). However, in
the period following well drilling, 1969-1973, the rate was lower than expected
(SMR = 50). Rates did begin to increase again between 1974-1978, but this
increase occurred over ten years after the opening of Wells G and H. The
increase could also have been a result of improved screening methods. Earlier
research also noted that childhood leukemia mortality was not elevated during
the period 1949-1958, but childhood leukemia mortality began to rise in 1959-
1963. Therefore, a factor introduced during the late 1950's would be a more
likely cause (MDPH, November 1981). Third, the odds ratios related to exposure
to water from Wells G and H in utero or in vivo were statistically insignificant, so
no direct association could be made.

Finally, the conclusion that the potential for exposure to the contaminated
water in utero was associated with childhood leukemia in Woburn was based
upon an extremely weak analysis. After finding no significant association with
direct exposure to well water, investigators parameterized the cases in three
exposure categories: Never, Least, or Most exposed. The adjusted odds ratios
for the Least and Most categories relative to the Never group were as follows:
Least (OR = 3.53; 95% Cl. 0.22, 58.14), Most (OR = 14.30; 95% Cl. 0.98,

195.60). Even though the Most category flirted with statistical significance, the

85

confidence interval was cavernous. Lacking statistically significant results utilizing
the trichotomous parameterization method, investigators turned to trend analysis,
finding a slightly significant trend across the three groups. Basing a conclusion
on a barely significant trend analysis of categorical variables that were not
statistically significant on their own, calls that conclusion into question. The
investigators also appear to have misused trend analysis by basing the trend on
only three data points. Consequently, I assert that the "conclusive" findings of
the MDPH in Woburn are anything but conclusive.

The findings in final case-control study in WCP are also based upon
somewhat inconclusive evidence. First, the investigators took an unfocused
approach in the study of the childhood leukemia cases in WCP and surrounding
Maricopa County. Investigators set out to test virtually every hypothesis they
could think of, with the exception of the infectious etiology hypothesis. Thus, it is
not surprising that they found a few weak associations. Given the number of
variables evaluated and analyzed, one would expect to find a few such
associations just by random chance, due to the inherent probability of detecting
errant associations in data analysis. The investigators made multiple
comparisons which can increase the overall probability of incorrectly rejecting a
null hypothesis, as statistical significance includes a probability of error (e.g. p =
0.05 or 5% chance of error). The weak associations identified were also not
necessarily biologically coherent, consistent with previous research, or intuitive.
For example, the finding that living within 3 or 5 miles of the gasoline tank farm

was weakly associated with childhood leukemia fails for the several reasons. (1)

86

It is inconsistent that there was less risk associated with living within 2 miles
versus 3 miles and that there was a risk for living within 5 miles but no risk
associated with living within 4 miles of the tank farm. (2) While investigators did
separate the cases into AML and non-AML in this particular analysis, their
findings were based on the aggregated childhood leukemia cases. (3) Benzene
is the Ieukemogenic agent in gasoline, but it has only been associated with AML
and not ALL, as noted earlier in this paper, and the findings were based on a
large number of ALL cases. The WCP case-referent study also had spurious
findings. Parental smoking by the father was associated with leukemia, but
smoking by the mother was not associated. The finding that use of power tools
was a protective factor (OR 0.33; 95% Cl 0.17 to 0.66) is seemingly nonsensical,
although it could be that this is some sort of marker for another protective factor
such as high socioeconomic status. These examples, along with the weak
association found with random food products and other risk factors, demonstrate
the shortcomings of examining a childhood leukemia cluster by making multiple
comparisons.

While studies in Arizona apparently took far too broad an approach to
examining the excess of leukemia among children in WCP, it is also clear that
studies in Woburn too quickly focused on specific environmental exposures, to

the detriment of other potential putative risk factors.

87

Chapter 6

CONCLUSIONS

Descriptive 8: Analytic Epidemiological Methods: Woburn 8. WCP Studies
The CDC developed a systematic and integrated approach for responding
to reports of cancer or other disease clusters in 1990. The method involves four
basic stages as follows:
Stage 1. Initial Contact and Response - The purpose of this preliminary stage is
to collect information from the person(s) or group(s) first reporting a perceived
cluster in a specific location.
Stage 2. Assessment - This stage involves the evaluation of two concurrent
issues: ( 1) whether an excess has actually occurred and (2) whether the excess
can be linked etiologically to some exposure. The first issue must be
satisfactorily evaluated in order to proceed to the second issue, which may or
may not be clear at this stage in the cluster investigation.
Stage 3. Major Feasibility Study - The purpose this portion of the investigation is
to determine the feasibility of performing an epidemiologic study linking the
health event and a specific putative exposure.
Stage 4. Etiologic Investigation - This final phase of the investigation, which is
followed by analysis of the data collected, is intended to elucidate the potential
disease-exposure relationship. This step is a standard epidemiologic study, for

which all the preceding effort has been preparatory. (CDC, 1990)

88

 

Evaluating the childhood leukemia studies in Woburn and West Central
Phoenix (WCP) in light of the CDC's cancer cluster guidelines reveals some of
the strengths and deficiencies of these studies. Health officials in both
Massachusetts and Arizona apparently satisfactorily implemented Stage 1.
There was a concerted response to the allegations of the presence of childhood
leukemia clustering in these areas. After proceeding to the assessment phase in
Stage 2 both Woburn and WCP investigators employed basic descriptive
techniques. The study of the distribution of childhood leukemia in Woburn and
WCP was conducted through evaluation of person, place and time relative to
clustering. The investigators did satisfactorily identify the types of childhood
leukemia cases and the spatial-temporal relationship of these cases in
determining if excesses of childhood leukemia did in fact exist. The investigators
in both study groups conducted multiple analyses of both childhood leukemia
incidence and mortality to determine the nature of the clustering. After
succeeding in determining the presence of elevated ALL rates in Woburn and
elevated AML rates in WCP, the methodology began to break down.

