The female reproductive axis is thought to respond to increasing physical exercise with a continuum of dysfunction, ranging from luteal-phase defects to anovulation to infertility and, finally, amenorrhea. 1 This paradigm is founded primarily on observations of agrarian women in subsistence economies and of accomplished athletes within the United States. Although unequivocal evidence indicates that intensive physical training disrupts menstrual cycling, it is not clear that moderate-intensity activity affects female reproductive function. 2,3 Furthermore, the data on moderate activity have been obtained on women as they initiate new exercise regimens; there are almost no observational studies of established, low-intensity, low-frequency exercise habits more typical of American women. 4
Conversely, evidence suggests that being moderately overweight increases the risk of ovulatory infertility. 5,6 Given the high and rising prevalence of overweight among young U.S. women, 7 this could contribute more to infertility than does underweight.
In a previous retrospective analysis, we observed a J-shaped association between adolescent body mass index (BMI) and relative risk of ovulatory infertility in adulthood. 6 In the present prospective analysis, we examine adult body mass index and physical activity as predictors of ovulatory disorder infertility in a large cohort of American women enrolled in the Nurses’ Health Study II.
The Nurses’ Health Study II is a questionnaire-based prospective cohort of 116,671 female registered nurses who were age 25–42 years at the cohort’s inception in 1989. The study is designed to investigate associations of lifestyle and diet with breast cancer and other major diseases. The current analysis is a prospective investigation of incident ovulatory disorder infertility among the 20,417 women who reported ovulatory infertility or a pregnancy between 1989 and 1995.
Ascertainment of Cases and Controls
On biennial questionnaires from 1989 to 1995, participants were asked whether they had tried to become pregnant for more than 1 year without success and to indicate which of the following caused their infertility: ovulatory disorder, endometriosis, tubal blockage, cervical mucus factor, spouse factor, cause not found, cause not investigated, and/or other. In a validation substudy of 90 women who reported a history of ovulatory infertility on the 1989 questionnaire, ovulatory disorder infertility was confirmed by receipt of appropriate diagnostic tests or treatment in 93% of participants. 6
For each participant, every 2-year questionnaire interval could contribute a potential case or control event. First reports of ovulatory infertility recorded on the 1991–1995 questionnaires were considered case events. Control events were defined as pregnancies of any duration reported on the 1991–1995 questionnaires. If both ovulatory infertility and a pregnancy were reported in the same 2-year interval, we assumed that the 12-month infertility episode preceded the pregnancy, and we counted the event as a case. Women could contribute both case and control events to the analysis. Once a woman indicated infertility of any kind, she was excluded from contributing subsequent events because her infertility history might affect her exposure status. The 21,376 women who reported a history of infertility at baseline in 1989 were excluded from this analysis. Participants contributed to case or control events only during intervals in which they reported either ovulatory infertility or a pregnancy; thus, the 69,019 women who reported neither a pregnancy nor ovulatory infertility on any questionnaire from 1991 through 1995 never entered this analysis. Women were excluded after reporting a tubal ligation, or when they did not indicate that they were still premenopausal.
Weight and physical activity levels were considered unknown if women were pregnant when responding to these exposure questions. We excluded data from intervals in which body mass index or level of physical activity was unknown, or for which women reported implausible levels of body mass index (defined as less than 16 kg/m2 or greater than 40 kg/m2) or physical activity (defined as more than 21 hours of vigorous activity or more than 28 hours of vigorous or moderate activity per week). In a secondary analysis, we used as the control group 511 women who reported infertility due to spousal factor.
