Vitamin D and uterine fibroid growth, incidence, and loss: a prospective ultrasound study

The Study of Environment, Lifestyle and Fibroids (SELF) is a prospective cohort study designed to evaluate fibroid growth and incidence ( 24 ). Due to earlier fibroid onset in Black/African American women in the United States ( 2 ), SELF enrollment was limited to those who self-identified as ‘Black or African American’ among a list of racial and ethnicity categories. Details of study recruitment and activities are provided in Appendix 1 . Briefly, recruitment was conducted in the Detroit, Michigan area from 2010 to 2012 in collaboration with Henry Ford Health (HFH). Eligible participants were aged 23–34, who reported no prior diagnosis of fibroids at recruitment. The participants completed baseline questionnaires and attended a clinic visit that included an ultrasound examination, anthropometric measurements, and non-fasting blood collection. Of 1693 participants enrolled, 1610 (95%) attended ≥1 of the three follow-up visits during which the same study activities were completed. The four study visits occurred approximately 20 months apart and were completed in 2018. Participants who missed a visit were encouraged to attend the next. Visits for pregnant participants were delayed until 3–4 months post-partum for better ultrasound imaging. Study retention was high; 91% of the enrolled cohort attended the final visit, 95% attended at least 2 visits, and 79% attended all 4 study visits ( Supplemental Figure 1 ). The 5% who only attended the enrollment visit tended to be older with lower income, lower body mass index and were more likely to have had a pregnancy. Those participants who attended all 4 visits tended to older and were more likely to be employed compared to those who attended only 2 or 3 visits, but otherwise the data showed no clear pattern of differences in baseline characteristics ( Supplemental Table 1 ). SELF was approved by the institutional review boards of the National Institutes Health and HFH. All participants provided written informed consent and received compensation. Serum from each visit was aliquoted and stored at −80C until assayed. Analysis of 25(OH)D was conducted in three batches (serum from baseline, follow-up 1, and follow-up 2). Details of the vitamin D measurement and quality control procedures are described in Appendix 2 . In brief, total serum 25(OH)D for the baseline visit was measured using LIAISON, a competitive chemiluminescence immunoassay ( 25 , 26 ). For subsequent visits serum 25-hydroxyvitamin D 2 (25(OH)D 2 ) and 25-hydroxyvitamin D 3 (25(OH)D 3 ) were measured with liquid chromatography-tandem mass spectrometry (LC-MS/MS) and summed to create a total 25(OH)D measure. Blinded quality control serum samples were included in all assays, and based on these samples all intra-assay coefficients of variation (CV) were <5% and inter-assay CVs were 9% for LIAISON and <5% for LC-MS/MS. The total 25(OH)D for 77 samples run on both platforms showed a mean difference of 0.6 ng/ml (25 th to 75 th percentiles: −1.8 to 1.1). We did not adjust for different assay methods. To account for seasonal variation in 25(OH)D, we estimated the overall annual average 25(OH)D using a cosinor model ( 27 ). We modeled the natural log of 25(OH)D using the sine and cosine of the day of the year of blood sample collection (first and second harmonic) as predictors ( 27 , 28 ). We then created individual values for annual average 25(OH)D by adding the residual for each participant/visit to the model intercept and transforming back to the original scale (ng/ml). This season-adjusted 25(OH)D is denoted 25(OH)D hereafter. For 90 samples, the serum volume was insufficient for the 25(OH)D assay (1.2–2.6% of samples at a given visit). We imputed the majority (N=80) of these values using the mean of measured values from adjacent visits, or carryover of the value from an adjacent visit if there was only one. The remaining observations (N=10) were dropped from the analyses. We used clinically relevant cut points based on recommendations from the Institute of Medicine (IOM) ( 29 ) and the Endocrine Society (ES) ( 30 ) to categorize individual season-adjusted 25(OH)D values for analyses: <12 ng/ml (vitamin D deficient) (IOM and ES), <20 ng/ml (vitamin D inadequacy (IOM) or deficient (ES)), <30 ng/ml (vitamin D insufficient (ES)). Experienced sonographers, trained on the SELF protocol, used transvaginal ultrasound to count, localize and measure fibroids ≥0.5 cm in diameter. Sonographers made three separate passes through the uterus, recording the 3 diameters of a given tumor at each pass. We calculated the fibroid volume for each pass with the ellipsoid formula, and the three volumes were averaged for analysis. Per protocol, sonographers noted any problems with visualization (e.g., calcifications, shadowing) that may have limited visualization. Video and still images were archived and an 8% sample for each sonographer, oversampled for fibroid cases, was reviewed every month by the lead sonographer (details in Appendix 3 ). The primary outcomes of interest were fibroid growth and fibroid incidence. Given the observed reduction in fibroid growth, fibroid loss was explored as a secondary outcome. Every fibroid outcome analysis had different eligibility criteria given that fibroid growth and loss require a fibroid to be detected and fibroid incidence is estimated among those who are fibroid-free at enrollment ( Supplemental Figure 2 ). The lead sonographer (TC) and one other author (DB) identified fibroids that could be seen across two successive visits using archived images and fibroid locations. The growth analyses used data from 434 participants (n=1357 interval growth measurements from successive visits). We defined