Plasma 25-hydroxyvitamin D and risk of premenstrual syndrome in a prospective cohort study

Approximately 8-20% percent of premenopausal women experience premenstrual syndrome (PMS), [1, 2] a disorder of moderate to severe symptoms in the luteal phase of the menstrual cycle that substantially interfere with normal life activities and interpersonal relationships [3]. Symptoms are diverse, and commonly include affective symptoms such as irritability, mood swings, anxiety and depression; physical symptoms such as breast tenderness, bloating, and headaches; and behavioral symptoms such as insomnia, changes in appetite, and difficulty concentrating. While pharmacologic treatments including anti-depressants and oral contraceptives are frequently used to treat PMS, these have substantial side effects and none has a reported efficacy greater than 60-70% [2, 3]. Because of these limitations, effective non-pharmaceutical strategies for preventing and treating PMS are needed.

Previous studies have evaluated the role of vitamin D in preventing and/or treating mood and gynecologic disorders that share common features with PMS. Multiple studies have reported inverse associations between vitamin D status and risk of depression, [4, 5] fibromyalgia, [6] dysmenorrhea, [7] and uterine fibroids [8, 9]. However, it remains unknown whether vitamin D may be useful for preventing or treating PMS. Our research group previously reported that women who consumed approximately 400 IU of vitamin D per day had a significant 40% lower risk of being diagnosed with PMS in the next 2–4 years, as compared to women consuming approximately 100 IU/day [10]. This association was present even after adjustment for calcium, which has been shown to beneficially impact PMS symptoms in some [11‐14] but not all studies [15]. While these results suggest that vitamin D may be related to the development of PMS, many questions remain and must be addressed before vitamin D can be recommended for PMS. Vitamin D status is the product of cutaneous vitamin D production and not just dietary intake. Correlations of dietary vitamin D intake with the main circulating vitamin D metabolite, 25-hydroxyvitamin D (25OHD), are relatively low, [16] indicating that many factors influence vitamin D metabolism and availability [17, 18]. Thus, assessing dietary vitamin D intake alone may be insufficient to accurately characterize a woman’s vitamin D status.

Few studies have evaluated how 25OHD levels are associated with PMS or with premenstrual dysphoric disorder (PMDD), and results have been inconsistent [19‐21]. Importantly, because previous studies have been limited to women already experiencing PMS at the time of their 25OHD measurement, results from these studies may be confounded by patients’ use of vitamin D or multivitamin supplements or other dietary changes to treat PMS symptoms, as is currently recommended [22, 23]. Prospective studies evaluating 25OHD levels in women prior to PMS development are needed to better understand whether vitamin D may be etiologically related to the development of PMS.

We have evaluated whether plasma levels of 25OHD, along with two vitamin D-related biomarkers (total calcium and parathyroid hormone), are associated with PMS among a subset of participants in the Nurses’ Health Study II (NHS2). We have assessed relations overall, and after taking the timing of 25OHD measurement vs. PMS diagnosis into consideration. Finally, because the physical, affective, and behavioral symptoms of PMS likely have different etiologies, we have assessed whether 25OHD levels are differently associated with specific symptoms of PMS.

Characteristics of PMS cases and controls measured at the time of blood collection are presented in Table 1. Overall, cases were more likely than controls to be current or former smokers (P < 0.001 for both). Cases had a marginally higher mean daily alcohol intake (4.6 vs. 3.7 g/day; P = 0.05), and were more likely to report significant trauma in childhood (18.1% vs. 5.6%; P < 0.0001). Though the study population was racially homogenous overall, cases and controls differed significantly in the proportion of participants reporting white race (P = 0.02). Both anti-depressant use and oral contraceptive use were more commonly reported by cases than controls (P < 0.001 for both), as was use of B vitamin supplements (P = 0.002). Cases and controls did not differ by age, BMI, age at menarche, or physical activity. Mean levels of vitamin D-related biomarkers did not differ between cases and controls.

In our main analyses of all PMS cases and their matched controls, we did not find plasma levels of 25OHD, total calcium or parathyroid hormone to be associated with PMS (Table 2). In unadjusted analyses, the OR for women in the highest quintile of 25OHD (median = 91.6 nmol/L) was 1.26 (95% CI = 0.79 – 2.02) compared to the lowest quintile (median = 42.4 nmol/L). Adjustment for other factors including BMI, smoking, and physical activity did not materially alter results. For example, after adjustment, women in the highest quintile of 25OHD had an OR of 1.16 (95% CI = 0.60 – 2.41). Results for 25OHD classified into broader categories indicating vitamin D deficiency vs. sufficiency (i.e., <50 vs. ≥50 nmol/L) were similarly null. Neither total calcium nor iPTH levels were associated with risk of PMS.

