x I . .m g h: 1 y. 3". amp . .. iii! X3": . ‘5 al_v1¢a...l ! .: :2... :9 2 .2... . .12 .z w... . #2: Eng ‘- Illa, vain. fan“. , 3‘ . u ‘ $5M .Ii: It s q .o > . ISLE». 3..) .. .. ; is .7). Fit. ....:J::..\ . . .. r 3 . .1 hf“: chi... I. 1... 33 it? {:5 ’53. 4.1!: $2... ; [$4113. 1 531.2»)? . 3.2 M 31.12... (it! :3; p it 9 1:. 5134.. . : rts. : .3. .1 25): .y 2...: . .. 7:28;: Illlllllllllllllll LIBRARY " r Michigan State University This is to certify that the thesis entitled BODY MASS INDEX AND TIME TO PREGNANCY IN EUROPEAN WOMEN OF CHILDBEARING AGE presented by Elisha Paul DeKoning has been accepted towards fulfillment of the requirements for Master's Epidemiology degree in 61/1421 9/4 Major professor Date June 7, 2000 0-7 639 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 mmwspu Body Mass Index and Time to Pregnancy in European Women of childbearing age BY Elisha Paul DeKoning 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 Body Mass Index and Time to Pregnancy in European Women of childbearing age By Elisha Paul DeKoning Analysis of a pregnancy-based, cross-sectional study was performed to assess the relationship between body mass index (BMI), its covariates, and fecundity as measured by time to pregnancy (TTP). This study suggests fecundity is highest within an ideal range of body composition. Membership in an ideal range of BMI can be predicted for the study population by the year at which menses began, the year at which women began having intercourse without doing anything to avoid pregnancy, and the number of cigarettes smoked at the starting time, defined as the time at which a couple began having sexual intercourse without doing anything to avoid pregnancy. Cigarette smoking, previous gynecological operations, and early age at starting time significantly reduce fecundity as measured by time to pregnancy. Path analysis suggests that the effects of age at starting time, level of completed education, and alcohol consumption at the starting time on time to pregnancy are mediated by body mass index in women with a body mass index greater than 18 kg/m". To Mom and Dad “For I know the plans I have for you, declares the LORD, plans for welfare and not for evil, to give you a future and a hope.” Jeremiah 29:11 ACKNOWLEDGMENTS I would like to thank my thesis advisor, Dr. Wilfried Karmaus, for his many hours of help and encouragement on much more than just this thesis. There are very few from whom I have learned so much. You are a great mentor, teacher, and friend; I will miss you. Many thanks to Dr. Michael Collins and Dr. Alka lndurkhya for serving on my committee. Your instruction and input were invaluable. Special thanks to ESIS, European Studies in Infertility and Subfecundity, for providing the data. Thanks to M. Calin Botezan for the friendly competition to finish the thesis—I’m gonna win! And finally to my roommates, Chris and Euge—the bon bons are on me. TABLE OF CONTENTS LIST OF TABLES ................................................................................. vii LIST OF FIGURES ............................................................................... vii LIST OF ABBREVIATIONS ..................................................................... ix INTRODUCTION ................................................................................... 1 CHAPTER 1 BACKGROUND .................................................................................... 2 Current issues in fertility and fecundity research ................................. 2 European Studies of Infertility and Subfecundity (ESIS) ........................ 6 CHAPTER 2 METHODS ........................................................................................... 8 The study population ..................................................................... 8 Questionnaire .............................................................................. 8 Statistical analysis ....................................................................... 10 CHAPTER 3 RESULTS .......................................................................................... 16 Selection and description of the study population ............................... 16 TTP modeling using survival analysis .............................................. 22 BMI-group membership modeling ................................................... 23 Final TTP survival analysis model ................................................... 25 Path analysis of BMI and TTP ........................................................ 25 CHAPTER 4 DISCUSSION ...................................................................................... 33 The study population .................................................................... 33 BMI and TTP .............................................................................. 37 Caffeine, alcohol, and cigarettes in relation to BMI and TTP .................. 39 Path analysis ............................................................................. 44 CHAPTER 5 METHODOLOGIC CONSIDERATIONS .................................................... 48 Does age at interview confound the effect of age at starting time? ......... 48 Pregnancy-based versus population-based sampling .......................... 50 CHAPTER 6 CONCLUSIONS .................................................................................. 51 APPENDIX A ESIS QUESTIONNAIRE ON PREGNANCY AND FERTILITY ........................ 54 REFERENCES ............................................................... ' ..................... 81 vi Table 1 Table 2 Table 3 Table 4 Table 5 Table 6a Table 6b Table 7 Table 8 Table 9 Table 10 Table 11 LIST OF TABLES Description of the study populations ....................................... 17 BMI stratified by Primary Predictor Variables—FOP / NPLB ........ 18 TTP stratified by Primary Predictor Variables—FOP / NPLB ....... 19 TTP stratified by Clinical/Medical History variables—FOP/ NPLB .............................................................................. 20 Fecundability ratios in Time to Pregnancy (TTP) modeling using proportional hazards regression, TIES = EXACT. Likelihood ratio = 41.3662, 4df, p<0.0001 .............................................. 23 Logistic model of membership in the low BMI group (<18kg/m2) versus the middle BMI groups (18-30 kg/mz). Likelihood ratio Chi- square = 41.1788, 2 df, p < 0.0001, n = 1357 ........................... 24 Logistic model of membership in the high BMI group (30+kg/m2) versus the middle BMI groups (18-30 kg/mz). Likelihood ratio Chi- square = 10.2468, 1 df, p = 0.0014, n = 1304 ........................... 24 Final TTP proportional hazards model, including covariates that were significant predictors of BMI group membership. Likelihood ratio Chi-square = 108.2776, 16 df, p < 0.0001 ......................... 26 Model stability of path analysis—stepwise covariate addition ....... 30 Model stability of path analysis—multi-case deletion .................. 31 Coding for caffeine-alcohoI-cigarette interaction terms ............... 43 Characteristics of deleted cases in multi-case deletion Diagnostics ....................................................................... 47 vii Figure 1 Figure 2 Figure 3 Figure 4 LIST OF FIGURES Hypothesized causation web for covariates explaining Body Mass Index (BMI) and Time to Pregnancy TTP) ................................. 7 Flowchart of analytical methods ............................................ 11 Beta coefficients and p-values (in parentheses) of path analysis models ............................................................................. 27 Spearman rank correlations between coffee, cigarette smoking, and alcohol ............................................................................. 40 viii LIST OF ABBREVIATIONS ALC ............................................................................................................. Alcohol BMI .............................................................................................. Body Mass Index CI ............................................................................................. Confidence Interval ESIS ............................................ European Studies of Infertility and Subfecundity FOP ................................................................................ First and Only Pregnancy FR ............................................................................................ Fecundability Ratio GYN. OP. ...................................................................... Gynecological Operations LogTTP ......................................................... Time to Pregnancy, log-transformed NPLB ........................................................................................ No Prior Live Births OR ......................................................................................................... Odds Ratio PID ............................................................................ Pelvic Inflammatory Disease ST ..................................................................................................... Starting Time STD ......................................................................... Sexually Transmitted Disease TTP ........................................................................................... Time to Pregnancy INTRODUCTION Worldwide variations in fertility, reproductive behavior defined by the number of childbirths, and fecundity, the biological ability to give birth or achieve a recognized pregnancy (Juul et al., 1999) are increasingly becoming areas of interest to biologists, gynecologists, epidemiologists, and population demographers. As advances in reproduction biology and technology (e.g. in vitro fertilization) simultaneously occur in an era of concern about worldwide population growth, fertility problems are a great public health issue considering its medical, social, and demographic implications. Epidemiology is especially well-suited for fertility (and infertility) research through critically reviewing previous studies, examining the role of medical and behavioral risk factors, and evaluating new therapies and prevention programs (Thonneau and Spira, 1990). In this investigation, data from a pregnancy-based, cross-sectional study of European women was analyzed to investigate the role of body mass index (BMI) and other correlates as they relate to time to pregnancy (TTP), a measure of fecundfiy. Chapter 1 BACKGROUND Current issues in fertility and fecundity research Previous and current research has investigated the effects on fertility and fecundity of: physical activity and obesity (Norman, 1998; Bongain, 1998), cigarette smoking (Suonio, 1990; Bolumar et al., 1996), caffeine intake (Grodstein et al., 1993b; Bolumar et al., 1997), alcohol consumption (Grodstein et al., 1994; Olsen et al., 1983), diabetes (Yeshaya et al., 1995; Pedersen et al., 1994; Gabbe et al., 1993; Charles et al., 1994), age at onset of menses (Yeshaya et al., 1995; Helm et al., 1995; Otor et al., 1998; Rockhill et al., 1998), and history of gynecological conditions such as sexually transmitted diseases, pelvic inflammatory disease, and endometriosis (Grodstein et al., 1993a; Rodriguez-Escudero et al., 1988; Westrom, 1994; Cates et al.,1994; Berube et al., 1998), to name a few. In addition to the effects of these individual factors, interrelationships and interactions exist between them and study designs vary (e.g. cross-sectional, longitudinal, and case-control), thereby complicating the interpretation of results. A lively area of debate has centered on the role of maternal nutrition in reproduction. While less disagreement exists over the effects of severe malnutrition, due to famine or conditions such as anorexia nervosa, there is substantial disagreement between physiologists and demographers on the role of mild to moderate undernutrition in fertility (Wood, 1994). The percentage of fat in the mature human female, as a measure of nutrition, has been argued as playing a causative role in reproductive ability (Frisch 1980; Frisch 1990, Rich-Edwards et al., 1994). Rose Frisch and colleagues have suggested and long maintained that a critical percentage of body fat may be necessary for the onset and maintenance of reproductive ability (menses) in females. This critical weight (fat) hypothesis suggests that the onset of menses at puberty (menarche) occurs at a body composition of about 17% fat. If secondary amenorrhea occurs, for example in women athletes who train heavily, a fat content of about 22% of body weight is required for the resumption and maintenance of menstrual cycles (Frisch 1980; Frisch 1990). From a biological standpoint, this theory argues that the ratio of fat-to-lean body mass is directly related to the endocrine changes of puberty and reproduction since adipose tissue is seen as an extragonadal source of estrogens, a source of estrogen over and above that produced by the gonads (Frisch 1980). Others investigators, however, have suggested that influences of central nervous system maturation or genetics may play a role (Scott and Johnson, 1982; Kaprio et al., 1995). Kaprio et al. (1995) studied Finnish twins from consecutive birth cohorts to study the variability of body weight and age at menarche due to genetic influences. Age at menarche was compared for 468 monozygotic girls, 378 girls from like-sex dizygotic pairs, 434 girls from opposite- sex pairs, and 141 older female siblings of the twins. The age at menarche was significantly higher for girls from opposite-sex pairs versus like—sex pairs. Bivariate twin analysis of BMI and age at menarche suggested a high degree of genetic influence (Kaprio et al., 1995). Less debatable are