LOO?

 

LIBRARY

Ii Mal. “kg-Q
IVIIUI "you! Vtutv

University

 

 

 

This is to certify that the
thesis entitled

BODY MASS INDEX AND ORAL CONTRACEPTIVE
FAILURE

presented by

Cristin M. Larder

has been accepted towards fulfillment
of the mquirements for the

MS. degree in Epidemim

 

 

Li {M L—a/

 

Major Professor’s Signature

May 9, 2009

 

Date

MSU is an Affinnative Action/Equal Opportunity Employer

-.—.—.-.---.-.-a--u-.----n-a-n-n-o-.-

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K'IProi/Acc&Prele|RCIDateDue.mdd

BODY MASS INDEX AND ORAL CONTRACEPTIVE FAILURE
by
Cristin M. Larder

A THESIS

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

MASTER OF SCIENCE
EPIDEMIOLOGY
2009

ABSTRACT
BODY MASS INDEX AND ORAL CONTRACEPTIVE FAILURE
by
Cristin M. Larder

There are suggestions that use of oral contraceptives is related to a higher risk of
birth control failure in women with increased weight. However, the few existing
studies investigating the association between body mass index (BMI) and oral
contraceptive (00) failure report inconsistent findings. To date, no research has
been conducted on the topic using a prospective design, and previous results
have been based on self-reported exposure and outcome variables. This thesis is
a grant proposal to apply for funding to conduct a prospective cohort study. The
objective of the proposed study will be to determine the combined effect of high
BMI and 0C use on birth control failure. In addition, the study will collect
information on exposure and outcome variables from sources other than a self-
report and investigate each the following possible confounding variables: 00
type and dose, compliance to OC regimen, frequency of intercourse, parity,
concurrent contraceptive methods, ethnicity, socioeconomic status, pre-existing
diseases and/or medical conditions, and desire for future children. The study
proposal includes sample size estimates needed to carry out the investigation, as

well as proposed statistical methods of analysis.

Copyright by
Cristin M. Larder
2009

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................... v
LIST OF FIGURES ................................................................................................. vi
LIST OF ABBREVIATIONS .................................................................................... vii
A. SPECIFIC AIMS ................................................................................................. 1
B. BACKGROUND ................................................................................................. 3
31 Association between efficacy and weight among subdermal
implant and transdermal patch users ................................................... 5
8.2. Current studies investigating the association between
efficacy and overweight/obesity among oral contraceptive users ....... 7

83. Possible biological explanations for the association between
overweight/obesity and increased hormonal contraceptive failure
rates ...................................................................................................... 11
8.4. Limitations of current research on the association between
overweight/obesity and oral contraceptive

failure .................................................................................................... 13
C. RESEARCH DESIGN AND METHODS ............................................................. 17
0.1. Study protocols .................................................................................... 17
0.1.1. Specific aim 1 ......................................................................... 17
0.1.2. Specific aim 2 ......................................................................... 24
0.1.3. Specific aim 3 ......................................................................... 27
0.2. Sample size and statistical power calculations .................................... 29
0.2.1. Step 1: Calculation of preliminary sample size ....................... 29
0.2.2. Step 2: Adjustment for prevalence of pre-existing
Conditions ................................................................. 30
0.2.3. Step 3: Correction for attrition ................................................. 31
0.2.4. Step 4: Account for 00 failures .............................................. 32
0.3. Statistical analysis ...................................................................... 33
D. STRENGHS AND LIMITATIONS ....................................................................... 35
REFERENCES ....................................................................................................... 37

LIST OF TABLES

Table 1. Published research examining the association between
BMI/weight and oral contraceptive failure ................................................. 10

Table 2. Covariates analyzed in ovenrveight/obesity and oral contraceptive
failure research ......................................................................................... 16

LIST OF FIGURES

Figure 1. Norplant failures by weight group ............................................................ 6

vi

KEY TO ABBREVIATIONS

OC: oral contraceptives

E2: ethinyl estadiol

OB/GYN: Obstetrics/Gynecology

POP: progesterone only pill

NHANES: National Health and Nutrition Examination Survey
BMI: body mass index

HMO: health maintenance organization

000: combined oral contraceptive

SHBG: sex hormone binding globulin

ALT: alanine aminotransferase

AST: aminotransferase

PCOS: Poly Cystic Ovary Syndrome

WHO: World Health Organization

MEMS: Medication Event Monitoring System
BRFSS: Behavioral Risk Factor Surveillance System

PHM: proportional hazards model

vii

A. Specific Aims

Oral contraceptives (OC) revolutionized family planning in the Twentieth
Century by providing a reliable and reversible form of contraception. The typical-
use failure rate of the 00 has been reported to be between 3.0% and 8.1%."1
However, over the years, the dose of ethinyl estradiol (E2) has been lowered with
every new generation of 00 in effort to reduce adverse side-effects such as
headache, nausea, and risk of venous thrombosis.

Additionally, prospective studies suggest that newer hormonal
contraceptives administered through subdermal implants and the transdermal
patch are less effective in ovenrveight and obese women.5'7 Reasons for the
reported increase in failure rate are unknown. Retrospective and case—control
studies have reported mixed findings regarding the same association with 00’s.

Furthermore, studies show that the prevalence of overweight and obesity
is rapidly increasing in the United States. As the dose of E2 decreases, and the
prevalence of overweight and obesity increases, there are concerns of an
increasing prevalence of unintended pregnancies among 00 users. Hence, I
plan to address this issue in a prospective cohort study with the following
hypothesis: overweight and obese women using oral contraceptives will have
significantly higher failure rates than normal weight and undenrveight women
using oral contraceptives.

Specific aim 1: I will gain access to the target population through doctor
and nurse contacts at local OB/GYN clinics in South Carolina. Four thousand,
four hundred, and sixty-six (4,266) women will be enrolled in the study through

these contacts after signing an informed consent document.

Specific Aim 2: I will assess compliance with the prescribed OC regimen
by obtaining used 00 cases at the end of each month. Participants will be
offered ten dollars for each case they return. I will also measure compliance
using a monthly telephone interview asking how many pills were missed during
the previous cycle.

Specific aim 3: In order to identify differences in failure rates among
women using high dose, low dose, and very low dose contraceptives, I will
determine OC brand from the health care providers’ records. I will then classify
brands according to the amount of E2 in each: brands containing 50119 or more
E2 will be classified as “high dose”, brands with 35pg or more, but less than 50pg
E2 as “low dose”, and brands containing less than 35ug E2 as “very low dose.”

Progesterone only pills (POP) will be classified as “no estrogen.”

8. Background

Unintended pregnancy has an enormous effect on social, financial, and
emotional well-being, and it is estimated that half of pregnancies in the United
States are unintended.‘ Thirty-two percent of women using reversible
contraceptive methods opt for oral contraceptives (OC’s),2which provide a
generally safe and effective method of birth control: typical use failure rates
(rates that include failures due to user error) are estimated to be up to 8%, and
most failures have been attributed to user compliance issues.3'4 However,
research shows that other hormonal contraceptive methods, such as the
subdermal implant and the transdermal patch, are not as effective in larger
women.5'7 Studies examining this effect in oral contraceptives are scarce, and to
date, no longitudinal investigations have been published.