fibroid growth as the change in the natural logarithm of the tumor volume (ln-volume). Change in ln-volume was scaled to a growth rate per 18-months (median time between visits=19 months; 25 th -75 th percentiles: 18–21) by calculating daily growth rates and multiplying by 540. To compare growth of fibroids from participants in high versus low categories of 25(OH)D we estimated percent difference in volume per 18 months ( Appendix 4 ). We identified incident fibroid cases (first appearance of any fibroids) among participants confirmed via ultrasound not to have fibroids at enrollment (N=1246). If sonographers noted factors which impaired detection of fibroids (e.g., calcifications or shadowing, only a transabdominal ultrasound) the data were excluded from analysis (~0.5% of ultrasounds), resulting in incidence data for 1232 participants. After excluding observations with insufficient serum for 25(OH)D, the incidence analysis included 1230 participants ( Supplemental Figure 2 ). Fibroid loss was assessed among participants with prevalent fibroids at enrollment that had not been clinically diagnosed previously, or participants who developed incident fibroids. For analysis of fibroid loss, defined as a decrease in fibroid number between two successive visits, we excluded intervals including or following a myomectomy, hysterectomy, or uterine artery embolization. When there are numerous fibroids, it can be difficult to accurately count tumors ( 31 ), therefore we restricted this analysis to the 539 participants who had ≥2 successive visits with ≤4 fibroids at the earlier visit and no sonographer report of difficulty with visualization ( Supplemental Figure 2 ). Fibroids that become undetectable by ultrasound between visits include those that shrink below the limit of ultrasound detection (0.5 cm for any diameter); therefore, this outcome cannot be interpreted as complete resolution of fibroids. We identified covariates of interest based on prior fibroid research ( 5 ), previous work in this cohort ( 32 , 33 ), and availability of SELF data. We measured height during the first clinic visit, and weight at every visit to calculate body mass index (BMI kg/m 2 ) for each visit. We collected other covariate data via telephone or computer assisted questionnaires at every visit including reproductive and hormonal variables (years since last birth, parity, age at menarche, years since last use of depot medroxyprogesterone acetate (DMPA)), demographic variables (household income, educational attainment, current employment), physical activity and current smoking. All factors of interest were explored in adjusted models and those that did not affect observed associations between 25(OH)D and the fibroid outcome were not included ( 34 ) in final models. Parameterization of covariates (categorization or continuous) for modelling is noted in the table footnotes for each fibroid outcome. Categorical covariates were usually modelled using indicator variables to allow for non-linear associations. Covariates were updated at the beginning of each interval, except for years since last birth and parity which were updated at the end of the interval between visits to incorporate events during the interval ( 35 ). There were minimal missing covariate data (<0.5%); therefore, when adjusting, we conducted complete-case analyses. We used linear mixed models to account for correlated growth among fibroids from the same participant and for the same fibroid over time as previously described ( 32 , 33 , 36 ). Minimally adjusted growth models included time-varying 25(OH)D, continuous age, and categorical values for fibroid volume and number. Fully-adjusted models included the minimal adjustment set plus age at menarche and time-varying measures of income, employment, BMI, years since last birth, and years since last DMPA use ( Appendix 4 ). We used Cox regression with age as the time-scale to estimate hazard ratios (HR) and 95% confidence intervals (95% CI) for associations between 25(OH)D categories and fibroid incidence. Participants were included from the age at enrollment until the age at fibroid detection, with censoring for loss to follow-up, non-fibroid-related hysterectomy or the final study visit, whichever occurred first. Age-adjusted models examining the association between 25(OH)D and fibroid incidence were estimated first without covariates, and then were further adjusted for BMI, current smoking, income, parity, years since last birth, and years since last DMPA use. There was no evidence that the proportional hazards assumption was violated. We used Poisson regression accounting for multiple observations per participant to estimate risk ratios (RR) and 95% CI with robust standard errors for fibroid loss. Minimally adjusted models included time-varying 25(OH)D, continuous age, months between visits, and categorical values for largest fibroid volume and number of fibroids. Fully adjusted models included the minimal adjustment set plus time-varying education, BMI, years since last birth, and years since last DMPA use. For all outcomes, we removed observations with imputed serum 25(OH)D. Because use of estrogen-containing oral contraceptives is associated with increased serum concentrations of 25(OH)D ( 28 ), we explored possible confounding by including current use (yes, no) as a covariate. For fibroid incidence we set the incidence of fibroids to the mid-point of the interval rather than end. For fibroid growth and loss we removed fibroid-number and fibroid-volume covariates to evaluate the possible influence of exposure-related differences in fibroid number and size. For fibroid growth we excluded statistical outliers (positive or negative growth beyond 3SD). (Details in Appendix 5 ). All analyses used SAS 9.4 (Cary, NC).

