Predictors of the Age at which Natural Menopause Occurs in Women with Type 1 Diabetes: The Pittsburgh Epidemiology of Diabetes Complications (EDC) Study

Abstract

Objective:

Women with type 1 diabetes (T1D) are thought to experience menopause earlier than women without diabetes, although not all studies agree. We assessed metabolic predictors of the age at which natural menopause occurs among women with T1D participating in the Epidemiology of Diabetes Complications study.

Methods:

Women with childhood-onset (<17 years) T1D who underwent natural menopause without use of hormone therapy during their menopausal transition were included in the analysis (n=105; mean baseline age, 29.5 and diabetes duration, 20.2 years). Self-reported reproductive history and the Women’s Ischemia Syndrome Evaluation hormonal algorithms were used to determine menopause status. Linear regression was used to ascertain whether time-weighted metabolic factors (e.g. BMI, lipids, HbA1c, insulin dose, albumin excretion rate (AER)) were associated with age at natural menopause.

Results:

Univariately, only insulin dose (β=−4.87, p=0.04) and log(AER) (β=−0.62, p=0.02) were associated (negatively) with age at natural menopause. Adjusting for BMI, smoking status, lipids, HbA1c, number of pregnancies and oral contraceptive use, each 0.1 unit increase in the daily dose of insulin per kilogram body weight was associated with 0.64 years younger age at natural menopause (p=0.01), while for every 30% increase in AER, age at natural menopause decreased by 0.18 years (p=0.03).

Conclusion:

Higher average levels of insulin dose and AER over time were significantly associated with a younger age at which natural menopause occurred among women with T1D. The biologic mechanisms underlying the observed associations between exogenous insulin dose and AER on reproductive health should be investigated among women with T1D.

Introduction

Natural menopause is the cessation of ovarian function and the end of women’s reproductive life, resulting from natural oocyte depletion. Age at natural menopause and its related factors have attracted great research interest due to the substantial impact menopause has on women’s health. Earlier age at menopause is associated with increased risks of cardiovascular disease ¹, osteoporosis ², and fracture ³ later in life. Importantly, earlier age at menopause is related to increased all-cause mortality ⁴ and also mortality from cardiovascular disease ^{1, 5}, atherosclerosis ⁶, and stroke ⁷, with a 2% increase in age-adjusted mortality per year decline in age at menopause ⁸.

Due to the important role of insulin in maintaining normal functioning of the female reproductive system, it is logical to postulate that women with type 1 diabetes who have a disruption in insulin regulation might experience earlier natural menopause compared to women without diabetes. Indeed, several studies have provided evidence to support this hypothesis, although not all studies agree ^{9, 10}. The Familial Autoimmune and Diabetes (FAD) study was the first to report that women with type 1 diabetes reached menopause at a younger age compared with their nondiabetic sisters or unrelated control participants ¹¹. The European Prospective Investigation into Cancer and Nutrition (EPIC) study also suggested that early-onset diabetes (onset before the age of 20 years) was associated with an earlier onset of menopause, compared with nondiabetic controls ¹². In our recent study, we noted that natural menopause occurred 2.0 years earlier in women with childhood-onset type 1 diabetes compared with non-diabetic women after adjustment for age at baseline, age at menarche, race, BMI, smoking status, blood pressure, HDL, and non-HDL cholesterol, number of pregnancies, and having ever taken oral contraceptives ¹³.

Nevertheless, data on determinants of an earlier age at menopause in women with type 1 diabetes are scarce. In the general population, it is well known that smoking accelerates natural menopause onset ^{14, 15}, whereas high parity is associated with later natural menopause onset ¹⁶. However, little is known about the effects of traditional risk factors, such as smoking and BMI, or diabetes-specific factors, such as glycemic control, insulin dose, and diabetes complications, on the age at which natural menopause occurs among women with type 1 diabetes. Previous work from the Diabetes Control and Complications Trial (DCCT) and its observational follow-up, the Epidemiology of Diabetes Interventions and Complications (EDIC) study examined the impact of intensive treatment, HbA1c, and microvascular complications on menopause onset among women with type 1 diabetes ¹⁷. However, in this study, the age at menopause did not differ by treatment group.

