Nurses in our study (n = 71,077) ranged in age from 28 to 45 years at the time they responded to the 1993 questionnaire (mean = 37.8, standard deviation = 4.3). Table 1 shows selected age-standardized characteristics of participants by rotating shift work categories. Women who worked rotating shifts tended to be younger and have a higher BMI than those working a permanent (non-rotating) shift. Of the nurses who reported working 20+ months of rotating night shift over the past two years, 68% worked as inpatient or emergency room nurses. Race/ethnicity differed by rotating shift work status, with more minority women reporting 20+ months of rotating shift work. Rotating shift workers were more likely to smoke and less likely to be physically active. Caffeinated coffee consumption was higher among rotating shift workers.

Table 2 provides information on the associations between months of rotating shift work and irregular pattern, adjusted for age (in 5-year categories), and then further adjusted for age at menarche, parity, race/ethnicity, smoking status, alcohol consumption, physical activity, and BMI as categorical variables. We observed associations between rotating shift work and irregular pattern, with the relative risks increasing modestly with increasing months of rotating shift work. The trend test shows a 13% increased risk per every 12 months of rotating shift work (95% confidence interval = 13%–13%). Younger and older age groups had increased risks of irregular patterns, and BMI showed a dose-response relationship with irregularity. Asian women, Hispanic women and women of “other” race/ethnicity were more likely than white participants to report irregular cycles. As expected from the frequent association of irregular cycles with subfertility, women with irregular cycles had lower parity.

Tables 3 shows age-adjusted models for the risk of short and long cycle length compared with 26–31 days. Rotating shift work was not associated with moderately short or long cycles, but was associated with extremely short or long cycles. As expected, as women aged, their cycles became shorter.

Table 4 shows the cycle-length analyses adjusted for lifestyle and demographic covariates. The adjusted associations between rotating shift work and cycle length are similar to the age-adjusted models shown in Table 3, although somewhat attenuated, primarily by BMI adjustment. A dose-response relationship of rotating shiftwork with cycles of 40 or more days was observed, with a 25% increased risk for every 12 months of rotating shift work (95% CI = 24%–26%). The positive association between shiftwork duration and extremely short cycles was still evident but less consistent than for extremely long cycles. Older women had shorter cycles, while later onset of menarche was associated with longer cycles. Current smoking was associated with shorter cycle length. There were strong associations of higher BMI with both short and long cycles. In addition, non-white women were more likely to have short cycles. Frequency of physical activity was not associated with cycle length. Alcohol consumption was strongly associated with fewer reports of short or long cycle lengths.

To remove any residual confounding of parity, we conducted a sub-analysis restricted to the 22% of the cohort that was nulliparous in 1993 (n = 15,558). In this subgroup, we observed that 20+ months of rotating shiftwork was not associated with menstrual irregularity (RR=1.06 [95% CI = 0.90–1.24]), but appeared to remain associated with both short (for <21 days, 1.58 [0.95–2.61]) and long (for 40+ days, 1.45 [0.90–2.34]) cycle length.

Because a recent pregnancy might affect both a woman’s report of her “current usual length or pattern” and the likelihood of working a rotating schedule, we conducted a sub-analysis that excluded the 13,426 women who reported having given birth during 1991–1993, or who reported being currently pregnant in 1993. The results were similar to our main analyses for all outcomes. The relative risks of working 20+ months of rotating shift work for irregular cycle, short cycle length (<21 days), and long cycle length (40+ days) were 1.21 (1.11–1.32), 1.26 (0.98–1.62), and 1.59 (1.24–2.04), respectively.

Although we did not collect data on permanent night shift work in the two years immediately preceding the menstrual cycle assessment of 1993, we were able to exclude the 7,639 women who reported on a subsequent questionnaire that they had worked at least 6 months of permanent night shifts between 1989–1994. The results were similar to our main analyses; the relative risks of working 20+ months of rotating shift work for irregular cycle, short cycle length (<21 days), and long cycle length (40+ days) were 1.20 (1.09–1.32), 1.28 (0.94–1.74), and 1.32 (0.98–1.79), respectively.

Our data a modest association between rotating shift work and menstrual regularity and cycle length. The associations were somewhat stronger with more months of rotating shift work in the previous two years, which could reflect a cumulative effect of shift work. However, because we did not know precisely when the rotating shift work occurred in the two-year time window, it is equally plausible that women who worked more months were more likely to be current shift workers at the time they responded to the questionnaire. Thus, the seeming dose-response associations could also reflect recency of exposure to shift work, suggesting a short-term, reversible effect. Overall, our results indicate a possible effect of shift work on adverse health outcomes related to irregular menstrual cycle pattern or length, such as subfertility.^6–10

Several studies have examined the relation between shift work and menstrual cycle patterns,^11–16 three of which were studies of nurses.^11–13 While most of these studies reported an association of shift work with menstrual cycle perturbations, four had small sample sizes (ranging from 12 to 151 participants), thereby limiting their ability to adjust for confounding factors.^12,13,15

One study examined 479 Canadian poultry workers who reported day-to-day variability in work schedule (defined as the day of work beginning at irregular or unpredictable times).¹⁴ The age range in this study was similar to ours, with 24% being 40 years or older. Though not a typical definition of “shift work,” schedule variability in that study was associated with long cycles (>33 days) and irregular cycles (>7 days variability), with odds ratios of 2.4 (95% CI = 1.0–5.5) and 2.0 (1.2–3.6), respectively, adjusted for temperature at work, smoking status, use of intrauterine device, parity, and age. The ORs in the study were higher than the RRs in our study, possibly due to differences in the definition of long cycles and shift work, or to possible overestimation of risk when using odds ratios to analyze outcomes are not that rare. A prospective study of Taiwanese 12-hour-shift factory workers who rotated between day and night shift every four months showed an increased risk of having either long (>35 days) or short (<25 days) cycle length in shift workers (n=280) compared with office workers (n = 49) (OR = 1.71 [95% CI = 1.0–2.9]).¹⁶ These results are somewhat difficult to interpret because long and short cycle lengths were combined as a single outcome. None of these studies, including ours, was able to distinguish always working at night from rotating night shifts.