At enrollment, among the 1610 participants who had ≥1 follow-up visit, mean age was 29.2±3.4 years, 78% had educational attainment beyond high school, 62% were employed, and 45% had a household income below $20,000. Almost a quarter (24%) had a BMI ≥40 kg/m 2 ( Table 1 ). Participants had low 25(OH)D (median 25(OH)D= 15.3 ng/ml, 25 th -75 th percentiles: 11.1–20.6). Compared with those with 25(OH)D <20ng/ml, those with 25(OH)D ≥20ng/ml had higher educational attainment and income, lower BMI, and were more likely to be using oral contraceptives and be non-smokers ( Table 1 ). There was no time trend in 25(OH)D concentrations across the study’s duration (2010–2016) ( Supplemental Figure 3 ). Median 25(OH)D concentrations were 15.3 ng/ml at enrollment, 14.8 ng/ml at follow-up 1, and 15.3 ng/ml at follow-up 2. Among 25(OH)D measures across the study, 32% were deficient (<12 ng/ml), 41% were 12-<20 ng/ml, 27% were ≥20 ng/ml. Only 7% exceeded the Endocrine Society ( 30 ) cut point for sufficiency (≥30 ng/ml) ( Supplemental Figure 4 ). Very few (1.7%) of the participants had this higher concentration at all three visits. Participants had a median length of study participation of 4.8 years (25 th -75 th percentiles: 4.7–5.0 years). Baseline characteristics were similar for the participants in each of the fibroid outcome groups (growth, incidence and loss analyses), except for age, percent nulliparous, recency of birth and use of DMPA. These are known factors associated with fibroid prevalence and are expected to differ because the growth and loss analyses require participants with fibroids while the incidence analysis requires participants without fibroids ( Supplemental Table 2 ). Most fibroids, including undiagnosed fibroids detected at enrollment and incident fibroids that developed during the study, were small. Median volume of incident fibroids at time of first detection was 0.56 cubic centimeters (cm 3 ) (25 th -75 th percentiles: 0.23–1.4 cm 3 ). Median volume for those followed for growth was 2.2 cm 3 (25 th -75 th percentiles: 0.7–8.6 cm 3 ). At enrollment, 23% of participants had ≥1 fibroid (median: 1 fibroid, 75 th percentile: 2 fibroids), while by the study’s end, 32% had ≥1 fibroid (median: 2 fibroids, 75 th percentile: 3 fibroids). Overall estimated fibroid growth averaged 77.0% (95% CI: 68.6%, 85.9%) volume increase per 18 months. Serum 25(OH)D ≥20 ng/ml was associated with a reduced fibroid growth rate. Fibroids among those with 25(OH)D ≥20 ng/ml had 9.7% less growth per 18 months (95% CI: −17.3%, −1.3%) than fibroids from participants with 25(OH)D <20 ng/ml ( Table 2 ). Fibroid incidence averaged 9.6% within each study interval. Only at 25(OH)D concentrations ≥30 ng/ml was there any indication of a decrease in incidence ( Table 3 ). Hazard ratios for serum 25(OH)D ≥30 ng/ml suggested an estimated 22% reduction in fibroid incidence (HR 0.78, 95% CI: 0.47, 1.30) when compared with serum 25(OH)D <30 ng/ml. Associations were imprecise given the small number of 25(OH)D measures ≥30 ng/ml. Fibroid loss during an interval was common (24.3% of eligible intervals had loss of at least one fibroid between study visits). Participants with 25(OH)D concentration ≥30 ng/ml had 32% higher fibroid loss in an observation interval than those with 25(OH)D <30 ng/ml (RR 1.32, 95% CI: 0.95, 1.83) ( Table 4 ). Sensitivity analyses showed very similar associations to those in the primary analyses ( Appendix 4 , Supplemental Table 3 ).

