Pre-diagnostic circulating follicle stimulating hormone (FSH) concentrations and ovarian cancer risk

Abstract

Gonadotropins have been indicted in ovarian carcinogenesis but direct evidence has been limited and inconsistent. The aim of this study was to determine the association between pre-diagnostic levels of follicle stimulating hormone (FSH) and subsequent development of invasive epithelial ovarian cancer. A nested case-control study was conducted using cases and controls drawn from three cohorts: CLUE I and CLUE II of Washington County, MD, and the Island of Guernsey Study, UK. In total, 67 incident invasive epithelial ovarian cancer cases were each matched to one to two controls on age, menopausal status, time since last menstrual period, current hormone use, and other relevant factors. FSH concentrations were classified into ranked thirds of low, medium, or high based on the distribution among controls. Conditional logistic regression was used to estimate the odds ratio (OR) across increasing thirds of FSH concentrations. Results of the analysis showed that ovarian cancer risk decreased with higher FSH concentrations (p-trend =0.005). Compared to the lowest third of FSH concentrations, the odds ratio among those in the middle and highest thirds were 0.45 (95% Confidence Interval (CI) 0.20– 1.00) and 0.26 (95% CI 0.10–0.70), respectively. Associations persisted after excluding cases diagnosed within five years of follow-up. In conclusion, a reduction in subsequent risk of invasive epithelial ovarian cancer was observed among women with higher circulating FSH concentrations. These findings contradict the hypothesized role of FSH as a risk factor in ovarian carcinogenesis.

Introduction

Exposure of the ovarian surface epithelium to elevated gonadotropin levels has been proposed to play a deleterious role in ovarian carcinogenesis¹. This view is based on the concept that post-ovulatory inclusion cysts formed from invagination of the ovarian surface epithelium (OSE) are exposed to excessive sub-stromal stimulation by gonadotropins and estrogen in the pre-menopausal period, resulting in increased cellular proliferation with eventual progression to malignant transformation. The hypothesis is largely based on observations in animal models in which various means of increased production of gonadotropins (follicle-stimulating hormone [FSH] and/or luteinizing hormone [LH]) was associated with ovarian tumor development¹. In humans, epidemiologic evidence in support of this hypothesis includes the decreased risk observed with oral contraceptive use, pregnancy, and possibly lactation. All of these factors, which suppress gonadotropin releasing hormone (GnRH) signaling for synthesis and secretion of both FSH and LH, are in agreement with this hypothesis².

Only two studies of actual pre-diagnostic circulating concentrations of gonadotropin in relation to ovarian cancer risk have been reported, and their results do not, overall, support the “gonadotropin hypothesis”^1;3. A small nested-case control study of 31 cases (of whom 14 were premenopausal and 17 postmenopausal) each matched to two controls, found a statistically significant decreased risk of ovarian cancer with elevated concentrations of FSH³. In another nested case-control study of 88 post-menopausal women who later developed ovarian cancer matched to 168 controls, investigators observed a non-statistically significant decrease in risk with increasing FSH levels⁴. As compared to women in the lowest third of FSH concentrations, the adjusted odds ratio for women in the highest third was 0.85 (95% CI 0.36–1.99) ⁴. These studies do not support the original proposition that increased FSH secretion increases the risk of subsequent ovarian cancer development. In fact, both studies suggest that gonadotropins may play a role in ovarian carcinogenesis by decreasing risk.

In the present study, the resources of three cohorts from two study sites, CLUE I and CLUE II of Washington Country, Maryland; and the Island of Guernsey Studies, United Kingdom, were combined to examine the role of pre-diagnostic circulating FSH and risk of subsequent development of epithelial ovarian cancer.