Ascertainment of Exposures
Body mass index (kg/m2) was calculated from self-reported height in 1989 and weight on the questionnaire preceding the infertility or pregnancy event. Previous studies have established validity of self-reported height and weight in this and similar cohorts. 8,9Physical activity was assessed in 1989 and 1991, when participants indicated the amount of time they had spent on average per week during the preceding year at each of the following activities: walking or hiking outdoors, jogging (slower than 10 minutes/mile), running (10 minutes/mile or faster), bicycling (including stationary machine), calisthenics/aerobics/aerobic dancing/rowing machine, tennis/squash/racquetball, lap swimming, and “other aerobic recreation (e.g., lawn mowing).” In 1989, participants also indicated their “usual walking pace outdoors” as: easy, casual (less than 2 mph); normal, average (2–2.9 mph); brisk pace (3–3.9 mph); very brisk/striding (4 mph or faster); or unable to walk. From this information, time spent in vigorous activity was derived by adding the total time spent at all of these activities except walking and “other aerobic recreation.” A score for moderate-intensity exercise was derived by adding the time spent walking at a brisk or very brisk/striding pace and time spent at other aerobic recreation. The category of vigorous activity included activities assigned a metabolic equivalent task value of at least 6 by Ainsworth et al, 10 and the moderate activity category included activity rated at metabolic equivalent task values between 4.0 and 4.5. As detailed information on physical activity and alcohol consumption was not collected on the 1993 questionnaire, infertility and pregnancy events occurring in the 1993–1995 cycle are assigned to 1991 activity and alcohol categories. Previous work in this cohort has established that the total physical activity measured by this questionnaire has a 2-year test-retest reliability of 0.59 and a validity of 0.79 compared with activity recalls and 0.62 compared with activity diaries. 11
We adjusted for age in 5-year categories. Parity was categorized as 0, 1, 2, and >2 births. Recency of oral contraceptive use was categorized as never-use, past use more than 2 years ago, past use within 2 years, current, and unknown. Alcohol intake was adjusted for in six categories, and cigarette smoking in five categories. All covariates reflect status reported on the questionnaire preceding the infertility or pregnancy event. Because physical activity and alcohol use were last reported on the 1991 questionnaire, exposure status in 1991 was used to represent exposure status in 1993.
The age-adjusted prevalence of potentially confounding and modifying variables was obtained by direct standardization to the overall age distribution in 5-year categories. We modeled the logit of ovulatory infertility using a nonparametric restricted cubic spline regression model 12 to avoid loss of information from categorizing the exposure variables and to investigate the nonlinear association between BMI and infertility, without imposing a priori assumptions about the shape of the curve. The spline function, a series of cubic polynomials, is smoothly joined at “knots” fixed at default percentiles by constraining the function and its first and second derivatives to be continuous at the knots. The BMI spline was centered at 21.0 kg/m2 as the reference point. A pointwise 95% confidence band was fit around the curves.
A logistic regression model was used to derive adjusted odds ratios of infertility by category of physical activity, as well as for continuous hours per week of physical activity. To distinguish moderate from vigorous intensity activity, we used a logistic regression model with indicator variables for each of nine activity categories formed by the cross-classification of moderate and vigorous activity.
The partial population attributable risk percentages 13 for body mass indices below 20.0 and at or above 25.0 were calculated from logistic regression coefficients representing the relative risk of BMI from 16.0 to <20.0 (“underweight”), and from ≥ 25.0 to 40.0 (“overweight”), respectively, compared with a referent group with BMI 20.0 to <25.0. Coefficients representing the additional relative risk of infertility associated with underweight and overweight were applied to the age, parity, and BMI distribution of women age 25–40 surveyed by the Third National Health and Nutrition Examination Survey between 1988 and 1994, 14 to estimate the partial population attributable risk percentage (PAR%) of ovulatory infertility attributable to under- and overweight in the United States. The 95% confidence intervals for the partial PAR% were calculated using the method developed by one of us (D. S.; see Appendix).
From 1989 to 1995, women reported 26,125 eligible pregnancies and 830 incident cases of ovulatory disorder infertility. Half (53%) of the cases occurred after a pregnancy. Infertility cases were similar in age to pregnancy controls (Table 1). As expected, cases were more likely to be nulligravid or nulliparous, to report irregular or abnormal length cycles in 1993, and to have used oral contraceptives within the last 2 years. There were no apparent differences in alcohol and cigarette consumption between ovulatory infertility cases and controls.
We observed a U-shaped association between BMI and relative risk of ovulatory disorder infertility. Compared with women with a BMI of 21.0, we observed increasing age-adjusted relative risks of ovulatory disorder infertility below a BMI of 20.0 and above a BMI of 24.0. These relative risk estimates were slightly strengthened by adjustment for parity. Further adjustment for recency of oral contraceptive use, physical activity, alcohol intake, or cigarette consumption made little difference to the relative risk estimates (spline curve from the “full” model including these variables is shown in Figure 1).
There was no association between vigorous activity and relative risk of ovulatory infertility in age-adjusted data (Table 2). However, after adjustment for parity, recency of oral contraceptive use, physical activity, alcohol intake, cigarette consumption, and moderate physical activity, a strong graded inverse association emerged between vigorous activity and risk of ovulatory infertility: each additional hour of vigorous exercise per week was associated with a 7% relative risk reduction (95% confidence interval = 4–10%) (Table 2).