Given evidence of an earlier menopause onset in type 1 diabetes, and the enormous health impact induced by early menopause, identifying modifiable factors which contribute to early menopause in type 1 diabetes would have great public health significance. Therefore, our objective was to identify metabolic factors (HbA1c, lipids, blood pressure, insulin dose, etc.) that are independently associated with age at natural menopause in a cohort of women with type 1 diabetes.

Methods

Study population

The Pittsburgh Epidemiology of Diabetes Complications (EDC) Study recruited childhood-onset (<17 years) type 1 diabetes patients diagnosed, or seen within one year of diagnosis, at Children’s Hospital of Pittsburgh between 1950 and 1980. All participants (n=658) attended a first clinical assessment in 1986-1988. The mean participant age at study entry was 28 years (ranging from 8 to 48 years) and their diabetes duration was 19 years (ranging from 8 to 37 years). These 658 participants (325 female and 333 male) were then prospectively followed via surveys and/or clinical examinations for up to 30 years. The EDC Study has been described in detail elsewhere ¹⁸. All study participants provided written informed consent prior to performing any study procedures. The EDC study protocol was approved by the University of Pittsburgh Institutional Review Board.

For this study, exclusion criteria comprised unavailability of data on plasma follicle stimulating hormone and estradiol due to missingness (n=53) or death (n=75), having had a hysterectomy or oophorectomy prior to menopause (n=35), or having used sex hormones during the menopausal transition (n=20). Additionally, 37 women who were premenopausal at their last available follow-up (baseline age 18.3±4.0 years; age at last follow-up 47.5±3.7 years) were also excluded. The data analyses therefore included 105 female EDC participants who had gone through natural menopause. Comparisons of women who were or were not included in analyses are presented in Supplemental Table 1.

Covariate Assessment

Risk factors were assessed at baseline and repeated at 2-, 4-, 6-, 8-, 10-, and 18-years of follow-up. Demographic information, as well as data on medical history and diabetes self-care were obtained by self-report. Anthropometric characteristics were assessed during the clinical visits: body mass index (BMI, in kilograms (kg) per meters squared (m2)) and waist to hip ratio (WHR). Blood pressure was assessed according to the HDFP protocol ¹⁹. Hypertension was defined as blood pressure ≥140/90 mm Hg or use of medications for high blood pressure. Stable glycosylated hemoglobin (HbA1) was initially (baseline-18 months) measured using Ion exchange chromatography (Isolab, Akron, OH), and via automated high-performance liquid chromatography (Diamat, BioRad, Hercules, CA) subsequently for 10 years. There was high correlation (r=0.95) between the two assays. These HbA1 values were standardized to DCCT HbA1c values as previously described ²⁰. At the 18-year follow-up, HbA1c was measured using the DCA 2000 analyzer (Bayer Healthcare LLC. Elkhart, IN) and converted to DCCT standard HbA1c values using an equation (DCCT HbA1c = [EDC HbA1c-1.13]/0.81) ²⁰. An estimate of glucose disposal rate (eGDR) was calculated using a regression equation with terms for WHR, hypertension and HbA1c ²¹.

From baseline through the 10 year follow-up, high-density lipoprotein cholesterol (HDL-C) was assessed via a precipitation technique (heparin and manganese chloride), modified from the Lipid Research Clinics method ^{18, 22} whereas an enzymatic method was used to measure total cholesterol and triglycerides ^{18, 22}. At the 18-year follow-up, serum lipids were measured using the Cholestech LDX (Cholestech Corp., Hayward, CA). Non-HDL cholesterol (non-HDL-C) was computed by subtracting HDL-C from total cholesterol. White blood cell count was measured using a counter S-plus IV. Serum and urinary albumin were measured by immunonephelometry ²³, and creatinine was assayed by an Ectachem 400 Analyzer (Eastman Kodak Co., Rochester, NY). Glomerular filtration rate was estimated (eGFR) by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine formula ²⁴

Menopause status assessment

Reproductive history information was self-reported. Natural menopause is defined as cessation of menstruation, which is not induced by surgical procedures (e.g., hysterectomy, oophorectomy) or medications (e.g., hormone medications). Having had a hysterectomy or an oophorectomy before menopause or having used sex hormone therapy during the menopausal transition therefore comprised exclusion criteria. Women younger than 45 years with regular menstrual cycles were classified as pre-menopausal; women older than 55 years who had not experienced menstrual periods for at least 12 months were classified as post-menopausal. In women falling outside the above classification, we measured plasma follicle stimulating hormone (FSH) and estradiol and used the Women’s Ischemia Syndrome Evaluation (WISE) hormonal and historical algorithms ²⁵ to assess menopausal status, without distinguishing peri- from pre-menopausal women (Figure 1).