It is apparent that as the investigators were in the analytic planning phase,
equivalent to Stage 3, a major bias was introduced. In order to make the studies
more feasible, investigators increased the sample sizes by aggregating ALL,
AML and even CML into the definition a childhood leukemia case. This
misclassification of caseness would later diminish the quality of the study results

as these subtypes of childhood leukemia have distinctive etiologies.

89

Fifima.

The subsequent etiologic investigation, Stage 4, in both Woburn and WCP
went even further astray of sound epidemiology and science that one would find
in most peer-reviewed studies. In spite of the advanced nature of the infectious
etiology hypothesis at the time the final conclusive studies were being conducted,
there was not a single a priori hypothesis to examine the potential of an
infectious origin. The investigators were also prone to inductive reasoning in
quickly focusing on environmental exposures to the detriment of other potential
putative agents. This was most pronounced in the Woburn studies, that were
very narrowly focused on the well water contamination. Not only was this a very
narrow approach to investigating the leukemia cluster, but it was also bad
science in that there was no biologically coherent evidence to suggest that TCE
exposure could even be Ieukemogenic.

Investigators in both Woburn and WCP failed to take a deductive
approach with strict hypothesis formulation and testing, and they also failed to
utilize statistical and mathematical modeling techniques (see example on page
101) that could have elucidated an undetected association. The haphazard
hypothesis testing in WCP and the narrow and unscientific hypothesis testing in
Woburn led to truly inconclusive results.

Support for the Infectious Etiology Hypothesis: Literature Review

Kinlen asserted that:

1) The influx of individuals into an isolated community produces conditions in

which leukemia is more likely to occur.

90

2) Isolated communities (susceptible individuals) newly exposed to migrants
(infected individuals) from elsewhere could experience an unusual exposure to
some hypothetical infectious agent or to otherwise common infections for which
no immunity happens to exist.

3) This exposure, in turn, increases the risk of childhood leukemia.

4) Consequently, isolated communities may experience "mini-epidemics”
because they are too small to maintain common infections in endemic form,
which could lead to a cluster of ALL (Kinlen, 1988).

Greaves asserted that ALL may be due to two separate genetic events:
1) spontaneous mutation of a pre-B-cell in utero
2) proliferation of the mutated pre-B-cell when exposed to a later antigenic
challenge (i.e. an infectious agent) (Greaves, 1988)

Freda Alexander advanced an Aberrant Response Model (ARM) that
unifies the Kinlen and Greaves models. The ARM states that a substantial
portion of ALL cases arise as a rare host response to certain patterns of
exposure to common infectious agents. (Alexander, 1993).

Support for an infectious etiology hypothesis was identified in numerous
studies. An association between population mixing and childhood leukemia was
observed in Hong Kong (Alexander et al., 1997), Greece and Italy (Kinlen and
Petridou, 1995), and the United Kingdom (Kinlen and Hudson, 1991) (Langford,
1991) (Alexander et al., 1990) (Alexander et al., 1992) (Alexander, 1992). Kinlen
also identified increased childhood leukemia trends in cities within the UK that

had greater commuting levels (Kinlen et al., 1991) (Kinlen, 1993). Recent

91

 

evidence has also supported the infectious etiology by indicating that childhood
leukemia epidemics arise regularly in moderately densely populated areas and
also sporadically in areas which are somewhat less densely populated
(Alexander et al., 1999). Also, in a more recent study there was highly significant
evidence of space-time clustering based on place of birth and time of diagnosis,
particularly for ALL cases age 0-4 years (Birch et al., 2000).

The potential infectious etiology of childhood leukemia is additionally
supported by observations of birth order and parental immune system function
(James, 1990) (Meelemkjaer, Alexander and Olsen, 2000). The role of viral
agents in the development of other human cancers and suspected ALL putative
infectious agents also supports infectious etiology hypothesis (Morris, Eddleston
and Crook, 1995) (T oren et al., 1996) (McCool, 1996) (Smith et al., 1997)
(Davies and Ross, 1997).

Support for the Greaves model has been found in the growing body of
evidence around the possibility of in utero infectious origin with the identification
of the potentially antigen induced non-hereditary TEL-AML1 gene fusion
abnormality in children with ALL (Greaves et al., 1997) (Felix, 1999) (Wiemels et
al., 1999) (Magalhaes et al., 2000) (Knox et al., 1983). Leukemogenic animal
models of infectious origin have also been elucidated (Goldman and Jarrett,
1984).

Evaluation of individual causal criteria demonstrate the rigor of the
infectious etiology in light of the evidence in the literature. The first criteria of

time order (infection precedes ALL) seems to be met in many of the ecological

92

 

studies. The proxies for exposure, such as population influx and population
mixing, have been demonstrably shown to precede "epidemics" of ALL in
children; and there is evidence of dose-response relationship through
identification of greater risk in areas with much higher population increases over
a defined period (Kinlen, 1988) (Kinlen et al., 1990) (Alexander et al., 1990)
(Kinlen & Hudson, 1991) (Kinlen et al., 1991) (Alexander et al., 1991) (Langford,
1991) (Alexander, 1992) (Alexander, 1993) (Greaves & Alexander, 1993) (Kinlen
& Petridou, 1995) (Alexander et al., 1996) (Alexander et al., 1997). These same
studies have an inherent weakness however, in that the applicability of the
findings to the individual level would be fallacious. The ecological nature of the
previously noted studies, while they demonstrate the potential mechanism of an
infectious origin, precludes the necessary conclusion that an individual case of
ALL could have resulted from an infection linked to population mixing.
Nevertheless, the consistency of the ecological studies within themselves and
with other biological infectious models would indicate that the hypothesis is valid.
The biologic and epidemiologic evidence that supports the infectious
hypothesis has been most clearly demonstrated in animal Ieukemogenic models
and other human cancer models (Morris, Eddleston and Crook, 1995) (Toren et
al., 1996) (McCool, 1996). The coherence and consistency of the studies and
evidence withstanding, the strength and specificity of association between
infection and childhood ALL does not provide conclusive support for the
hypothesis. The fundamental problem is the lack of understanding of the

potential infectious agent(s) that precipitates the disease. Without knowing what

93

 

the specific agent(s) is, no specificity of association can even be considered, and
likewise strength of association is nearly non-existent when the specific exposure
is a mystery. The continued research to identify a specific agent(s), such as the
JC virus or an EBV-type virus, should rectify this in the future.