Furthermore, obesity in the United States has reached a historically high
prevalence rate in women. According to data from the 1999—2002 National
Health and Nutrition Examination Survey (NHANES), 62% of American women
>19 years of age are either ovenrveight or obese.8 If the elevated risk of failure for
overweight and obese women observed in the implant and patch is indeed
present in 003, the public health consequences would be substantial,
particularly if the prevalence of ovenrveight and obesity continues to rise.

Unintended pregnancy has been associated with such adverse
consequences as high cost to both the individual and the public and
psychological distress. Trussell, et al estimated that compared to the $422 cost
of one year of oral contraceptive use, unintended pregnancies cost significantly

more for medical services if carried to term-$8619 per pregnancy for individuals,

or $3623 for the public if the individual is on Medicaid.9 Orr and Miller found that
mothers whose pregnancies are unintended are more likely to have depressive
symptoms during pregnancy than mothers with planned pregnancies.10
Additionally, unintended pregnancy is associated with insufficient prenatal care,
smoking and alcohol use during pregnancy, low birth weight, elevated infant
mortality, and poor health and development of the child.11 Ultimately, women
experiencing an unintended pregnancy not only face an immense financial cost,

but are also at risk for increased burdens during and after pregnancy.

8.1. Association between efficacy and weight among subdermal implant
and transdermal patch users

The transdermal patch and subdermal implants operate in a manner
similar to 00’s. Specifically, both use a combination of estrogen and
progesterone to prevent pregnancy. Studies have shown that as weight
increases, the efficacy of the Norplant® subdermal implant decreases. In a
clinical trial conducted in China, Gu, et al found that after 5 years, Norplant users
weighing 70kg or more had a 4.58% risk of becoming pregnant, compared to a
.77% risk among women weighing less than 50kg.5 After seven years of use, the
percentage of failure grew to 6.62% in women weighing 70kg or more, while the
percentage for women under 50kg rose to only 1.18%. Since Norplant® is
implanted underneath the skin and does not require daily compliance by the user
and thus invites little room for user error, these findings suggest that there may
be a biological explanation for the elevated failure rate among women weighing
70kg or more. Similarly, in a trial conducted among 17 countries, Grubb et al
reported increasing failure rates with each consecutive 10kg weight group they
examined (Figure).6 Although relative risks were not reported, both studies
display a disturbing trend among heavy women.

The Ortho EvraTM/ EvraTM transdermal patch has also been shown to be
less effective in heavy women. Zieman et al pooled data from three clinical tn'als
conducted to determine the efficacy of Ortho Evra'l‘M/ Evra'FM and found in a post
hoc analysis that among 3,319 participants and a total of 22,160 menstrual
cycles, 15 pregnancies occurred.7 Of the 15 pregnancies, 5 took place in women

who weighed 90kg or more, a subgroup that represented only 3% of the total

participants. Statistical analysis was not performed on these events, most likely
because of the small number of pregnancies that occurred in the pooled data.
However, the results of Zieman’s analysis did lead to the following warning on
the Ortho EvraTM/ EvraTM label:

Health Care Professionals who consider ORTHO EVRATM for women at or above
198 lbs. should discuss the patient’s individual needs in choosing the most
appropriate contraceptive option.12

Research investigating the efficacy of these hormonal contraceptives in
ovenrveight or obese women has consistently shown that the pregnancy rate
among larger users is higher than among users weighing less. These findings

are suggestive of an association between overweight/obesity and 00 failure.

Figure 1. Norplant failure rates by weight group”
12

10 /-
X
/

 

 

 

 

 

+Grubb et al“

 

 

 

   

 

 

Percent Failures

 

 

 

2 -9

 

 

_. ____‘.._~JL _ 4

5059 60-69 >70
Weight (kg)
“Failures for weight groups below 50kg were less than 1% in both studies.

“Difference between 60-69 and >70 in Grubb et al is not significant (p=.26). All other differences
are significant. (P<.01).

8.2. Current studies investigating the association between efficacy and
overweight/obesity among oral contraceptive users

Four studies have been published examining the relationship between oral
contraceptive failure and either weight or body mass index (BMI), all of which are
retrospective cohort or followed a case-control in design (Table 1). Two studies
by Holt and colleagues (2002 and 2005) reported an increased risk among
women with a BMI 227.3 or weighing 270.5kg, respectively (Table 1).""14 Both
studies used a study population of women enrolled in a health maintenance
organization (HMO) in Washington state, which may not be representative of
women outside of the HMO if these women are collectively different from the
general population.

A retrospective cohort study by Brunner 8. Hougue found a significant
crude hazard ratio of 1.8 (CI 1.01, 3.20) for women with a BMI 230 compared to
women in their 20-29 BMI group, indicating that obese women were 1.8 times
more likely to experience 00 failure than non-obese women. But, the difference
dropped below significance when confounding variables were added to the
model (Table 1).15 The study is a secondary analysis using data from the 1995
National Survey of Family Growth, for which all relevant data was collected by
face-to-face interviews where women were asked to recall information from over
two years prior to the survey. Therefore, the analysis may be subject to recall
bias.

Finally, a retrospective cohort study by Vessey & Painter examined
participants in the Oxford Family Planning Association Family Planning Study.

The participants were married women who were enrolled between 1968 and

1974.16 Vessey and Painter reported a non-significant difference in failure
between oral contraceptive users weighing 282kg compared to those weighing
less. However, only 4 women in the sample actually weighed more than 77kg,
and the number of women over 82kg was not reported. The lack of significance
may be due Type II error since there may not have been enough women in the
282kg group to detect a significant difference in the failure rate. Additionally, the
study included only married women using either combined oral contraceptives
(000’s) or POP’s. POP’s are recommended for use while breastfeeding, so it is
possible that POP use is overrepresented among married women. This group of
women may breastfeed more often because they have partners to help with
household duties, which frees up time for breastfeeding to take place, and they
may take pills less diligently than the general population of 00 users since they
have a family structure that is conducive to caring for children. The formulations
of Norplant®, Ortho EvraTM/ EvraTM contain both estrogen and progesterone, and
are therefore more similar to the combined oral contraceptive.7 There is no other
published research regarding POP efficacy among BMI or weight groups, which
contributes to the lack of knowledge about any biological mechanism(s) that may
be involved with the association between ovenrveight/obesity and increased risk
of OC failure.

Though limited research involving increased 00 failure among
ovenrveight/obese women exists, there is disagreement between studies as to
whether the relationship is significant. The conflicting results may be due to
issues involving the design of the existing studies: none use a prospective design

to measure pregnancy while using 003, and the low prevalence of obesity in

8

Vessey & Painter’s study resulted in inadequate statistical power to test the

hypothesis.