Figure 1.

Classification of menopausal status within the EDC study

Statistical analyses

Time-weighted mean values of continuous metabolic factors were constructed based on clinical assessments at baseline and at 2-, 4-, 6-, 8-, 10-, and 18-years of follow-up. They were used for all continuous metabolic factors of interest to represent their average levels over time up to the exam cycle prior to the onset of menopause. The time-weighted mean values were calculated as a sum of the products of average values from two consecutive follow-up visits multiplied by the time interval (years) between the two visits and then divided by the total follow-up time until the examination cycle prior to reaching natural menopause. E.g., time-weighted BMI = (((BMI1 + BMI2)/2)*(time2 - time1)+((BMI2 + BMI3)/2)*(time3 - time2)+((BMI3 + BMI4)/2)*(time4 - time3)+⋯)/((time2 - time1)+(time3 - time2)+(time4 - time3)+⋯) . Risk factors evaluated included BMI, blood pressure, total cholesterol, HDL-C, non-HDL-C, LDL-C, triglycerides, HbA1c, eGDR, insulin dose, albumin excretion rate (AER), eGFR, and white blood cell count (WBC). The distribution of AER was skewed, thus natural log transformation was used for this variable. Univariate and multivariable linear regression models (PROC REG) were used to assess whether time-weighted factors were associated with age at natural menopause. In the multivariable models, covariates were selected if p<0.2 in univariate models or if they were previously associated with age at menopause (e.g. smoking status, BMI, lipids, contraceptive use, and pregnancy etc.) ¹⁶. To assess the importance of insulin resistance, models were also constructed with eGDR as a covariate. However, since eGDR is derived from WHR, hypertension, and HbA1c, these three variables were excluded in models with eGDR. SAS version 9.4 (SAS Institute, Cary, NC) was used for all analyses.

Results

Among the 105 EDC women in the present study, mean baseline age was 29.5 years and duration of type 1 diabetes, 20.2 years. All these 105 women had been pre-menopausal at the baseline assessment. The vast majority (96.2%) were non-Hispanic white. A summary of the study population’s characteristics at the baseline assessment is presented in Table 1. By the last available follow-up (after an average of 29.2±7.1 years from baseline), 63.5% of the study participants reported having used oral contraceptives and 72.4 % had been pregnant. Among those who reported at least one pregnancy (n=76), the mean number of pregnancies was 2.4 and the mean number of live births was 1.4. The average age at natural menopause in the study population was 49.5 years. A comparison of women who were or were not included in analyses were presented in Supplemental Table 1.

Table 1.

Characteristics of EDC female participants at the baseline assessment, 1986-88 (n=105)

Characteristics
Baseline age, in years	29.5 ± 6.2
Age at diabetes onset (years)	9.3 ± 3.9
Type 1 diabetes duration, baseline, in years	20.2 ± 7.0
Race
Non-Hispanic White	96.2 (101)
Black	3.8 (4)
Marital status
Currently single and never married	37.1 (39)
Married or living with a partner as if married	48.6 (51)
Divorced	7.6 (8)
Other (e.g., separated, widowed)	6.7 (7)
Education (n=103)
Less than high school or high school graduate	35.0 (36)
At least some college or college graduate	56.3 (58)
Post bachelor’s degree education	8.7 (9)
Smoking status
Never smoked	67.6 (71)
Used to smoke	13.3 (14)
Currently smokes	19.1 (20)
BMI (kg/m2)	22.9 ± 2.9
Waist to hip ratio (n=104)	0.77 ± 0.05
Systolic blood pressure (mmHg)	107.9 ± 11.7
Diastolic blood pressure (mmHg)	68.5 ± 9.6
Hypertension (n=105)	8.6 (9)
Blood pressure medication use (n=103)	6.8 (7)
Total cholesterol (mg/dl)	181.0 (162.0, 197.0)
HDL-C (mg/dl)	61.0 (50.6, 73.0)
LDL-C (mg/dl) (n=100)	100.9 (87.0, 127.4)
Triglycerides (mg/dl) (n=102)	74.0 (51.0, 100.0)
Use of lipid medications (n=104)	0 (0)
HbA1c (%, n=104)	8.4 ± 1.2
HbA1c (mmol/mol, n=104)	68 ± 13.1
eGDR (mg*kg-1*min-1, n=103)	8.9±1.5
Insulin dose (units/day/kg, n=100)	0.71 ± 0.19
AER (μg/min)	10.4 (6.2, 25.7)
eGFR (mL/min/1.73 m2)	102.3 ± 27.9
WBC x 103/mm2	6.4 ± 1.9
At the last available follow-up
Percent ever contraceptive use (n=104)	63.5 (66)
Percent ever pregnant	72.4 (76)
No. of pregnancies (n=76)	2.4 ± 1.3
No. of live births (n=76)	1.4 ± 0.8
Age at last available follow-up (years)	58.7 ± 6.1
Age at natural menopause (years)	49.5 (4.1)