Parity was a strong confounder of the inverse association between vigorous activity and infertility, because time spent in vigorous activity decreases with increasing parity, and the risk of ovulatory disorder infertility is low among parous as compared with nulliparous women. The 523 cases and 7,553 controls who were nulliparous when they attempted to conceive experienced an 8% (95% CI = 5–12%) relative risk reduction per hour of weekly vigorous activity (from the full model described in Table 2) (data not shown).
Vigorous activity had a similar association with primary and secondary ovulatory infertility. The estimated relative risk reductions per hour of weekly vigorous activity, adjusted for the full model, were 8% (95% CI = 4–12%) for primary infertility and 5% (95% CI = 0–9%) for secondary infertility.
After further adjustment for BMI, which is partially determined by physical activity, each hour of vigorous activity per week was associated with a 5% (95% CI = 2–8%) reduction in relative risk of ovulatory infertility (Table 2). The odds ratios (estimated relative risks) for infertility per hour of weekly vigorous activity were 0.96 (95% CI = 0.90–1.02) for women with BMI <20 kg/m2, 0.92 (95% CI = 0.88–0.97) for women with BMI 20 to <25 kg/m2, and 0.98 (95% CI = 0.92–1.04) for women with BMI ≥25 kg/m2. When we examined specific vigorous activities separately, the largest reductions in estimated relative risk were observed for running (34% reduction per hour per week) and jogging (22%), and smaller estimated relative risk reductions were observed for racquet sports (12%), lap swimming (5%), aerobics/calisthenics (5%), and biking (5%).
In contrast to vigorous activity, we observed no association between moderate-intensity activity and relative risk of ovulatory infertility (Table 3). The multivariate odds ratio for each additional hour of moderate-intensity exercise per week was 1.01 (95% CI = 0.99–1.03, after adjustment for factors included in the full model in Table 2).
When we compared the ovulatory infertility cases with 511 spousal infertility controls, we observed a similar U-shaped association of BMI with ovulatory infertility and a more modest 3% reduction in estimated relative risk of ovulatory infertility for every hour of weekly vigorous activity (95% confidence interval of 8% reduction in relative risk to 2% increase in relative risk, based on the full model, not including BMI).
In this cohort of women age 25–46, the estimated relative risk of ovulatory infertility began to increase among women who were only modestly overweight and continued to increase with higher BMI. At the other end of the scale, a low BMI was also associated with an increased relative risk of ovulatory infertility. Given the far greater prevalence of overweight than underweight among the U.S. population, we estimate that there is twice as much ovulatory infertility attributable to being overweight (BMI ≥25.0) as to being underweight (BMI <20.0), independent of age and parity. The partial PAR% for ovulatory infertility associated with being overweight is 25% (95% CI = 20–31%) and for being underweight is 12% (95% CI = 7–20%).
Previous studies have implicated extremes of body mass index in ovulatory infertility. 5,6,15 Obesity is believed to disrupt ovarian function by depressing sex hormone-binding globulin, 16,17 increasing insulin resistance, 17,18 and raising free androgens. 16–18 Insulin resistance and consequent hyperandrogenism has been implicated in polycystic ovarian syndrome, which may represent an extreme endpoint of a more prevalent asymptomatic endocrine imbalance underlying ovulatory infertility. 19,20 At the other end of the scale, weight loss and extreme leanness has been associated with reduced estradiol levels and suspension of ovarian activity. 1,21 Both weight loss among obese women and weight gain among lean women have been associated with increases in fecundity. 22,23
Vigorous activity was associated in the present study with a reduction in the relative risk of ovulatory infertility, in part because of its impact on adiposity. However, after adjustment for BMI, we still observed a 5% reduction in the relative risk of ovulatory infertility for every hour per week of vigorous activity. This suggests that physical activity may protect ovarian function through a mechanism independent of BMI, such as increasing insulin sensitivity. 24,25 Vigorous activity appeared to have a stronger protective association among women in the normal weight range (BMI 20–<25 kg/m2) than for women who were underweight or overweight, although relative risk estimates were in the same direction in each BMI stratum. In contrast to vigorous activity, we saw no evidence that moderate-intensity physical exertion (largely comprising brisk walking) had any effect on ovulatory infertility.