the effects of behaviors such as cigarette smoking, caffeine, and alcohol consumption. In a sample of pregnant European women, Bolumar et al. (1996) found an association between female smoking (2 11 cigarettes per day) and subfecundity, defined as more than 9.5 months of unprotected intercourse until conception (odds ratio (OR) = 1.7, 95% Confidence Interval (CI) 1.3-2.3). Women who smoked were also more likely to consume greater amounts of coffee and alcoholic beverages. Using a similar population- based sample of European women, Bolumar et al. (1997) found a significantly increased OR of 1.45 (95% CI 1.03-2.04) for subfecundity (more than 9.5 months of unprotected intercourse until conception) in the first pregnancy of women who drank more than 500 mg of caffeine per day. This effect was stronger for smokers (OR = 1.56, 95% CI 0.92-2.63) than for non-smokers (OR = 1.38, 95% CI 0.85-2.23). In another study of TTP and smoking in 2198 mothers interviewed at the 20th week of pregnancy, Suonio et al. (1990) found that the longer the conception delay, the more influential the effect of even light smoking. The effect of smoking on conception delay seemed to be dose-dependent. Sexually transmitted disease-associated genital infections are known to be capable of causing permanent damage to the reproductive tracts of both men and women (Westrom, 1994; Gates et al. 1994). A strong association exists between sexually transmitted diseases (e.g. infection, pelvic inflammatory disease (PID)), and infertility, primarily tubal infertility (Cates et al., 1994). Experts agree that post-infection infertility is diagnosed more often in women than in men (Westrom, 1994). Reproductive events were studied in a cohort of 1309 pregnancy-seeking women 35 years of age or younger after acute salpingitis and in 451 controls. Salpingitis refers to cases of visually or histopathologically confirmed inflammation of the fallopian tubes (Holmes, 1998). Among these women tubal factor infertility was diagnosed in 12.1% of the case patients versus 0.9% of the controls. Ectopic pregnancy was also more common in the case group. Variables of independent importance for infertility in this study also included the number and severity of infections (Westrom, 1994). Pelvic inflammatory disease refers to an ascending infection of the endometrium and/or fallopian tubes (Holmes, 1998). The post-infectious scarring of the PID healing process can result in infertility when bilateral tubal adhesions prevent the movement of sperm and/or ova by either damaging the mucosa and cilia of the female reproductive tract or by blocking the fallopian tubes. Occlusion of the fallopian tubes is commonly associated with prior chlamydial infection (Cates et al., 1994). Endometriosis, the presence of functional endometrial tissue outside the uterus, has also traditionally been associated with infertility and is assumed to lower pregnancy rates with increasing severity of disease. Research in women with mild-to-minimal endometriosis, however, suggests that fecundity is not significantly reduced in these women (Berube et al., 1998; Rodriguez- Escudero et al., 1988). The preceding introduction briefly introduces some of the issues in the field of reproductive epidemiology and demonstrates that the etiology of infertility and sub-fecundity and the interactions of its causes are extremely complex. Considering the above debate surrounding the role of BMI in fertility and fecundity, the objective of this thesis is to further investigate the web of causation for covariates explaining both body mass index and time to pregnancy in a sample of European women of childbearing age (Figure 1). Further, the roles of other conditions and behaviors (e.g. smoking, clinical history, etc.) were investigated as they relate to TTP: do they exert their influence through BMI or independent of it? The central hypothesis is that women with a higher BMI before their first and only pregnancy have a decreased waiting time to pregnancy. Additionally, the following sub-hypotheses were examined: 1. increased adult BMI is positively associated with early age at menarche; 2. increased adult BMI is positively associated with early age at first concepfion; 3. low level of education is positively associated with increased BMI; 4. maternal smoking is negatively associated with increased BMI. European Studies of Infertility and Subfecundity (ESIS) In 1990 a European study group was formed (ESIS, European Studies of Infertility and Subfecundity) to conduct comparable studies of fertility and fecundity in European countries. A cross-sectional, pregnancy-based survey of women from Denmark, Germany, Italy, Sweden, and France was conducted by ESIS in 1992. Women were approached immediately following delivery at a hospital or birth clinic or at an antenatal care visit after 20 weeks gestation. The institutions were chosen because they served geographically well-defined populations. All pregnant women during a defined data collection period were invited to participate. The participating women were asked to complete a highly- structured questionnaire that covered such areas as health and education, reproductive history, and pregnancy planning (Juul et al., 1999). Early age at Early age at first menarche conception N140 /.(+) Smoking at _ BMI 4 Low. Starting 4, H ' ‘ 3_ (+) Education Time (-) Central Hypothesis Waiting Time To 0 Pregnancy (TTP) *(+) denotes positive association (-) denotes negative association FIGURE 1: Hypothesized causation web for covariates explaining Body Mass Index (BMI) and Time to Pregnancy (TT P). Chapter 2 METHODS The study population The study was structured as a pregnancy-based, cross-sectional survey— all women who were recruited to participate were either currently pregnant or very recently pregnant. Women were recruited from the countries of Denmark, France, Germany, Italy, and Sweden in 1992. The 4035 participants in the initial sample with information on time to pregnancy were grouped into three populations based on their pregnancy history. The first group (1340) included women for whom this pregnancy was the first and only pregnancy (FOP); the second group (417) included women with no prior live births (NPLB); the third group included all others (2278). Time to pregnancy ('I'I'P, months), the amount of time it took for conception to occur was determined as a continuous variable; body mass index (BMI, kg/mz), also continuous, was obtained by dividing an individual’s weight (kg) by the square of their height (m). Questionnaire A highly structured questionnaire (Appendix A) was administered to pregnant or recently pregnant women at an antenatal care visit after 20 weeks gestation or at a hospital or birth clinic following delivery. The format of the questionnaire was structured by reproductive experience: the groups of questions asked depended on whether or not the pregnancy was the woman’s first, pregnancy and the circumstances of conception (e.g. never used birth control; result of birth control failure; intentionally discontinued birth control), and/or menstrual status at conception (menses had returned since the previous pregnancy, menses had not returned since the previous pregnancy). The content of the questionnaire covered such areas as health and education, reproductive history, starting time and waiting time to pregnancy (TTP) for the current or most recent pregnancy, exposures around the starting time, and pregnancy planning. “Starting Time" was defined as the time at which a couple began having sexual intercourse without doing anything to avoid pregnancy. In particular, questions were asked about the woman’s age at onset of menses, her height and weight at starting time, age at starting time, and completed level of education. Questions were also asked about: the woman’s clinical history (past presence of pelvic inflammatory disease; infection with chlamydia, gonorrhea, or another sexually transmitted disease; the presence of fibroids or myomas/endometriosis; past curettage or other operations of the uterus, tubes, or ovaries; oral contraceptive use; and diabetes) and exposures at or near the starting time (smoking status: cigarettes, cigars, and/or pipes; alcohol consumption; beer, wine, liquor; and caffeine: coffee, tea, cola). The exact questions are included in Appendix A (page 55). TI'P was defined as the length of time from the “Starting Time” until conception. This question was phrased “How long was it from that ‘starting time’ until you became pregnant? (the date you became pregnant is the date you conceived)” (Appendix A). Women could respond in terms of weeks, months, and/or years. The questionnaire was designed for self-administration. In two Italian centers, however, it was administered by female interviewers but there were no marked differences attributable to administration style (Juul et al. 1999). Statistical Analysis Figure 2 is a step-by-step flowchart of the methods used in this investigation. Univariate analyses were performed and the median and geometric mean of TTP, BMI, age at menarche (in years), and age at starting time (in years) were compared for the three groups of women (Figure 2, step 1). The Kruskal-Wallis test for non-parametric distributions was used to calculate the significance of differences in the medians. Spearman rank correlations and 10 Orig'nd Step 7. I Step 10. Test model stability. multi-wse deletion I Step 10. Test model stability. mtlti-(ase deletion Figure 2: Flowchart of analytical methods 11 ——H First&Only oreg. N=1340 Find study Step BI'BwModel populafion I of not I I (FOP + NPLB): linear No prior live births. N: 417 n=1757 , Others. N: 2278 l x J x J Step 4' Model TTP Y Y using Step1.Coerarethethree Step2 Desoribethefinal suvival populations on the basis of study population interns analysis TTP, BM, age at d BM, TTP, behaviors, N=1757 rnenarohe, age at starting and clinical history. obs. time. Model 'th Step 6' m Exclude Low WI men'bership in Fiji ' sidificart ‘ . vs. Md BMI group BMI soup ‘ oovanates using log'stic reg Step 5. Create 2 sm- populations Model with Step 6. Model Exclude ”9" significant merrbership in Low ‘ covariates I‘ vs. Md BMI grow ‘ BMI 90“? I using Iog'stic reg Find ‘lTP Path andysis model that ‘ . done includes separately for BMI—gap sw- populations created in step 5. Step 8. Path Analysis: Step8. Path Analysis: Low and Md BM Fiji and Md BMI (exclude Iigw) (exclude low) Step 9. Test model Step 9. Test model stability covariate stability covariate addtion addtion partial Spearman correlations were used to examine correlations between covariates. Univariate analysis and the Kruskal-Wallis test was used to describe the final study population (n = 1757) (Figure 2, step 2) stratified by age at menarche (grouped into three categories: <10, 11-14, 215 years), age at starting time (grouped into six categories: £20, 21 -25, 26-30, 31-35, 36-40, and 241 years old), BMI (grouped into four categories: <18, 18-25, 25-30, and 230 kg/mz), and levels of completed education (leaving school at or before the age of 15, between the ages of 16 and 17, at age 18 or older, completed manual trade training, professional training, a university degree, and other). Previous analysis of this data reported regional differences in time to pregnancy. Therefore, nationality was examined as a possible confounder of the relationship between BMI and TTP. After establishing the final study population, linear regression (PROC GLM in SAS) was employed to model BMI as a continuous variable with age at menarche, age at starting time, smoking status at starting time, and level of completed education as covariates (Figure 2, step 3). Survival analysis, utilizing proportional hazards regression (PROC PHREG in SAS), was employed to model TTP, in months, as a continuous outcome variable (Figure 2, step 4). There is no censoring of TTP in this sample of women because of the pregnancy-based structure of the study—all women conceived and therefore had the event. In the proportional hazards regression, the TIES=EXACT method was used instead of the Breslow method to break ties. 12 The Breslow method produces estimates in breaking ties that occur when two or more individuals have the same data points. The EXACT method does not utilize estimates (SAS, 1999). The hazard ratio in the output of proportional hazards regression represents the Fecundability Ratio (FR). Fecundability is defined as the monthly probability of conception: a value less than one represents decreased fecundability; a value greater than one represents increased fecundity. Confounders were determined by sequentially removing covariates and noting the effect of their removal on the BMI FRs. A covariate was retained in the model if its removal resulted in at least a 10% change in any of the BMI fecundability ratios. The final sample of women (n = 1757) was split into two sub-populations based on the categorical grouping of BMI: one group excluded the highest BMI category (230 kg/mz), the other excluded the lowest BMI category (<18kg/m2 ) (Figure 2, step 5). Logistic regression was used to model membership in the extreme BMI categories within the sub-populations (Figure 2, step 6). Confounders were assessed by sequentially dropping covariates from a starting model and comparing the reduced model to the starting model by means of a Chi-square test with degrees of freedom = (dfstamng mode. -— dfreduced rnocm); alpha = .05. Covariates that were significant in the Chi-square test were retained in the model. Those covariates that were significant (p < 0.05) in the BMI-group membership logistic models were entered into the TTP survival analysis model from step 4 to produce a final TTP survival analysis model (Figure 2, step 7). 