The other concern from the hypothesis supporting studies is the
laboratory-based studies. Sample sizes in these studies are critically lacking.
Reliance on very few subjects has provided the evidence we have to date on
potential agent(s) and possibilities of in utero exposure (Smith et al., 1997)
(Greaves et al., 1997) (Felix, 1999) (Wiemels et al., 1999). However, while we
lack conclusive evidence of specific infectious agents, the hypothetical modeling
of ALL clusters when certain population mixing criteria have been met seems to
intuitively point a causal finger at an infectious origin.

Support for the Infectious Etiology_flypothesis: Woburn 8. WCP Studies

 

Review of the study results from Woburn and WCP indicate that there is at
least some anecdotal evidence in support of the infectious etiology hypothesis
present in these reports.

In Woburn, one of the earlier studies revealed that four of the ALL cases
to date had been born in the same 6-month period (CDC, 1981). This indicates
that there was potential for some common in utero exposure. While investigators
in Woburn would have interpreted this as the potential for exposure to some sort
of toxin, it could also be argued that the in utero exposure could have been an
antigenic challenge in concordance with Greaves' hypothesis. Another finding in

Woburn that potentially implicates a challenge of the child's immune system was

94

the statistically significant association identified between developing childhood
leukemia and being breast-fed as a child (OR = 10.17; Cl. 1.22, 84.58) (MDPH,
1997). At first glance it may seem counter-intuitive that breast-feeding would be
a risk factor, in that if there was an infectious agent responsible for the cases of
ALL, breast-feeding should be protective as a result of the antibodies passed
from mother to child. However, since the infectious etiology of ALL could be the
result of a combination of antigenic challenges or even a lack thereof, it is
plausible that breast-feeding could increase risk by reducing the necessity of the
child's immune system to respond to new antigenic challenges. In this scenario,
the antibodies conferred to the child from breast-feeding may actually make the
child susceptible to leukemogensis because the child did not uniquely experience
the antigen or antigens and develop antibodies on their own.

In the WCP studies it was found that children who spent less than one
hour per day outdoors seemed to have an elevated risk for childhood leukemia
(OR 1.23; p=0.044) (ADHS, 1997). This finding suggests a potential lack of
antigenic challenges as a child, or at the theoretical critical times, due to minimal
mixing with peers through playing outside. This finding could also be looked at
from another direction by saying that the child would experience more
opportunity for exposure to potentially Ieukemogenic infectious agents due to
spending more time indoors in crowded situations with less ventilation. Another
interpretation of this result, from an infectious etiology point of view, would be
that the child is generally less healthy due to a more sedentary lifestyle (or

sedentary childhood could be a marker for a sickly child who could not go out to

95

 

play), which could make the child's immune system more susceptible to an
infectious agent.

Admittedly, these findings and the subsequent interpretations of them are
anecdotal and conjectural, but this new approach could lead to future hypothesis
generation in the exploration of the infectious etiology hypothesis.

The Infectious EtiologjLI-Iypothesis - Evaluation of Listed Objectives in

 

l_._ight of Woburn and WCP

 

2.1) Evaluate if Woburn and West Central Phoenix were relatively isolated from
newcomers and did not experience a great deal of population mixing prior to the
ALL epidemic.

This hypothesis was not supported by any information or data collected

within the Woburn and WCP studies. The two communities are both suburban
and quite close to major urban centers, so it is unlikely that indigenous residents
could have been isolated to any appreciable extent. The City of Woburn
currently only has 35,943 residents, but it resides in Middlesex County, which
has a population of 1,417,868. Likewise, WCP is within Maricopa County, which
has a population of 2,696,198. Thus, it is unlikely that either community was
isolated prior to the excesses of childhood leukemia in the late 1960's.
2.2) Evaluate if Woburn and West Central Phoenix experienced (a) an influx of
newcomers and (b) experienced increased incidence of ALL in children (age 0 to
19) following an influx of newcomers.

(a) There were no data collected or evidence available to suggest that an

appreciable influx of newcomers took place prior to the increases in ALL

96

incidence in WCP. However, in Woburn there was a significant increase in
population between 1940 and 1960. The population increased by 58% in this
two decade period, jumping from 19,751 in 1940 to 31,241 by 1960.

Presumably, the majority of this increase was a result of urban dwellers in Boston
and other metropolitan areas relocating to a more rural location. (b) A significant
increase in ALL did occur in Woburn, but only AML was significantly increased in
WCP compared to national averages.

2.3) Evaluate if Woburn and West Central Phoenix experienced significant
mixing of newcomer and indigenous populations prior to and during the elevated
incidence of ALL in children (age 0 to 19).

No data were available in the subject studies to support this hypothesis,
and there is no way to determine if newcomers did mix with indigenous
populations, as these two groups are not quantifiable. However, a crude proxy
for mixing in general is the current population density. In Woburn, there are
2,835 people per square mile and in Maricopa County there are 293 people per
square mile (this number is low due to the large land mass of counties in
Arizona). Thus, in Woburn it is reasonable to assume that a fair amount of
population mixing occurs and has occurred in the past. Also. in 1960 the Woburn
section of route 93 opened which allowed more commuting and influx of non-
Woburn residents. This too would have likely led to greater population mixing
than previously experienced and this occurred prior to the epidemic of childhood
ALL. In WCP it is also likely that population mixing does and has occurred much

more than in rural areas.

97

 

 

2.4) Evaluate if the cases of ALL occurred primarily in children (age 0 to 19) born
in Woburn and West Central Phoenix and whose parents were indigenous to the
community (susceptibles) or vice versa in the case of susceptibles moving into
an infected population (urban to rural or rural to urban).