Data shown are means (SD) for normally distributed continuous variables, median (IQR) for non-normally distributed continuous variables and percent (n) for categorical variables.

AER, albumin excretion rate; BMI, body mass index; eGDR, estimated glucose disposal rate; eGFR, estimated glomerular filtration rate; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; and WBC, white blood cell count.

In univariate linear regression models (Table 2), insulin dose (β±Standard Error (SE)= −4.87 ± 2.3, p=0.04) and ln(AER) (−0.62 ± 0.27, p=0.02) were significantly inversely associated with age at natural menopause. Each 0.1 unit increase in daily insulin dose per kilogram body weight was associated with 0.487 years younger age at natural menopause, whereas for every 30% increase in AER, age at natural menopause decreased by 0.16 years (back transformation: −0.62 * log (1.30) = −0.16). No other statistically significant associations were observed. Insulin dose, ln(AER), and HbA1c were thus included in multivariable linear regression models as their p values were <0.2 and smoking status, BMI, lipids, contraceptive use, and pregnancy were also included as they were associated with age at natural menopause based on previous studies.

Table 2.

Univariate associations between time-weighted values of metabolic factors of interest and age at natural menopause (n=105)

Time-weighted values	Change in age at natural menopause	95% C.I.	P values
BMI	0.001	−0.24 to 0.25	0.99
Systolic blood pressure	−0.01	−0.09 to 0.07	0.85
Diastolic blood pressure	−0.04	−0.15 to 0.06	0.43
Total cholesterol	−0.01	−0.04 to 0.01	0.36
HDL cholesterol	−0.01	−0.08 to 0.06	0.76
Non-HDL cholesterol	−0.01	−0.04 to 0.02	0.45
LDL cholesterol	−0.01	−0.04 to 0.02	0.58
Triglycerides	−0.002	−0.02 to 0.02	0.83
HbA1c	−0.61	−1.41 to 0.18	0.13
eGDR	−0.22	−0.78 to 0.33	0.42
Insulin dose	−4.87	−9.46 to −0.28	0.04
AER (log)	−0.62	−1.15 to −0.10	0.02
eGFR	−0.02	−0.05 to 0.02	0.35
WBC	0.06	−0.38 to 0.51	0.78

AER, albumin excretion rate; BMI, body mass index; eGDR, estimated glucose disposal rate; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; LDL, low-density lipoprotein; WBC, white blood cell count.

The time-weighted mean values were calculated as a sum of the products of average values from two consecutive follow-up visits multiplied by the time interval (in years) between the two visits and then divided by the total follow-up time until the examination cycle prior to reaching natural menopause. E.g., time-weighted BMI = (((BMI1 + BMI2)/2)*(time2 - time1)+((BMI2 + BMI3)/2)*(time3 - time2)+((BMI3 + BMI4)/2)*(time4 - time3)+⋯)/((time2 - time1)+(time3 - time2)+(time4 - time3)+⋯) .