Table 1. Published research examining the association between BMI/weight
and oral contraceptive failure

 

 

 

 

 

 

 

 

 

 

 

 

 

Vessey 8. Painter, Holt, et al, Holt, et al, Brunner &
2001 2002 2005 Hogue, 2005
Design Eggorspectwe (R:'t;orspect1ve Case-control Ejttgorspecfive
wohfiihmaagfli‘is- W°men ages 18' English speaking ”'3' w°me"'
39 using 000 or 39’ enrolled in women age 18+ ages 1544'
Study POP enrolled in ovarian CVSt study enrolled in Group enrolled in the
Population Oxfor d-FP A Stud through Group Health National
England & y, Health Cooperative HMO Survey Of
Scotland 1968- C°°perafive ”MO in WA 1998-2001 Fami'y Gmw‘h'
1994 ’ in WA, 1990-1994 ’ 1993-1995
Sample 280 cases
Size 133 618 536 controls 1763
0 BMI <20
(26%)
BMI/Weight . <77kg (97%) o <70.5kg (75%) 0 BMI <27.3 (65%) 0 BMI 20-24.9
Categories (52%)
(% of o Referent: . Ref: 270.5kg 0 Ref: BMI 227.3 0 Ref: BMI 25-
sample) 277kg (3%) (25%) (35%) 29.9 (16%)
0 BMI 230
(7%)
One Face-to-Face
interview covering
lifetime history up 22:52:33;
Data Several interviews One Face-to-Face Lining-1031f covering time
Collection covering entire interview covering (cases!) or period from
Method (s) Leggiductive :insttige reproductive randomly matched 31:)ng t1‘95): to
ry ry reference month interview in
(controls), medical 1995
and pharmacy
records
Analysis Not reported COX proportional unconditional Sriiortional
hazards model logistic regression hazards model
. o Weight 270.5kg . o BMI 230 (20-
'I’Qifiifé‘Iitii'B9as (jg-39:93,” (33' $3,111....» 29:93:"),
n = . _' cru e = .
per 100 woman 1.1.2.4) OR-1.62 (0| (CI 1.01,
years 1.02-2.57) 3 20)
. . . 0 Very low dose
Findings 0 gfiflgtiiggw RR=4'5 (CI 1"" .cAonmSTs'tgnt users ° non-
' 14.4) BMI >27 3 ’ significant in
Ch t (27—3? ' f t multivariate
° . a?” "to 0 Low Dose OR-2 :78 glen ) model HR
3'9’“ 'Ca" "° p' RR=2.6 (or 1.2, ' ° I 1.51 (or 0.81,
value reported 59) 1.38, 3.41) 232)

 

10

 

8.3. Possible biological explanations for the association between
overweight/obesity and increased hormonal contraceptive failure rates

There is more than one potential biological explanation for increased
hormonal contraceptive failure among ovenrveight/obese women. First, Edman,
et al found that morbidly obese women took 12 times longer to excrete estrone
than women of normal weight.17 Coupled with Lukanova, et al’s findings of an
inverse correlation between sex hormone binding globulin (SHBG) and levels of
free estradiol,18 this suggests that ovenrveight/obese women may sequester
estrogens in their adipose tissue. Notably, ovenrveight/obese women have been
found to have lower levels of SHBG than normal weight women.“""20 Ultimately,
ovenrveight/obese women may simply have more fat tissue with which to “trap”
estrogens.

Some evidence suggests that SHGB has a positive correlation with
estrogen.21 This relationship is complex, however, and other studies suggest that
there are also non-steriodal factors that determine levels of SHBG.”23 Bates
and Whitworth reported that E2 is “markedly depressed” in obese girls.2‘The
relationship between obesity, SHGB, and estrogen levels may provide insight to
the BMI/OC failure hypothesis after further research has disentangled the
underlying association and causality.

A second possible explanation involves the metabolism in the liver. Oral
contraceptives, like many other drugs, go through a “first pass” metabolism.
During first pass metabolism, the liver and/or intestinal wall metabolize the drug
before it is able to circulate throughout the body. Differences in metabolism may

lead to different levels of bioavailability.25 Oveniveight/obese individuals have

11

been shown to have higher levels of liver enzymes than those of normal
weight.”27 For example, Golik, et al. found that over 28% of their 60 healthy,
obese participants had elevated blood levels of the hepatic enzymes alanine
aminotransferase (ALT) and aminotransferase (AST).28 This suggests that
ovenrveight/obese women may experience liver damage or dysfunction, which
may lead to decreased metabolism of OC steroids; therefore, the 00 may be
prone to fail more often than in normal weight women.

Additional research by Butte et al indicates that ovenrveight/obese
individuals have a higher basal metabolic rate, due to elevated fat free body

t.29 In a study designed to determine

mass compared to those at a normal weigh
the energy requirements of women, Jequier found that basal metabolic rate was
predicted by height and weight: women with high BMI had a higher basal

I.30 This difference in metabolism could

metabolic rate than those with lower BM
result in a shortened time in which the 00 is functional in the bodies of obese
women.

Biomechanisms explaining the relationship between obesity and OC
metabolism have not been thoroughly investigated, mainly because research
regarding the effect of obesity on 00 failure has not been conclusive to date.

These potential biomechanisms can be studied further if prospective research

confirms the relationship between obesity and 00 failure.

12

8.4. Limitations of current research on the association between
overweight/obesity and oral contraceptive failure

The first issue with the current research regarding ovenrveight/obesity and
oral contraceptive failure is that, to date, no prospective studies have been
conducted. Consequently, the results are susceptible to differential
misclassification, the most important example being recall bias.

The four previous studies used mostly questionnaire data to assess
important variables, including height, weight, and 00 brand, with the exception of
Holt’s 2005 study, which collected this information from medical records”. Their
methods could easily be improved upon by using clinic measurements for height
and weight and abstracting 00 brand from medical records. Moreover, three of
the studies only administered questionnaires once, at baseline participation. This
is a major shortcoming since many of the covariates that potentially affect the
outcome of 00 failure are not static, but have the capacity to change over time.
For example, a participant may smoke 20 cigarettes per day at baseline, but quit
smoking completely by the study’s conclusion. Therefore, the effect of tobacco
on her risk of 00 failure would change over the course of observation. Only
prospective study designs can eliminate these types of error by collecting data on
exposure and possible confounding variables repeatedly before the outcome (00
failure) occurs.

Furthermore, due to the retrospective design of all but one study among
the BMI/weight and oral contraceptive failure publications, many potential
confounding variables were not controlled for; hence, the studies could not adjust

for these confounders (Table 2). To begin with, frequency of intercourse may

13

confound the relationship between BMI/weight and oral contraceptive failure;
however, only Holt et al (2005) included it in their analysis. Some research
suggests that women with a higher waist-to-hip ratio have fewer sexual
partners;31 Rhodes, Simmons, and Peters concluded that body averageness
(average body size, as opposed to under or overweight) and symmetry had no
relationship with sexual behavior in women.32 Due to these mixed findings,
assessing intercourse frequency may or may not be significantly different in
women of different weight categories, and data should be collected and analyzed
in future studies to be certain.