Material and Methods

The details of the study cohorts and the general approach to the study design have been previously described ⁵. The present study is comprised of a subset of a larger consortium of four cohorts from three study sites that analyzed the association between several circulating biomarkers and ovarian cancer. Consortium participants include CLUE I and CLUE II (named based on campaign slogan, “Give us a CLUE to cancer and heart disease”) of Washington County, Maryland, the Columbia, Missouri Serum Bank, and the Island of Guernsey Prospective Study, United Kingdom. Only a subset of the original study group had sufficient serum available for the FSH assay. The analytic subset was comprised of study participants from CLUE I, CLUE II, and the Island of Guernsey Prospective Study. Each of the cohorts in the analysis enrolled community-based volunteers who were cancer-free at study entry. CLUE I enrolled participants during a four month period in 1974. CLUE II actively recruited participants from May to October 1989. Enrollment of community-based study participants in the Island of Guernsey Prospective Study occurred in two recruitment phases, the first in 1977–1985 and the second in 1986–1990. Follow-up for this study continued through February 2000. The parent multi-center prospective study included 166 women who developed ovarian cancer matched to an average of two controls⁵. For this study, 67 women who developed ovarian cancer and 109 matched controls had sufficient serum for the FSH assay. Of those, 41 cases and 76 controls were contributed by CLUE I, 7 cases and 8 controls were from the Guernsey studies, and 19 cases and 25 controls were from CLUE II.

Each case-patient included in this analysis had at least one matched control. In total, 25 case-control sets contain one control (37%) and 42 contain 2 controls (63%). Matching criteria included age at study entry, menopausal status as defined by one year or more since last natural menstrual period, time since last menstrual period if pre-menopausal, current oral contraceptive use, current hormone replacement therapy, , cohort of origin, and time of blood draw. Three ovarian cancer patients underwent hysterectomy prior to menopause and state of ovarian function was thus unclear. These women were matched to control-women who had also undergone pre-menopausal hysterectomy at similar age/date.

Written informed consent was provided by all participants in each of the cohorts at baseline. The Institutional Review Boards at the Johns Hopkins University Bloomberg School of Public Health, the National Cancer Institute, the Guernsey Board of Health, and the International Agency for Research on Cancer reviewed and approved the study.

Laboratory Measurements

A subset of the CLUE I cohort cases were included in a previous study of sex hormones and ovarian cancer risk³. To conserve serum, the measures taken for the 1995 study were used in the present analysis. This group consists of 22 cases and 38 controls. FSH concentrations in this group were assayed using radioimmunoassay (RIA) kits (Ciba Corning, East Walpole, MA). The remaining FSH concentrations were measured by a direct immunoradiometric assay from DSL (Diagnostic System Laboratories, Texas).

The concentrations of sex hormones were used in the analysis to adjust for components of the hormonal milieu. For the older subset of CLUE I, androstenedione, and estrone were measured by RIA³. Dehydroepiandrosterone-sulfate (DHEA-S) was measured using RIA kits (Wein Laboratories, Succasunna, NJ). Sex hormones measured in the remaining samples were assessed as follows: testosterone and DHEA-S were measured by direct RIA (Immunotech, Marseille, France). Androstenedione and estrone were measured by a direct double-antibody RIA (Diagnostic System Laboratories[DSL], Texas). Sex hormone binding globulin (SHBG) was measured by a direct “sandwich” immunoradiometric assay (IRMA; Cis-Bio, Gif-sur-Yvette, France). Luteinizing hormone (LH) was assayed with a direct RIA (DSL, Texas).

Each case-control set was run adjacently in the same assay batch. Pooled quality control (QC) aliquots, approximating a 5% of the total, were also arranged in triplet clusters and interspersed among the case-control sets. Laboratory personnel were masked to case-control and quality control status. The sub-set of CLUE I measured in 1995 were also assayed in adjacently run matched case-control sets with a QC set included for every 10 sets³. The intra-set coefficients of variation were 17% in plasma, 6% in serum. The intra-set coefficient of variation for FSH in the CLUE I subset measured in 1995 was 0.8% ³.

Measures of CA-125, an ovarian cancer tumor marker, were available for a subset of 20 cases and 36 controls from CLUE I as part of a prior study ⁶.

Statistical Analysis

The data for FSH were transformed using the natural logarithm to account for the right-skewed distribution and minimize the concomitant departure from normality. FSH concentrations of case-patients were compared with controls by calculating the difference in concentrations between each case-patient and her matched control. The mean difference between the two groups was calculated by averaging the individual differences between each of the matched pairs. This method of comparison accounts for variability between cohorts in assay methods and storage details as differences were calculated within sets of matched pairs,. Comparisons of age at study entry, menstrual status, oral contraceptive use, hormone replacement therapy, and current smoking were compared in the same manner. Potential confounding variables were assessed among the control population comparing covariate data across ranked thirds of FSH, as defined below, using the chi-square test for categorical variables and the non-parametric Kruskal-Wallis test for continuous variables.