We then evaluated associations separately by timing of 25OHD measurement vs. PMS diagnosis (Table 3). 25OHD levels were higher in cases who had already been diagnosed with PMS as the time of their blood collection (67.9 nmol/L) as compared to controls (64.8 nmol/L), and 25OHD was positively associated with PMS risk. Specifically, each 10 nmol/L increase in 25OHD was associated with a 13% higher risk of PMS (95% CI = 1.03 – 1.25). In contrast, 25OHD levels were not associated with risk in cases with blood collected before diagnosis.

Relations of 25OHD levels with individual menstrual symptoms also varied by timing of blood collection. Among women with 25OHD measured before they were diagnosed with PMS, plasma 25OHD levels were inversely related to risk of developing breast tenderness, fatigue, diarrhea and/or constipation, and depression (all P < 0.05; Table 3). For example, each 10 nmol/L increase in 25OHD level was associated with a significant 21% lower risk of developing breast tenderness (P = 0.02), and a significant 19% lower risk of depression (P = 0.047). In contrast, among women who already had PMS at the time of blood collection, 25OHD was positively associated with risk of diarrhea/constipation, food cravings, appetite changes, and a variety of affective symptoms (all P < 0.05). For example, each 10 nmol/L increase in 25OHD level was associated with a significant 19% higher risk of anxiety and irritability (P < 0.01 for both).

For all biomarkers, results from sensitivity analyses excluding women reporting a diagnosis of depression during follow-up were similar to the main analyses. For example, the OR for each 10 nmol/L change in 25OHD was 1.08 (95% CI = 0.98 – 1.19).

Overall, we found no association between plasma levels of 25OHD and risk of PMS in our longitudinal study. However, results differed somewhat based on the timing of blood measurement in relation to PMS diagnosis. Among women who had already been diagnosed with PMS by the time of blood collection, 25OHD levels were positively associated with PMS risk. In contrast, among women diagnosed with PMS after blood draw, 25OHD levels were not associated with future risk of developing PMS.

The observed positive association of 25OHD with PMS risk was contrary to our hypothesis, and reasons for this finding are unclear. One potential explanation is that women experiencing PMS may treat their symptoms with calcium and vitamin D supplements, or multivitamins containing these nutrients, as is currently recommended by the American Congress of Obstetricians and Gynecologists and other groups [22, 23]. Consequently, 25OHD levels would elevated in PMS cases due to differences in supplementation practices, and this type of confounding by indication would bias results. In our study, mean 25OHD levels were indeed significantly higher in women who had PMS at blood collection (67.9 nmol/L) compared to those not yet diagnosed (63.2 nmol/L; P = 0.04), suggesting that supplementation, diet and/or behaviors influencing vitamin D status differed between case groups.

We further explored this hypothesis in post hoc analyses. At the time of blood collection, calcium and vitamin D supplement use was reported by 43.5% of women already experiencing PMS, compared to 36.6% of women not yet diagnosed (P = 0.19). Multivitamin use also varied slightly between groups (50.% vs. 44.7%, respectively; P = 0.27). This finding is consistent with results from Thys-Jacobs et al.[20], who reported non-significantly higher use of calcium supplements (46% vs. 32%) and vitamin D supplements (41% vs. 30%) among women with PMDD compared to symptom-free controls [20]. Among calcium supplement users in our population, 56% reported taking 400–900 mg/day, while 36% reported taking >900 mg/day; this level of supplementation may have considerable impact on associations.

When we took supplementation practices into consideration among cases already diagnosed with PMS at the time of blood collection, we found the positive association to be limited to women using supplements (OR for each 10 nmol/L change = 1.19; P = 0.002). In contrast, among non-users of supplements, 25OHD levels were unrelated to risk (OR = 1.03; P = 0.67). These results support the hypothesis that women experiencing PMS may be self-medicating with dietary supplements. This interpretation is also consistent with results from a previous study of B vitamin supplementation and PMS in our population, in which we found vitamin B6 supplement use to be more common among women already experiencing PMS than among women not yet diagnosed with PMS [33]. The likelihood of confounding by indication suggests a need for caution in interpreting results from case–control and cross-sectional studies of nutritional factors and PMS. It also underscores the importance of evaluating these relations prospectively whenever possible, to minimize the chance of bias.

Over 100 different menstrual symptoms may occur in PMS, [34] and the specific symptoms experienced by individual women may differ substantially. Because physical, affective, and behavioral symptoms likely have different etiologies, it is may be important to evaluate the associations of etiologic factors with individual symptoms and symptom clusters separately. This hypothesis is supported by randomized trials of pharmacologic therapies for PMS, as reviewed by Halbreich et al.[35], which demonstrate substantial variation in treatment response between symptom profiles. For example, selective serotonin reuptake inhibitors appear most beneficial to women experiencing mood symptoms, especially irritability, and behavioral symptoms, but are less effective for treating physical symptoms.

