Oral Contraceptive and Menopausal Hormone Therapy Use and Risk of Pituitary Adenoma: Cohort and Case-Control Analyses

Abstract

Context

No prospective epidemiologic studies have examined associations between use of oral contraceptives (OCs) or menopausal hormone therapy (MHT) and risk of pituitary adenoma in women.

Objective

Our aim was to determine the association between use of OC and MHT and risk of pituitary adenoma in two separate datasets.

Methods

We evaluated the association of OC/MHT with risk of pituitary adenoma in the Nurses’ Health Study and Nurses’ Health Study II by computing multivariable-adjusted hazard ratios (MVHR) of pituitary adenoma by OC/MHT use using Cox proportional hazards models. Simultaneously, we carried out a matched case-control study using an institutional data repository to compute multivariable-adjusted odds ratios (MVOR) of pituitary adenoma by OC/MHT use.

Results

In the cohort analysis, during 6 668 019 person-years, 331 participants reported a diagnosis of pituitary adenoma. Compared to never-users, neither past (MVHR = 1.05; 95% CI, 0.80-1.36) nor current OC use (MVHR = 0.72; 95% CI, 0.40-1.32) was associated with risk. For MHT, compared to never-users, both past (MVHR = 2.00; 95% CI, 1.50-2.68) and current use (MVHR = 1.80; 95% CI, 1.27-2.55) were associated with pituitary adenoma risk, as was longer duration (MVHR = 2.06; 95% CI, 1.42-2.99 comparing more than 5 years of use to never, P trend = .002). Results were similar in lagged analyses, when stratified by body mass index, and among those with recent health care use. In the case-control analysis, we included 5469 cases. Risk of pituitary adenoma was increased with ever use of MHT (MVOR = 1.57; 95% CI, 1.35-1.83) and OC (MVOR = 1.27; 95% CI, 1.14-1.42) compared to never.

Conclusion

Compared to never use, current and past MHT use and longer duration of MHT use were positively associated with higher risk of pituitary adenoma in 2 independent data sets. OC use was not associated with risk in the prospective cohort analysis and was associated with only mildly increased risk in the case-control analysis.

cohort, case control, epidemiology, menopausal hormone therapy, oral contraceptives, pituitary adenoma

Pituitary adenomas are common lesions of the pituitary gland, with an estimated prevalence as high as 20% based on autopsy and radiological studies, but often remain undiagnosed for years before discovery (1, 2). They can cause debilitating symptoms including visual field deficits, headache, hypopituitarism, and symptoms attributable to hyperfunction for hormone-secreting adenomas (1-5).

Identification of individuals at risk of pituitary adenoma may allow for early and effective treatment (5). Currently, a large proportion of lesions are incidentally discovered (3). In addition, diagnosis is often made among young women during evaluation of amenorrhea or infertility, as pituitary adenomas can cause these disorders (5).

A potential link between oral contraceptive (OC) or menopausal hormone therapy (MHT) use and pituitary adenoma incidence is plausible, given that estrogen stimulates lactotroph cells in the pituitary gland (6). Previous case-control studies have examined the association between OC use and incidence of pituitary adenoma with mixed results, but that design is subject to bias (7-12). It is possible, for example, that women experiencing menstrual irregularity or other symptoms as the result of an as-yet undiagnosed pituitary adenoma are then treated with OC or MHT, leading to an observed association in case-control studies despite the lack of a causal relationship. To our knowledge, no prospective studies have examined associations between OC or MHT and incidence of pituitary adenoma.

In this study, we conducted 2 complementary analyses in parallel. In the Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHSII) cohorts, we prospectively assessed the association between OC and MHT use and risk of pituitary adenoma (13). In addition, we performed a case-control analysis using data from the Mass General Brigham Research Patient Data Registry (RPDR) to examine the same associations. We hypothesized that past or current use of OC or MHT, and longer duration use of these medications, would be associated with higher risk of pituitary adenoma. We further hypothesized that, owing to reverse causation, these associations would be attenuated in lagged analyses, particularly for OC use.