Our study has several limitations stemming from the fact that the Nurses’ Health Study II is not a cohort of women recruited as they began an attempt to become pregnant. We do not know that the control pregnancies were intentional, nor do we know that controls would have sought an infertility diagnosis had they gone 12 months without conceiving. The primary concern is that medically diagnosed cases may represent a relatively health-conscious group compared with pregnant controls. In particular, physically active women may be more likely to seek infertility diagnosis; however, this tendency would have biased the findings toward a positive association between activity and infertility, rather than the inverse association we observed.
Although our secondary analysis comparing ovulatory infertility cases with spousal infertility controls also yielded an inverse association between vigorous activity and relative risk of ovulatory infertility, it was considerably weaker than the association we observed when using pregnant controls. This may be due to random error because of the small number of spousal infertility controls or it may reflect an unidentified bias in one of these two control groups.
To our knowledge, the only previous observational study of physical activity and ovulatory infertility is a case-control study of 346 women with ovulatory infertility, compared with pregnant controls. 4 In that study, women who recalled less than 60 minutes of daily vigorous activity in the year before they attempted to conceive had a reduced risk of ovulatory infertility compared with those who never exercised. Women who recalled exercising 60 minutes or more had an increased risk of infertility. However, only 13 participants engaged in 60 minutes or more of daily exercise.
A larger body of literature has addressed the association between physical activity and ovarian function, as measured by menstrual patterns or hormone levels. In experimental studies, large and abrupt increases (to several hours per day) in vigorous activity disrupt menstrual cycling, particularly when accompanied by weight loss. 2 Although this level of activity is considerably higher than that observed in our cohort, others have documented luteal-phase suppression among habitual recreational runners with apparently normal menstrual cycles whose activity levels fall in the upper range observed in our cohort. 26 Another intervention trial of more gradually introduced vigorous activity (to several hours per week, comparable to levels observed in this cohort) demonstrated no changes in menstrual cycle length. 27 Finally, the ovarian function of agricultural laborers in the developing world is highly sensitive to seasonal fluctuations in workload and caloric intake. 28–30 These observations led Ellison et al31 to propose that the ovary is sensitive to nutritional status; to energy expenditure; and, in particular, to energy flux (the direction of weight change). Thus, we too hypothesized either no association or a direct association between vigorous activity and risk of ovulatory infertility rather than the inverse association we observed.
In contrast to the previous studies of ovarian function, our study did not record activity during the menstrual cycles in which women attempted to conceive. Rather, we measured a more consistent habitual pattern in the years preceding the test of fertility. It is possible that physically active women, who can easily reduce activity, had a greater opportunity to “throw the switch” into positive energy flux, increasing their chances of conceiving. 31 We cannot test this possibility with our data. Alternatively, our results may diverge from the ovarian-function studies because we observed a different end of the spectrum of human ecology. In contrast to laborers in marginal agrarian economies, the participants in our study, like most Americans, are at greater risk of caloric surfeit than shortage. In this context, a long-term habit of physical activity may prevent a syndrome that includes persistent weight gain, increasing insulin resistance, and impaired fertility. Our data suggest that, short of competitive athletic training, a habit of vigorous activity in the years before a woman attempts to conceive does not reduce, and may well enhance, fertility.
This study was several times larger than those that have preceded it, allowing more detailed examination of BMI and physical activity, including the distinction of moderate from vigorous activity. When placed in the context of previous studies that have established disrupted menstrual cycling among competitive athletes, our study suggests that there may be a U-shaped association of physical fitness with reproductive function, with optimal levels falling in the range achievable by many American women.
We are grateful for the expertise contributed by Stephanie Smith-Warner, Linlin Chen, and Jennifer Burlington, and the thoughtful input of Peter Ellison and Lisa Chasan-Taber. This study is founded upon the consistent, voluntary effort of the participants of the Nurses’ Health Study II.
1. Ellison PT. Human ovarian function and reproductive ecology: new hypotheses. Am Anthropol 1990; 92: 933–952.
2. Bullen BA, Skrinar GS, Beitins IZ, von Mering G, Turnbull BA, McArthur JW. Induction of menstrual disorders by strenuous exercise in untrained women. N Engl J Med 1985; 312: 1349–1353.
3. Bonen A. Recreational exercise does not impair menstrual cycles: a prospective study. Int J Sports Med 1992; 13: 110–120.
4. Green BB, Daling JR, Weiss NS, Liff JM, Koespsell T. Exercise as a risk factor for infertility
with ovulatory dysfunction. Am J Public Health 1986; 76: 1432–1436.
5. Green BB, Weiss NS, Daling JR. Risk of ovulatory infertility
in relation to body weight. Fertil Steril 1988; 50: 721–726.