Using prospectively collected ultrasound data and repeated measures of serum 25(OH)D, we observed slower fibroid growth in participants with 25(OH)D ≥20 ng/ml. For participants with serum 25(OH)D ≥30 ng/ml, we found evidence for a greater likelihood of fibroid loss, but only limited evidence for reduced incidence given the wide confidence interval. In this cohort, there were few 25(OH)D observations ≥20 ng/ml (27%) or ≥30 ng/ml (7%). Only 1.7% of participants maintained 25(OH)D concentrations ≥30 ng/ml for all three measures. Thus, we had limited power to detect associations at higher concentrations of 25(OH)D. The optimal range of 25(OH)D is debated ( 37 ). While the Institute of Medicine has identified ≥20 ng/ml as needed for skeletal health ( 29 ), the Endocrine Society argues that concentrations ≥30 ng/ml are needed ( 30 ). Observational and clinical trials for improved reproductive outcomes including the success of fertility treatment ( 38 ), fecundability in natural conceptions ( 39 ), preterm birth ( 40 ), menstrual cycle characteristics ( 41 ), and breast cancer risk ( 42 ) suggest that concentrations of 25(OH)D ≥40 ng/ml may be required. Although most fibroids in this cohort were small and might not warrant immediate clinical attention, our study still has important clinical implications. Symptoms ( 43 ) and major interventional treatments ( 44 ) are more likely with larger fibroids. Therefore, minimizing fibroid growth and increasing fibroid loss when fibroids are small can delay severe morbidity and prevent the need for surgical or radiologic treatments. Our findings for reduced growth of fibroids from participants with higher 25(OH)D are consistent with results of small clinical trials that treated participants with vitamin D supplements ( 19 – 22 ), though most of those studies ( 19 , 21 , 22 ) included participants with 25(OH)D concentrations ranging up to 30 ng/ml at enrollment, and treated with supplementation that would have resulted in higher concentrations than generally seen in our sample. Fibroid loss was not included as an outcome in the trials of fibroid growth. However, if our finding of reduced fibroid growth with higher vitamin D results in some fibroids shrinking below the ultrasound detection limit, increased fibroid loss would be expected. Although the association between vitamin D and fibroid incidence has not been previously studied, a recent systematic review and additional observational studies, report lower concentrations of 25(OH)D among women with prevalent fibroids compared with those without fibroids ( 6 , 45 – 47 ). The observational studies rely on self-reported fibroid status, and this results in undiagnosed fibroids being included in the “non-case” group ( 5 ) as well as temporal misclassification of 25(OH)D concentrations. A large, randomized control trial using prospective ultrasound examinations is underway in China which will investigate the effects of daily vitamin D supplementation over 2 years on fibroid incidence in more than 2000 women, and fibroid growth in 360 women ( 48 ). Trials in other populations are warranted. In vitro and animal models support our finding that vitamin D limits fibroid growth (reviewed in Vergara ( 10 )). In vitro models including those using immortalized human uterine leiomyoma cells (HuLM) and cultured human cells from fibroid and adjacent myometrial tissue show that treatment with 1,25(OH) 2 D inhibits their growth through changes in cell proliferation, apoptosis, extracellular matrix composition, Wnt/β-catenin and TGFβ3 expression, and down regulation of estrogen and progesterone receptors ( 12 , 13 , 16 , 18 , 49 – 51 ). Treatment with 1,25(OH) 2 D or an analogue in mice and Eker rats with fibroid tumors also results in decreased tumor size through such mechanisms ( 17 , 50 , 52 ). There have been fewer laboratory studies related to fibroid initiation and vitamin D. Most fibroids harbor specific somatic mutations in HMGA2 or MED12 ( 53 ). Fibroids have elevated DNA damage and vitamin D treatment of HuLM cells reduced evidence of DNA damage with upregulation of DNA-repair proteins ( 14 ). Improved detection and repair of DNA damage could prevent the proliferation of cells with mutations and thus limit fibroid initiation ( 54 ). An inherent limitation to studying the growth of individual fibroids is that they must be observed over time, and lost fibroids are not included; to compensate, we examined fibroid loss as a separate outcome. There is also measurement error in ultrasound assessment of fibroid size, and very small fibroids may be missed. However, taking three separate passes through the uterus will maximize fibroid detection, and using the mean of 3 separate volume measures has less measurement error than a single measure. In addition, measurement error is greater for smaller fibroids ( 55 ), but we address this limitation by accounting for the differential error in our growth model ( 33 ). The most important limitation