00 compliance is also an important covariate in the relationship between
ovenrveight/obesity and 00 failure. Only Holt et al’s 2005 study included
compliant behavior. There may be a difference in compliance between different
BMI groups, and it is also imperative to determine whether oral contraceptive
failures are due to increased BMI or merely an artifact of non-compliance.

An issue closely related to compliance is that of the desire for children.
According to Trussell et al, some women using oral contraceptives may actually
want children, or feel ambivalent about the possibility.33 Therefore, carefully
designed questions regarding women’s intentions to become pregnant would be
useful in determining exactly how many accidental pregnancies can actually be
considered unintended pregnancies.

Not all oral contraceptives contain the same amount of estrogen, and
some use progesterone only. Holt el al (2002) found that the risk of oral
contraceptive failure among women _>.70.5kg were elevated in a sub-analysis of

women taking oral contraceptives with low and very low doses of estrogen (Table

14

1).13 Therefore, results should be stratified according to oral contraceptive type
and dose.

Using two contraceptive methods simultaneously undoubtedly lowers the
risk of contraceptive failure, yet only two of the four published studies collected
data regarding dual contraceptive use and were therefore able to control for this
variable. Obtaining information from clinical records or questionnaire data
regarding secondary birth control measures would substantially decrease error
surrounding this issue.

Caffeine consumption, cigarette smoking, and alcohol use have all been
shown to reduce fertility in women."“"36 As suggested by Sowers et al, smoking,
caffeine consumption, and wine drinking can alter the metabolism of estrogen.37
Smoking was taken into account in two of the four studies examining the
association between BMI/weight and oral contraceptive failure (Table 2), but
none of the published research includes information on caffeine or alcohol intake.
None of the previous studies were prospective in nature, and data was collected
for purposes other than investigating the association between BMI and 00
failure: it is likely that the information on confounders was not sufficient to
adequately measure this association. Questionnaire data requiring information
about the use of these substances throughout the study period will provide
information that then can be taken into account in by stratification or controlling

during statistical analyses.

15

Table 2. Covariates analyzed in overweight/obesity and oral contraceptive
failure research

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Brunner & Holt et al, Vessey &
Covariate Hogue, 2005 2005 1:33;”; :33: Painter, 2001
(Retrospective (Case- 00h?) rt) (Retrospective
Cohort) Control) Cohort)
Frequency of ‘1
Intercourse
00 Compliance ‘1
Type/Dose of OC «1 \J ‘J
Pre-existing ‘1
Conditions
Desire for Children «J
Dual Contraception ‘1 \/
Methods
Medication \J
Marital Status \I )1
Parity v \I \l \/
Race 1 \l \I
Socioeconomic
Status ‘1 ‘1 ‘1
Caffeine
Consumption
Alcohol Use
Smoking 4 \/
Age v v v

 

 

 

 

 

 

 

16

0. Research Design and Methods
0.1. Study Protocols
0.1.1. Specific aim 1

I will gain access to the target population through physician and nurse
contacts at local OB/GYN clinics in South Carolina. Four thousand, four hundred,
and sixty-six (4,266) women will be enrolled in the study through these contacts
after signing an informed consent document.

Rationale: The best way to access the target population is by
collaboration with clinics specializing in women’s health because the proportion
of patients using OC’s is higher than at general practitioners’ offices, which
accept a wider spectrum of patients. Recruitment through OB/GYN clinics is also
more beneficial than attempting to contact participants directly since practitioners
have an established relationship with their patients, and women would be more
inclined to participate in research knowing that their practitioners are also
involved. Furthermore, communication with OB/GYN clinics will provide access
to changes in contact information in the event participants move during the
study’s follow up period.

Procedures for recruitment: The target population for this study is oral
contraceptive users in the state of South Carolina. The study sample will be
drawn from a collection of private and publicly funded obstetrics and gynecology
(OB/GYN) clinics. The decision to include both private and pubic practices was
made in order to increase external validity by recruiting women from a variety of
socioeconomic demographics. Subjects will be recruited at their regularly

scheduled gynecology appointments. At this time, clinic nurses will use an

17

eligibility form to determine whether patients qualify for the study. Women of all
BMI categories will be included, who have childbearing capacity, desire a
prescription for DOS, and have no contraindications to 00 use. Eligibility will be
also depend on the following criteria:

Age: Patients must be at least 18 years old to participate in the study.
Minors will be excluded because they are not old enough to give consent without
a parent or guardian’s permission.

Duration of 00 Use: In order to reduce loss to follow up, women who do
not plan to use 00’s for at least 12 months will not be included in the study.

Pre-existing Conditions: Chronic diseases may affect women’s
metabolism or hormone levels. Consequently, patients with diabetes mellitus or
any thyroid disorder will be excluded. Additionally, women with polycystic ovary
syndrome (PCOS) will also be excluded. The reasoning for this is that research
has shown that PCOS is associated with obesity as well as altered glucose and
hormone levels.38 Since the syndrome and its association with reproductive
function are complex, and many questions regarding PCOS remain unanswered,
including subjects with the syndrome would be beyond the scope of this study.

Pregnancy: An eligibility requirement of the study will be women’s intent to
return to the clinic where they were enrolled in the case of pregnancy during the
course of the study. This purpose of this requirement is to ensure that 00
failures will be available for abstraction from medical records.

Pharmacy: Participants in the study must be willing to obtain 00
prescriptions at participating pharmacies. Details regarding pharmacy

involvement are provided under Specific Aim 2.

18

Phone: Only women with valid telephone numbers (either landline or
cellular) will be eligible for the study since monthly follow up interviews will be
conducted via telephone.

After determining eligibility, potential participants will be given a pamphlet
describing the study and asked if they would like to participate. All women who
answer yes will sign an informed consent document administered by the nurses.
The nurses will also provide a study overview with contact information in case of
any questions about study procedures.

Clinic records will be abstracted to determine the patient’s height and
weight without shoes at the time of the initial study visit, as measured by the
clinic nurses, as well as the type and dose of OC prescribed. The records will be
abstracted by study personnel at a later date.

After the 00’s are prescribed, clinic nurses will then inform participants of
nearby pharmacies participating in the study. Local pharmacies will be recruited
to participate through contacts with pharmacists. The study will offer to pay for
participants’ 00 prescriptions for the duration of enrollment as compensation for
their time. When study participants bring their 00 prescriptions to be filled,
participating pharmacies will bill the cost of the prescription to the study. If a
participant has valid health insurance, their co-pay will be billed to the study.