In multivariable linear regression (Table 3, model 1), adjusting for BMI, smoking status, HDL-C, non-HDL-C, triglycerides, HbA1c, number of pregnancies, having ever taken oral contraceptives and AER, each 0.1 unit increase in daily insulin dose per kilogram body weight was associated with 0.64 years younger age at natural menopause (p=0.01). Moreover, after adjustment for the above-mentioned time-weighted covariates, including daily insulin dose, one natural log microgram per minute increase in AER was associated with 0.67 years earlier age at natural menopause (p=0.03). For every 30% increase in AER, age at natural menopause decreases by 0.18 years (back transformation: −0.67 * log (1.30) = −0.18). In the multivariable model with eGDR (Table 3, model 2), while insulin sensitivity itself (eGDR) was not associated with age at natural menopause (p=0.21), insulin dose (p=0.01) and ln(AER) (p=0.02) both maintained their significant inverse association.

Table 3.

Multivariable linear regression analysis for the prediction of age at natural menopause (n=105)

Variables	Model 1		Model 2
	Change in age at natural menopause (95% C.I.)	p-value	Change in age at natural menopause (95% C.I.)	p-value
Ever smoked	0.50 (−1.85 to 2.85)	0.68	0.41 (−1.96 to 2.78)	0.74
Time-weighted BMI	−0.12 (−0.37 to 0.13)	0.39	−0.07 (−0.32 to 0.18)	0.57
Time-weighted HDL-C	−0.06 (−0.14 to 0.02)	0.11	−0.05 (−0.13 to 0.03)	0.25
Time-weighted non-HDL-C	−0.01 (−0.05 to 0.03)	0.52	−0.01 (−0.05 to 0.03)	0.47
Time-weighted triglycerides	0.02 (−0.02 to 0.06)	0.21	0.01 (−0.03 to 0.05)	0.53
Time-weighted HbA1c	−0.76 (−1.62 to 0.10)	0.09	NA	NA
Time-weighted eGDR	NA	NA	−0.43 (−1.12 to 0.26)	0.21
Time-weighted insulin dose	−6.35 (−11.23 to −1.47)	0.01	−6.3 (−11.22 to −1.38)	0.01
Time-weighted AER (log)	−0.67 (−1.28 to −0.06)	0.03	−0.72 (−1.33 to −0.11)	0.02
Ever contraceptive use	0.28 (−1.39 to 1.95)	0.74	0.07 (−1.62 to 1.76)	0.93
No. of pregnancies	−0.41 (0.28)	0.14	−0.39 (0.28)	0.17

AER, albumin excretion rate; BMI, body mass index; eGDR, estimated glucose disposal rate; HbA1c, hemoglobin A1c; HDL-C, high-density lipoprotein cholesterol

Note:

In Model 1, predictors were selected if their p values were less than 0.2 in Table 2 (e.g. HbA1c, insulin dose, AER) or based on previous literature (e.g. smoking status, BMI, HDL, non-HDL, triglycerides, contraceptive use, and number of pregnancies).

In Model 2, the importance of insulin resistance was assessed; thus, this model was constructed with eGDR as a covariate. However, since eGDR is derived from WHR, hypertension, and HbA1c (eGDR (mg/kg/min)=24.395-(12.971*WHR)-(3.388*Hypertension)-(0.601*HbA1c)), HbA1c was excluded from Model 2.

Discussion

In this study, we evaluated predictors of the age at which natural menopause occurs among women with long duration type 1 diabetes. We observed that higher average levels of insulin dose and AER over time were significantly associated with an earlier age at natural menopause after multivariable adjustments, including for HbA1c. Interestingly, despite insulin dose being a predictor after accounting for HbA1c levels, insulin sensitivity (eGDR) did not predict age at natural menopause in a separate model. These findings suggest, that the adverse effect of insulin dose does not reflect insulin resistance but rather may suggest that the higher exogenous insulin concentrations have a direct, deleterious effect on ovarian aging in women with type 1 diabetes.

Type 1 diabetes is characterized by endogenous insulin deficiency and thus exogenous insulin administration is needed for survival. However, unlike insulin produced by the pancreas, injected exogenous insulin does not go through hepatic first-pass metabolism and clearance; rather, it goes into systemic circulation directly, and thus peripheral tissues, including the ovary, are exposed to excessive insulin levels in type 1 diabetes and its potential sequalae²⁶. Evidence provided from experimental studies suggest that insulin stimulates the transition from primordial to primary follicles ^27–29. It is thus possible that increased exposure of the ovary tissue to excessive insulin in women with type 1 diabetes leads to excessive, early follicle maturation, resulting in a premature depletion of the primordial follicle pool and an earlier age at menopause.

