We evaluated the relation of smoking with SHBG and the hormones FSH, LH, estradiol, and progesterone during the menstrual cycle among women in the BioCycle Study. In models adjusted for potential confounding factors, we observed higher follicular phase levels of LH and FSH in smokers compared to nonsmokers. Non-statistically significant higher luteal phase estradiol levels and lower peak progesterone levels were seen among smokers compared with nonsmokers. No differences in SHBG were observed.
Our findings support those of Windham et al
. who found elevated FSH levels in smokers of at least 10 cigarettes/day compared to nonsmokers around the time of transition from the luteal to follicular phase, using daily urinary hormone measurements.4
Our results differ from those of Zumoff et al
who followed eight nonsmokers and eight smokers (> 1 pack/day for at least 3 years) over a single cycle and found significantly higher serum levels of estradiol and progesterone and lower LH in the follicular phase for smokers compared to nonsmokers. Differences could be due to timing of sample collection, or the fact that confounding was not addressed in their analyses which used t-tests to assess day-specific and an overall follicular phase smoking effect.18
We utilized mixed models to assess day-specific effects and to address repeated measures from study participants, with IPT weights to handle confounding. While cigarette smoking has been shown to affect levels of some pituitary hormones, previous studies have not observed altered levels of LH and FSH, but these were among male smokers.11
The differences observed in LH and FSH between smokers and nonsmokers have possible implications for fertility. Elevated FSH in current smokers compared to nonsmokers among women 38 – 49 years of age may be associated with a shortened transition to menopause.21
In the context of assisted reproductive technologies (ART), baseline FSH has been suggested to reflect fertility and ovarian reserve.37,38
Among women with cancelled cycles, baseline FSH was significantly higher than among those completing treatment (day 3 FSH = 10.67 IU/I vs. 8.2 IU/I, p
The ratio of basal FSH to LH has been evaluated as a predictor of outcomes in ART, where elevated FSH/LH has been associated with lower uterine responsiveness to fertility treatments.39,40
It is unclear what effects the altered levels of FSH and LH we observed may indicate for a younger, non-ART cohort of women.
In BioCycle, serum was collected longitudinally using fasting, early-morning draws and short-term specimen storage. Fertility monitors aided timing of visits and were used to standardize cycle data. This resulted in missing data that may have impacted power; however, it allowed for comparison of biospecimens collected at the same phase of the menstrual cycle to correct for variation in cycle length. This approach minimizes measurement error, to which self-reported cycle length is prone.27,41
Some limitations for the current study are noted. We relied upon self-report for smoking data; however, studies have observed high validity between smoking determination by serum cotinine and self-report.42,43,44
A further limitation regards the amount of smoking among the exposed and misclassification of smoking status. Our definition of smoking (≥0.5 cigarettes/day) included light smokers in the smoking group (and some very light smokers in the nonsmoking group). The propensity for young women to smoke on an irregular, infrequent basis was considered when developing the exposure classification.45
We addressed possible exposure misclassification through sensitivity analyses, which produced similar results to our main analysis.
Finally, we were limited by the small number of smokers in the study. This resulted in some loss of precision, particularly in sensitivity analyses, and also limited covariate adjustment. However, IPT score weighting allowed consideration of many established potential confounders.4
Analysis with confounding factors individually included in adjusted mixed models yielded similar results to weighted models, suggesting adequacy of IPT weights for confounding.
Comparing biospecimens of smokers from those of nonsmokers collected at multiple times during the menstrual cycle, we observed small statistically significant differences in levels of LH and FSH early in the follicular phase. Relations between smoking and hormone levels have implications for fertility and reproductive health. Further work among women with higher levels of smoking would help clarify this relation.