6. Rich-Edwards JW, Goldman MB, Willet WC, et al
. Adolescent body mass index
caused by ovulatory disorder. Am J Obstet Gynecol 1994; 171: 171–177.
7. Flegal KM, Carroll RJ, Johnson CL. Overweight and obesity in the United States: prevalence and trends, 1960–1994. Int J Obes Relat Metab Disord 1998: 22: 39–47.
8. Troy LM, Hunter DJ, Manson JE, Colditz GA, Stampfer MJ, Willet WC. The validity or recalled weight among younger women. Int J Obes Relat Metab Disord 1995; 19: 570–572.
9. Rimm EV, Stampfer MJ, Colditz FA, Chute CG, Litin LB, Willet WC. Validity of self-reported waist and hip circumferences in men and women. Epidemiology 1990; 1: 466–473.
10. Ainsworth BE, Jacobs DS, Leon AS, et al
. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993; 25: 71–80.
11. Wolf AM, Hunter DJ, Colditz GA, et al
. Reproducibility and validity of a self-administered physical activity
questionnaire. Int J Epidemiol 1994; 23: 991–999.
12. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med 1989; 8: 551–561.
13. Bruzzi P, Green SB, Byar DP, Brinton LA, Schairer C. Estimating the population attributable risk for multiple risk factors using case-control data. Am J Epidemiol 1985; 122: 904–914.
14. U.S. Dept Health and Human Services National Center for Health Statistics, National Center for Health Statistics. Third National Health and Nutrition Examination Survey, 1989–1994, NHANES II Laboratory Data File(CD-ROM). Hyattsville, MD: Centers for Disease Control and Prevention, 1996.
15. Grodstein F, Goldman MB, Cramer DW. Body mass index
and ovulatory infertility
. Epidemiology 1994; 5: 247–250.
16. Harlass FE, Plymate SR, Fariss BL, Belts RP. Weight loss is associated with correction of gonadotropin and sex steroid abnormalities in the obese anovulatory female. Fertil Steril 1984; 42: 649–652.
17. Kiddy DS, Sharp PS, White DM, et al
. Differences in clinical and endocrine features between obese and non-obese subjects with polycystic ovary syndrome: an analysis of 263 consecutive cases. Clin Endocrinol 1990; 32: 213–220.
18. Pasquali R, Antenucci D, Casimirri F, et al
. Clinical and hormonal characteristics of obese amenorrheic hyperandrogenic women before and after weight loss. J Clin Endocrinol Metab 1989; 68: 173–179.
19. Yen SSC, Jaffe RB. Reproductive Endocrinology, 3rd ed. Philadelphia: WB Saunders, 1991.
20. Nestler JE. Insulin regulation of human ovarian androgens. Hum Reprod 1997; 12: 53–62.
21. Lipson SF, Ellison PT. Comparison of salivary steroid profiles in naturally occurring conception and non-conception cycles. Hum Reprod 1996; 11: 2090–2096.
22. Bates GW, Whitworth NS. Effect of body weight reduction on plasma androgens in obese, infertile women. Fertil Steril 1982; 4: 406–409.
23. Bates GW, Bates SR, Whitworth NS. Reproductive failure in women who practice weight control. Fertil Steril 1982; 37: 373–378.
24. Devlin JT. Effects of exercise on insulin sensitivity in humans. Diabetes Care 1992; 15: 1690–1693.
25. Henriksson J. Influence of exercise on insulin sensitivity. J Cardiovasc Risk 1995; 2: 303–309.
26. Ellison PT, Lager C. Exercise-induced menstrual disorders. N Engl J Med 1985; 313: 825–826.
27. Bonen A. Recreational exercise does not impair menstrual cycles: a prospective study. Int J Sports Med 1992; 13: 110–120.
28. Ellison PT, Peacock NR, Lager C. Ecology and ovarian function among Lese women of the Ituri Forest, Zaire. Am J Phys Anthropol 1989; 78: 519–526.
29. Panter-Brick C, Lotstein DS, Ellison PT. Seasonality of reproductive function and weight loss in rural Nepali women. Hum Reprod 1993; 8: 684–690.
30. Jasienska G, Ellison PT. Heavy workload impairs ovarian function in Polish peasant women. Am J Phys Anthropol 1993; 16: 117–118.
31. Ellison PT, Panter-Brick C, Lipson SF, O’Rourke MT. The ecological context of human ovarian function. Hum Reprod 1993; 8: 2248–2258.