The data from the final Woburn study neither support nor refute this
hypothesis. At least 10 of the 21 cases of childhood leukemia were born in
Woburn. This is based on the number of cases that were exposed to water from
Wells G and H in utero, but it does not rule out the possibility that the remainder
of the cases were born in other areas of Woburn. Data from the final case-
control study in WCP does not support this hypothesis in that only 10 of 23 cases
diagnosed in WCP were actually born in WCP. Of course it is not specified what
number of the 10 born in WCP or the 13 not born in WCP were actually ALL
cases. Also, these 23 cases were not all of the cases in WCP as they were only
the cases used in one of the residential comparisons (this was the only source of
this type of data in the WCP studies).

2.5) Evaluate if the increased incidence of ALL in children (age 0 to 19) Woburn
and West Central Phoenix was a result of an aberrant response to an infectious
agent(s) (in utero or in vivo) introduced by the newcomers.

This hypothesis is not generally supported by any of the study results, with
the exception of the previously noted evidence under the "Support for the

Infectious Etiology Hypothesis" heading.

98

2.6) Evaluate whether susceptible indigenous children exposed to the infectious
agent(s) at the ages of 3-4 were more likely to develop ALL according to the
typical peak in ALL incidence at that age.

This hypothesis does have some support from the Woburn and WCP
studies. In Woburn, 9 of the 21 childhood leukemia cases (43%) occurred in the
0-4 age group. This was the largest percentage for any of the 5-year age groups
(Table 14). In WCP, there were 18 childhood leukemia cases identified that were
born in WCP, and of these seven (39%) were diagnosed at the age of 3 or 4
(Table 15).

Table 14:

Childhood Leukemia Cases in Woburn from 1169 to 8I89 by Age Group

 

at Cases
0-4 rs 9
5-9 rs 4

10-14 7
1 5-19 1
Total 21

 

99

Table 15:

Childhood Leukemia Cases Born in WCP 1965 to 1990 by Age

t nosis Cases

{Dm‘lm‘fl-FNNAO

f8

 

0
1
0
2
5
1
0
2
1
1
1
1
0
1
0
0
0
1
1
0
18

Total

Conclusions and Recommendations
To date, no one has able to demonstrate the Kinlen model of rapid

exposure of susceptibles, observed in the UK, Hong Kong, Italy and Greece, or
the inverse of it in North America. It could be that there are cases of infected
individuals moving into susceptible communities in North America as well, but it
is quite possible that infectious model follows the inverse approach in North
America. It would appear that both WCP and Woburn were close enough to
urban centers that, rather than the communities having a high proportion of

susceptibles at a critical age for leukemogenesis, they could have had a high

100

 

 

proportion of the Ieukemogenic infection(s) at a non-critical age. Thus new age-
speclfic susceptibles could have moved into these communities and later
developed the aberrant response of ALL. This model seems to fit well with the
Greaves model of an in utero antigenic exposure or exposures. Clusters of cases
are observed in suburban areas, to which susceptibles from rural locations would
be likely to migrate. Comprehensive population-based studies and mathematical
modeling could make answering these questions possible. The following
description of my proposed mathematical simulation model is an example of the
epidemiological techniques that could have been utilized in Woburn and WCP
and that can be implemented in the future:

The mathematical simulation model would be developed to mirror the
processes of a typical infectious agent. The reference agent would be the
Epstein-Barr Virus, which is hypothesized to be similar to the ALL agent if not the
actual agent. The model would generate an over-representation of cases in the
2 to 5 range to mimic the known ALL peak incidence. This would be done to
determine if the modeled peak coincrdes with the actual peak in the population
given the assumed time of exposure. The model would assume that rural
populations would have low level endemnicity of the EBV-type agent and urban
populations have high endemnicity of the EBV-type agent. This allows the urban
children to be infected prior to the critical age (1-3 years) that predisposes a child
to ALL, while rural children remain susceptible. This would allow the model to

determine age-specific susceptibility.

101

The model would be compartmentalized to predict the dynamic interaction
between parasitic and host populations. The host population would be
segregated into susceptible, latent but not yet infectious, infected, and recovered-
and-immune individuals. The basic reproductive rate, known threshold host
densities, and direct & indirect transmission routes of EBV would be factored into
the mathematical model as well. The model would also take into consideration
the threshold host densities for maintenance of EBV in a population through
deterministic means (influx of population) and stochastically from either epidemic
or endemic fade-out.

Ultimately the model would consist of dynamical equations that would
produce a predicted outcome of ALL cases based on the relevant
epidemiological and demographic parameters selected for the analysis. These
parameters include but are not limited to the following: baseline community
population, childhood ALL Incidence statistics, percentage of population increase
in 5-year increments, percentage change in population density, number of births
and families in a community, poverty level and per capita income, education
level, and commuting levels. The model would conform to the guidelines set
forth in Infectious Diseases of Humans: Dynamics and Control by Roy M.
Anderson and Robert M. May, with specific reference to Chapter 6 (The basic
model: dynamics).

The mathematical simulation model would produce an expected ecological
outcome and a statistical comparison would be made between the expected and

observed ecological outcome. The comparison for the observed cases in the

102

 

 

 

final analysis would be the predicted “cases” according to the mathematical
model. If no significant difference exists when the observed and predicted data
are compared then it would be interpreted as support. If a difference does exist
then results would be viewed as support for the null hypothesis rather than
merely a conclusion of a fault in the model. The use of this or other
mathematical models could significantly advance the elucidation of the childhood
ALL infectious etiology hypothesis in North America and elsewhere.

In addition to demonstrating the need to use modeling to evaluate the
infectious etiology hypothesis, the childhood leukemia studies conducted in
Woburn and West Central Phoenix illuminate the shortcomings of cluster
analysis. Childhood leukemia cluster studies are hampered by deficient sample
sizes for statistical analysis, as well as strong sociological biases within the
community and political pressures to find an answer quickly. Also, childhood
leukemia cluster analyses in Woburn, WCP, and elsewhere have all failed to
identify a single putative agent. To circumvent these negative aspects of cluster
analysis, large population-based case-control studies must be implemented to
increase sample sizes and test the infectious etiology hypothesis. Additionally,
the erroneous results from making multiple comparisons in the WCP studies
demonstrates the need to generate strict a priori hypotheses for testing in future
studies. Future studies must also avoid aggregating ALL with AML and other
leukemia cell types, to account for the separate and distinct etiologies.