Materials and Methods

Nurses’ Health Study Analysis

Study participants

The methods of the NHS and NHSII have been previously described (14). NHS was started in 1976 with an enrollment of 121 701 female nurses aged 30 to 55 years; NHSII began in 1989 with an enrollment of 116 686 female nurses aged 25 to 42 years. Participants completed a baseline questionnaire and subsequent biennial follow-up questionnaires. Questionnaires were expanded across the course of the study to include detailed information on dietary factors, physical activity, and health behaviors, in addition to assessment of health outcomes. Follow-up rates in the cohorts have been higher than 90% (15). The study protocol was approved by the institutional review boards of the Harvard T. H. Chan School of Public Health and Brigham and Women’s Hospital, and those of participating registries, as required.

Assessment of oral contraceptive use

At baseline, participants reported their use of OC. Use was categorized as never, past, or current, and was updated every 2 years from baseline until 1982 in the NHS (at which point fewer than 500 participants reported current use) or until the end of follow-up in the NHSII. Participants also reported overall duration of OC use, which was categorized as never, 0 to 1 years or less, more than 1 to 2 years or less, more than 2 to 3 years or less, more than 3 to 4 years or less, more than 4 to 5 years or less, or more than 5 years. Self-report of OC use was validated among 215 participants in the NHSII, with 99% agreement between self-report and a telephone interview using a structured life events calendar for ever use, and a correlation of 0.94 between mean durations of use (42.7 mo by telephone interview vs 44.6 mo by self-report) (16).

Assessment of menopausal hormone therapy use

At baseline, participants reported their use of MHT. Use was categorized as never, past, or current, and was updated every 2 years from baseline until the end of follow-up in both cohorts. Participants also reported duration of MHT use, which was categorized as never, 1 year or less, more than 1 to 5 years or less, or more than 5 years. Beginning in 1978 in the NHS and at baseline in the NHSII, participants also reported the type of MHT, categorized as estrogen only or estrogen plus progestin.

Identification of pituitary adenoma cases

Each biennial questionnaire included a list of diagnoses and a space to write in any other important diagnosis. All cases of pituitary adenoma were identified by participants who reported their diagnosis in the write-in space on the biennial questionnaires. The date of diagnosis was taken to be 12 months before the return of the questionnaire where the diagnosis was identified. We did not seek medical records to confirm cases of pituitary adenoma.

Covariate assessment

Smoking status was reported by participants on each biennial questionnaire and was categorized as never, past, or current. Body mass index (BMI) was calculated for each time period by using the reported height at baseline and the updated weight, and was categorized according to the World Health Organization categorization as less than 25, 25 to 30, or more than 30. Age at menarche was reported at baseline and was categorized as younger than 11, 11 to 14, or older than 14 years.

To account for potential diagnostic or surveillance bias, we used biennial data to create a variable representing recent health care use. On each questionnaire, all participants reported whether they had undergone a recent physical exam, sigmoidoscopy/colonoscopy, or breast exam within the most recent questionnaire cycles.

Statistical analyses

We began follow-up time at the date of return of the baseline questionnaire and continued to the date of pituitary adenoma diagnosis, death from another cause, date of return of last questionnaire, or the end of follow-up (June 2018 for NHS and Health Professionals Follow-Up Study; June 2017 for NHSII), whichever came first. We included 118 993 participants in the NHS and 115 044 in the NHSII. We constructed Cox proportional hazards models to calculate multivariable-adjusted hazard ratios (MVHRs) and 95% CIs to evaluate risk of pituitary adenoma by OC and MHT use, using months as the time metameter and age and calendar year as stratification variables. For missing values for these variables, we carried forward prior responses up to 4 years. If data were still missing, those participants no longer contributed person-time to the analysis. We adjusted for smoking in multivariable-adjusted Cox models because it may be related to likelihood of diagnosis. To assess the independent effects of OC and MHT use, we further conducted an analysis that cross-classified both by OC and MHT use. In this analysis, users were categorized as never OC/MHT users (reference), ever MHT/never OC users, never MHT/ever OC users, and ever OC/ever MHT users.