This approach to recruiting participants is ideal for several reasons. Most
importantly, collaboration with OB/GYN clinics will afford the opportunity to
review participants’ medical records in order to confirm information such as
height and weight, 00 brand, existing medical conditions, and 00 failure (i.e.

pregnancy). Secondly, women will be able to ask questions and voice any

19

concerns regarding the study directly to their practitioners, with whom they have
an existing rapport. As mentioned previously, study personnel will be able to
obtain changes in addresses and phone numbers of participants from clinics in
order to successfully complete the follow up period. Additionally, women who
visit specialty clinics, rather than their general practitioners, for 00’s are likely to
be more serious about using 00’s for birth control, which would increase the
number of participants in the study who comply with the OC regimen.

A possible objection to this approach is the fact that 00 users who obtain
00 prescriptions from their general practitioners will be excluded from the study.
However, since OB/GYN clinics are specialized, they will provide the best
environment for recruitment. General practitioners have a broad range of
patients to attend, so they may not be as willing to participate in research
regarding women’s reproductive health and would not have the opportunity to
recruit as many women as OB/GYN clinics. Additionally, women who visit
OB/GYN clinics for DOS would be more likely to return to the same clinic for pre-
natal care than those who see general practitioners: the latter group may be
referred to a specialty clinic if a pregnancy is detected. The ability to clinically
confirm pregnancies is important in order to acquire accurate information
regarding 00 failure.

After recruitment, and after their scheduled gynecologic appointment is
finished, participants will be asked to fill out a written questionnaire in a private
area inside the clinic. Subjects will hand completed questionnaires to the clinic

receptionist, who will store them in a locked file until retrieved by study

20

personnel. The questionnaire will include questions regarding basic
demographics and each of the following variables:

Medication: Research suggests that drugs and herbal supplements such
as anticonvulsants, antibiotics, and St. John’s Wort may affect the efficacy of
oral contraceptives.”‘42 Due to these interaction issues, participants’ medical
records will be reviewed to obtain information on administration of the
aforementioned drugs and supplements, and subjects themselves will be asked
in the initial questionnaire if they have taken anticonvulsants, St. John’s Wort, or
any antibiotics, and if so, to write the name of the drug or supplement.
Medications will then be reviewed on a case-by-case basis to determine if a
possible interaction does in fact exist.

Desire for Children: The importance of participants’ desire for children has
been described earlier, and this variable will be measured in the proposed study.
This will be done by using a Lickert-type scale to assess how burdensome a
having a baby would be in participants’ life, and how happy they would be to
have a child, if they found out that they become pregnant within the 30 days prior
to enrollment.

Frequency of Intercourse: As mentioned earlier, intercourse frequency
may vary across BMI categories.31 Furthermore, it is an important variable to
consider because a more frequent exposure to sperm will increase the risk of
oral contraceptive failure. Frequency of intercourse will be measured by the
initial questionnaire, on which participants will check the number of times per

week, on average, they engage in sexual intercourse.

21

Dual Contraception Methods: A monthly telephone interview will ask
participants how many times on average they use a second birth control method
out of the number of times they reported having sexual intercourse. The
rationale for collecting this information is to reduce confounding associated with
the added protection against unintended pregnancy that accompanies dual
contraception method use.

Marital Status: Given that the National Center for Health Statistics
reported that during the time period between 1990 and 2000, pregnancy rates
among married women were consistently higher than unmarried women, this
factor will be taken into account in the proposed study.43 The question regarding
marital status will include the following response options: married, single and
living with partner, single and not living with partner, separated, divorced, and
other. Marital status will be measured separately from frequency of intercourse
because although married women experience more pregnancies, this may also
be due to other factors, such as pregnancy wantedness.

Age, Gravidity, Race, and Socioeconomic Status: As these variables are
closely related to pregnancy rates,43 questions regarding age, pregnancy history,
and race and will be included in an initial questionnaire given at enrollment by
clinic nurses. Gravidity will be captured in a question asking how many times
participants have been pregnant, regardless of the outcomes of the pregnancies.
Gravidity will be used instead of parity, which only refers to the number of
pregnancies a woman has carried to term, in order to account for pregnancies

that ended in spontaneous and induced abortions. Socioeconomic status will be

22

determined by education level and health insurance coverage, both of which will
be included in the questionnaire.

Smoking, Alcohol, and Caffeine: Due to their association with decreased
fertility and estrogen metabolism, smoking, alcohol use, and caffeine
consumption will be determined. This information will be obtained by asking
subjects to record the number of cigarettes, alcoholic and caffeinated beverages
consumed on average at the time of enrollment. The monthly telephone interview
also includes items on smoking, caffeine, and alcohol consumption to determine

any changes in these variables over time.

23

0.1.2. Specific Aim 2

I will assess compliance with the prescribed OC regimen by obtaining
used 00 cases at the end of each month. Participants will be offered ten dollars
for each case they return to the pharmacy. I will also measure compliance using
a monthly questionnaire asking how many pills were missed during the previous
cycle.

Rationale: Measurement of compliance to the prescribed OC regimen is
essential in order to determine whether differences in failure rates can be
contributed to missed pills. The World Health Organization’s (WHO) practice
recommendation for missed 00 pills states that if only one active pill per cycle is
missed, the missed pill should be taken as soon as possible, and then the regular
one pill per day regimen should be resumed. If more than one active pill is
missed, the subsequent missed pills may be discarded. Additionally, the
recommendation cautions that the risk of pregnancy is higher when active pills
are missed on the day immediately preceding or the day immediately following
the 7-day inactive pill interval.“ This recommendation is central to the following
methods of measuring OC compliance.

A monthly telephone interview will be conducted in order to capture new
medications and diagnoses, as well as any changes in weight, intercourse
frequency, marital status, and tobacco/alcohol/caffeine consumption.
Implementation of the monthly interview will allow for more accurate data
collection for third variables than with the initial questionnaire alone since
information will be available regarding these variables for each menstrual cycle

for the entire duration of the study.

24

Procedures for measurement of OC compliance: Two methods will be
used to measure 00 compliance--a monthly telephone questionnaire and
collection of 00 cases at the end of each prescription by the participating
pharmacy. The monthly questionnaire will ask how many pills were missed
during the previous month, the date(s) missed, and whether the participant took
the pill at a later time or skipped it completely. The starting date of the
prescription and the starting date of the last menstrual period will also be
collected in order to determine which day of the menstrual cycle pill(s) were
missed, thereby indicating if the missed pill(s) were during the inactive or high-
risk days. The questionnaire will also include questions regarding dual
contraception use, new medical diagnoses, medications taken,
tobacco/alcohol/caffeine consumption, desire for children, and change in marital
status within the past month.

The second method of measuring 00 compliance will be the collection of
used OC cases at the end of each prescription will be employed. Participants will
receive instructions from the pharmacy. They will be advised to leave all missed
pills after the first missed one in the case (they will be asked to take the first
missed pill as soon as possible after remembering it was skipped per the WHO’s
recommendation)“. The pharmacist will also give participants written.
instructions on what to do when they miss a pill at this time.

Participants will then be instructed to return the pill case to the pharmacy
upon refill of the OC prescription, when they will receive a gift card to the
pharmacy as compensation for the time and effort required to keep track of

missed pills.