13 The PROC CALIS (Covariance Analysis of Linear Structural equations) procedure in SAS was used for path analysis to model multiple linear paths with BMI and TTP (log-transformed) as continuous endogenous variables (Figure 2, step 8). The CALIS procedure has the advantage of being able to model intervening effects in the regression; proportional hazards regression does not permit this. For the purpose of path analysis, age at starting time in years, age at menarche in years, smoking (the number of cigarettes smoked per day) and alcohol consumption (the number of alcoholic beverages consumed per week) at starting time were continuous covariates; level of completed education was replaced by Blom-transformed ranks. Gynecological operations included past curettage or other operations of the uterus, tubes, or ovaries (Methods, questionnaire, page 9). The following linear structural equations were used in parallel invocations of the CALIS procedure: BMI = [31(age at starting time) + [32(age at menarche) + [33(Cigarettes) + [34(Alcohol) + [35(Education level) + £le LogTTP = [35(age at starting time) + 87(age at menarche) + [33(Cigarettes) + 89(Alcohol) + 810(Education level) + B11(any gynecological operations) + 5123”" + ELogTTP One-sided t-tests were used to calculated the significance level of regression coefficients produced by the CALIS procedure. 14 The stability of the models produced by path analysis was tested by two methods. First, the models were re-constructed by adding one covariate at a time and noting the influence of its addition on the other B-coefficients (Figure 2, step 9). Variables which, upon addition to the model, changed B-coefficients by 20% or more indicated an unstable model or possible confounding. Secondly, covariate addition was repeated after multi-case deletion to assess the influence of outliers on model stability (Figure 2, step 10). All analyses were performed using the SAS statistical package, version 8.0. 15 Chapter 3 RESULTS Selection and description of the study population Table 1 describes women for whom the pregnancy was their first and only (FOP) and women who had no prior live births (NPLB) in terms of time to pregnancy, body mass index, age at menarche, and age at starting time (Figure 2, step 1). The significance of differences in the medians of the primary predictor variables (age at menarche, age at starting time, smoking at starting time, level of completed education) were tested using the Kruskal-Wallis test. The FOP and NPLB populations did not differ in the median of TTP (p=.0951), BMI (p=.8105), or age at menarche (p=.7623). They did differ, however, in the median age at starting time (FOP = 25 years, NPLB = 26 years, p < 0.0001). The FOP and NPLB populations were combined, excluding pregnancies that were the result of contraceptive failures (n=236 for the FOP population), yielding a final study population of 1757 women with information on time to pregnancy who had no prior live births or for whom this pregnancy was their first and only pregnancy. Tables 2-4 describe the final study population (n = 1757) (Figure 2, step 2). The median of BMI was statistically significantly different when stratified by age at menarche (p<.0001), age at starting time (p=.0184), completed level of education (p=.0073), and nationality (p=.0011) (Table 2). BMI was highest for those women in the youngest grouping of age at menarche and it decreased with increasing age at menarche. BMI tended to increase with older age at starting 16 Table 1: Description of Study Populations TTP (months) stratified by Population Population n median geometric mean 5% 95% First and only pregnancy 1340 3 3.38 0.5 36 No prior live births 417 3 4.11 0.2 60 total 3438 BMI (Kg/m"2) stratified by Population Population n median geometric mean 5% 95% First and only pregnancy 1317 21.26 21.63 17.99 28.08 No prior live births 411 21.26 21.72 17.91 28.73 total 3952 Age at Menarche (years) stratified by Population Population n median geometric mean 5% 95% First and only pregnancy 1253 13 12.88 11 15 No prior live births 397 13 12.86 11 15 total 3776 Age at Starting time (years) stratified by Population Population n median geometric mean 5% 95% First and only pregnancy 1322 25* 25 19 32 No prior live births 409 26* 26 19 36 total 3406 *p-value (difference in medians) <.0001 17 TABLE 2: BMI stratified by Primary Predictor Variables—FOP I NPLB Age at menarche .<_1O 11-14 215 Age at starting time £20 21-25 26-30 31 -35 36-40 241 Completed education left < 15 left 16-17 left 18+ manual trade prof train. univ. degree other no answer missing Country Denmark Germany Sweden N Italy S Italy France N 151 1376 201 227 685 605 171 37 176 139 350 1 97 466 320 37 33 1 0 216 570 317 184 210 231 Smoking at starting time yes no no answer 608 1115 5 MED 21.95 21.30 20.57 20.94 21.30 21.33 21.08 21.48 28.40 21.64 21.14 21.48 21.83 21.08 20.82 21.22 21.50 20.90 21.76 21.20 21.48 21.09 21 .26 20.70 21.17 21.30 22.03 5% 18.52 18.07 17.37 17.36 17.91 18.07 18.69 19.37 26.85 17.58 17.47 18.07 18.08 17.93 18.03 17.63 18.78 18.73 18.25 17.51 18.67 17.75 17.67 17.91 17.58 18.08 18.37 18 95% 28.42 28.39 25.35 29.67 28.39 27.74 27.43 25.82 29.88 30.10 31.05 26.30 28.39 28.03 26.97 27.44 31.14 27.68 29.38 29.38 27.51 26.57 28.52 26.56 29.38 27.77 24.49 Kruskal-Wallis p<0. 0001 p=0.0184 :0. 0073 :0. 001 1 TABLE 3: TTP stratified by Primary Predictor Variables—FOP I NPLB N Mg) 5% 95% Smoking at starting time yes 618 4.00 0.50 48.00 no 1133 3.00 0.50 36.00 no answer 6 1.25 0.00 8.00 BMI group <18 125 4.00 0.50 42.00 18-24 1410 3.00 0.30 39.00 25-29 169 3.00 0.50 36.00 230 53 4.00 0.50 83.00 Completed Education left <15 176 3.00 0.50 48.00 left 16-17 144 4.00 0.50 52.00 left 18+ 354 3.00 0.50 38.00 manual trade 205 3.00 0.20 36.00 prof training 469 4.00 0.50 42.00 univ. deg. 322 2.20 0.20 31.00 other 39 3.00 0.50 54.00 no answer 37 3.00 0.00 60.00 missing 11 2.00 0.20 120.00 Age at menarche $10 160 3.00 0.25 35.00 11-14 1393 3.00 0.50 45.00 215 204 3.00 0.50 36.00 Age at starting time £20 238 5.00 0.30 84.00 21-25 698 3.00 0.50 42.00 26-30 610 3.00 0.50 36.00 31-35 171 3.00 0.50 28.00 36-40 37 2.50 0.00 27.00 241 3 1 .40 1 .00 32.00 KruskaI-Wallis p<0. 0001 p=0. 0635 p<0. 0001 =0. 9524 :0. 0002 TABLE 4: TTP stratified by Clinical/Medical History variables—FOP I NPLB N MED 5% 95% Kluskgl-Wallis PID no 1602 3.00 0.50 36.00 yes 151 5.00 0.50 81.00 p=0.0003 Chlamydia no 1635 3.00 0.50 42.00 yes 118 3.00 0.20 33.00 p:0.5746 Gonorrhea no 1722 3.00 0.50 39.00 yes 31 5.00 0.20 96.00 p=0.0205 Other STD no 1623 3.00 0.50 39.00 yes 130 3.00 0.30 48.00 p=0. 5297 Ovarian Cyst no 1620 3.00 0.40 36.00 yes 132 5.00 0.50 96.00 p<0.0001 Fibroidslmyomas no 1715 3.00 0.50 42.00 yes 38 4.00 0.20 43.00 p=0.7105 Endometriosis no 1729 3.00 0.50 36.00 yes 23 36.00 0.90 96.00 p<0.0001 Diabetes no 1739 3.00 0.50 42.00 yes 13 2.00 0.50 12.00 =0.2329 Curettage no 1545 3.00 0.50 36.00 yes 208 3.60 0.20 54.00 p=0. 0543 Other Gyn. operations no 1679 3.00 0.50 36.00 yes 73 12.00 0.50 96.00 p<0. 0001 Oral contraceptive use no 365 3.00 0.50 72.00 yes 1379 3.00 0.50 36.00 =0. 8393 20 time. BMI was highest in those women who left school before the age of 15 and in those women with manual trade training; women with a university degree had the lowest BMI. Body mass index was highest for women of Danish nationality and lowest for French women. No statistically significant differences in BMI were detected when stratified by smoking status at starting time (p=.26). Women who smoked had a significantly increased median TTP (p<.0001) (Table 3). TTP was the longest for women with a BMI at the extremes of the distribution, <18 kg/m2 and 230 kg/mz, but differences in the medians were not statistically significant (p=.0635). When the middle two groups are combined (BMI = 18-30 kg/mz) and the significance of differences in medians are again compared, the low BMI group (median TTP = 4 months, 125 observations) emerged as having a significantly longer median TI'P than the two middle groups combined (median TI'P = 3 months, 1579 observations; p = .024). However, there is no statistically significant difference in median TTP for the high BMI group (median TTP = 3 months, 1579 observations) when compared to the combined middle two groups (median ‘I'I'P = 4 months, 53 observations; p=.113). TTP was longer for women who left school between the ages of 16 and 17 and with professional training; women who left school after the age of 18 or with a university degree had the shortest TTP (p<.0001) (Table 3). There were no statistically significant differences in 'l'l'P when stratified by age at menarche (p = .95). The median TTP was statistically significantly different for different levels of age at starting time (p = .0002), decreasing with increasing age. 21 Differences in the median TTP were also calculated for clinical history variables as potential confounders, namely pelvic inflammatory disease (PID), chlamydia, gonorrhea, other STDs, ovarian cysts, fibroids/myomas, endometriosis, curettage, other gynecological operations, diabetes, and oral contraceptive use (Table 4). Time to pregnancy was significantly longer for women with a history of PID (p=.0003), gonorrhea (p=.0205), ovarian cysts (p<.0001), endometriosis (p<.0001), and other operations of the uterus, tubes, or ovaries (p<.0001). TTP modeling using survival analysis Using proportional hazards regression (survival analysis), TTP was modeled using time of conception as the event and both the primary predictor and clinical history variables as covariates (Figure 2, step 4). The hazard ratio in the output corresponds to the fecundability ratio. After confounder assessment, the TTP survival analysis model (Table 5) included, in addition to the BMI dummy variables, all past gynecological operations (past curettage and other operations of the uterus, tubes, or ovaries): likelihood ratio = 41.3662, 4 df, p<.0001. Membership in the low BMI group (<18 kg/mz), in the high BMI group (230 kg/mz), and previous gynecological operations all reduced fecundity at the statistically significant level (p < 0.05): fecundibility ratio (FR) = 0.820, 0.757, 0.693, respectively. 22 Table 5: Fecundability ratios in time to pregnancy modeling using proportional hazards regression, TIES=EXACT. Likelihood ratio = 41.3662, 4df, p<0.0001. Variable Fecundability Ratio p 95% C.l. BMI <18 kg/m2 0.82 0.0339 (0.683, 0.985) BMI 18-24 kg/mz' 1.000 - - BMI < 25-29 kg/m2 1.028 0.7327 (0.876, 1.207) BMI 230 kg/m2 0.757 0.0476 (0.575, 0.997) any gynecological operation” 0.693 <0.0001 (0.607, 0.793) *referent group ”past curettage or other operations of the uterus, tubes, or ovaries BMI-group membership modeling As membership in the tails of the BMI distribution was significant in the TTP survival analysis model, logistic regression was used to separately model membership in the low and high BMI categories (Figure 2, steps 5-6). To model membership in the low BMI group (<18 kg/mz), a population was created that excluded those individuals in the high BMI group (230 kg/mz). Membership in the low BMI group served as the dichotomous outcome in the regression model, using the combined middle two groups (18-30 kg/mz) as the reference. A starting model, based on 1357 observations, was constructed that included age at interview, age at starting time, age at menarche, number of cigarettes smoked per day at starting time, number of alcoholic beverages consumed per week at starting time, caffeinated coffee consumption (cups per day) at starting time, completed level of education, and presence of maternal diabetes. To test for 23 confounders, covariates were removed successively from the starting model (Likelihood ratio Chi-square = 51.4046, 12 df, p<.0001). The final predictive model for membership in the low BMI group included age at starting time and age at menarche (Likelihood ratio Chi-square = 41.1788, 2 df) (Table 6a). The model for membership in the high BMI group, based on 1304 observations, was constructed in a manner identical to that used for the low BMI group, with the middle two groups serving as the reference group (Table 6b). The same initial covariates were included in the starting model (Likelihood ratio Chi- square = 23.0537, 12 df, p=.0273). Only one covariate, number of cigarettes smoked per day at starting time, remained in the final model (Likelihood ratio Chi- square = 10.2468, 1 df, p=0.0014). Table 6a: Logistic model of membership' In the low BMI group (<18kg/m2 ) versus the middle BMI groups (18-30 kglm2 ). Likelihood ratio Chi-square- - 41.1788, 2df, p<0.,0001 n: 1357. Covariate Estimate Chi-square p age at menarche 0.256 18.169 <.0001 age at starting time -0.121 22.555 <.0001 Table 6b: Logistic model of membership In the high BMI group (30+kg/m2 ) versus the middle BMI groups (18-30 kglm2 ). Likelihood ratio Chi-square_ - 10. 