To evaluate duration relationships, we assigned the median of each duration category to all members of that category, and tested for the trend across categories in a multivariable-adjusted Cox model.

To address reverse causation—the possibility that OC or MHT use among pituitary adenoma cases may be affected by prediagnostic disease—we applied the main analysis variables in 4- and 8-year lagged analyses, resulting in the exclusion of the first 4 and 8 years of follow-up in each case. In a secondary analysis, we stratified by BMI (< 25 vs ≥ 25) to assess whether the association between OC and MHT and pituitary adenoma differs by BMI.

In a secondary analysis, we aimed to account for the possibility that women taking MHT or OC may be more regular health care users, and therefore be more likely to experience an incidental diagnosis of a pituitary adenoma. In this analysis, we limited to women who reported recent health care use within the 2 most recent questionnaire cycles.

Analyses were performed separately in each cohort and then combined by fixed-effect meta-analysis. All statistical analyses were performed using the SAS 9.4 statistical package (SAS Institute), and all P values were derived from 2-sided tests; the statistical significance threshold was set to .05.

Mass General Brigham Research Patient Data Registry Case-Control Analysis

Data source

The Mass General Brigham health care system established the RPDR in 2002 to provide centralized data storage for demographic, administrative, and clinical data from electronic medical record systems. This data source has been used for studying a variety of other health exposures and outcomes (17-20). In 2012, a validation study of RPDR compared the registry to a surgeon-collected case log and demonstrated that after accounting for mismatches, RPDR captured 96% of primary joint arthroplasties from 2000 to 2009 (17).

Study participants

Using RPDR’s web-based query interface, we identified cases of pituitary adenoma diagnosed within our hospital system among women from 2010 to 2020. We used a comprehensive search including both International Classification of Diseases, Revision 9 (ICD-9; 227.3) and Revision 10 (ICD-10) codes (D35.2). In the overall analysis, the date of diagnosis was taken to be the date when a code for pituitary adenoma first appeared in the medical record. These cases were frequency matched to controls from the same institutional data repository, selected at random from the larger cohort of all women with a medical record during these years. Matching factors were age (within 10 y), race, sex, and overall health care use (a composite metric based on the number of encounters, diagnoses, and visits in the medical record). The index date for controls was the date of a randomly selected encounter in the medical record. In additional analyses, we restricted to cases that had co-occurring hyperprolactinemia (ICD-10: E22.1) diagnosed at the same encounter as the pituitary adenoma, in an effort to attempt to identify patients with prolactinoma from the overall pool of pituitary adenomas.

Assessment of oral contraceptive and menopausal hormone therapy use

The data repository provided a list of all medications that appeared in each participant’s medical record, with associated dates of administration. From this list of medications, we identified OC and MHT, and further identified whether participants did or did not receive these medications before their diagnosis of pituitary adenoma (for cases) or their index date (for controls). We followed medication use from the time of first entry into the data repository until the index date of control selection or date of diagnosis for cases.

Covariate assessment

Covariates were drawn from the medical record and included age (y, continuous) and race (White vs non-White). These matching factors were adjusted for in the final analysis to improve precision and limit possible biases in control selection.

Statistical analyses

We constructed logistic regression models, adjusted for age and race, to estimate multivariable-adjusted odds ratios (MVORs) and 95% CIs for ever use of MHT and OCs vs never use. We also constructed a single model that mutually adjusted MHT use for OCs, and vice versa. We further conducted a sensitivity analysis using a 1-year lag, in which both the index date for controls and diagnosis date for cases were taken to be 365 days before their original diagnosis or index date, to take into account the possibility that cases were diagnosed earlier at another site and later appeared in the medical record on presentation for a specialty referral. All statistical analyses for this portion of the analysis were performed using R version 3.6.1, and all P values were derived from 2-sided tests. All supplementary tables are available in a data repository (21), and all data are available from the investigators on reasonable request.

Results

Nurses’ Health Studies Analysis

During 6 668 019 person-years of follow-up, 331 participants (132 in NHS, 199 in NHSII) reported a new diagnosis of pituitary adenoma (Table 1). Participants in the NHSII were younger on average than those in the NHS (34.8 vs 42.9 y) and reported a longer duration of OC use but a shorter duration of MHT use. In addition, participants in NHSII were substantially younger at the time of diagnosis than those in NHS (42.4 vs 60.3 y).