The Woburn and WCP clusters were exhaustively studied for over a

decade and at great expense, but no truly conclusive results were obtained. The

103

 

data from these studies should therefore be made available to other researchers
for viable hypothesis testing. The datasets should not be unavailable, rather they
should be downloaded to some sort of centralized clearinghouse such as SEER.
Future studies need to be directed away from specific clusters and need to focus
on the population as a whole. Careful retrospective and prospective studies
need to be undertaken to assess the potential infectious etiology of ALL, with
comprehensive collection of data on residence history, mobility, population
density, and level of interaction or mixing with others in the community. This will

facilitate the analysis of the infectious etiology of childhood ALL.

104

APPENDICES

105

 

APPENDIX A

LOCATION OF WOBURN, MA AND WCP, AZ ON U.S.MAP

Figure 1: Location of Woburn, MA and West Central Phoenix, AZ on U.S. Map

 

106

 

APPENDIX B

MAPS OF WOBURN, MASSACHUSETTS

Figure 2: Location of Woburn, MA on Massachusetts Map

 

 

.1“! was
.'v

 

Ar 5..

' g‘. ,
a. ‘ . I I." . ( "
i.’ V ’\ %§I£IE°\. 1"

107

 

APPENDIX C

MAPS OF WEST CENTRAL PHOENIX, ARIZONA

Figure 4: Location of West Central Phoenix, AZ on Arizona Map

 

{ UT H ‘ OI in)

I NEVADA \ ‘
_ _‘ lake Mead ‘

{um-e- .
’ b
. ,J ,
y s. U
M?

   
   
 
 
 
 
  
   

. .’
0 I ‘ Mlle:
j._..—_,—_i
0 IWKflo our!

'2 «5'

 

  
  

CALIFO' "IA
,. ‘ . N WME ' C0
aim—1.1. ‘ > , 3'». .63.“. .. ‘
._‘ , " ,‘ .ivb ij‘l ',.",’ .Jl'n} ‘
SI. . -- ‘. ,- 7 3‘31. ~ "
53”,?" :-‘ E ‘1'}; ‘--" ' 14.5%”. “I, I
i. .i- a i. 1. , ‘41}lT’ w“: J
‘rfil‘fl‘aftlr

1 1 Q
xi:
I a
— 19“,“)
N
X

'J ”01"" O-‘LW‘W?

Q- 2‘ 1:53:13“

Figure 5: West Central Phoenix, Arizona Map

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

at" -/ .1
is“ - ii "i .
i .' I. 1
0 Dr I ‘ ' “‘ ' E
:95 “I .2510," o c m
w t ,' 0
., Uzud ' iiii __J'
. Bull- . _
Von to m
j m! i I/
m _ -
°<ltriin , :_ 51 T4
ulnrnul'inih " I u
'mu?h‘?m m; “1 ' on r e
Yum , g g {l WCP
Perry» e G r cMIeId _ “3
/"T j E.
, eatewpe m I _ __... - °.
r z 5 2 woifder mo 1 kl 3
Co-op VIllueo : m r [Til l—
E L I
Salntjohni:lriIbn4 MODEL. 9 .

 

 

€41)

l

l
’Faiea

gullihowvuley 5"“‘I7mio "LEM
‘\ - Harinqua sun,

3 __, ___, .._ ,
naketowno
6Mulwpl all:

108

l3}—

#5

 

 

LITERATURE CITED

109

Literature Cited

Aickin M, Chapin CA, Flood TJ, Englender SJ, Caldwell, GG. (1992).
Assessment of the spatial occurrence of childhood leukemia mortality using
standardized rate ratios with a simplelinear Poisson model. International Journal
of Epidemiology, 21, 649-655.

Alexander FE. (1999). Clusters and clustering of childhood cancer: a review.
European Journal of Epidemiology,15(9):847-52.

Alexander FE, Boyle P, Carli PM, Coebergh JW, Ekbom A, Levi F, McKinney PA,
McWhirter W, Michaelis J, Peris—Bonet R, Petridou E, Pompe-Kirn V, Plesko l,
Pukkala E, Rahu M, Stiller CA, Storm H, Terracini B, Vatten L, Wray N. (1999)
Population density and childhood leukaemia: results of the EUROCLUS Study.
European Journal of Cancer, 35(3):439-44.

Alexander, F.E., Chan, L. C., Lam,, T. H., Yuen, P., Leung, N. K., Ha, S. Y.,
Yuen, H. L., Li, C. K., Li, C. K., Lau, Y. L., and Greaves, M. F. (1997). Clustering
of childhood leukaemia in Hong Kong: association with the childhood peak and
common acute lymphoblastic leukaemia and with population mixing. British
Journal of Cancer, 75 (3), 457-463.

Alexander, F. E., Leon, D. A., and Cartwright, R. A. (1996). Isolation, car
ownership, and small area variation in incidence of acute lymphoblastic
leukaemia in children. Paediatric and Perinatal Epidemiology, 10, 411-417.

Alexander FE. 1993. Viruses, clusters and clustering of childhood leukemia: a
new perspective? European Journal of Cancer, 29 (10), 1424-1443.

Alexander FE. 1992. Space-time clustering of childhood acute lymphoblastic
leukaemia: indirect evidence for a transmissible agent. British Journal of Cancer,
65, 589-592.

Alexander FE, McKinney PA, Moncrieff KC, Cartwright RA. 1992. Residential
proximity of children with leukaemia and non-Hodgkin's lymphoma in three areas
of Northern England. British Journal of Cancer, 65, 583—588.

Alexander FE, Cartwright R, McKinney PA, Ricketts TJ. 1990. Investigation of
spatial clustering of rare diseases: childhood malignancies in North Humberside.
Journal of Epidemiology and Community Health, 44, 39-46.

Alexander FE, Ricketts TJ, McKinney PA, Cartwright RA. 1990. Community

lifestyle characteristics and risk of acute lymphoblastic leukaemia in children.
The Lancet, 336, 1461-1465.