2468, 1df, p: O. 0014, n: 1304. Covariate Estimate Chi-square p number of cigarettes 0.0527 1 1.959 0.0005 24 Final TTP survival analysis model The covariates that were retained in each of the BMI-membership models were re-entered into the TTP proportional hazards model from Table 5, yielding a model that included age at starting time, BMI group, age at menarche, all past gynecological operations, and the number of cigarettes smoked per day at the starting time (Figure 2, step 7) (Table 7). Of the covariates in the model, the following were statistically significant: age at starting time less than 21 years (p < 0.0001), age at menarche 10 years or younger (p = 0.025), 11-15 cigarettes per day (p = 0.009), 16-20 cigarettes per day (p = 0.0003), more than 20 cigarettes per day (p = 0.016), and any past gynecological operation (p < 0.0001). Only age at menarche 10 years or younger was significantly associated with increased fecundability (FR > 1.0); all other significant covariates were associated with decreased fecundability (FR < 1.0). Path analysis of BMI and TTP Figure 3 is a diagrammatic representation of the results of path analysis (Figure 2, step 8) of the effects of the covariates from Table 7 (with the addition of alcohol and education) on BMI and TTP. Path analyses were completed separately for the low BMI group versus the middle two groups and for the high group versus the middle (as in the logistic regression). The term BIzBMI (Methods, page 12) is modeled as a potential intervening variable in predicting LogTTP. Both age at menarche and age at starting time are statistically significant predictors of body mass index in path analysis at the p = 25 Table 7: Final TTP proportional hazards model, including covariates that were significant predictors of BMI group membership. Likelihood ratio Chi-square = 108.2776, 16 df, p < 0.0001. Variable Wald Chi-Siflare p Hazard Ratio 95% C.l. Age at starting time (yrs) <21 25.667 <0.0001 0.651 (0.551, 0.769) 21-25 3.5525 0.060 0.896 (0.800, 1.004) 26-30 - - 1.000 - 31-35 0.9683 0.325 0.916 (0.768, 1.091) 36-40 0.0975 0.755 0.947 (0.672, 1.334) >40 2.5236 0.112 3.116 (0.767, 12.665) BMI (kg/m2) <18 3.1607 0.075 0.833 (0.681, 1.019) 18-25 - - 1.000 - 25-30 0.3511 0.554 1.052 (0.889, 1.246) >30 2.7882 0.095 0.781 (0.585, 1.044) Age at menarche (yrs) 510 5.0548 0.025 1.373 (1.042, 1.811) 11-14 - - 1.000 - >14 0.3187 0.572 1.045 (0.897, 1.217) Cigarettes (number/day) 0 - - 1.000 - 1-5 0.0115 0.915 0.985 (0.751, 1.293) 6-10 1.5657 0.211 0.891 (0.744, 1.067) 11-15 6.8798 0.009 0.798 (0.675, 0.945) 16-20 13.2751 0.000 0.662 (0.530, 0.826) >20 5.8087 0.016 0.793 (0.656, 0.958) Any past Gync. operation 26.9244 <0.0001 0.690 (0.600, 0.794) 26 High vs. Mid BMI group (Goodness of fit adjusted for degrees of freedom: 0.9946: Root mean square residual: 0.0064) -.0257 @190) AGE at ST .332 (.022)* .0819 (0.07) MENARCHE 1.024 (.015)* .0193 (.09) SMOKING .0573 (.09) d v y l i BMI .0268 (0.11) > LOGTTP R2=.969 R2=.461 1 -.078 (.89) F ALCOHOL .0160 (0.17) T .528 (0.065) -.407 (.90) EDUCATION '0678 (0'18) GYN. OP. Low vs. Mid BMI group (Goodness of fit adjusted for degrees of freedom: 0.9995; Root mean square residual: 0.0017) -.025 (0.90) AGE at ST .343 (.018)* .112 (0.053) MENARCHE . .962 (.013)* .021 (.085) SMOKING .033 (.13) v v i i BMI .010 (0.27) LOGTTP R2=.975 R2=.4608 -.056 (0.87) A ALCOHOL K -015 ((113) T I .498(0.069) -.434 (0.92; EDUCATION ~04“ (026) GYN. OP. Figure 3: Beta coefficients and p-values (in parentheses) of path analysis using PROC CALIS. 27 .05 level (Figure 3). There is virtually no effect of BMI on IogTTP in the Low vs. Mid model (8:.010) as opposed to the High vs. Mid model (8:.0268). Figure 3 indicates that, for every unit increase (kg/m2) in body mass index over 18 kg/m2 (the High vs. Mid group), time to pregnancy increases eB months. For example, an increase of 1 kg/m2 in BMI results in an increase in 'ITP of 1.03 months. Similarly, for women with a BMI greater than 18 kg/m2, every 1-year increase in '0257 months, or .97 months. These age at starting time decreases TTP by e' effects are multiplicative; the greater the increase, the greater the effect on TTP. The B-coefficients of all covariates are larger in magnitude in the paths to BMI than to logTTP. With the exception of the influence of BMI on log'l'l'P, path analysis indicates that in the extreme groups (low and high BMI groups), relationships between the predictor variables and BMI and logTTP are similar. To test the stability of the models obtained in path analysis, the models were re-constructed by adding one covariate at a time to the model and noting the influence of its addition on the B-coefficients (Figure 2, step 9) (Table 8). The addition of age at menarche to the model changed B-coefficients by more than 20% in every model (Table 8a-d) while addition of age at starting time had an effect only in the High vs. Mid BMI path (Table 80). Model stability was also tested using multi-case deletion diagnostics (Table 9). Five individuals were temporarily dropped to test if the instability of the models, indicated by the effect of age at menarche, was due to outliers. An age at menarche greater than 18 years qualified as an outlier. The addition of age at menarche still changed the B-coefficients by more than 20% (Table 9, bold). 28 While addition of the age at menarche term still produced instability in the model, the B-coefficients of all terms in the High vs. Mid BMI group models (Tables 90 and 9d) were virtually unchanged when compared to the coefficients before case deletion (Tables 8c and 8d). This was not true for the Low vs. High BMI group ~ models (Tables 9a and 9b). After case deletion, the B-coefficients were different for age at starting time (increased), age at menarche (decreased), smoking 29 Table 8: Model stability of path analysis—stepwise covariate addition a: Low vs. Mid BMI group CALIS ModelsuBMl as endogenous variable Variable added [5 AGE at ST [3MENARCHE BSMOKING BALCOHOL [3EDU AGE at ST 0.8207 - - - - MENARCHE 0.3339 0.9847 - - - SMOKING 0.3352 0.9713 0.0336 - - ALCOHOL 0.3451 0.9652 0.0296 -0.0529 - EDUCATION 0.3427 0.9631 0.0315 -0.0548 -0.4343 *logTTP 0.3431 0.9624 0.0329 -0.0561 04340 *as endogenous variable R2 = 0.96 b: Low vs. Mid BMI group CALIS Models-logTTP as endogenous variable Variable added BBMI fiGYN. OP DSMOKING (3 AGE at ST BMENARCHE BALC. BEDU BMI 0.0539 - - - - - - GYN.OP 0.0504 0.5415 - - - - - SMOKING 0.0463 0.4933 0.0238 - - - - AGE at ST 0.0467 0.4936 0.0238 -0.00034 - - - MENARCHE 0.0072 0.4974 0.0217 -0.0207 0.1080 - - ALCOHOL 0.0085 0.4937 0.0204 -0.0235 0.1090 0.0153 - EDUCATION 0.0098 0.4978 0.0207 -0.0254 0.1106 0.0151 0.0403 R2 = 0.46 c: High vs. Mid BMI group CALIS ModelsuBMl as endogenous variable Variable added JSMOKING B AGE at ST BMENARCHE BALCOHOL BEDU SMOKING 1.3820 - - - - AGE at ST 0.1211 0.8255 - - - MENARCHE 0.0641 0.3059 1.0710 - - ALCOHOL 0.0594 0.3230 1 .0510 -0.0676 - EDUCATION 0.0554 0.3310 1.0257 -0.0766 -0.4084 *logTTP 0.0573 0.3319 1 .0239 -0.0783 -0.4069 ‘as endogenous variable R2 = 0.97 d: High vs. Mid BMI group CALIS Models-logTTP as endogenous variable Variable added DBMflGYN. OP BSMOKING [3 AGE at ST BMENARCHE DALC. BEDU BMI 0.0528 - - - - - - GYN.OP 0.0491 0.5659 - - - - - SMOKING 0.0455 0.5419 0.0214 - - - - AGE at ST 0.0473 0.5163 0.0213 -0.00152 - - - MENARCHE 0.0240 0.5249 0.0201 -0.0198 0.0776 - - ALCOHOL 0.0254 0.5206 0.0187 -0.0226 0.0785 0.0161 - EDUCATION 0.0268 0.5281 0.0193 -0.0257 0.0819 0.0160 0.0678 R = 0.46 30 Table 9: Model stability of path analysis—after multl-case deletion a: Low vs. Mid BMI group CALIS Models-BMI as endogenous variable Variable added (3 AGE at ST BMENARCHE BSMOKING BALCOHOL BEDU AGE at ST 0.8657 - - - - MENARCHE 0.5856 0.5569 - - - SMOKING 0.5765 0.5581 0.5222 - - ALCOHOL 0.5769 0.5668 0.0471 -0.0502 - EDUCATION 0.5722 0.5657 0.0589 -0.0795 -0.4047 *IogTTP 0. 5717 0.5666 0.0602 -0.0803 -0.3835 *as endogenous variable R2 =0.98 b: Low vs. Mid BMI group CALIS Models-~logTTP as endogenous variable Variable added BBMI 8GYN. OP [SSMOKING (3 AGE at ST BMENARCHE BALC. (SEDU BMI 0.0505 - - - - GYN.OP 0.0470 0.5693 - - - - - SMOKING 0.0435 0.5307 0.0206 - - - - AGE at ST 0.0374 0.5263 0.0207 -0.00545 - - - MENARCHE 0.0281 0.5350 0.0212 -0.0156 0.0573 - - ALCOHOL 0.0294 0.5361 0.0204 -0.0183 0.0582 0.0157 - EDUCATION 0.0307 0.5410 0.0208 -0.0207 0.0604 0.016 0.0536 R2 = 0.44 c: High vs. Mid BMI group CALIS Models—BMI as endogenous variable Variable added BSMOKING B AGE at ST BMENARCHE BALCOHOL BEDU SMOKING 1.382 - - - AGE at ST 0.1211 0.8255 - - MENARCHE 0.0642 0.3056 1 .071 - - ALCOHOL _ 0.0595 0.3226 1.052 -0.0673 - EDUCATION 0.0555 0.3305 1.026 -0.0763 -0.4058 JogTTP 0.0575 0.3314 1.025 -0.0780 04043 as endogenous variable R2 = 0. 97 d: High vs. Mid BMI group CALIS ModelsulogTTP as endogenous variable Variable added BBMI fiGYN. OlflSMOKING [3 AGE at ST BMENARCHE BALC. BEDU BMI 0. 0528 - - - - - GYN.OP 0.0491 0.5528 - - - - - SMOKING 0.0455 0.5004 0.0217 - - - - AGE at ST 0.0471 0.5017 0.0216 -0.00146 - - - MENARCHE 0.0236 0.5095 0.0204 -0.0199 0.0785 - - ALCOHOL 0.0250 0.5049 0.0190 -0.0228 0.0794 0.0165 - EDUCATION 0.0265 0.5123 0.0195 -0.026 0.0830 0.0164 0.0701 R2 = 0.46 31 (increased), alcohol (increased), and education (decreased) (Table 98 versus Table 8a). In Table 9b, only the coefficients for BMI (increased), gynecological operations (increased), and age at menarche (decreased) are changed from Table 8a (before case deletion). 32 Chapter 4 DISCUSSION The results suggest that fecundity, as measured by time to pregnancy, is highest within an ideal range of body composition (Table 3, Table 5). Membership in the left tail of the BMI distribution (<18 kglm2) can be explained by the year at which menses began and the year at which women began having intercourse without doing anything to avoid pregnancy. The number of cigarettes smoked at the starting time predicts membership in the right tail of the distribution (230 kg/m2) (Table 6). Cigarette smoking, previous gynecological operations, and early age at starting time are associated with significantly reduced fecundity (Table 7). The results of path analysis suggest that age at starting time, alcohol consumption at starting time, and completed level of education may exert their influence on time to pregnancy through an intervening effect of body mass index (Figure 3). The study population In order to increase sample size and power, the first and only pregnancy (FOP) and no prior live births (NPLB) populations were combined. In so doing, it was assumed that women who have had no prior live births (NPLB) have not carried (due to spontaneous abortion, miscarriage, etc.) their pregnancies long enough to experience the changes in body composition associated with pregnancy (Gunderson and Abrams, 1999). A significant difference was detected 33 for age at starting time between women for whom the pregnancy was their first and only (FOP) and women with no prior live births (NPLB). The null hypothesis is that the age at starting time is the same for the NPLB and FOP populations. The null hypothesis is rejected. Women with previous pregnancy failures (the NPLB group) are likely to be older than their counterparts who have never before been pregnant (the FOP group); the NPLB group tends to try repeatedly to carry a pregnancy to term. This justifies the combining of the FOP and NPLB populations into the final study population (n = 1757). The finding that BMI was highest for women with age at menarche in the lowest category (Table 2) lends credence to sub-hypothesis 1 (page 6) and seems to support the theory that women with a higher fat content tend to begin menstruating at an earlier age. This finding should be interpreted with caution; it is possible that changes in body composition could have occurred between puberty and the starting time. BMI for this study population was calculated from height and weight at the starting time. However, excess weight in adolescence does often persist into young adulthood (Wada and Ueda, 1990; Srinivasan et al., 1996) and suggests that a reverse association may exist, namely that adult BMI can be used as a proxy for BMI in adolescence. Table 2 also indicates that BMI was statistically larger in women who were older at starting time, supporting sub-hypothesis 2 (page 6), using age at starting time as a proxy measure for age at first conception. Others have found that fat content increases with age (Sarlio-Lahteenkorva and Lahelma, 1999). The finding that BMI was lowest in the highest category of completed education 34 (Table 2) supports sub-hypothesis 3 (page 6) and is bolstered by the findings of Sarlio-Lahteenkorva and Lahelma (1999). In their study of body mass index and social disadvantage in a representative sample of Finnish men and women aged 25-64, the authors found that the percentage of women with only a basic education (3 9 years) increased with increasing body mass index. They concluded that obese women in particular tend to face multiple social and economic disadvantages. Level of completed education was chosen in this current analysis as a proxy for socio-economic status. Women with professional training had a significantly shorter TTP than women with other levels of completed education, suggesting an association between social advantage and fecundity. Rachootin and Olsen (1982), in a study of the socioeconomic correlates of subfecundity, also found that women without a college education were more likely to exhibit primary subfecundity than college- educated women (p < 0.05). The differences in body mass indexby nationality (Table 2) are likely explained by social and cultural differences. Regional differences in 'I'TP detected in previous analysis of this data were not explainable in terms of regional differences in BMI (Juul et al. 1999). TI'P was shortest for women in the oldest category of age at starting time (Table 2) and could lead one to believe that old age at starting time increases fecundity (Table 7). However, this is likely to be a spurious association: older women who do not conceive right away may be more likely to stop trying sooner and thus would not be included in a pregnancy-based sample such as this. This 35 type of selection-out will result in only highly fecund older women being included in a pregnancy-based sample. The finding that median TTP was significantly longer for women with a history of PID, gonorrhea, ovarian cysts, endometriosis, and other gynecological operations (Table 4) was expected and agrees with the published literature (Cates et al., 1994). Table 5 also indicates that women in the low (<18 kglm2) and high (230 kg/m2) BMI groups and women with any previous gynecological operation (past curettage or other operations of the uterus, tubes, or ovaries) all have a significantly longer time to pregnancy (p=0.0339, p=0.0476, and p<0.0001, respectively). PID is a broad category of conditions and has an imprecise diagnosis (Holmes, 1998). Therefore it is likely that there is substantial overlap in these categories. A subfecund woman with a history of PID is more likely to have a history of infection (either gonorrhea or chlamydia), is more prone to endometriosis, and may seek medical attention in the form of curettage or another gynecological operation. Sexually transmitted diseases are capable of causing permanent damage to the reproductive tract (Westrom, 1994; Cates et al., 1994); endometriosis and/or therapeutic gynecological operations could confound an association between STDs and increased 'I'I'P. The analyses support the consistent finding of an association between clinical history variables (e.g. STDs, endometriosis, and the like), and decreased fertility and fecundity. 