Table 1.

Baseline characteristics of the Nurses’ Health Study (1976) and Nurses’ Health Study II (1989) participants

	NHS		NHSII
	Full cohort(N = 118 993)	Cases(n = 132)	Full cohort(N = 115 044)	Cases(n = 199)
Age, y	42.9 (7.2)	41.1 (4.1)	34.8 (4.7)	34.4 (4.5)
BMI (median, SD)a	23.8 (4.2)	24.6 (2.7)	24.1 (5.1)	25.0 (4.0)
Smoking status (%)
Never	51 731 (43)	53 (40)	73 656 (64)	143 (72)
Past	27 717 (23)	32 (24)	28 781 (25)	48 (24)
Current	39 212 (33)	44 (33)	10 963 (10)	8 (4)
Missing	333 (1)	3 (2)	1644 (1)	0
Raceb
White	78 945 (66)	88 (67)	103 639 (90)	174 (87)
Non-White	22 144 (19)	29 (22)	9638 (8)	22 (11)
Missing	17 904 (15)	15 (11)	1767 (2)	3 (2)
OC usec
Never	62 777 (53)	58 (44)	18 993(17)	38 (19)
Past	49 188 (41)	65 (49)	82 621 (72)	140 (70)
Current	7028 (6)	9 (7)	13 204 (11)	19 (10)
OC use duration (mo, median, SD)	46.8 (42.2)	44.0 (20.9)	54.6 (43.7)	48.4 (27.6)
MHT usec
Never	101 455 (85)	109 (83)	101 379 (88)	158 (79)
Past	6840 (6)	13 (10)	8497 (7)	19 (10)
Current	9505 (8)	10 (8)	4352 (4)	20 (10)
MHT use duration (mo, median, SD)	49.5 (48.3)	63.1 (39.5)	19.5 (25.5)	26.3 (24.6)
Time to diagnosis from baseline (y, median, SD)		19.0 (11.0)		8.0 (7.1)
Age at diagnosis (y, median, SD)		60.3 (12.5)		42.4 (8.6)

Abbreviations: BMI, body mass index; MHT, menopausal hormone therapy; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; OC, Oral contraceptive.

^aAge-adjusted.

^bAvailable from baseline in the NHSII and from 1992 in the NHS. All NHS participants who did not survive to complete 1992 questionnaire are listed as missing.

^cMay not sum to total because of missing baseline values.

We observed no strong associations between OC use and pituitary adenoma (Table 2). Compared to never use, neither past (MVHR = 1.05; 95% CI, 0.80-1.36) nor current OC use (MVHR = 0.72; 95% CI, 0.40-1.32) was associated with risk of pituitary adenoma. Similarly, longer duration of OC use was not associated with risk (MVHR = 0.95; 95% CI, 0.68-1.31 comparing > 5 y to never use, P trend = .43).

Table 2.

Risk of pituitary adenoma in the Nurses’ Health Study and Nurses’ Health Study II by oral contraceptive and menopausal hormone therapy use