110

Anderson RM, May RM. 1991. Infectious Diseases of humans: Dynamics and
Control. Oxford University Press.

Arizona Department of Health Services. (1997). Follow-up of childhood leukemia
incidence rate in Maricopa County, standardized incidence ratio of Area F to
Area R for 1991-1995.

Arizona Department of Health Services. (1993). Follow-up of childhood leukemia
incidence rate in Maricopa County, standardized incidence ratio of Area F to
Area R for 1987-1990.

Auvinen A, Linet MS, Hatch EE, Kleinerman RA, Robison LL, Kaune WT,
Misakian M, Niwas S, Wacholder S, Tarone RE. (2000). Extremely low-
frequency magnetic fields and childhood acute lymphoblastic leukemia: an
exploratory analysis of alternative exposure metrics. American Journal of
Epidemiology, 152(1), 20-31 .

Beral, V. (1990). Childhood leukemia near nuclear plants in the United
Kingdom: the evolution of a systematic approach to studying rare disease in
small geographic areas. American Journal of Epidemiology, 132 (1), $63-$68.

Birch JM, Alexander FE, Blair V, Eden OB, Taylor GM, McNally RJ. (2000)
Space-time clustering patterns in childhood leukaemia support a role for
infection. British Journal of Cancer, 82(9):1571-6.

Caldwell GG. (1990). Twenty-two years of cancer cluster investigations at the
Centers for Disease Control. American Journal of Epidemiology, 132(1
Suppl):S43-7.

Cave, H., van der Werff ten Bosch, J., Suciu, S., Guidal, C., Waterkeyn, C.,
Otten, J., Bakkus, M., Thielemans, K., Grandchamp, B., and Vilmer, E. (1998).
Clinical significance of minimal residual disease in childhood acute lymphoblastic
leukemia. European Organization for

Research and Treatment of CanceruChildhood Leukemia Cooperative Group.
New England Journal of Medicine, 339(9), 591-598.

Centers for Disease Control. Guidelines for Investigating Clusters of Health
Events. MMWR 1990; 39(No. RR-11): 1-16.

Centers for Disease Control: Chronic Disease Division, Center for Environmental
Health. (1981). CDC: Cancer in Woburn.
http:llwww.state.ma.us/dph/beha/woburn.htm

Cline, M. J. (1994). The molecular basis of leukemia. The New England Journal
of Medicine, 330 (5), 328-336.

111

Conglio, S., Blackman, J. (1995). Developmental outcome of childhood
leukemia. Topics in Early Childhood Special Education, April 1, 1995, 19.

Cutler JL.(1984). Childhood leukemia, Woburn, MA: an update.
http://www.state.ma.us/dph/beha/woburn.htm

Daniels, J., Olshan, A., and Savitz, D. (1997). Pesticides and childhood
cancers. Environmental Health Perspectives, 105(10), 1068-1077.

Davies, S. M. and Ross, J. A. (1997). ALL - ls JC virus to blame? Causes of
Childhood Cancer Newsletter, 8 (3)

Draper, 6., Little, M., Sorahan, T., Kinlen, L., Bunch, K., Conquest, A., Kendall,
G., Kneale, G., Lancashire, R., Muirhead, c., O’Connor, 0., and Vincent, T.
(1997). Cancer in the offspring of radiation workers: a record linkage study.
British Medical Journal, 315, 1181-1188.

Felix CA. 1999. Evidence of an early start for common acute lymphoblastic
leukaemia. The Lancet, 354, 1486-1487.

Flood TJ, Meaney FJ, Vertz PD, Laubham KA, Porter RS, Aickin M. (1997).
Case-referent study of childhood leukemia in Maricopa County, Arizona, 1965-
1990 Arizona Department of Health Services.

Flood TJ, Aickin M, Lucier JL, Petersen NJ. (1990). Incidence study of childhood
cancer in Maricopa county: 1965-1986, final report. Arizona Department of
Health Services.

Flood TJ, Chapin CA. (1988). Report on mortality in Maricopa County 1966-1986.
Arizona Department of Health Services.

Gilman EA, McNally RJ, Cartwright RA. (1999) Space-time clustering of acute
lymphoblastic leukaemia in parts of the UK (1984-1993). European Journal of
Cancer, J35(1):91-6.

Goldman, J. M. and Jarrett, O. (1984). Mechanisms of Viral Leukaemogenesis,
Churchill Livmgstone: Edinburgh, London, Melbourne and New York.

Greaves, M. F. and Chan, L. C. (1984). Epidemiology of Leukaemia and
Lymphoma, Pergamon Press: Oxford, New York, Toronto. Sydney and Frankfurt.

Greaves MF et al. Dec. 8, 1997. Proceedings of the National Academy of
Sciences.

Greaves MF. 1988. Speculations on the cause of childhood acute lymphoblastic
leukemia. Leukemia, 2 (2), 120-125.

112

Greaves MF, Alexander FE. 1993. An infectious etiology for common acute
lymphoblastic leukemia in childhood? Leukemia, 7(3), 349-360.

Hatch, M. and Susser, M. (1990). Background gamma radiation and childhood
cancers within ten miles of a U.S. nuclear plant. lntemational Journal of
Epidemiology, 19 (3), 546-552.

Heath CW. The epidemiology of leukemia. In: D Schottenfeld, ed. Cancer
Epidemiology and Prevention. Springfield IL: Charles C. Thomas, 1974:318-
350.

Henshaw, D. L., Eatough, J. P., and Richardson, R. B. (1990). Radon as a
causative factor in induction of myeloid leukemia and other cancers. The Lancet,
335, 1008-1012.

Hjalmars U, Gustafsson G. 1999. Higher risk for acute childhood lymphoblastic
leukaemia in Swedish population centres 1973-94. British Journal of Cancer, 79
(1), 30-33.

Hoffmann, W., Dieckmann, H., and Schmitz, F. (1997). A cluster of childhood
leukemia near a nuclear reactor in northern Germany. Archives of Environmental
Health, 52(4), 275-280.