The distribution among all controls with FSH measures available was used to assign tertile cut-points, within cohort, and by menopausal status. Within-cohort rankings were pooled across cohorts. Tests for trend in normally distributed continuous variables were assessed across categories of FSH concentrations by thirds using linear regression with robust variance estimates to account for the matched design. Conditional logistic regression was used to calculate matched odds ratios (OR) as estimates of the relative risk and corresponding 95% confidence intervals (95% CI). Odds ratios for risk with increasing thirds of FSH concentration were assessed by using the scores as indicator variables, separately comparing the middle and highest thirds against the lowest ranked third as the referent group. Tests for trend in ovarian cancer risk were assessed using the quantitative scores for the pooled thirds of FSH in the model as an ordinal variable. For those observations with LH available, ranked thirds were assigned in the same manner as described for FSH. In order to assess for polycystic ovary syndrome (PCOS), a putative risk factor for ovarian cancer⁷, the LH to FSH ratio was calculated and analyzed as both a continuous variable and dichotomized at >=2 vs <2, which is the clinical cut-point used to assess for PCOS⁸.

A random effects model was used to adjust for influence on the data by cohort, using the “meta” command in STATA 8.2 on pooled odds ratios and standard errors computed from individual cohort data. The data became too sparse for this model when using the score for ranked thirds of FSH as an indicator variable. Thus, results for the overall effect of FSH by thirds using the random effects model are reported. Test for heterogeneity between cohorts include Cochrane’s Q statistic and the I² statistic⁹.

Many of the factors that might potentially confound the relationship between FSH and ovarian cancer risk were addressed through matching on age at study entry, cohort of origin, current hormone use, menopausal status, days since last menstrual period if pre-menopausal, and time of day at blood draw. Stratified analyses were conducted by menopausal status, smoking at time of blood draw, and body mass index (BMI). BMI was available for a subset of 26 cases and 33 controls. Adjustment for BMI did not alter associations and this variable was not further considered.

Analyses were performed excluding cases diagnosed within two and five years of blood draw to evaluate possible effects of occult tumors on hormone concentrations. Among the sub-set of CLUE I with FSH and CA-125 data available, Spearman correlation coefficients were calculated to assess the association between log transformed FSH and CA-125, an established tumor marker for ovarian cancer.

Results

Case-patients and controls were similar in age, race (all Caucasian), menopausal status, and exogenous hormone use due to matching (Table 1). Average FSH concentrations, taken within case-control matched pairs, were 1.22 IU/mL higher among controls than women who developed ovarian cancer (p=0.006). The direction of this association was consistent in both pre-menopausal (1.4 IU/mL, p=0.03) and postmenopausal women (1.1 IU/mL, p=0.11).

Table 1.

Baseline characteristics of women with ovarian cancer and controls

Variable	Case	Control	Average within-set difference between case and control	P-Value
N	67	109
Age, mean yrs (SD)	54.2 (12.5)	53.3 (12.6)	−0.11	0.90
Current Oral	0 (0.0)	0 (0.0)	0	1.0
Contraceptive Use
Past Oral	6 (9.0)	6 (5.5)	0	1.0
Contraceptive use
Current HRT Use	6 (9.0)	11 (10.1)	0	1.0
Current Smoking	9(13.6)	23 (21.3)		0.26
Menstrual Status
Pre-Menopausal	24 (35.8)	43(39.5)	0	1.0
Post-Menopausal	43 (64.2)	66 (60.6)
Menstrual Phase (Pre-menstrual only*)			0	1.0
0–12 Days	8(36.4)	14 (34.2)
13–16 Days	4(18.2)	9(23.0)
17–35	10(45.5)	18 (43.9)
Time since Menopause			0	1.0
2–4 years	5(11.6)	8 (12.1)
5–9 years	9 (20.9)	14 (21.2)
10–14 years	10 (23.3)	17 (25.8)
>= 15 years	16 (37.2)	24 (36.4)
Premenopausal hysterectomy	3(7.0)	3 (4.6)

Comparisons were made using individual inter-set differences between case-patients and controls using sign test to test for equality between matched groups.