		NHS (n = 132)			NHSII (n = 199)			Total (N = 331)
		Casesa	MV HRb	95% CI	Casesa	MV HRb	95% CI	Casesa	MV HRb,c	95% CI
OC use	Never	58	Ref.		37	Ref.		94	Ref.
	Past	74	1.33	0.91-1.95	144	0.84	0.58-1.21	219	1.05	0.80-1.36
	Current	0	–		16	0.73	0.40-1.32	16	0.72	0.40-1.32
OC use duration, y	Never	58	Ref.		37	Ref.		95	Ref.
	0-1	25	1.58	0.96-2.58	38	1.07	0.68-1.69	63	1.28	0.92-1.79
	1-2	7	0.87	0.39-1.96	17	0.70	0.39-1.24	24	0.75	0.47-1.20
	2-3	6	0.98	0.41-2.33	19	0.87	0.50-1.51	25	0.90	0.56-1.43
	3-4	8	1.57	0.73-3.39	12	0.68	0.35-1.31	20	0.96	0.59-1.59
	4-5	5	1.20	0.47-3.05	13	0.80	0.42-1.50	18	0.90	0.53-1.53
	> 5	23	1.33	0.80-2.21	56	0.75	0.49-1.15	79	0.95	0.68-1.31
	P trend			.44			.12			.43
MHT use	Never	58	Ref.		122	Ref.		180	Ref.
	Past	48	1.80	1.11-2.93	49	2.12	1.48-3.04	97	2.00	1.50-2.68
	Current	26	1.40	0.83-2.37	25	2.19	1.38-3.49	51	1.80	1.27-2.55
MHT use duration, y	Never	58	Ref.		122	Ref.		180	Ref.
	0-1	9	1.25	0.60-2.60	31	1.93	1.29-2.88	40	1.74	1.23-2.48
	1-5	24	1.93	1.14-3.28	21	2.13	1.29-3.53	45	2.04	1.41-2.93
	> 5	41	1.55	0.94-2.55	19	2.94	1.68-5.16	60	2.06	1.42-2.99
	P trend			.19			< .001			.002
OC and MHT cross-classification	Never MHT/OC user	28	Ref.		23	Ref.		51	Ref.
	Ever MHT/Never OC user	30	1.42	0.82-2.47	16	3.98	2.07-7.65	46	2.18	1.43-3.33
	Never MHT/Ever OC user	30	1.05	0.60-1.82	102	0.96	0.61-1.52	132	1.00	0.70-1.42
	OC/MHT ever user	44	2.09	1.22-3.58	58	1.81	1.09-3.00	102	1.93	1.34-2.80

Abbreviations: BMI, body mass index; HR, hazard ratio; MHT, menopausal hormone therapy; MV, multivariable; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; OC, oral contraceptive; Ref., reference.

^aCase counts may not sum to total because of missing values.

^bAdjusted for age, smoking status (never vs past vs current), BMI (< 25 vs 25-30 vs > 30), and age at menarche (< 11 y vs 11-14 y vs > 14 y).

^cCalculated by fixed-effect meta-analysis of both cohorts.

In contrast, strong positive associations were observed between MHT use and higher risk of pituitary adenoma. In the pooled analysis, compared to never use, both past (MVHR = 2.00; 95% CI, 1.50-2.68) and current use (MVHR = 1.80; 95% CI, 1.27-2.55) were associated with higher risk. Similarly, longer duration of MHT use was associated with a higher risk of pituitary adenoma (MVHR = 2.06; 95% CI, 1.42-2.99 comparing > 5 y to never use, P trend = .002). When cross-classified by both OC and MHT use, strong positive associations were observed for participants ever using MHT but never using OC (MVHR = 2.18; 95% CI, 1.43-3.33) and ever using MHT and ever using OCs (MVHR = 1.93, 95% CI, 1.34-2.80), compared to never MHT/OC use. No such association was identified for ever OC use in the absence of MHT use (MVHR = 1.00; 95% CI, 0.70-1.42).

When examined by type, positive associations were observed for both estrogen only and estrogen and progestin MHT (Table 3). For estrogen only, use for 1 to 5 years was associated with a higher risk of pituitary adenoma compared to never use (HR = 1.77; 95% CI, 1.15-2.73), but with no trend for duration (P trend = .24). For estrogen/progestin, use for 1 to 5 years was similarly associated with higher risk (HR = 1.68; 95% CI, 1.05-2.68), also with no trend for duration (P trend = .64), although overall case counts for analyses by type were limited.

Table 3.