Hoffmann, W., Kranefeld, A., and Schmitz-Feuerhake, I. (1993). Radium-226-
contaminated drinking water: hypothesis on an exposure pathway in a population
with elevated childhood leukemia. Environmental Health Perspectives
Supplements, 101 (3), 113-115.

Jackson, E. W., Norris, F. D., and Klauber, M. R. (1969). Childhood leukemia in
California-bom twins. Cancer, 23 (4), 913-919.

James, W. H. (1990). Further evidence for the hypothesis that one cause of
childhood leukaemia is infection. Paediatric and Perinatal Epidemiology, 4, 113-
1 17.

Kinlen, L. J. (1988). Evidence for an infective cause of childhood leukaemia:
comparison of a Scottish New Town with nuclear reprocessing sites in Britain.
The Lancet, 1323-1327.

Kinlen, L. J. (1993). Population mixing and excess of childhood leukaemia.
British Medical Journal, 306, 930.

Kinlen, L. J. (1997). Infection and childhood leukaemia near nuclear sites. The
Lancet, 349, 1702.

113

Kinlen, L. J., Clarke, K., and Balkwill, A. (1993). Paternal preconceptional
radiation exposure in the nuclear industry and leukaemia and non-Hodgkin’s
lymphoma in young people in Scotland. British Medical Journal, 306, 1153-1158.

Kinlen, L. J., Clarke, K., and Hudson, C. (1990). Evidence from population
mixing in British New Towns 1946-85 of an infective basis for childhood
leukaemia. The Lancet, 336, 577-582.

Kinlen, L. J. and Petridou, E. (1995). Childhood leukemia and rural population
movements: Greece, Italy, and other countries. Cancer Causes and Control, 6,
445-450.

Kinlen LJ, Hudson C. 1991. Childhood leukaemia and poliomyelitis in relation to
military encampments in England and Wales in the period of national military
service, 1950-63. British Medical Journal, 303, 1357-1362.

Kinlen LJ, Hudson CM, Stiller CA. 1991. Contacts between adults as evidence
for an infective origin of childhood leukaemia: an explanation for the excess near
nuclear establishments in West Berkshire?

Kleinerman RA, Kaune WT, Hatch EE, Wacholder S, Linet MS, Robison LL, Niwa
S, Tarone RE. (2000). Are children living near high-voltage power lines at
increased risk of acute lymphoblastic leukemia? American Journal of
Epidemiology, 151(5), 512-515.

Knox, E. G. and Gilman, E. (1992). Leukemia clusters in Great Britain. 1.
Space-time interactions 2. Geographical concentrations. Journal of
Epidemiology and Community Health, 46, 566-576.

Knox, E. G., Marshall, T., and Barling, r. (1984). Leukaemia and childhood
cancer in twins. Journal of Epidemiology and Community Health, 38, 12-16.

Knox EG, Stewart AM, Kneale GW. 1983. Foetal infection, childhood leukaemia
and cancer. British Journal of Cancer, 48, 849-852.

Kotelchuck M, Parker GS. (1979). Woburn: Health data analysis 1969-1978.
http:l/www.state.ma.usldph/beha/woburn.htm

Langford l. 1991. Childhood leukaemia mortality and population change in
England and Wales 1969-73. Soc. Sci. Med, 33 (4), 435-440.

Laurier D, Bard D. (1999). Epidemiologic studies of leukemia among persons

under 25 years of age living near nuclear sites. Epidemiologic Reviews, 21(2),
188-206.

114

Leiss, J. K. and Savitz, D. A. (1995). Home pesticide use and childhood cancer:
a case-control study. American Journal of Public Health, 85 (2), 249-252.

Li, C., Lee, W., and Lin, R. (1998). Risk of leukemia in children living near high-
voltage transmission lines. Journal of Occupational and Environmental Medicine,
40(2), 144-147.

Linet, M. S. (1985). The Leukemias: Epidemiologic Aspects, Oxford University
Press: New York and Oxford.

Linet, M. S., Hatch, E. E., Kleinerman, R. A., Robison, L. L., Kaune, W. T.,
Friedman, D. R., Severson, R. K., Haines, C. M., Hartsock, C. T., Niwa, S.,
Wacholder, S., and Tarone, R. E. (1997). Residential exposure to magnetic
fields and acute lymphoblastic leukemia in children. The New England Journal of
Medicine, 337 (1), 1-7.

Locasciulli, A., Testa, M., Pontisso, P., Benvegnu, L., Fraschini, D., Corbetta, A.,
Noventa, F., Masera, G., and Alberti, A. (1997). Prevalence of Hepatitis C
infection in patients cured of childhood leukemia. Blood, 90 (11), 4628-4633.

Lyon, J. L., Klauber, M. R., Gardner, J. W. Udall, K. S. (1979). Childhood
leukemias associated with fallout from nuclear testing. The New England Journal
of Medicine, 300 (8), 397-402.

MacMahon B. 1992. Is acute lymphoblastic leukemia in children virus-related?
American Journal of Epidemiology, 136 (8), 916-924.

Magalhaes IQ, Pombo-de-Oliveira MS, Bennett CA, Cordoba JC, Dobbin J, Ford
AM, Greaves MF. (2000) TEL-AML1 fusion gene frequency in paediatric acute
lymphoblastic leukaemia in Brazil. British Journal of Haematology, 111(1):204-7.