is that we had few participants with 25(OH)D concentrations above 30 ng/ml which limits our power to find associations at the higher vitamin D concentrations recommended by the Endocrine Society. Our prospective study design with multiple standardized ultrasounds is the first that allows for the timely detection of incident fibroids, thus reducing the misclassification of fibroid status which is present in prior observational studies. Our multiple measures of serum 25(OH)D reduce the bias introduced when a single measure of 25(OH)D is assumed to be relevant for later timepoints. In addition, our focus on Black women allows us to examine individuals at high risk for both vitamin D deficiency and fibroid burden, although vitamin D deficiency in reproductive-aged women in the United States is widespread ( 7 ). Vitamin D is well known to be critical for bone health ( 29 ) with growing evidence that higher concentrations of 25(OH)D improve reproductive outcomes ( 56 ) and reduce breast cancer risk ( 42 ). Our findings add support to the existing literature suggesting vitamin D may also reduce fibroid development. As compared with existing medical and surgical treatments for fibroids that have significant side-effects and impact fertility, vitamin D is safe and compatible with pregnancy. The high prevalence of vitamin D deficiency in the SELF cohort and reproductive-aged U.S. women in general ( 47 ) suggests that further public health messaging about vitamin D is needed. An additional issue for U.S. women is that elevated BMI is associated with lower 25(OH)D concentrations ( 57 ). Current dietary and supplement recommendations for vitamin D intake may not be sufficient for individuals with high BMI who require 2–3 times higher vitamin D supplementation to attain desired serum concentrations ( 58 ). In conclusion, 25(OH)D concentrations above 20 ng/ml were associated with reduced fibroid growth. Results for analyses examining 25(OH)D concentrations above 30 ng/ml provided suggestive evidence for increased likelihood of fibroid loss and a possible reduction in fibroid incidence. Intervention trials with vitamin D supplementation to attain and maintain sufficient 25(OH)D will be needed to further assess the impact of vitamin D on fibroid development.

Uterine fibroids are non-cancerous tumors of the myometrium that cause significant morbidity including menorrhagia and pelvic pain, often requiring medical or surgical intervention ( 1 ). Estimates based on ultrasound screening suggest that fibroids develop in over 70% of women. Black women have a 10-year earlier tumor onset compared with White women ( 2 ); they also have larger tumors, more debilitating symptoms, and an increased need for surgery ( 3 ). Existing medical and surgical treatments for fibroids can impact childbearing goals and, apart from hysterectomy, most treatments do not preclude fibroid recurrence ( 1 ). Most individuals with fibroids prefer non-invasive treatments and preservation of the uterus ( 4 ). Despite the high prevalence of fibroids, few modifiable risk factors have been identified. None explain differences in disease severity between Black and White individuals ( 3 , 5 ). Vitamin D concentration, which is lower on average among self-identified Black women, has been proposed as a possible modifiable risk factor for the development of fibroids ( 6 – 9 ). In vitro and animal studies support a protective effect of vitamin D. In these models, vitamin D reduces fibroid tissue proliferation ( 8 , 10 , 11 ), changes the expression of estrogen and progesterone receptors ( 12 ), alters gene expression in proliferation and apoptosis pathways ( 13 – 15 ), and reduces expression of extracellular matrix proteins ( 16 – 18 ). Human observational studies report lower fibroid prevalence in women with higher vitamin D ( 6 , 11 ). Most ( 19 – 22 ) but not all ( 23 ) human intervention studies document reduced fibroid growth when participants are treated with vitamin D, but studies were small (30 to 205 participants). Although highly suggestive, the existing laboratory and epidemiologic literature has limitations. Animal and in vitro studies usually study the active form of vitamin D [1,25-dihydroxyvitamin D (1,25(OH) 2 D)]. These models may not reflect tissue concentrations in humans who obtain vitamin D through ultraviolet-B (UV-B) light exposure (D 3 ), diet (D 2 or D 3 ), or vitamin supplements (D 2 or D 3 ) with conversion to 25-hydroxyvitamin D (25(OH)D) in the liver and final conversion to 1,25(OH) 2 D in the kidney or target tissue. In addition, most epidemiologic studies have relied on clinical diagnosis or ultrasound detection of prevalent fibroids, but initial fibroid onset occurs many years before clinical detection ( 6 , 11 ). We hypothesize that higher concentrations of serum 25(OH)D will be associated with reduced fibroid growth and incidence. We investigated this hypothesis with four repeated standardized ultrasound examinations over five years.