Among controls, FSH concentrations were greater among post-menopausal women with a median of 53.9 IU/mL (95% range 6.2–124.4) vs pre-menopausal women (median 5.3 IU/mL; 95% range 1.0–45.4, p-value 0.0001). Cohort-specific mean and median FSH values were similar to that of all groups combined (p=0.16 among pre-menopausal, p=0.20 among post-menopausal, p=0.16, adjusting for age). FSH increased with age but was not statistically associated with smoking status, menstrual cycle phase, time since menopause, although mean/median concentrations varied as consistent with physiologic expectation (data not shown).

Ovarian cancer risk decreased with higher concentrations of circulating FSH (Table 3). Compared to women categorized in the lowest third of FSH concentrations, the odds ratio decreased from 0.45 (95% CI 0.20–1.00) to 0.26 (95% CI 0.10–0.70) in the middle and highest thirds, respectively (p-trend=0.005). The results for analyses that were further adjusted for potential residual confounding by age and for BMI, among those with this information available, were not appreciably different. Estimates adjusted only by matching criteria are reported.

Table 3.

Odds ratios of ovarian cancer risk by ranked thirds of FSH

	Low-third	Mid-third OR (95% CI)	High-third OR (95% CI)	P trend
Cases	34	21	12
Controls	35	41	33	0.005
OR (95% CI)	1.0	0.45 (0.20–1.00)	0.26 (0.10–0.70)
Pre-Menopausal
Cases	11	8	5
Controls	14	16	13
OR (95% CI)	1.0	0.61 (0.19–1.98)	0.28 (0.05–1.50)	0.12
Post-Menopausal
Cases	23	13	7
Controls	22	25	20
OR (95% CI)	1.0	0.34 (0.11–1.06)	0.24 (0.07–0.82)	0.02
Not Currently Using Hormones
Cases	31	20	10
Controls	29	40	29
OR (95% CI)	1.0	0.42(0.18–0.96)	0.23 (0.08–0.66)	0.004
Not Current Smokers
Cases	25	20	12
Controls	27	31	27
OR (95% CI)	1.0	0.51 (0.20–1.28)	0.36 (0.12–1.07)	0.05
Excluding cases diagnosed within 2 years of follow-up
Cases	33	21	11
Controls	33	41	32
OR (95% CI)	1.0	0.42 (0.19–0.96)	0.22 (0.08–0.63)	0.003
Excluding cases diagnosed within 5 years of follow-up
Cases	25	16	10
Controls	27	30	28
OR (95% CI)	1.0	0.46 (0.18–1.17)	0.24 (0.08–0.76)	0.01

Matching criteria included age at study entry, menopausal status, time since last menstrual period if pre-menopausal, current hormone use, cohort of origin, and time of day of blood draw.

As seen in the stratified analyses presented in Table 3, the inverse association between FSH concentrations and ovarian cancer risk was internally consistent with groups defined by menopausal status, hormone use, cigarette smoking, and time to diagnosis. The potential influence of occult cancers undetected at baseline was assessed by stratifying data by time to diagnosis from baseline blood draw. The relationships persisted after excluding case-patients and their matched controls who were diagnosed within two years and within five years of study entry. (Table 3).

An inverse correlation of borderline significance was observed between log-transformed FSH and CA-125 among case-patients (Spearman r = −0.39, p=0.09) but no correlation was evident among controls (Spearman r = −0.003, p=0.98). CA-125 is a known tumor marker for ovarian cancer.

An analysis was conducted that excluded the 21 case-patients and 38 controls previously analyzed in the 1995 study³. Results continued to show a protective association (mid-third OR 0.39, 95% CI 0.15–1.05, high third OR 0.35 95% CI 0.12–1.05, p-trend 0.04). Analyses excluding women who underwent pre-menopausal hysterectomy, and thus had unknown ovarian function, also showed no significant changes in estimates. (mid-third OR 0.47, 95% CI 0.21–1.07, high third OR 0.27 95% CI 0.10–0.72, p-trend 0.006).