25

It may be argued that this two-fold method of measuring 00 compliance is
inferior to electronically monitoring compliance. A study by Potter et al found that
women actually missed three times as many pills as they reported on a monthly
diary card.41 Information regarding missed pills was collected using an electronic
device called the Medication Event Monitoring System (MEMS). MEMS records
the dates and times that the OC case is opened, and the data is transferred to a
computer after each prescription runs out. Due to the high cost of the MEMS
device and its low success rate (35% of MEMS used by Potter et al were either
not returned or the batteries had expired), the device will not be used in the
current study. Furthermore, the use of the MEMS device may introduce selection
bias. The group of women who failed to return the devices are likely to be non-
compliant 00 users due to irregular schedules. Collecting electronic data for
only compliant 00 users would not be helpful since the information would not be

generalizable to the entire population of 00 users.

26

0.1.3. Specific Aim 3

In order to identify differences in failure rates among women using high
dose, low dose, and very low dose contraceptives, I will determine 00 brand
from the health care providers’ records. I will then classify brands according to
the amount of E2 in each: brands containing 50pg or more E2 will be classified
as “high dose”, brands with 35pg or more, but less than 50pg E2 as “low dose”,
and brands containing less than 35pg E2 as “very low dose.” Additionally,
progesterone only pills (POP) will be classified as “no estrogen.”

Rationale: Holt, et al (2002) reported higher risks of failure among users
of low dose and very low dose OC’s than all 00 users combined (relative risks
1.6, 2.6, and 4.5, respectively, Table 1)13 These results indicate that it may be
beneficial to carry out a subanalysis with stratification based on the dose of E2 in
each 00 brand. Furthermore, since POP’s provide a slightly different
mechanism of avoiding pregnancy than COC’s, it is reasonable to anticipate a
difference in pregnancy risk between the two types of 00’s.

Procedures for data collection: For the purpose of this study, oral
contraceptive failure will be defined as a pregnancy occurring while a subject is
taking OC’s. Non-compliant women will be excluded since pregnancies in this
group may reflect user error instead of 00 failure for biological reasons. This
information will be obtained by two methods: self-report and the health care
provider’s medical records.

BMI will be determined both by clinic measurements at recruitment and by
self-report of height and weight from a monthly telephone questionnaire. Using

both sources of information will help gain a more accurate measure of height and

27

weight, and it will also aid in determining the extent of over and underreporting
among oral contraceptive users. BMI will then be calculated using body weight
(kg) over height (m) squared. Four BMI groups will be created using the WHO’s
definition of overweight and obesity: undenrveight (320), normal weight (>20 and
s25), ovenrveight (<25 and :30), and obese (<30).45 These guidelines were
chosen to maximize external validity, as WHO is an internationally established
organization. BMI will then be grouped into two categories before statistical
analysis—under/normal weight and ovenrveight/obese. If a significant difference
in 00 failure is detected between the two groups, a sub-analysis will be
performed using the four WHO categories to determine whether or not the trend

isfinear

28

C.2. Sample Size and Statistical Power Calculations
C.2.1. Step 1: Calculation of Preliminary Sample Size

A Preliminary sample size was calculated using the survival option in PS:
Power and Sample Size Calculation© software."6 The alpha level was set to .05,
and the power to detect a significant difference between groups to .80, which are
both standard values in Epidemiologic studies. The relative risk was estimated at
1.6, and the median survival time in the referent group (undenrveight/normal
weight) at 112 months: both were based on the results of Holt et al, 2002.13

Accrual time was set to 12 months, in order to keep the recmitment period
to a reasonable length for participating clinics and pharmacies. Follow up time
was also set to 12 months to allow enough time for women starting 005 at
enrollment to become accustomed to the pill taking routine, while avoiding high
rates of attrition that are associated with longer follow up periods.

The final parameter, the ratio of the referent group to the experimental
group (overweight/obese), was estimated based on data from the 2006
Behavioral Risk Factor Surveillance System (BRFSS), which reported that 60%
of South Carolinian women are either ovenrveight or obese and 40% were either
underweight or normal weight." 8y dividing .40 by .60, the ratio of .67 was
obtained.

After entering all parameters into the survival option in PS, the resulting
sample size estimates were 1,612 for the referent group and 1,080 for the

experimental group—a total of 2692 women.

29

0.22 Step 2: Adjustment for Prevalence of Pre-Existing Conditions

The preliminary sample size was then adjusted to account for comorbid
conditions that may confound the relationship between BMI and 00 failure and
render women ineligible for participation. Population based prevalence estimates
of the relevant disorders are as follows: PCOS—6.6%,"8 Diabetes Mellitus—
4.0%,49 Hypothyroidism—4.6%,50 and Hyperthyroidism—13%.50 Adding these
percentages gives a total 16.5% of women approached who would be ineligible
for the study based on pre-existing medical conditions. After dividing the
preliminary sample size estimates by .835 (1 - .165), the new estimates were
1,931 for the referent group and 1,293 for the experimental group—3,224 women

total.

30

0.2.3. Step 3: Correction for Attrition

The next step in calculating the study sample size was to adjust for
attrition. Based on previous oral contraceptive research, loss to follow up for a 12
month period per women has been estimated at 25%.“52 By dividing the
estimates from Step 2 by .75 (1 - .25), the estimated sample size after
accounting for attrition were obtained: 2,575 for the referent group and 1,724 for

the experimental group—4,299 total.

31

0.2.4. Step 4: Account for 00 Failures

Finally, in order to obtain an adequate sample size, it was necessary to
account for loss to follow up due to 00 failure. Fu, et al estimated that between
3% and 8% of women using OC’s became pregnant within 12 months of use.2
Using the conservative prevalence of 3%, the referent group sample sizes
determined in Step 3 was divided by .97 (1 - .03). The experimental group was
divided by .952 (1 - .03*1.6), based on the relative risk estimate of 1.6 from Step
1. The final sample size estimates were 1,811 for the experimental group and

2,655 for the referent group—4,466 women total. -

32

0.3. Statistical Analysis

The main analysis will compare two groups of women:
undenrveight/normal weight women (referent group) to ovenrveight/obese women
(experimental group). In addition, two sub-analyses will be performed. The first
will use all four of the WHO’s BMI groups in order to check for a non-linear trend.
The second will include four OC groups based on E2 levels (high dose, low dose,
very low dose, no estrogen) to determine whether or not the dose of E2 depicts a
linear relationship.

The effect of BMI on risk of oral contraceptive failure will be determined by
a Cox proportional hazards model (PHM) for survival analysis, with the hazard
ratio between BMI groups as the outcome variable. This model is appropriate for
analyzing the data collected from this study for several reasons. It allows
variables that change over time to be analyzed as functions of time, so the
changes will be captured in the statistical model. Furthermore, the model makes
no assumptions about the underlying distribution of the population. The model
does assume that the risk of pregnancy in each group is proportional over time.