Massachusetts Department of Public Health, Bureau of Environmental Health
Assessment. (1997). Woburn childhood leukemia follow-up study, Volume I
Analyses, Final Report. July 1997. http:llwww.state.ma.us/dphlbehalwoburn.htm

Massachusetts Department of Public Health. (1995). Health assessment -
lndustri-Plex site in Woburn. http:llwww.state.ma.usldph/beha/wobum.htm

Massachusetts Department of Public Health. (1995). Health assessment
addendum - Wells G&H, Woburn. http:llwww.state.ma.usldph/beha/woburnhtm

Massachusetts Department of Public Health, Bureau of Environmental Health
Assessment; U.S. Centers for Disease Control and Prevention, Division of Birth
Defects and Developmental Disabilities; The Massachusetts Health Research
Institute. (1994). Final report of the Woburn Environment and Birth Study.
http:llwww.state.ma.us/dph/beha/woburn.htm

115

Massachusetts Department of Public Health. (1989). Health assessment Wells
G&H, Woburn. http:lew.state.ma.usldph/beha/woburn.htm

Massachusetts Department of Public Health, Bureau of Environmental Health
Assessment. (1985). Final report of the Woburn advisory panel to the
Massachusetts Department of Public Health.
http:llwww.state.ma.usldph/behaMoburn.htm

Mauer, A. M. (1990). The Biology of Human Leukemia, The John Hopkins
University Press: Baltimore and London.

McCool, G. (1996). Viruses may cause 15 percent of cancers - Study. Reuters,
6 Aug 1998.

Meinert R, Schuz J, Kaletsch U, Kaatsch P, Michaelis J. (2000). Leukemia an
non-Hodgkin's lymphoma in childhood and exposure to pesticides: results of a
register-based case-control study in Germany. American Journal of
Epidemiology, 151(7), 639-646.

Mellemkjaer L, Alexander F, Olsen JH. (2000) Cancer among children of parents
with autoimmune diseases. British Journal of Cancer, 82(7):1353-7.

Michaelis, J., Schuz, J., Meinert, R., Zemann, E., Grigat, J., Kaatsch, P.,
Kaletsch, U., Miesner, A., Brinkmann, K., Kalkner, W., and Karner, H. (1998).
Combined risk estimates for two German population-based case-control studies
on residential magnetic field and childhood acute leukemia. Epidemiology, 9(1),
92-94.

Morris, J. D. H., Eddleston, A. L. W. F., and Crook, T. (1995). Viral infection and
cancer. The Lancet, 346, 754-758.

Murphy PJ. (1990). Exposure to Wells G and H in Woburn, Massachusetts.
http:llwww.state.ma.us/dph/behalwoburn.htm

Neglia, J. P., Meadows, A. T., Robison, L. L., Kim, T. H., Newton, W. A.,
Ruymann, F. B., Sather, H. N., and Hammond, G. D. (1991). Second
neoplasms after acute lymphoblastic leukemia in childhood. The New England
Journal of Medicine, 325, 1330-1336.

Nordlinder, R. and Jarvholm, B. (1997). Environmental exposure to gasoline
and leukemia in children and young adults: an ecology study. lntemational
Archives of Occupational and Environmental Health, 70(1), 57-60.

Parker GS, Rosen SL. (1981). Woburn: Cancer incidence and environmental
hazards 1969-1978. http:llwww.state.ma.us/dph/beha/woburn.htm

116

 

Petridou, E., Trichopoulos, D., Kravaritis, A., Pourtsidis, A., Dessypris, N.,
Skalkidis, Y., Kogevinas, M., Kalmanti, M., Koliouskas, D., Kosmidis, H.,
Pantagiotou, J., Piperopoulou, F., Tzortzatou, F., and Kalapothaki, V. (1997).
Electrical power lines and childhood leukemia: a study from Greece.
International Journal of Cancer, 73(3), 345-348.

Rowe RW. (1981). Retrospective study of Woburn 1949-1968.
http:llwww.state.ma.us/dph/beha/woburn.htm

Severson, R. K., Buckley, J. D., Woods, W. G., Benjamin, 0., and Robison, L. L.
(1993). Cigarette smoking and alcohol consumption by parents of children with
acute myeloid leukemia: an analysis within morphological subgroups— A report
from the Childrens Cancer Group. Cancer Epidemiology, Biomarkers &
Prevention, 2, 433-439.

Shu, X. 0., Ross, J. A., Pendergrass, T. W., Reaman, G. H., Lampkin, B., and
Robison, L. L. (1996). Parental alcohol consumption, cigarette smoking, and
risk of infant leukemia: a Childrens Cancer Group study. Journal of the National
Cancer Institute, 88 (1), 24-31.

Shu XO, Clemens J, Zheng W, Ying, DM, Ji BT, Jin F. (1995). Infant
breastfeeding and the risk of childhood lymphoma and leukaemia. International
Journal of Epidemiology, 24 (1), 27-32.

Smith M. (1997). Journal of the National Cancer Institute, 1542-1544.

Stevens, W., Thomas, D. C., Lyon, J. L., Till, J. E., Kerber, R. A., Simon, S. L.,
Lloyd, R. D., Elghany, N. A., Preston, M. S. (1990). Leukemia in Utah and
radioactive fallout from the Nevada test site: A case-control study. Journal of the
American Medical Association, 264 (5), 585-591.

Surveillance, Epidemiology and End Results: Cancer Query System on the Web
(1973-1 995). http:llwww-seer.ims. nci. nih.gov/scientificsystemsl.

Telles NC. (1981). Cancer mortality in Woburn: a three decade study 1949-1978.
http:llwww.state.ma.usldph/beha/woburn.htm

Toren A, Ben-Bassat l, Rechavi G. (1996). Infectious agents and environmental
factors in lymphoid malignancies. Blood reviews, 10, 89-94.

Voelker, R. (1998). Vitamin K controversy continues. Journal of the American
Medical Association, 279, 420.

117

Wang, P. P. and Haines, C. S. (1995). Childhood and adolescent leukaemia in
a North American population. lntemational Journal of Epidemiology, 24(6),
1 100-1 109.

Westergaard, T., Andersen, P., Pedersen, J., Olsen, J., Frisch, M., Sorensen, H.,
Wohlfahrt, J., and Melbye, M. (1997). Birth characteristics, sibling patterns, and

acute leukemia risk in childhood: a population-based cohort study. Journal of the
National Cancer Institute, 89(13), 939-947.

Wiemels JL, Cazzaniga G, Daniotti M, Eden OB, Addison GM, Masers G, Saha

V, Biondi A, Greaves MF. (1999). Prenatal origin of acute lymphoblastic
leukaemia in children. The Lancet, 354, 1499-1503.

118