Similarly, associations of vitamin D with specific symptoms may vary. In the present analysis, among women not yet diagnosed with PMS at blood collection, we found 25OHD level to be associated with significantly lower risk of developing breast tenderness, diarrhea/constipation, fatigue and depression (P < 0.05) and possibly with swelling of extremities and bloating (P ≤ 0.11). It is possible that some of these findings may be due to chance, given the relatively large numbers of comparisons we made. Alternatively, differences in observed associations suggest that vitamin D may be associated with some symptoms but not others. For example, clinical studies of PMS suggest that dysfunction of the renin-angiotensin-aldosterone system (RAAS) is involved in premenstrual edema symptoms such as abdominal bloating, swelling of extremities, and breast tenderness [36]. Vitamin D deficiency is also associated with increased RAAS function, contributing to increased fluid balance, blood pressure changes, and hypertension [37, 38]. Furthermore, an inverse relation between 25OHD levels and premenstrual depression would be consistent with the literature suggesting that vitamin D may lower risk of unipolar depression [39]. It therefore seems plausible that vitamin D may specifically influence menstrual depression and symptoms related to fluid balance, but not other symptoms. This should be explored in future studies.

Our study has several potential limitations requiring consideration. First, our ability to assess the relation of vitamin D-related biomarkers and PMS may have been limited by access to only a single blood sample collected several years before or after PMS diagnosis. It is possible that the 25OHD level measured in this sample does not reflect long-term vitamin D status. Multiple studies have evaluated the stability of 25OHD within individuals over time [40‐43]. For example, in the Nurses’ Health Study, the intraclass correlation comparing three measurements of 25OHD levels collected over 3 years was 0.72 (95% CI = 0.62-0.80) [40]. However, most studies of vitamin D stability in women have been conducted in postmenopausal women or have not directly evaluated variation by menopausal status [40, 42, 43]. In one study evaluating blood samples drawn before and after menopause in 34 women (median time between samples = 5.8 years) the correlation for 25OHD was 0.39 (P < 0.05), suggesting that levels may be stable over many years, even in the context of substantial changes in levels of estradiol and other sex steroid hormones [41]. While these data suggest that a single 25OHD measurement can accurately represent vitamin D status over multiple years, additional studies specifically evaluating relations in premenopausal women are needed.

Additionally, we were unable to consider the effect of vitamin D binding protein concentrations, which recent studies suggest may importantly affect the bioavailability of 25OHD, and thus relation between vitamin D measurements and health outcomes [44]. Future studies of vitamin D status and PMS risk should incorporate this important parameter. It is important to note that for our prospective analyses, misclassification resulting from the use of a single 25OHD measurement to represent long-term vitamin D status is likely to be non-differential with respect to PMS diagnosis. Consequently, this type of bias would lead to an underestimation of the relation of vitamin D and PMS, as opposed to exaggerating associations. Additional prospective studies, including clinical trials, are needed to clarify the impact of timing of 25OHD measurement on the association of vitamin D and the development of menstrual symptoms and PMS.

Due to the collection and storage protocol used in the NHS2, we were unable to measure ionized calcium, which is a preferred measure of calcium status in clinical populations, and to measure protein stores as they impact calcium status [45]. This may have contributed to some misclassification of the association of calcium status with PMS risk. While we adjusted for a variety of dietary and lifestyle factors that could confound the association of vitamin D and PMS risk, we were unable to separately evaluate indoor and outdoor physical activity (which could differently affect 25OHD levels), or comprehensively assess other aspects of sun exposure. Though we included race/ethnicity, geographic region of residence, and number of moles on lower leg in our multivariable regression models, these factors are likely imprecise measures of sun exposure and some residual confounding may persist.

Our study of PMS is nested within a large, ongoing prospective cohort study of over 116,000 women. Consequently, we were unable to use prospective symptom diaries to classify PMS, as is recommended by the International Society for Premenstrual Disorders [46]. However, we used established criteria to identify women meeting strict criteria for moderate to severe PMS [25] as well as to define a comparison group of women experiencing no menstrual symptoms or few symptoms with no personal impact. Women meeting neither case nor control criteria were excluded from analysis. Thus, the likelihood of misclassification between cases and controls is minimal and is unlikely to bias our findings. Additionally, when we excluded women reporting a diagnosis of depression during follow-up, results were very similar to the main analysis. This suggests that associations were robust to potential misclassification of depression as PMS, or to the impact of comorbid depression on the vitamin D – PMS relation.

In summary, we did not find 25OHD levels consistent with current IOM guidelines for vitamin D sufficiency to be associated with overall risk of PMS risk. The positive association of 25OHD and risk we observed among women already experiencing PMS at the time of vitamin D measurement is likely due to confounding by indication resulting from the use of dietary supplements to treat menstrual symptoms. Results from prospective analyses, which were likely less influenced by this bias, suggest that higher 25OHD levels may be inversely related to the development of specific menstrual symptoms. Additional prospective studies of vitamin D for the treatment and prevention of PMS are warranted.