FSH concentrations and biological activity are affected in a regulatory manner by other components of the hormonal milieu. Adjusting for continuous, log transformed concentrations of estrone, androstenedione, and DHEA-S did not materially affect the association between FSH and ovarian cancer (Table 4). There were 45 and 44 case-control sets with SHBG and testosterone concentrations available, respectively, as the 1995 sub-set of CLUE I did not assay these markers. Adjusting for either marker, individually and simultaneously, had minimal effect on associations. Measures of LH concentrations were available among 52 case-controls sets. Adjusting for LH in this group had a marginal effect in strengthening the odds ratios for FSH (Table 4). In analyzing for independent effects of LH, no significant associations were observed with increasing thirds among the subset of case-control sets with measures of LH available (mid-third OR 1.06, 95% CI 0.41–2.75, high third OR 0.89 95% CI 0.31–2.57, p-trend 0.79).

Table 4.

Risk of ovarian cancer by ranked thirds of FSH, adjusted for components of hormonal milieu

	Case-control sets	Mid-third† OR (95% CI)	High-Third† OR (95% CI)	P-value for trend
Androstenedione	66	0.47 (0.21–1.06)	0.26 (0.10–0.70)	0.005
DHEA-S	66	0.47 (0.21–1.07)	0.26 (0.10–0.71)	0.006
Estrone	66	0.47 (0.20–1.08)	0.26 (0.09–0.77)	0.01
LH	52*	0.48 (0.19–1.21)	0.30 (0.10–0.91)	0.03
	52	0.43 (0.16–1.14)	0.24 (0.07–0.85)	0.02
SHBG	45*	0.39 (0.15–1.05)	0.35 (0.12–1.05)	0.04
	45	0.35 (0.14–0.97)	0.30 (0.09–0.94)	0.02
T	44*	0.38 (0.14–1.05)	0.25 (0.07–0.84)	0.01
	44	0.37 (0.13–1.03)	0.22 (0.06–0.79)	0.01
T and SHBG	44	0.35 (0.12–0.99)	0.20 (0.06–0.74)	0.009

^*adjusted only for matching criteria and excluding case-controls sets without LH/SHBG/T serum measures. Subsequent row shows ORs adjusted for LH, SHBG, T, respectively.

^†Mid-third and high-third ORs estimated using low-third as referent group.

Using the LH to FSH ratio as a surrogate marker for PCOS, no association was observed with ovarian cancer risk (OR 1.00 95% CI 0.72–1.39). An increased, non-statistically significant risk was observed when the dichotomized ratio (>=2 vs <2) was entered into the model (OR 3.41, 95% CI 0.67–17.3).

Using a random effects model to adjust for effects attributed to the individual cohorts did not appreciably change inferences. Given the sample size of the individual cohorts, only the overall odds ratio for increasing thirds of FSH concentrations was estimable under the random effects model. The overall odds ratio for the fixed effects model was 0.49 (95% CI 0.29–0.83, p-trend 0.008). The overall odds ratio under the random effects model was 0.51 (95% CI 0.25–1.05, p-trend 0.07). The change in precision around the estimates under the random effects model may reflect the loss of power inherent to this analytical approach. Cochrane’s Q statistic for heterogeneity between cohorts was 5.005 (p=0.17) and the I² statistic was 40%, suggesting low to moderate heterogeneity between groups⁹. This may reflect the instability inherent to the small sample sizes within the cohort specific analyses, as the Guernsey-specific odds ratio was estimated from 7 case-control sets. Exclusion of the Guernsey case-control sets did not appreciably affect odds ratios (mid-third OR 0.40, 95% CI 0.17–0.95, high third OR 0.16, 95% CI 0.05–0.53, p-trend 0.001).

Discussion

This prospective study, which combined resources of three cohorts from two study sites, demonstrated that risk of ovarian cancer decreased with higher circulating concentrations of pre-diagnostic FSH. These findings do not provide support for the hypothesis that gonadotropin exposure is a risk factor for ovarian cancer as originally proposed in 1983¹. The results suggest that FSH may play a role in ovarian carcinogenesis but in a direction opposite to that proposed. Thus, the hypothesis with regard to FSH should be re-visited. Our observed associations also seem consistent with data supporting the increase in risk observed with HRT use¹⁰, as HRT use would suppress endogenous FSH levels among post-menopausal women.