36 BMI and TTP The central hypothesis of these analyses surrounds BMI as the main predictor variable, namely that women with a higher BMI before their first and only pregnancy have a decreased waiting time to pregnancy (page 6). The finding that the median TTP was longer for the tails of the BMI distribution (p = .0635, Table 3, page 18) and that both BMI < 18 kg/m2 and BMI 2 30 kglm2 significantly reduced TTP (Table 5, page 22) suggests that fecundity is highest within an ideal range of body composition (18 kg/m2 < BMI < 30 kg/m2), lending at least partial support to the central hypothesis. Since body mass index is calculated from self-report of height and weight at the starting time, there is a potential for differential recall bias. Overweight women may be more likely to underestimate their weight. This differential misclassification, however, would tend to dilute the observed effect of BMI on TTP. If a body mass index at the tails of the distribution increases waiting time to pregnancy, what predicts BMI? The logistic membership models indicate that every year of later age at menarche increases the log odds of membership in the low BMI group by 0.2560 (p < 0.0001); every year of later age at starting time decreases the log odds of membership in the low BMI group by 0.121 (p < 0.0001). For every 1-cigarette increase in the number of cigarettes smoked per day, the log odds of membership in the high BMI group (2 30 kg/m2) increased by 0.0527 (p < 0.0005), suggesting that sub-hypothesis 4 (page 6) does not apply to this population. The overall model is highly significant as are the number of cigarettes smoked per day at the starting time (p=0.0005). 37 Those covariates that were significant predictors of BMI membership (Table 6) when added to the covariates included in the initial TI’P proportional hazards model (Table 5), yielded a final TTP proportional hazards model (Table 7, page 25). As the age at starting time increased, there was a decrease in the degree to which fecundity was lowered; the FR increases with increasing age, however it does not exceed one until the age at starting time is greater than 40 years of age. Again, the effect in the oldest age at starting time group is likely due to selection-out (page 33). Membership in both the low and high BMI groups serves to lower fecundity and women who had onset of menses before age 10 had increased fecundity. As the number of cigarettes smoked per day increased, the more fecundity was decreased, with the exception of the highest exposure group. Significance is reached for 11 or more cigarettes per day. While Table 7 suggests that the effect of smoking at starting time on TTP is diminished in the highest level of exposure (>20 cigarettes), this is likely due to misclassification of exposure due to self-report of exposure status. Individuals who truly belong in the highest category of exposure may, upon self-report, prefer to consider themselves in the next lowest category. This would serve to artificially inflate the effect of the penultimate exposure classification and deflate any effect of the highest classification. These results suggest that not only are age at starting time, age at menarche, and cigarette smoking at starting time associated with BMI, but they also influence fecundity as measured by TTP. 38 Caffeine, alcohol, and cigarettes in relation to BMI and TTP Bolumar et al. (1996) in a sister-study (population-based) to this one (pregnancy-based) found a similar association between female smoking at starting time and subfecundity as that found in Table 3, namely an increase in ‘I'I'P for women smokers. The population-based sample, drawn from Denmark, Germany, Italy, Poland, and Spain detected the association in each individual country and in all countries together. Women at the upper level of exposure (211 cigarettes per day) were more likely to have had a TTP of more than 9.5 months (OR = 1.7, 95% CI 1.3-2.1) (Bolumar et al., 1996). Indeed, most studies have found decreased fertilization potential associated with cigarette smoking (Stillman et al.,1986; Olsen et al., 1983). In preliminary analysis of the data, a linear model was constructed to model BMI as a continuous variable (data not shown). While the model indicated that BMI is not linear, an interrelation was noted between caffeinated coffee, the number of cigarettes smoked per day, and alcohol consumption. Cigar and pipe smoking was not included because only 1 woman was a cigar smoker and none of the women smoked pipes. In Spearman rank correlation, caffeinated coffee was significantly correlated with cigarettes (r=.28, p<.0001) and with alcohol (r=.21, p<.0001), and alcohol was significantly correlated with cigarettes (r=.09, p=.0002) (Figure 4). 39 r=.28, p<.0001 Coffee ¢ > Cigarettes r=.21, p<.000\ /r:09, p=.0002 Alcohol Figure 4: Spearman rank correlations between coffee, cigarette smoking, and alcohol. To assess if one of the variables was responsible for the interrelation, partial Spearman correlations were obtained. Spearman correlations are based on ranks and are calculated for data that is not normally distributed (Rosner 1995). If an association between two of the variables disappeared while controlling for the third, the controlled variable explains the association. The relationships remained unchanged except when caffeinated coffee consumption was controlled. Partialing out caffeinated coffee caused the association between cigarettes and alcohol to disappear (r=.04, p=0.11). BMI was not correlated with either alcohol, caffeine, or cigarettes; controlling for BMI did not change the effect of coffee on the relationship between alcohol and cigarettes. By creating a frequency matrix of the categorical variables BMI, coffee, alcohol, and cigarettes, it was noted that 53.3% of the womenldid not smoke at the starting time, 23.9% of the women neither smoked nor drank caffeinated coffee, and 14.6% of the women drank neither coffee or alcohol nor smoked cigarettes. To verify the role of caffeinated coffee in the interrelation, partial correlations were obtained separately when coffee consumption was zero, when 40 alcohol consumption was zero, and when cigarette smoking was zero; the relationship remained. The questionnaire also obtained information on the consumption of other caffeinated beverages, namely teas and colas. To determine if the relationship existed for only caffeinated coffee or for caffeine in general, all caffeinated beverages were added together and the correlations re-assessed. The interrelations remained. When controlling for caffeinated beverages, the relationship between cigarettes and alcohol disappeared (r=0.042, p=.11). Based on the partial Spearman correlations, the interrelation between alcohol consumption and cigarette smoking seems to be mainly explained by caffeine as a third variable. To further explore the relationship between caffeine, alcohol, and cigarettes, four separate proportional hazards regression models were created with different constructs of 2-variable interaction terms. Alcohol, caffeine, and cigarettes were each categorized into three levels of use: none (0) , moderate, and considerable. Moderate alcohol consumption was defined as 1-3 beverages per week; considerable consumption was 4 or more beverages per week. Moderate caffeine consumption was defined as 1-4 servings of caffeinated coffee, tea, or cola per day; considerable consumption was defined as 5 or more servings per day. Moderate cigarette use was defined as 1-10 cigarettes per day; considerable use was 11 or more cigarettes per day. Four sets of two-variable interaction terms (e.g. caffeine-alcohol interaction, caffeine-cigarette interaction) were created: 2 two-level interactions, 1 41 three-level interaction, and one 4-level interaction (Table 10). Separate proportional hazards regression models were constructed, with time to pregnancy as the event for each set of caffeine-cigarette-alcohol interaction and including the covariates in the starting TTP model (Figure 2, step 4) (all primary predictor and clinical history variables). The zero-value terms served as the referent group. In none of the four models were any of the interaction terms close to statistical significance (p >> 0.11, data not shown). This suggests that interaction does not explain the relationship between caffeine, cigarettes, or alcohol; the relationship seems to be explained by caffeine consumption as a moderator variable. If caffeine consumption is not controlled in an analysis of the effects of cigarette smoking and alcohol consumption on time to pregnancy, a spurious association could appear between alcohol consumption and cigarette smoking. It might be that consumption of greater quantities of alcoholic beverages by an individual may serve to “train” the liver into metabolizing caffeine more efficiently. 42 Table 10-Coding for caffeine-alcohol-cigarette interaction terms a. 2-level interactions none moderate considerable Exposure 1 b. 2-level interactions none moderate considerable Exposure 1 c. 3-level interactions none moderate considerable Exposure 1 d. 4-Ievel interactions none moderate considerable Exposure 1 Exposure 2 none moderate considerable 0 0 0 0 0 0 0 0 1 Exposure 2 none moderate considerable 0 0 0 0 1 1 0 1 1 Exposure 2 none moderate considerable 0 0 0 0 1 1 0 1 2 Exposure 2 none moderate considerable 0 0 0 0 1 2 0 2 4 43 Path analysis Path Analysis is a relatively recent addition to the set of statistical tools in the clinical sciences. The fact that the CALIS procedure in SAS can model intervening effects makes it appealing to researchers. Age at menarche (in addition to age at starting time) was retained in the final logistic model predicting membership in the low (<18 kg/m2) BMI group (Table 6a) and may indicate that the effect of age at menarche on time to pregnancy is mediated by body mass index. This data on TTP proposes an interesting methodologic issue: BMI appears to have an effect on TI'P at the extremes of the BMI distribution. To investigate this U-shaped effect of BMI on TTP, path analysis was completed separately for two groups of the study population, namely the same two groups used in the logistic regression modeling (low vs. mid; high vs. mid). Figure 2 indicates that both age at menarche and age at starting time are statistically significant predictors of body mass index. Women with delayed menarche do have increased cycle variability (r = 0.063, p = 0.018), but there is no direct correlation between BMI (continuous or grouped) and cycle length variability (r = 0.028, p = 0.29). The presence of an intervening effect was tested by removing BMI as both and endogenous and exogenous variable from the path analysis. An intervening effect is assumed if the parameter estimates are reduced when BMI is removed. Removal of BMI from the Low versus Mid model did not produce changes in the parameter estimates; there is little intervening effect of BMI in this model. 44 Removal of BMI from the High versus Mid model did produce changes in the parameter estimates. These changes suggest that BMI has an intervening effect on: age at starting time and LOGTTP ((3 = -.0257 reduced to B = -.0168 upon removal of BMI); on completed level of education and LOGTTP ((3 = .0678 reduced to [3 = .057 upon removal of BMI); and on alcohol consumption and LOGTTP (B = .016 reduced to B = .0139 upon removal of BMI). The parameter estimate of age at menarche increased upon removal of BMI ([3 = .0819 increased to 8 = .1094), suggesting that BMI is a moderator of the effect of age at menarche on LOGTTP. ‘As a type of linear regression, the CALIS procedure in SAS assumes that the endogenous (dependent) variable or its log and any intervening variables are normally distributed (a bell-shaped curve), and that the data fit a linear model. To test these assumptions, 8 linear model of TTP using the covariates in the final proportional hazards model was created (data not shown). Those covariates that were significant predictors of TI'P in the survival analysis model (age at starting time < 20, age at menarche < 10, gynecological operations, and greater than 10 cigarettes per day at the starting time) were also the only variables that reached significance in the linear model; the linear model assumption is fulfilled. The univariate tests for normality (Shapiro-Wilk, Kolmogorov-Smirnov) were all significant (Ho: the distribution is normal). However, these tests are very sensitive and the histogram of the residuals of log‘l'l'P and the normal probability plot both suggest that the distribution of log'l‘I'P is approximately normal (data not shown). The plot of residual vs. predicted plots suggests homoskedasticity. 