Risk of pituitary adenoma in Nurses’ Health Study and Nurses’ Health Study II by menopausal hormone therapy use, by hormone type

		NHS (n = 124)			NHSII (n = 199)			Total (N = 323)
		Casesa	MV HRb	95% CI	Casesa	MV HRb	95% CI	Casesa	MV HRb,c	95% CI
E only duration, y	Never E user	79	Ref.		186	Ref.		265	Ref.
	0-1	4	0.89	0.32-2.45	1	0.71	0.10-5.08	5	0.85	0.34-2.09
	1-5	18	1.76	1.04-3.00	8	1.80	0.86-3.73	26	1.77	1.15-2.73
	> 5	23	1.36	0.82-2.23	4	0.92	0.33-2.55	27	1.26	0.80-1.97
	P trend			0.18			0.65			0.24
E + P duration, y	Never E + P user	91	Ref.		187	Ref.		278	Ref.
	0-1	8	1.97	0.93-4.18	2	1.02	0.25-4.19	10	1.71	0.88-3.31
	1-5	15	1.58	0.88-2.86	8	1.86	0.86-3.98	23	1.68	1.05-2.68
	> 5	10	1.00	0.49-2.03	2	1.34	0.32-5.63	12	1.06	0.56-2.00
	P trend			0.92			0.22			0.64
Any MHT duration, y	Never MHT user	61	Ref.		178	Ref.		239	Ref.
	0-1 E	2	0.69	0.17-2.86	1	0.81	0.11-5.84	3	0.73	0.23-2.31
	1-5 E	13	2.30	1.23-4.29	5	1.45	0.58-3.59	18	1.98	1.19-3.32
	> 5 E	15	1.28	0.69-2.37	3	0.85	0.26-2.73	18	1.17	0.68-2.02
	0-1 E + P	5	2.22	0.86-5.74	2	1.23	0.30-5.09	7	1.39	0.66-2.92
	1-5 E + P	6	1.41	0.58-3.43	4	1.18	0.42-3.35	10	1.31	0.66-2.57
	> 5 E + P	7	1.29	0.54-3.05	2	1.74	0.41-7.36	9	1.17	0.68-2.02

Abbreviations: BMI, body mass index; E, estrogen; HR, hazard ratio; MHT, menopausal hormone therapy; MV, multivariable; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; P, progestin; Ref., reference.

^aCase counts may not sum to total because of missing values.

^bAdjusted for age, smoking status (never vs past vs current), BMI (< 25 vs 25-30 vs > 30), and age at menarche (< 11 y vs 11-14 y vs > 14 y).

^cCalculated by fixed-effect meta-analysis of both cohorts.

In 4-year (n = 235) and 8-year (n = 179) lagged analyses, there was a mild attenuation of the estimated MVHRs for the association between MHT use and pituitary adenoma risk, but overall findings remained similar, with positive associations for MHT and null associations for OCs (21). In the 4-year lagged analysis, OC use duration of less than 1 year was associated with an increased risk of pituitary adenoma (MVHR = 1.60; 95% CI, 1.08-2.36) (21).

In stratified analyses, we found no substantial differences in the overall findings by BMI category for OCs or MHT (21). Restriction of the analysis to women with recent health care use resulted in overall similar results, with null findings for OCs and positive associations between MHT and pituitary adenoma incidence (21). Compared to those in this group who never used MHT, those with more than 5 years of MHT use had an MVHR of 2.08 (95% CI, 1.40-3.08, P trend = .004).

Mass General Brigham Research Patient Data Registry Case-Control Analysis

We identified 5469 cases of pituitary adenoma diagnosed in the Mass General Brigham RPDR from 2010 to 2020, to which we frequency-matched 6262 controls on age (10-y intervals), race (White vs non-White), female sex, and health care use (Table 4).

Table 4.

Participant characteristics in the case-control analysis^a

	Cases (n = 5469)	Controls (n = 6262)	Total (N = 11 731)
Age (y, mean, SD)	44.8 (16.5)	46.5 (16.8)	45.7 (16.7)
With hyperprolactinemia (n, %)	882 (16.1%)	818 (13.1%)	1700 (14.5%)
Female sex (n, %)	5469 (100)	6262 (100)	11 1 (100)
White race (n, %)	4031 (73.7)	4738 (75.7)	8769 (74.8)
Ever MHT use (n, %)	408 (7.5)	313 (5.0)	721 (6.1)
Ever OC use (n, %)	768 (14.0)	679 (10.8)	1447 (12.3)

Abbreviations: MHT, menopausal hormone therapy; OC, oral contraceptive.