Adjusting for other hormones, such as LH, estrone, androstenedione, DHEA-S, SHBG and testosterone did not substantially affect the risk estimates in the subset of case-controls sets with information on those hormones available. LH was not associated with increased risk in this study sample but non-statistically significant elevations in risk with LH:FSH >=2 suggest the role for LH requires more rigorous exploration. A total of 4 case-patients and 3 controls had an elevated ratio, limiting the precision of the current analysis.

Several aspects of the results suggest the observed relationship between FSH and ovarian cancer risk may be a true association. The measurement of FSH concentration was based on blood samples drawn prior to follow-up for ovarian cancer. A significant dose-response relationship was observed across increasing thirds of FSH concentrations, which was robust to adjustment for known confounders and analyses stratified by smoking status, menopausal status, and hormone use.

The observed associations do not appear to be due to the presence of occult cancers undiagnosed at baseline, as the association was observed after excluding cases diagnosed within two years and then five years of follow-up. Log transformed FSH was inversely correlated with tumor marker CA-125 among cases, suggesting that the effects of FSH may be acting remotely from tumor development. Among case-patients, FSH levels appear to decrease as CA-125 concentrations increase. These data raise the possibility that elevated FSH levels confer a protective effect before significant tumor development occurs as indicated by increasing CA-125 concentrations. However, one cannot exclude the possibility that, as ovarian tumors progress, pituitary FSH secretion is increasingly reduced as a result of perturbed feedback mechanisms between the ovary and the hypothalamic/pituitary system. The lack of correlation among controls is expected as ovarian tumor development is restricted to the case-patients only.

There are several challenges intrinsic to pooling resources across cohorts. The FSH concentrations from sub-set of CLUE I that participated in the 1995 study were assessed in a separate laboratory. The differences were controlled for by matching, so that relative differences were preserved and compared within case-control groups. Associations persisted after excluding the older sub-set of CLUE I with previously published findings and the trend in risk remained statistically significant. Further, analyses accounting for testosterone and SHBG concentrations necessarily excluded this sub-group due to missing data, yet the strength and consistency of the inverse associations remained consistent. Several factors that could obscure the associations of interest were addressed by close matching. Blood samples were drawn pre-diagnostically, thereby reducing the likelihood that FSH levels were due to presence of an occult tumor. Another potential issue includes differences in storage method and number of freeze-thaw cycles. Case-patients and control were matched on storage methods and the CLUE cohorts were matched on number of freeze-thaw cycles. However, FSH concentrations in serum have been shown to be stable through up to ten freeze-thaw cycles and at storage temperatures of both −20°C and −70°C ¹¹. FSH concentrations did not differ significantly between cohorts after adjusting for age. The associations remained relatively unchanged under the random effects model and after exclusion of the Guernsey data suggesting that between-cohort heterogeneity is not exerting undue influence on study findings. These observations are guarded by the limitations in sample size.

It is thus reassuring that the findings presented in this report corroborate the protective association first observed in the CLUE I subset in 1995³. Another collaborative cohort study also examined the role of FSH in ovarian cancer risk using levels measured in pre-diagnostic serum⁴. There were 88 incident ovarian cancer cases identified in that study, from cohorts from New York City, USA, Umea, Sweden, and Milan, Italy. Study participants were restricted to post-menopausal women not currently using exogenous hormones. Cases were each matched by age, cohort, and enrollment date. Tertile cut-points were determined by the distribution among cases and controls combined. In that study, the odds ratio for women in the highest third of FSH concentration, as compared to women in the lowest, was 0.85 (95% CI 0.36–1.99) after adjusting for potential confounders. We re-analyzed our data using their study restrictions and assigning cut-points by thirds with cases included in the distribution. With these restrictions, our study was limited to 38 cases and 57 controls. However, the odds ratios for women in the middle and highest thirds were 0.29 (95% CI: 0.08–0.98) and 0.19 (95% CI: 0.05–0.75), respectively (p-trend= 0.02).

While these findings do not support the gonadotropin hypothesis as originally proposed, they do suggest that FSH may play an important protective role in ovarian carcinogenesis. The associations are similar among pre and post menopausal women, although estimates were not statistically significant among pre-menopausal women. This suggests that the relationship between FSH and ovarian cancer risk may be independent of menstrual cycling. The fluctuations in FSH characteristic of this cycling likely add noise to an effect related to basal levels of FSH, thus reducing precision of estimates among pre-menopausal women. We were able to match on time since onset of menopause for all but two case-control sets, thus reducing random differences between cases and controls that might be caused by variability in FSH levels associated with time from onset of menopause ^12–15.