45 Because body mass index is not linear, but rather U-shaped, two logistic models were used to model BMI. Within each of the logistic models, there is a linear trend between BMI and TTP. The multivariate normal distribution assumption is not fulfilled for BMI or its log. Inability to fulfill the assumption of normality for BMI may produce poor estimates in PROC CALIS (SAS, 1999). Testing the stability of the path analysis models indicates that in each instance that age at menarche is added to the model, B-coefficients change by more than 20% in each model (Table 8, bolded coefficients), suggesting that the models may be unstable. The fact that multi-case deletion does not alter the models in the high vs. middle BMI groups is likely due to the fact that, of the five individuals that were temporarily dropped, 2 were in the low BMI group and 3 were in the middle two groups (the reference category). The two cases that had the longest value of TTP for the five deleted cases also had the lowest BMI (<18 kg/m2), the earliest age at starting time (21 and 17 years), the lowest level of education (both left school before age 16), and smoked 10 and 20 cigarettes per day at the starting time, respectively (Table 11). For the High vs. Mid BMI group, age at menarche still has its effect on the B-coefficients in the models, but the effect is not due to outliers (Tables 9c and 9d versus Tables 8c and 8d). The difference noted for the Low vs. Mid BMI group (Tables 9a and 9b versus Tables 8a and 8b) suggests that the effect of age at menarche on the stability of the model is due, at least in part, to the presence of outliers. 46 Table 11: Characteristics of deleted cases Age at Age at Smoking at Menarche Starting Tlme BMI Starting Time Completed Level TTP Case (yrs) (yrs) (kgm2) (Cigarettes/day) of Education (months) 1 19 21 17.10 10 left before age 16 94.0 2 19 17 17.69 20 left before age 17 8.0 3 21 29 20.57 0 university degree 2.1 4 22 22 21.72 3 professional training 2.0 5 23 24 24.84 0 professional training 0.5 47 Chapter 5 METHODOLOGIC CONSIDERATIONS Does age at interview confound the effect of age at starting time? In general, the earlier a woman begins trying to achieve pregnancy the earlier she tends to succeed (Table 2). In this regard, current age (age at interview) could potentially confound a relationship between age at starting time and BMI if it is correlated both with body mass index and time to pregnancy. The age at which BMI was obtained is unavailable. The questionnaire merely asks "What was your height and weight before this pregnancy" (Appendix A). The Spearman rank correlation for age at interview and BMI was r = 0.065, p=.0071, based on 1728 observations. Though the correlation is small, it is statistically significant. Typically, this confounding could be determined in linear regression by evaluating the change in regression parameters as age at interview is added to the model. Indeed, using linear regression to model BMI as a continuous variable, age at interview does confound the association between starting time and BMI. However, for the purposes of this analysis two logistic models were constructed with BMI as the dependent variable: one that models membership in the left tail (<18 kg/m2) versus a middle group (18-30 kglm2), and one that models membership in the right tail (230 kglm2) versus the middle group. The potential confounding effect of age at interview was assessed by re-creating these logistic models with both age at interview and age at starting time in the 48 starting model and then without age at interview in the final model, yet retaining the variables that previously had remained in the model after confounder assessment. As before, the difference in the likelihood ratios follows a Chi- square distribution with df = dfslamng mode. - dfreduced model. For both of these models, removal of age at interview does not yield a significant p-value for the Chi-square test; in both instances .25< p <50 and suggests that age at interview can be removed from the model. Age at interview and age at starting time are more closely correlated in the middle of the BMI distribution but begin to diverge at the tails of the BMI distribution (and the TTP distribution). The results of the analyses carried out in this project have demonstrated that membership in the middle BMI groups does not seem to affect TTP. The median time to pregnancy in the left tail (<18 kg/m2) is significantly different from the middle (Table 3); the same is not true, however, for the right tall (230 kg/m2). Using the 18-25 kglm2 group as the reference, the fecundability ratio for the 25-30 kg/m2 BMI group in the proportional hazards model is close to 1 (FR=1.028), and not statistically significant (p=0.73). Therefore, because this analysis: 3.) uses logistic regression in the modeling of BMI; b.) is concerned only with the tails of the distribution of BMI (and thus the middle of the distribution serves as the reference in modeling); 0.) demonstrates that age at interview falls out of both BMI models using a Chi-square test for confounders; and d.) demonstrates that only the lowest two groups of age at starting time (<20, and 21 -25) are significant in the TTP proportional hazards 49 model, age at interview confounds neither the association between age at starting time and BMI nor the association between age at starting time and TTP. Pregnancy-based versus population-based sampling As a study of subfecundity, the study population consists only of pregnant or recently pregnant women and therefore, by design, does not include those women who never actually conceived. It is not expected that the “exposures” examined in this analysis lead to sterility. Indeed, this analysis would not be capable of detecting such effects. Rather, it is hypothesized that they have an impact on the period of time it takes for a couple to conceive. A pregnancy-based sample Is capable of detecting such shifts in the data distribution (Bolumar et al. 1996) A pregnancy-based survey has the advantage of reducing information bias. The pregnancy is a recent (or even on-going event) and memories surrounding conception will likely be fresh. Additionally, because all women in the study conceived, there is no differential recall for women who did conceive versus those who did not. 50 Chapter 6 CONCLUSIONS This pregnancy-based, cross-sectional study suggests fecundity, as measured by time to pregnancy, is highest within an ideal range of body composition. Membership in this ideal range of BMI can be predicted for the study population by the year at which menses began, the year at which women began having intercourse without doing anything to avoid pregnancy, and the number of cigarettes smoked at the starting time. Cigarette smoking, previous gynecological operations, and early age at starting time significantly reduce fecundity as measured by time to pregnancy. Path analysis suggests that the effects of age at starting time, completed level of education, and alcohol consumption on time to pregnancy in the high BMI population (2 30 kg/m2) are mediated by body mass index. It is estimated that in the year 2000, there will be approximately 5.13 million women in the United States with impaired fecundity, defined by the National Center for Health Statistics to include women who: are unable to have a baby due to reasons other than surgical sterilization; report difficulty conceiving or delivering a baby or that they had been told a pregnancy was dangerous to them and/or the baby; and/or were continuously married and did not conceive after 36 months of intercourse without contraception (Stephen, 1996). Additionally, the prevalence of obesity in the United States (15%) is significantly higher than in European countries (7% in France and 9% in the United Kingdom) 51 Women are also more commonly very obese (more than 50% overweight) (Laurier et al., 1992). Only 53 of the 1757 women (3%) in this study had a BMI 2 30 kg/m2, yet their time to pregnancy was significantly increased. While the population utilized in this investigation is not representative of all European women, the results do suggest that obesity may particularly important in fertility and fecundity; this role is likely amplified in a heavier population such as that found in the United States. 52 APPENDICES APPENDIX A ESIS QUESTIONNAIRE ON PREGNANCY AND FERTILITY 54 European Studies of Infertility and. S ulfecundz'fy Questionnaire on Pregnancy and Fertility (for women who have recently given birth) The purpose of these questions is to learn more about how easily women who want to have children get pregnant. Also we intend to find out whether the time it takes to become pregnant is related to factors such as working conditions, life-style or medical reasons. We can thus try to identify any avoidable risks ofinfertility. Most questions relate to the time leading up to your recent pregnancy. Please fill in the questionnaire as best as you can. We appreciate your help. The information you will give is anonymous and strictly confidential. Your name is not recorded together with your answers. Many women will be interviewed and your answers will be totalled up with theirs. Your participation is, of course, voluntary. It will take approximately 10 to 15 minutes to complete this questionnaire. Sign. Note: This questionnaire is intended for women who have recently given birth to a child. If you have not recently given birth, please give the questionnaire back. 55 Instructions for filling in the questionnaire For some questions, please put a cross or a tick in the box next to the answer that best describes you or your experience. For example: G7. What was your pattern of work? daytime (with or without flexi-time) ............ D 1 evening ................................................................ CI 2 night .................................................................... El 3 shift-work (changing or rotating) ................. C] q. Or enter dates or durations (Year/ month) of events. For example: A7. When were you born? Month: Year: 19 ( meaning August 1962) A8. How long did it take you to become pregnant? Weeks: Months: Years: (meaning 3% months) Sometimes your answer will allow you to skip certain questions or sections of questions. Please read the "GO TO -> " statements carefully to make sure you answer all the appropriate questions. For example: GO TO —-> question 0.5 on page 7 56 SECTION A: Your recent pregnancy Al. Please write today's date here: Day: Month: Year: A1D,A1M,A1Y A3. When was the baby born? Day: Month: Year: A3D,A3M,A3Y A4. How many weeks or months were you pregnant? Pregnant for: weeks: and/ or Months: A4W,A+M A5. Where do you live now? (just give the town or city, not the street) A5 The following questions consider the interval before this pregnancy 57 SECTION B: Pregnancy and contraception Below are some statements about the way you became pregnant. Choose the one that best describes how you became pregnant. Then tick the box and skip to the page indicated. This was your first pregnancy, and: You had been pregnant before this pregnancy, and: You have never used a birth control method (such as the pill, condoms, or rhythm method). D 1 GO TO -——> Section D, page 5.9 After your previous pregnancy, your menstrual periods started again. Since then, you have used no birth control methods (such as the pill, condoms, or rhythm method). Cl 4- 00 T0 —-> Section F, page 6’] You became pregnant while using a birth control method (regularly or irregularly). Cl 9 GO TO —> Section C, page .58 You became pregnant while using a birth control method (regularly or irregularly). Note: Breast-feeding is not a birth control method. CI 5 GO TO -—> Section C, page .58 You used to use birth control, and you became pregnant since you gave it up. Cl 3 GO TO —> Section D, page 5.9 After your previous pregnancy your menstrual periods started again. Since then you used birth control for a time, gave it up, and then became pregnant. [I 6 GO TO —> Section D, page 5.9 Your periods did not start again since your previous pregnancy. You were not using birth control when you became pregnant. CI 7 GO TO —> Section E, page 60 You should have ticked one of the above boxes. If not, please try again. 58 BI SECTION C Questions for you if you became pregnant in spite of using birth control (regularly or irregularly). If this is not true, go back to page 57 and check your answer. C1. What kind of birth control were you using around the time you became pregnant? (Tou may mark more than one) Rhythm method (safe periods) ....................... CI 1 CIA Withdrawal (coitus interruptus) ................... CI 1 C13 Coil (intra-uterine device) ............................... El 1 CIC The Pill (oral contraceptive) ........................... CI 1 C1D Condom ............................................................... Cl 1 C112 Cap (diaphragm) ................................................ El 1 CIF Contracpetive injection or implant ............... Cl 1 ClG Jelly, cream or foam .......................................... U 1 CIH Other .................................................................... Cl 1 ( Please write below) CII C2. Were you using the birth control in a regular and consistent manner when you became pregnant? No, not quite regularly .................................... CI 1 C2 Yes, regularly and consistently ..................... Cl 2 C3. For how long were you using the birth control up to your pregnancy? (Any of the methods you ticked above) Months:______ and/or Years:_ C3M,C8Y C4. Now write in the box below the month and year your pregnancy started. We call this the "STARTING TIME" STARTING TIME: C4M,C4Y Month: Year: 19 NOW GO TO: —> Section G on page 6.9 59 SECTION D Questions for you if you were not using any birth control when you became pregnant. If that is not true, go back to page 57 and check your answer. D1. Leading up to this pregnancy, when was it that you started having sexual intercourse without using any birth control to prevent pregnancy? We call this the "STARTING TIME" STARTING TIME: D1M,D1Y Month: Year: 19 D2. How long was it from that "STARTING TIME" until you became pregnant? (The date you became pregnant is the date you conceived). How long? Weeks: and/ or Months: and/ or Years: D2W,D2M,D2Y D3. To put it differently: How many periods did you have between the "STARTING TIME" and you becoming pregnant? No periods ..................................................... C] 0 D8 1 period .......................................................... El 1 2 periods ........................................................ CI 9 3 periods ........................................................ Cl 3 More than 3 periods ................................... El .1 NOW GO TO: —> Section G on page 63 60 SECTION E: Questions for you if your periods had not started again when you became pregnant. If this is not true, go back to page 57 and check your answer. E1. Now write in the box below the month and year your pregnancy started. We call this the "STARTING TIME" STARTING TIME: E1M,E1Y Month: Year: 19 NOW GO TO: —> Section G on page 6'3 61 SECTION F: Questions for you if you used no birth control method since your previous pregnancy. If this is not true, go back to page 57 and check your answer. Fl. When was it that you started having sexual intercourse after your previous pregnancy? Month: Year: 19 F1M,F1Y F2. F3. At that time (when having sexual intercourse again): Had your menstrual periods restarted? D 1 Yes V We call the time you started having sexual intercourse again after your previous pregnancy the "STARTING TIME". l D 2 NO F2 V Now work out approximately the month and year when your periods returned after your previous pregnancy. We call this the "STARTING TIME". STARTING TIME: F2M,F2Y Month: Year: 19 How long was it from that "STARTING TIME", until you became pregnant? (The date you became pregnant is the date you conceived). How long? Weeks: and/ or Months: and/or Years: F3W,F3M,F3Y 62 F4. To put it differently: How many periods did you have from the "STARTING TIME" up to you becoming pregnant? No periods ............................... Cl 0 1 period .................................... Cl 1 2 periods .................................. Cl 9 3 periods .................................. D 3 More than 3 periods ............. El .1 NOW GO TO: —> Section G, page 63 F4 63 SECTION G: Questions about life and work at the "STARTING TIME" G1. You have just stated what we call your "STARTING TIME". Please write again this date in the box: STARTING TIME: G1M,GlY Month: Year: 19 G2. Did you have a paid job at the "STARTING TIME"? (Do not considerjobs started only since the "STARTING TIME"). Ge N o .............................................................. Cl 1 GO TO —> question G11, page 67 YES .............................................................. D 2 G3. In what kind of industry or business were you working? (Please be as precise as possible) G4. What was your job-title? (Please be as precise as possible: Do not fill I'ny'ust 'nurse', write psychiatric nurse') G5. What kind of work did you do? (Please be as precise as possible) G5A,GSB G6. For how many hours per week did you work on average? Hours per week: ................................... G6 64 G7. What was your pattern of work? Daytime (with or without flexi-time) ...................... Cl 1 G7 Evening ......................................................... CI 2 Night .............................................................. I] .1 Shift-work (changing or rotating shifts) ..................... CI .1 G8. Did your job at the "STARTING TIME" involve any working with VDUs (computers or word processors with a screen)? NO ............................................................ D 00 Yes How many hours per week? (average) Gs 65 G9. How often did you come into contact with the following exposures in your job? (You should tick one box/or each erposure) GSA-P from time most of to time your occasio— each working never nally day week a. Paints, varnish, lacquer ................................ 1 2 3 4 b. Dyes, pigments, inks ...................................... 1 2 3 4 c. Solvents .............................................................. 1 2 3 4 d. Degreasing or drycleaning agents ............ 1 2 3 4 e. Resins, adhesives ............................................. 1 Q 8 4 f. Petrol, petrochemicals .................................. 1 2 8 4 g. Cutting, lubricating oils ............................... 1 2 3 4 h. Welding fumes ................................................. 1 2 3 4 i. Metal dusts, fumes ......................................... 1 2 3 4 j. Engine exhaust ................................................ 1 2 3 4 k. Pesticides, fungicides, insecticides, weedkillers ............................... 1 2 3 4 1. Wood preserving materials ......................... 1 2 3 4 m. Anesthetic gases .............................................. 1 2 3 4 n. Radioactivity or x-rays ................................. 1 2 8 4 o. Sterilizing gases (ethylenoxide etc.) ........ 1 Q 3 4 p. Loud noise ......................................................... 1 2 3 4 Others (please sped/y): Briefly describe how you came in contact with these substances, giving if possible their names: [ Item I: may be split in three. Ezplanation attached] 66 G 10. The following questions are about how you experienced your job situation at the "STARTING TIME". Please answer each question by ticking the one box that best fitted your job situation. Sometimes none of the answers fits exactly. Please choose the answer that comes closest. GIOA-N Strongly Strongly disagree Disagree Agree agree a. My job required that I learnt new things ................................ l 2 3 4 b. My job involved a lot of repetitive work ................................................. 1 2 3 4 0. My job required me to be creative ............ 1 2 8 4 d. My job required a high level of skill ....................................................... l 2 3 4 e. On my job, I had very little freedom to decide how I did my work ...................... 1 2 3 4 f. I had a lot of say about what happened on my job ....................................... 1 2 3 4 g. My job required lots of physical effort .................................................. 1 2 3 4 h. I was not asked to do an excessive amount of work ............................ 1 2 8 4 i. I had enough time to get the job done ...................................................... 1 2 8 4 j. I was often required to move or lift heavy loads on my job ....................... 1 2 3 4 k. My job was very hectic .................................. 1 2 3 4 l. I was often required to work for long periods with my body in physically awkward positions ..................... l 2 3 4 m. On my job I was often told that I was doing a good job .................................. 1 2 3 4 n. On my job I was often treated unfairly by another person ........................... 1 2 3 4 67 G 1 1. At the "STARTING TIME": Did you smoke? Cl 1 Yes i What? How many per day? Cigarettes: __ per day Cigars: __ per day Pipe tobacco: __ per day [19 No G11 GllA-C Did you smoke before then? C], No UgYes GnD When did you quit smoking? Month: Year: 19 G11M,G11Y G 12. At the "STARTING TIME": Were you exposed to other people's cigarette smoke? At work? ....................................................... [31 Yes D 2 No (312A Outside work? ............................................ Cl 1 Yes CI 2 No G128 G 13. For each of the following drinks, how much did you drink at the "STARTING TIME"? Caffeinated coffee: ................ __ cups per DAY (313A Decaffeinated coffee ............... __ cups per DAY 0133 Tea: ................................ __ cups per DAY G13C Cola: ............................................... __ glasses/cans per DAY 013D Beer: ............................................... __ glasses/ bottles per WEEK GISE Wine: ............................................... _ glasses per WEEK G13F Spirits: ............................................. __ glasses per WEEK (3130 Aperitifs/ sherry/ port: ................. __ glasses per WEEK GlsH G14. Where did you live at the "STARTING TIME"? (just town or city, not street) G14 68 G 15. Since when have you lived in this area? Month: Year: 19 [G15M,G15Y] 69 SECTION H: Questions about the father of your child. The following questions refer to his life and work atyour "S TAR TIN G TIZWE". H1. At the "STARTING TIME": Did your child's father smoke? Cl 1 Y s U 9 No Hi I U 9 Don't know What? How many per day? Cigarettes: __ per day H1A Cigars: _ per day HIB Pipe tobacco: __ per day h HlC H2. For each of the following drinks, around the "STARTING TIME", how much did he drink? Beer: ............................................... __ glasses/ bottles per WEEK HQA Wine .............................................. _ glasses per WEEK H28 Spirits: .................................................................. __ glasses per WEEK HQC Aperitifs/ sherry/ port: ...................................... __ glasses per WEEK HQD Don't know ................................................. U 99 H3. Was he practising any sport or hobby for which he exercised physically? El 99 No D Yes For how many hours per week on the average? __ hours per week [H3] H4. When was he born? Year: 19___ ................................................ H4 Don't know ................................................. C] 99 7O H5. In which town and country was he born? Don't know .................................................... C19 115 H6. Is he presently under education? Cl 9 No Cl Yes H6 Which education? Still in school ................................................ D 1 Post school training in a manual trade (eg apprenticeship) ........................... U 9 Write education below Post school training in a profession (cg nursing) .............................. Cl 9 Write education below University education .................................. Cl 4 Write education below Other education ........................................... D 5 Write education below V H7. What is the last education or training he has completed? Left school at age 15 or before, no further education ................................... U 1 Left school at 16 or 17, no further education ................................... Cl 9 Left school at 18 or older, no further education ................................... Cl 9 Post school training in a manual trade (eg apprenticeship) ........................... Cl 4 Post school training in a profession (eg nursing) .............................. E] 5 University degree ........................................ Cl 6 Other .............................................................. Cl 7 Don't know ................................................... E] 9 H7 Write education below Write education below Write education below Write education below 71 H8. Did he have a paid job at your "STARTING TIME"? YCS .............................................................. D 1 H8 No .............................................................. El 9—> GO TO section 11’, page 74 Don't know ................................................... El 9—> GO TO section K, page 74 H9. In what kind of industry or business did he work? ( Please be as precise as possible) H10. What was his job-title? (Please be as precise as possible: Do not‘justfill in 'welder', write 'steel—welder') H1 1. What kind of work did he do? (Please be as precise as possible) H11A,H11B H12. For how many hours per week did he work on average? Hours per week: ................................... H12 H13. What was his pattern of work? Daytime (with or without flexi-time) ...................... Cl 1 H13 Evening ......................................................... D 9 Night .............................................................. D 3 Shift work (changing or rotating) ................................ U 4 72 H14. At the "STARTING TIME", did your partner drive a vehicle regularly? D o No Cl Yes H14 What kind of vehicle? Car or taxi ............................... C] 1 Bus ......................................... Cl 9 Van ......................................... El 5 Lorry ......................................... Cl 4, Other ......................................... El 5 Please specify: For how many hours a day did he drive on the average? V Hours per day: ................... H1+A H15. Was his work at the "STARTING TIME" mostly seated or standing? Mostly seated in a chair ............................. U 1 H15 Mostly standing .......................................... Cl 9 Both seated and standing .......................... El 5 Mostly seated in a vehicle ......................... U .1. Other .............................................................. Cl 5 Please specify: 73 H16. Which of the following exposures did he come into contact with in his job? (Ton should tick one box for each exposure) H 16A-Q No Yes Don't know a. Paints, varnish, lacquer ................................ 1 2 3 b. Dyes, pigments, inks ...................................... 1 2 3 c. Solvents .............................................................. 1 2 3 d. Degreasing or drycleaning agents ............ 1 2 3 e. Resins, adhesives ............................................. 1 2 3 f. Petrol, petrochemicals .................................. 1 2 3 g. Cutting, lubricating oils ............................... 1 2 3 h. Welding fumes ................................................. 