^aCases and controls frequency-matched on age (in 10-y intervals), sex, and race (White vs non-White).

Ever use of MHT was associated with a higher risk of pituitary adenoma (MVOR = 1.57; 95% CI, 1.35-1.83), as was ever use of OCs (MVOR = 1.27; 95% CI, 1.14-1.42, Table 5). With the application of a 1-year lag, results were similar (MVOR = 1.86; 95% CI, 1.54-2.24 for ever vs never use of MHT, MVOR = 1.37; 95% CI, 1.20-1.56 for ever vs never use of OCs). Mutual adjustment of MHT for OCs and vice versa did not result in a substantial attenuation of the estimated MVORs (data not shown). In the analysis restricted to patients with a co-occurring diagnosis of hyperprolactinemia, results remained statistically significantly positive for the association between MHT and OCs and pituitary adenoma risk.

Table 5.

Odds ratios of pituitary adenoma by oral contraceptive and hormone therapy use in the case-control analysis

		First recorded diagnosis date			1 y before recorded date
Any pituitary adenoma
		Cases	MV ORa	95% CI	Cases	MV ORa	95% CI
MHT use	Never	5061	Ref.		4710	Ref.
	Ever	408	1.57	1.35-1.83	297	1.86	1.54-2.24
OC use	Never	4693	Ref.		4430	Ref.
	Ever	776	1.27	1.14-1.42	577	1.37	1.20-1.56
Pituitary adenoma with hyperprolactinemia
		Cases	MV ORa	95% CI	Cases	MV ORa	95% CI
MHT use	Never	805	Ref.		796	Ref.
	Ever	77	1.42	0.98-2.06	57	1.82	1.16-2.94
OC use	Never	651	Ref.		683	Ref.
	Ever	231	1.30	1.04-1.63	170	1.40	1.08-1.83

Abbreviations: MHT, menopausal hormone therapy; MV, multivariable; OC, oral contraceptive; OR, odds ratio; Ref., reference.

^aMatched by age (in 10-y intervals), sex, and race (White vs non-White), and additionally adjusted for these factors.

Discussion

Despite being relatively common lesions, few risk factors for pituitary adenoma have been identified (5, 22). Most studies investigating risk factors for these tumors used a case-control method, which is complicated by the possibility of protopathic bias (7, 9-11). These tumors are often incidental findings and may be present for years before diagnosis. The possibility of reverse causation is substantial for OC, because women with menstrual irregularities or other symptoms caused by an undiagnosed adenoma may be treated with OC.

In this study, we relied in part on prospective cohort data, with repeated assessment of OC and MHT use before outcome, limiting the possibility of reverse causation and selection bias. At the same time, we conducted a parallel case-control analysis using data from RPDR, identifying more than 5000 cases of clinician-diagnosed pituitary adenoma and matched controls, which also allowed for a lagged analysis to limit the possibility of reverse causation bias.

The prospective cohort analysis demonstrates a strong positive association between use of MHT and pituitary adenoma, with more than a doubling of risk for more than 5 years of use compared to never-users; this association persisted in lagged analyses and in analyses designed to limit potential diagnostic bias. In contrast, we did not demonstrate any substantial association for OC use and pituitary adenoma incidence. When mutually accounting both for OC and MHT use, strong positive associations were identified between MHT use and pituitary adenoma risk, with or without use of OCs, but no such associations were identified for OC use in the absence of MHT use. Although this analysis included self-reported cases, estimates from other epidemiologic studies of pituitary adenoma incidence suggest that across 6 668 019 person-years of follow-up, we should have expected approximately 172 to 493 cases of pituitary adenoma (23, 24), similar to our observed 331 cases, suggesting that self-report may not be a major limitation of this portion of the analysis.