The menstrual cycle is defined by cyclical patterns of circulating reproductive hormones. Shifts in the circadian rhythms, either through sleep disturbances or altered melatonin production, may affect regulation of the reproductive hormones that control the menstrual cycle, although the mechanism is not clear. Produced in the pineal gland during dark hours, melatonin is thought to regulate several physiologic mechanisms including sleep and core body temperature. ^19,20

Effects on animal reproductive behavior of variation in light hours from seasonal and lunar phases, presumably through altered melatonin production, has been well-studied.^2,20 In humans, however, a relation between hours or intensity of light exposure and menstrual cycle parameters, such as cycle length, remains unclear. ^20,21 Moreover, while urinary melatonin metabolite levels have been associated with shift work, parity, and amenorrhea,^22–28 an association between melatonin and reproductive hormones has not been established in humans.¹⁹ In one study, plasma melatonin levels remained consistent throughout the menstrual cycle, though melatonin levels were nearly twice as high in amenorrheic women as controls.²⁶ However, smaller studies have suggested an inverse association between exogenously-administered melatonin and estrogen levels.²⁹

Sleep appears to inhibit pituitary luteinizing hormone (LH) secretion^30,31; therefore, it is possible that altered sleep-wake patterns, as seen in shift workers who work rotating or night shifts, could modify LH secretion and thus change the regularity or length of the menstrual cycle. It is a limitation of our study that we did not collect information on sleep behaviors. Therefore, we cannot distinguish between the effects of lack of sleep and light at night, nor can we comment on whether efforts by participants to recover sleep altered the effects of rotating shift work.

In our study, there was evidence of a dose-response relationship between self-reported BMI and menstrual cycle outcomes. In previous studies, extremes of weight have been associated with the risk of long, irregular,^8,32–34 and anovulatory cycles^6,34 typical of polycystic ovarian syndrome. Moderate overweight has been associated with ovulatory disorder infertility in the NHS-II cohort^17,35 and others.⁷ Our study also suggests a dose-response association between alcohol intake and reduced risk of menstrual cycle aberrations. Previous studies have reported reduced cycle length variability with moderate or heavy alcohol intake,^36,37 and a reduced risk of long menstrual cycles among moderate drinkers. ^36,38

The results of our study were fairly robust when we excluded recent pregnancies; however, when we restricted our cohort from 71,077 to the 15,558 nulliparous women, the association of rotating shift work with cycle irregularity was attenuated, though associations with cycle length were not. The attenuated association of shiftwork with menstrual irregularity among the nulliparous could be a chance consequence of smaller numbers. Alternatively, it might indicate the elimination of confounding by factors associated with parity, or it may reflect a higher background risk of cycle irregularity in a group of nulliparous women with infertility.

The assessment of shift work in our study was defined as at least three nights per month in addition to other days and evenings in that month, and was limited to the report of the number of months the nurse worked a rotating shift in the previous two years. A better assessment of the number of nights worked per month would have refined our results by allowing us to investigate a threshold or dose-response effect.

Although permanent night work was not assessed in our study, non-rotating night work can result in a rotation of sleep—wake cycles if the worker reverts to normal hours on their days off. We did not collect sufficiently detailed information on permanent night work to distinguish its effects from the effects of rotating shift work. Unpublished data from a subset of approximately 60 nurses from this cohort suggest that permanent night workers work twice as many nights per month as rotating night workers (12.3 versus 6.4). Since the adverse effects of night work might be more severe among rotating night workers than among permanent night workers, we presume that classifying night workers as having no rotating shift work would have biased our results towards the null. Another possible source of misclassification could result from permanent night workers incorrectly classifying themselves as rotating shift workers. From previous Nurses’ Health Study analyses, we have found that about 10% of all reported rotating shift workers were actually permanent night workers (unpublished data). When we excluded nurses who had classified themselves as permanent night workers on subsequent questionnaires, the results were unchanged.

Women reported usual cycle characteristics without the aid of menstrual diaries. A more definitive study of this topic would use prospectively collected work records and menstrual diaries to examine whether change in work schedule induces changes in menstrual cycling. Validation studies of self-reported shift work data, especially among health care workers, are needed. Though it is possible women misclassified their cycle length, the fact that we found the expected differences in cycle length with age, age at menarche, smoking, and BMI suggests that women’s self-reported data were adequate. Moreover, because our data are over only a 2-year period, we expect misclassification from self-reported data to have been minimized.

With a sample size of over 70,000 women, our study is by far the largest study of its kind to examine the association between shift work and menstrual patterns. A consequence is that our study had the ability to look at more extreme definitions of short and long cycles than previous studies, and to adjust for BMI and stratify by age.

In conclusion, we found an increased risk for short and long menstrual cycle length and cycle irregularity in nurses who worked rotating shifts, which may have implications for infertility. Future research could clarify the interrelations between sleep-wake patterns, sleep behavior, melatonin production, exposure to light at night, and the reproductive hormones that regulate the menstrual cycle and ovulation. Because night work will continue to be necessary in the nursing profession, intervention studies may be needed to see whether shift work affects fertility.