The following measures will be employed to ensure that the model is
appropriate: covariates will be assessed for linearity, the proportional hazards
assumption will be checked, and the Wilcoxon test will be used to assess the fit
of the PHM.

If the proportional hazards assumption is not met, or if the number of
events is small (is zero inflation), which may occur with rare events such as 00
failure, one of two other statistical methods can be applied. A zero inflated

Poisson regression model may be used as long as the number of events at each

33

level of the covariates has a variance that equals the mean and the data follows
a Poisson distribution.53 However, overdispersion (when the variance is greater
than the mean) often occurs with zero inflation, and although correction terms
can be added to Poisson models to adjust the standard error, the correction
terms often lack efficiency since they add variability to the model.54 When this
occurs, a negative binomial regression model may be a better fit, as it is a
generalization of the Poisson model and can accommodate the overdispersion
using a dispersion term.

If the Cox PHM model is not found to be adequate to analyze the study
data due to overdispersion or zero inflation, both a Poisson model and a negative
binomial model will be executed in order to determine the best fit. The Scaled
Pearson Chi-Square test statistic will be used to determine the fit of the Poisson
model and Akaike’s information criterion will be used for the negative binomial

model. The most appropriate model will then be used for the final analysis.

34

D. Strengths and Limitations

The proposed study has several strengths, most importantly the
prospective design, which will help reduce recall bias. Furthermore, the study
will analyze third variables that have not been included in previous research and
measurement of third variables will be improved through clinic measured
baseline height and weight and monthly assessment of changes in third
variables. Finally, the study will use two methods of measuring 00 compliance,
namely the monthly telephone interview and the collection of OC cases by the
pharmacy.

Like all research, this study has limitations. First, only 00 users attending
OB/GYN specialty clinics will be enrolled in the study. However, this limitation is
not expected to be a hindrance to the selection of participants since differences
between these women and women who obtain 00 prescriptions from general
practitioners is expected to be negligible.

Since only women with valid telephone numbers will be included in the
study, there is a possibility for selection bias; however, this is a concern inherent
to longitudinal studies that use telephone interviews for follow up. The
percentage of women approached who do not have a phone will be collected in
order to quantify the number excluded for this reason.

Although overweight and obese women outnumber the under and normal
weight women in South Carolina, there may still be a stigma attached to being
overweight. To address this issue, I will name the study “Determinants of Oral
Contraceptive Failure” (DOCE Study), which does not specifically target

ovenrveight and obese women. This may also help mitigate any hesitation

35

OB/GYN’s may have to prescribe 00’s to these women based on the study
hypothesis by providing a more general description of the study’s aim.
Finally, previous 00 research has shown that younger women (under age

20), may have higher rates of failure than older women, most likely due to
unstable daily schedules and inexperience with the pill regimen.2 This should not
be a concern in this study since the Alan Guttmacher Institute found that the
majority of OC users (73%) are over age 205‘5 Furthermore, the conservative
sample size of the study will be able to account for O0 failures due to user

failure.

36

10.

11.

REFERENCES

Henshaw SK. Unintended pregnancy in the United States. Fam Plann
Perspect 1998;30(1):24-9, 46.

Fu H, Darroch JE, Haas T, Ranjit N. Contraceptive failure rates: new
estimates from the 1995 National Survey of Family Growth. Fam Plann
Perspect 1999;31(2):56-63.

Trussell J. Contraceptive failure in the United States. Contraception
2004;70(2):89-96.

Trussell J, Vaughan 8. Contraceptive failure, method-related
discontinuation and resumption of use: results from the 1995 National
Survey of Family Growth. Fam Plann Perspect 1999;31(2):64-72, 93.

Gu S, Sivin l, Du M, Zhang L, Ying L, Meng F, Wu S, Wang P, Gao Y, He
X, et al. Effectiveness of Norplant implants through seven years: a large-
scale study in China. Contraception 1995;52(2):99-103.

Grubb GS, Moore D, Anderson NG. Pre-introductory clinical trials of
Norplant implants: a comparison of seventeen countries’ experience.
Contraception 1995;52:287-96.

Zieman M, Guillebaud J, Weisberg E, Shangold GA, Fisher AC, Creasy
GW. Contraceptive efficacy and cycle control with the Ortho Evra/Evra

transdermal system: the analysis of pooled data. Fertil Steril 2002;77(2
Suppl 2):S13-8.

Hedley AA, Ogden CL, Johnson CL, Carroll MD, Curtin LR, Flegal KM.
Prevalence of ovenrveight and obesity among US children, adolescents,
and adults, 1999-2002. Jama 2004;291(23):2847-50.

Trussell J, Leveque JA, Koenig JD, London R, Borden S, Henneberry J,
LaGuardia KD, Stewart F, Wilson TG, Wysocki S, et al. The economic
value of contraception: a comparison of 15 methods. Am J Public Health
1995;85(4):494-503.

Orr ST, Miller CA. Maternal depressive symptoms and the risk of poor
pregnancy outcome. Review of the literature and preliminary findings.
Epidemiol Rev 1995;17(1):165-71.

Sable MR, Spencer JC, Stockbauer JW, Schramm WF, Howell V, Herman
AA. Pregnancy wantedness and adverse pregnancy outcomes:
differences by race and Medicaid status. Fam Plann Perspect
1997;29(2):76-81.

37

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

Ortho-McNeil Pharmaceutical l. Proposed text of the laeling for Ortho Evra
(norelgestromin/ethinyl estradiol transdermal system). 2001;Raritan, New
Jersey(08869).

Holt VL, Cushing-Haugen KL, Daling JR. Body weight and risk of oral
contraceptive failure. Obtet Gynecol 2002;99(2)part1z820-7.

Holt VL, Scholes D, Wicklund KG, Cushing-Haugen KL, Daling JR. Body
mass index, weight, and oral contraceptive failure risk. Obstet Gynecol
2005;105(1):46-52.

Brunner LR, Hogue CJ. The role of body weight in oral contraceptive
failure: results from the 1995 national survey of family growth. Ann
Epidemiol 2005;15(7):492-9.

Vessey M. Oral contraceptive failures and body weight: findings in a large
cohort study. J Fam Plann Reprod Health Care 2001;27(2):90-1.

Edman CD, MacDonald PC. Effect of obesity on conversion of plasma
androstenedione to estrone in ovulatory and anovulator young women.
Am J Obstet Gynecol 1978;130(4):456-61.

Lukanova A, Lundin E, Zeleniuch-Jacquotte A, Muti P, Mure A, Rinaldi S,
Dossus L, Micheli A, Arslan A, Lenner P, Shore RE, Krogh V, Koenig KL,
Riboli E, Berrino F, Hallmans G, Stattin P, Toniolo P, Kaaks R. Body mass
index, circulating levels of sex-steroid hormones, lGF-l and lGF-binding
protein-3: a cross-sectional study in healthy women. Eur J Endocrinol
2004;150(2):161-71.