Another possible explanation would be that women who required higher insulin doses early in the course of the disease may have had worse beta cell function and were in fact exposed to inadequate insulin doses. However, given the long duration of type 1 diabetes in this analysis, essentially all patients have complete loss of beta cell function. It is also possible that higher doses of insulin led to an increased occurrence of hypoglycemic events, and therefore greater exposure to counterregulatory hormones, which, in turn, could have affected ovarian function. Nonetheless, in contrast to our results, findings from the DCCT/EDIC study suggested that greater insulin dose was associated with lower menopause risk in women with type 1 diabetes, although age at menopause per se was not evaluated as an outcome in this study ¹⁷.It is possible, however, as study investigators note, that the DCCT/EDIC was a chance finding given the relatively high number of comparisons in the study ¹⁷.

It has been suggested that increased oxidative stress is associated with significant adverse effects on women’s reproductive function, including ovarian vascular endothelium damage and abnormalities in follicular growth, oocyte maturation, corpus luteum formation, and embryonic growth ³⁰. While there was no direct evidence in type 1 diabetes, HbA1c (r = −0.51, P < 0.001), and fasting blood glucose (r = −0.69, P < 0.001) were found to be negatively correlated with antral follicle count (AFC) in women with type 2 diabetes ³¹. We thus hypothesized that poor glycemic control would be associated with premature ovarian aging in type 1 diabetes, given the potential ovarian vascular damage caused by advanced glycation end-product (AGE) induced oxidative stress ^{32, 33}. However, the average level of HbA1c over time was not associated with age at natural menopause after multivariable adjustments. Although in the present study the non-significant finding regarding HbA1c may result from the small sample size (n=105) or the small variability in time-weighted HbA1c (8.6±1.0), our findings are consistent with results from the DCCT/EDIC study ¹⁷ and a Finnish study ¹⁰ which also reported no HbA1c association with age at menopause .

Within the EDC study, AER was also an independent predictor of menopause onset in women with type 1 diabetes, even after adjustment for diabetes-specific confounders (e.g. HbA1c, insulin sensitivity and insulin dose). A general population study previously suggested that women with chronic kidney disease (CKD) tend to experience menopause earlier than women from the general population (47.2 vs. 47.8 years) ³⁴. Although a direct impact of CKD on the hypothalamic-pituitary-ovary axis is speculative, CKD could lead to cellular senescence and premature aging through the effects of uremic toxins, oxidative stress and persistent inflammation ³⁵, and premature ovarian aging could be one of the aging phenotypes. Our findings therefore provide further insight on the potential effects of kidney disease on ovarian reserve among women with type 1 diabetes. Further evidence of a role of kidney disease on age at menopause comes for the above-mentioned Finnish study, in which end-stage renal disease (ESRD) was associated with early age at menopause, despite only nine ESRD cases in this analysis ¹⁰. On the contrary, no association between nephropathy and menopause was observed in the DCCT/EDIC study ¹⁷. Should further studies confirm that diabetic kidney complications play a role in ovarian function, women with these two risks of early menopause, type 1 diabetes and kidney complications, could be targeted for timely interventions to prevent the health consequences of premature ovarian aging.

Limitations of the present study were the relatively small sample size (n=105) and lacking racial diversity. Another limitation relates to the exclusion of women who died or dropped out before menopause occurred, potentially leading to selection or survival bias. One strength of the present study was that the analysis was based on longitudinal repeated measurements, which could better reflect the cumulative effect of metabolic factors on menopause compared with using baseline assessments only.

Conclusion

Higher insulin dose (after accounting for HbA1c levels or insulin sensitivity) and increased AER independently predicted age at natural menopause among women with type 1 diabetes. While these results require validation in large cohorts of women with type 1 diabetes, our findings raise significant questions relating to a potentially deleterious effect of high exogenous insulin doses, in addition to that of kidney disease, and may (upon replication) constitute a useful source in clinical reproductive counseling for women with type 1 diabetes.

Supplementary Material

SDC

Supplemental Table 1. Descriptive characteristics of female study participants by eligibility for analyses.