FSH could possibly exert a direct protective effect on tumor development. It could also be that increased FSH levels are indicative of decreased levels of directly acting hormones or factors that also serve to regulate gonadotropin levels. Studies involving FSH and/or LH administration to normal and malignant human OSE cultures have had conflicting results regarding the role of FSH in cellular proliferation. The FSH receptor is expressed on both normal and malignant human OSE¹⁶. In one study of normal human OSE cell cultures, FSH levels were found to suppress proliferation while LH exerted no effect¹⁷. In another study, FSH was found to be mitogenic in both normal and malignant cell cultures, with similar effects observed with LH administration¹⁶. In yet a third study, FSH-induced mitogencity was inhibited by LH in cells with benign abnormalities and in malignant cells¹⁸. Alternatively, proliferation was not induced by FSH alone and was suppressed when combined with human chorionic gonadotropin (hCG)¹⁹ in hormonally sensitive benign and malignant cell lines. Finally, FSH concentrations may have differing effects on gene expression regulation of putative tumor suppressor/oncogenes on OSE cells as compared to ovarian cancer cell lines²⁰ While these findings fail to clarify potential relationships, it should be noted that OSE and ovarian cancer lines are notoriously difficult to culture and may be limited in application as a model of the human ovarian surface epithelium²¹. The associations observed with LH and the LH:FSH ratio in this study, while certainly inconclusive, raise further questions regarding the complexity of the relationship of FSH to LH with regard to ovarian carcinogenesis.

In summary, this prospective study demonstrated that, among women who were cancer-free at baseline, higher FSH concentrations were associated with decreased risk of subsequent development of ovarian cancer. These observations raise interesting new questions about a potential protective role for gonadotropins. A better understanding of the link between FSH and ovarian cancer could lead to the identification of promising avenues for preventive intervention. However, further and larger studies are needed to finally elucidate the role of FSH in ovarian cancer

Table 2.

Association between baseline characteristics and ranked thirds of FSH among controls (count and percentage unless otherwise specified)

	Low-third	Mid-third	High-third	P-value
Age at blood donation, mean (SD)	49.4 (12.1)	54.4 (13.3)	55.9 (11.4)	0.03
BMI, mean (SD)	25.8 (4.6)	23.8 (3.1)	24.7 (3.5)	0.48
Pre-Menopausal	14 (40.0)	16 (39.0)	13 (39.4)	0.99
Post-Menopausal	21 (31.8)	25 (37.9)	20 (30.3)
Current Smoking	7 (20.6)	10 (24.4)	6 (18.2)	0.8
Current HRT Use	7 (17.1)	1 (2.4)	4 (12.1)	0.10
Menstrual Phase
0–12 Days	3(21.4)	7(43.8)	4 (36.4)	0.31*
13–16 Days	2(14.3)	3 (18.8)	4(36.4)
17–35 Days	9 (64.4)	6 (37.5)	3(27.3)
Years Since Menopause
2–4 years	2 (9.5)	2 (8.0)	4 (20.0)	0.81*
5–9 years	5(23.8)	5 (20.0)	4(20.0)
10–14 years	6(28.6)	6(24.0)	5 (25.0)
>= 15 years	6(28.6)	11(44.0)	7 (35.0)
Premenopausal hysterectomy	2 (9.5)	1(4.0)	0 (0)

^*P-values were calculated chi-square test for trend for categorical variables and Kruskal-Wallis test for continuous variables across thirds of FSH concentration.

ABREVIATIONS

BMI

body mass index

CI

Confidence Interval

DHEA-S

Dehydroepiandrosterone-sulfate

FSH

follicle stimulating hormone

GnRH

gonadotropin releasing hormone

HRT

hormone replacement therapy

hCG

human chorionic gonadotropin

IU

international units

LH

luteinizing hormone

mL

milliliter

OR

odds ratio

OSE

ovarian surface epithelium

PCOS

polycystic ovary syndrome

QC

quality control

RIA

radioimmunoassay

SHBG

sex hormone binding globulin