Glass Ar, Reed W. Endocrine aspects of obesity. Medical Clinics of North
America. 73(1):139-160.

Siiteri PK. Adipose tissue as a source of hormones. Am J Clin Nutr
1987;45(1 Suppl):277-82.

Moutos D, Smith S, Zacur H. The effect of monophasic combinations of
ethinyl estradiol and norethindrone on gonadotropins, androgens and sex
hormone binding globulin: a randomized trial. Contraception
1995;52(2):105-9.

Koritnik DR, Marschke KB. Sex steroid hormone binding globulin levels

and free 17 beta-estradiol and testosterone in cynomolgus monkeys
during different reproductive states. J Steroid Biochem 1986;25(1):135-41.

38

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

Toscano V, Balducci R, Bianchi P, Guglielmi R, Mangiantini A, Sciarra F.
Steroidal and non-steroidal factors in plasma sex hormone binding
globulin regulation. J Steroid Biochem Mol Biol 1992;43(5):431-7.

Bates GW, Whitworth NS. Effects of obesity on sex steroid metabolism. J
Chronic Dis 1982;35(12):893-6.

Goldzieher JW. Pharmacology of contraceptive steroids: a brief review.
Am J Obstet Gynecol 1989;160(5 Pt 2):12604

Himmerich H, Kaufmann C, Schuld A, Pollmacher T. Elevation of liver
enzyme levels during psychopharmacological treatment is associated with
weight gain. J Psychiatr Res 2005;39(1):35-42.

Lam GM, Mobarhan S. Central obesity and elevated liver enzymes.
Nutrition Reviews. 2004;62:394-9.

Golik A, Rubio A, Weintraub M, et al. Elevated serum liver enzymes in
obesity: a dilemma during clinical trials. Int J Obes 1991;15(12):797-801.

Butte NF, Treuth MS, Mehta NR, et al. Engery requirements of women of
reproductive age. American Journal of Clinical Nutrition 2003;77(3):630-
38.

Jequier F. Energy metabolism in human obesity. Soz Praventivmed
1989;34:58-62.

GG HSG. Sex differences in morphological predictors of sexual behavior:
Shoulder to hip and waist to hip ratios. Evolution and Human Behavior
2003;24:173-178.

Rhodes G, Simmons LW, Peters M. Attractiveness and sexual behavior:
Does attractiveness enhance mating success? Evolution and Human
Behavior 2005;26:186-201.

Trussell,J, Vaughan 8, Stanford, J. Are all oral contraceptive failures
unintended pregnancies? Evidence from the 1995 National Survey of
Family Growth. Farn Plan Perspec. 1999;31:246-7,260.

Greenlee AR, Arbuckle TE, Chyou PH. Risk factors for female infertility in
an agricultural region. Epidemiology 2003;14(4):429-36.

Grodstein F, Goldman MB, Cramer DW. Infertility in women and moderate
alcohol use. Am J Public Health 1994;84(9):1429-32.

Jensen TK, Hjollund NH, Henriksen TB, Scheike T, Kolstad H, Giwercman
A, Ernst E, Bonde JP, Skakkebaek NE, Olsen J. Does moderate alcohol

39

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

consumption affect fertility? Follow up study among couples planning first
pregnancy. ij 1998;317(7157):505-10.

Sowers MR, Crawford S, McConnell 08, et al. Selected diet and lifestyle
factors are associated with estrogen metabolites in a multiracial/ethnic
population of women Nutritional Epidemiology 2006;135(6):1588-95.

Balen A, Michelmore K. What is polycystic ovary syndrome? Are national
views important? Hum Reprod 2002;17(9):2219-27.

Siew C, Gruninger SE, Burrell KH. Antibiotic interference with oral
contraceptive steroid efficacy. Lda J 1991 ;50(1):6, 8.

Back DJ, Breckenridge AM, Orme ML. Drug interactions with oral
contraceptive steroids. IPPF Med Bull 1983;17(1):1-2.

Bialer M, Doose DR, Murthy B, Curtin C, Wang SS, Twyman RE,
Schwabe S. Pharmacokinetic interactions of topiramate. Clin
Pharmacokinet 2004;43(12):763-80.

Zhou S, Chan E, Pan SQ, Huang M, Lee EJ. Pharmacokinetic interactions
of drugs with St John's wort. J Psychopharmacol 2004;18(2):262-76.

Ventura SJ, Abma JC, Mosher WD, Henshaw 8. Estimated pregnancy
rates for the United States, 1990-2000: an update. Natl Vital Stat Rep
2004;52(23):1-9.

World Health Organization. Selected practice recommendations for
contraceptive use. www.who.int.

World Health Organization. Obesity: Preventing and managing the global
epidemic. Geneva: WHO, 2000;7-9.

Dupont WD, Plummer WD. PS power and sample size program available
for free on the internet. Controlled Clin Trials 1997;18:274.

South Carolina Department of Health and Environmental Control. BRFSS
survey results 2006 for South Carolina: obesity risk factors. Columbia, SC.
http://scdhec.net/HS/epidata/BRFSS/2006/ bmi4c.html.

Azziz R, Woods KS, Reyna R, Key TJ, Knochenhauer ES, Yildiz BO. The
prevalence and features of the polycystic ovary syndrome in an
unselected population. J Clin Endocrinol Metab 2004;89(6):2745-9.

Boyle JP, Huneycutt AA, Narayan KM, Hoerger TJ, Geiss LS, Chen H,
Thompson TJ. Projection of diabetes burden through 2050: impact of

40

50.

51.

52.

53.

54.

55.

changing demography and disease prevalence in the US. Diabetes Care.
2001;24(11):1936-40.

Hollowell JG, Staehling NW, Flanders WD, Hannon WH, Gunter EW,
Spencer CA, Braverman LE. Serum TSH T(4), and thyroid antibodies in
the United States population (1988 to 1994): National Health and Nutrition
Examination Survey (NHANES III). J Clin Endocrinol Metab.
2002;87(2):489-99.

Potter L, Oakley D, de Leon-Wong E, Canamar R. Measuring compliance
among oral contraceptive users. Fam Plann Perspect 1996;28(4):154-8.

Oddesson K, Leifels-Fescher B, de Melo NR, et al. Efficacy and safety of a
contraceptive vaginal ring (NuvaRing) compared with a combined oral
contraceptive: a 1-year randomized trial Contraception 2005;71:176-182.

Byers AL, Allore H, Gill TM, Peduzzi PN. Application of negative binomial
modeling for discrete outcomes: a case study in aging research. J Clin
Epidemiol 2003;56(6):559-64.

Abdel-Aty MA, Radwan AE. Modeling traffic accident occurrence and
involvement. Accid Anal Prev 2000;32(5):633-42.

Alan Guttmacher Institute. Family planning anual report: 2004 summary.

Office of Population Affairs,‘U.S. Department of Health and Human
Services, 2005.

41

   

M111111119111flijlflilljlljijljililli“