1 2 3 i. Metal dusts, fumes ......................................... 1 2 3 j. Engine exhaust ................................................ 1 2 3 k. Pesticides, fungicides, insecticides, weedkillers ............................... 1 2 3 1. Wood preserving materials ......................... 1 2 3 m. Anesthetic gases .............................................. 1 2 3 n. Radioactivity, x-rays ...................................... 1 2 3 o. Sterilizing gases (ethylenoxide etc.) ........ 1 2 3 p. Loud noise ......................................................... 1 2 3 Heat .................................................................... 1 2 3 Others (please specg’i below) ............................. 1 2 3 Briefly describe - if you can - how he came in contact with these substances, giving if possible their names: [Item k may be split in three. Explanation attached] 74 SECTION K: Health factors K1. Around your "STARTING TIME": How long was it from the start of one menstrual bleeding to the start of the next bleeding? Number'ofdays: ................................ __ or K1A,KiB Between ...................... _ and __ days No bleeding at all ................................... C] 0099 I can't remember, don't know .............. U 9999 K2. At your "STARTING TIME", how often did you have sexual intercourse? Daily [:1 1 .......................................................... K2 At least once a week ................................... C] 9 Two to four times a month ....................... Cl 5 Less than twice a month ............................ Cl .1 I can't remember .......................................... El 5 I don't want to answer ............................... Cl 5 K3. Did you plan to have a baby at that time? Cl 1 Yes Cl 9 Undecided Cl 5 No K3 l 1 K4. Did you or your partner seek any medical or professional advice to help you to become pregnant? Cl 1 Yes Cl 9 No K4 K5. How long had you been attempting to become pregnant when you sought this advice? Months:_____ and/or Years:_ K5M,K5Y V V 75 K6. Have you given birth to any children before this pregnancy? NO ............................................................ D 00 K6 ch ............................................. How many? (number of livcborn children) K7. Did you ever have a cesarean section before this pregnancy? NO .............................................................. D 0 K7 Yes ............................................. How many? (number of cesarean sections) K8. Have you ever had any miscarriages? (Don't include miscarriages you are uncertain about) NO ............................................................ D 00 K8 Yes .............................................. How many (number of miscarriages) K9. Have you ever had any pregnancies outside the uterus (ectopic pregnancies)? NO .............................................................. D 0 K9 Yes .............................................. How many (number of ectopic pregnancies) K110. Have you ever had any stillbirths? No .............................................................. Cl 0 K10 Yes ............................................. How many? (number of stillbirths) K1 1. Did any of your children die within the first 7 days after birth? No .............................................................. El 0 K11 Yes .............................................................. U 1 K12. Have you ever had any induced abortions (terminations)? NO ............................................................ D00 K12 Yes ............................................. How many? ' (number of induced abortions) K13. All in all: How many times have you been pregnant (including your last pregnancy)? I have been pregnant .................. times K13 76 K14. How old were you, when you had your first menstrual periods? Age: ....................................... years old K 14 I never had menstrual periods ................ D 00 I can't remember, I don't know .............. Cl 99 K15. Have you ever been told by a doctor that you have had any of the following infections, diseases or operations? Tick the N O or YES boxjor each one. If TES, please give the year for the first time. No Yes Year first time PID: Pelvic inflammatory disease (eg. infection in fallopian tubes or ovaries) .......................................................... Cl 1 Cl 9 19 K15A K15AY Chlamydia infection ....................................... Cl 1 Cl 9 19 K15B K15BY Gonorrhea infection ....................................... D 1 Cl 9 19 etc. Other sexually transmitted diseases ........ Cl 1 U 9 19 Ovarian cysts .................................................... Cl 1 Cl 9 19 Fibroids, myomas ............................................ Cl 1 El 9 19 Endometriosis .................................................. U 1 Cl 9 19 Thyroid disease ............................................... Cl 1 C] 9 19 Diabetes ............................................................. D 1 D 9 19 Removal of appendix (appendectomy) El 1 D 9 19 Pelvic infections after former pregnancies ......................................... Cl 1 El 9 19 Chemical or radiation therapy because of cancer ............................................ Cl 1 El 9 19 Curettage ........................................................... El 1 E] 9 19 Other operations of the uterus, K 15N tubes or ovaries ............................................... Cl 1 D 9 19 K 15NY 77 K16. Have you ever used the IUD coil for birth control? Cl 9 No, never Cl Yes K16 l How many times inserted? How long in total? V Months: and/0r Years: K16M,K16Y K17. Have you ever used the pill for birth control? CI 9 No, never Cl 1 Yes K17 When did you stop using the pill for the last time? (please be as precise as possible) j Month: Year: 19 K17M,K17Y K18. What was your height and weight before this pregnancy? Weight: ............................................. Height: ............................................. cm K18A kg K188 78 SECTION L: General questions L1. What is your year of birth? Year: 19 ................................................ L1 L2. Are you married? No, unmarried .............................................. Cl 1 L2 Yes, married ................................................. El 9 L3. In which town or country were you born? L3 L4. Do you belong to any religion? Cl 0 N 0 El Yes [L4] Which religion? L5. Regarding your attitude towards marriage, partnership and sexuality: Do you follow your religion's recommendations? No .............................. El 1 [L5] Yes ............................ CI 9 ; Don't know ............. Cl 5 79 L6. Are you presently under education? D 9 No [3 Yes L6 i Which education? Still in school ................................................ Cl 1 Post school training in a manual trade (eg apprenticeship) ........................... El 9 ”’rite education below Post school training in a profession (eg nursing) .............................. Cl 9 Write education below University education .................................. U 9 Write education below Other education ........................................... Cl 5 Write education below V L7. What is the last education or training you have completed? Left school at age 15 or before, no further education ........... [:11 L7 Left school at 16 or 17, no further education ................................... C] 9 Left school at 18 or older, no further education ................................... Cl 9 Post school training in a manual trade (e g apprenticeship) ........................... D .1 Write education below Post school training in a profession (eg nursing) .............................. D 5 Write education below University degree ........................................ U 9 Write education below Other .............................................................. D 7 Write education below END: Thank you very much for the information and for your patience 80 REFERENCES References Berube S, Marcoux S, Langevin M, Maheux R, et al. Fertil Steril 1998;69:1034- 1041. Bolumar F, Olsen J, Boldsen J, et al. Smoking reduces fecundity: a European multicenter study on infertility and subfecundity. Am J Epidemiol 1996; 1 43:578-587. Bolumar F, Olsen J, Rebagliato M, Bisanti L, et al. Caffeine intake and delayed conception: a European multicenter study on infertility and subfecundity. Am J Epidemiol 1997;145:324-334. Bongain A, lsnard V, Gillett JY. Obesity in obstetrics and gynaecology. Eur J Obstet Gynecol Reprod Biol 1998;77:217-28. Cates W, Wasserheit JN, Marchbanks PA. Pelvic inflammatory disease and tubal infertility: the preventable conditions. Ann N Y Acad Sci 1994;709: 179-95. Charles MA, Pettitt DJ, McCance DR, Hanson RL, Bennett PH, Knowler WC. Gravidity, obesity, and non-insulin-dependent diabetes among Pima Indian women. Am J Med 1994;97:250-5. Frisch RE. Pubertal adipose tissue: is it necessary for normal sexual maturation? Evidence from the rat and human female. Federation Proc. 1980;39:2395- 2400. Frisch RE. The right weight: body fat, menarche, and ovulatioh. Baillieres Clin Obstet Gynaecol 1990;4z419-39. 82 Gabbe SG. Pregnancy in women with diabetes mellitus. The beginning. Clin Perinatol 1993;20:507-15. Grodstein F, Goldman MB, Cramer DW. Infertility in women and moderate alcohol use. Am J Public Health 1994;84:1429-1432. Grodstein F, Goldman MB, Ryan L, Cramer DW. Relation of female infertility to history of sexually transmitted diseases. Am J Epidemil 1993a;137:577- 84. Grodstein F, Goldman MB, Ryan L, Cramer DW. Relation of female infertility to consumption of caffeinated beverages. Am J Epidemiol 1993b;137:1353- 1360. Gunderson EP, Abrams. Epidemiology of gestational weight gain and body weight changes after pregnancy. Epidemiol Rev 1999;21:261-275. Helm P, Munster K, Scmidt L. Recalled menarche in relation to infertility and adult weight and height. Acta Obstet Gynecol Scand 1995;74:718-22. Holmes KK. Pelvic Inflammatory Disease. In: Harrison's Principles of Internal Medicine, 14'h edition. Ed: Fauci AS, Martin JB, Braunwald E, Kasper DL, lsselbacher KJ, Hauser SL, Wilson JD, Longo DL. McGraw-Hill, New York. 1998; 812-817. Juul S, Karmaus W, Olsen J, et al. Regional differences in waiting time to pregnancy: pregnancy-based surveys from Denmark, France, Germany, Italy, and Sweden. Human Reproduction 1999;14:1250-1254. Kaprio J, Rimpela A, Winter T, Viken RJ, Rimpela M, Rose RJ. Common genetic influences on BMI and age at menarche. Hum Biol 1995;67:739-53. 83 Norman RJ, Clark AM. Obesity and reproductive disorders: a review. Reprod Fertil Dev 1998;10:55-63. Olsen J, Rachootin P, Schiodt AV, Damsno N. Tobacco use, alcohol consumption and infertility. Int J Epidemiol 1983;12:179-184. Otor SC, Pandey A. Puberty and the family formation process in Sudan: age-at- menarche differential fecundity hypothesis revisited. Soc Biol 1998;45:246-59. Pedersen KK, Hagen C, Sand-Pedersen SH, Eshj O. Infertility and pregnancy outcome in women with insulin-dependent diabetes. An epidemiological study. Ugeskr Laeger 1994;156:6196-6200. Rachootin P, Olsen J. Prevalence and socioeconomic correlates of subfecundity and spontaneous abortion in Denmark. Int J Epidemiol 1982;1 11245-249. Rich-Edwards JW, Goldman MB, Willett WC, Hunter DJ, Stampfer MJ, Colditz GA, Manson JE. Adolescent body mass index and infertility caused by ovulatory disorder. Am J Obstet Gynecol 1994;171:171-7. Rockhill B, Moorman PG, Newman B. Age at menarche, time to regular cycling, and breast cancer (North Carolina, United States). Cancer Causes Control 1 998;9:447-53. Rodriguez-Escudero FJ, Neyro JL, Corcostegui B, Benito JA. Does minimal endometriosis reduce fecundity? Fertil Steril 1988;50:522-524. Rosner B. Fundamentals of Biostatistics, 4'h edition. Wadsworth Publishing Company, Nelson, CA. 1995. 84 Sarlio-Lahteenkorva S, Lahelma E. The association of body mass index with social and economic disadvantage in women and men. Int J Epidemiol 1999;28:445-449. SAS, 1999. Research Triangle Park, NC. Scott EC, Johnston FE. Critical fat, menarche, and the maintenance of menstrual cycles. Journal of Adolescent Health Care 1982;2z249-260. Srinivasan SR, Bao W, Wittigney WA, Berenson GS. Adolescent overweight is associated with adult overweight and related multiple cardiovascular risk factors: the Bogalusa Heart Study. Metabolism 1996;45:235-40. Stephen EH. Projections of impaired fecundity among women in the United States: 1995-2020. Fertil Steril 1996;66:205-9. Stillman RJ, Rosenberg MJ, Sachs BP. Smoking and reproduction. Fertil Steril 1986;46:545-566. Suonio S, Kauhanen O, Metsapelto A, Terho J, Vohlonen 1. Smoking does affect fecundity. Eur J Obstet Gynecol Reprod Biol 1990;34:89-95. Templeton A, Fraser C, Thompson B. The epidemiology of infertility in Aberdeen. Br Med J 1990;301:148-52. Thonneau P, Spira A. Prevalence of infertility: international data and problems of measurement. European Journal of Obstetrics & Gynecology and Reproductive Biology 1990;38:43-52. Wada J, Ueda K. Correlation between changes in obesity from adolescence to young adulthood and family obesity—the results of cross sectional and 85 longitudinal studies (abstract). Nippon Koshu Eisei Zasshi 1990;37:837- 42. Westrom LV. Sexually transmitted diseases and infertility. Sexually Transmitted Diseases 1994;21 Supp|:S32-S37. Wood J. Maternal nutrition and reproduction: why demographers and physiologists disagree about a fundamental relationship. Ann N Y Acad Sci 1994; 709: 101-16. Yeshaya A, Orvieto R, Dicker D, Karp M, Ben-Rafael Z. Menstrual characteristics of women suffering from insulin-dependent diabetes mellitus. Int J Fertil Menopausal Stud 1995;40:269-73. 86 nrcuxan srnre UNIV. LIBRQRIES (Illillllllli”Hill“lllllillllllllllWilli 31293020488502