In the case-control analysis, we again demonstrated an increased risk for pituitary adenoma among users of MHT and also demonstrated smaller increases in risk of pituitary adenoma in individuals who had ever used OCs. Taken together, the results from these distinct data sets suggest a strong positive association between MHT use and risk of pituitary adenoma, and a possible smaller increased risk with use of OCs, though the latter findings were inconsistent between studies. In general, these results do not appear to be substantially affected by recall, reverse causation, or diagnostic bias, although these biases cannot be absolutely excluded. In particular, surveillance bias remains a potential issue despite our sensitivity analyses, given that OCs and, to a lesser extent, MHT, may cause headache, which may lead to increased rates of neuroimaging and therefore increased rates of pituitary adenoma diagnosis among users as compared to nonusers (25, 26).

Previous investigations of OC use and pituitary adenoma risk to our knowledge have been limited to case-control studies. In a case-control study of 9 pituitary adenoma cases and 36 controls, Coulam et al (11) reported a nonstatistically significant association for ever vs never use of OC (odds ratio [OR] = 0.5; 95% CI, 0.1-2.2). In 1980, Teperman et al (12) reported a nonstatistically significant relative risk of 4 for OC use (P = .25) in a case control study of 20 hyperprolactinemia cases with pituitary adenoma and matched appendectomy patients. Maheux et al (10) reported no association for OC use comparing 70 patients with prolactin-secreting adenomas with 3 different control groups. In 1983, the Pituitary Adenoma Study Group case-control analysis of 212 women with prolactin-secreting pituitary adenomas reported an OR of OC use of 1.33 (95% CI, 0.81-2.22) comparing those with pituitary adenoma to matched controls, with no dose-response relationship (9). A study by Shy et al (8) the same year compared patients with pituitary adenoma to matched controls and accounted for the indication for OC prescription. For those who used OCs for birth control, the OR for ever vs never use was 1.3 (95% CI, 0.7-2.6), but when OCs were prescribed for menstrual regulation, the OR was 7.7 (95% CI, 3.7-17.0), providing clear evidence of a protopathic bias due to an undiagnosed adenoma.

More recently, in 2009, a case-control study of 157 patients with pituitary adenoma and 336 controls also found no association between OCs and pituitary adenoma (OR = 0.8; 95% CI, 0.4-1.4) (7). For MHT, that same study reported an OR of 1.5 (95% CI, 0.8-2.6) for ever vs never use, a statistically significant OR of 3.3 (95% CI, 1.3-8.6) for less than 1 year of use but no duration trend (P trend = .6). The authors interpreted this latter finding as hormone use as a consequence of a pituitary tumor, again providing evidence of a protopathic bias (7). We are unaware of other studies of MHT and risk of pituitary adenoma. The use of MHT and risk of other cancers has been studied in detail, however, with evidence that use of estrogen plus progestin, particularly long duration use, is associated with an increased risk of breast cancer (27-30), but no association with bladder cancer (31), lung cancer (32), or renal cell cancer (33). Importantly, estrogen plus progestin MHT has also been associated with a lower risk of endometrial cancer (34), and estrogen-only MHT was associated with lower breast cancer risk in the randomized Women’s Health Initiative (35).

Taken together with the existing literature, the results here are concordant with the overall finding of a possible small increase in risk of pituitary adenoma with prior OC use. In the prospective cohort analysis, we demonstrated a positive association between less than 1 year of OC use duration and pituitary adenoma in the 4-year lagged analysis (but not the 8-y lagged analysis), which was similar to the finding of increased risk in the lagged case-control analysis. The lagged analyses and the analysis among those with recent health care use in the prospective cohort mitigate the possibility of substantial reverse causation or diagnostic bias affecting the results, unlike in prior studies of these associations.

On the other hand, our results suggest, for the first time, a strong positive association between MHT use and pituitary adenoma risk, with evidence of a duration effect. This finding is unlikely to be the result of reverse causation, but we cannot rule out diagnostic bias or confounding by another as-yet unidentified causal risk factor for pituitary adenoma. Nonetheless, the prognostic value of an association between MHT use and incidence of pituitary adenoma in risk-stratifying patients remains.

An association between circulating estrogens and pituitary adenoma incidence is plausible, given convincing evidence that estrogen stimulates lactotroph cells in the anterior pituitary (36-39). Among premenopausal women, prolactinomas arising from the lactotroph cells of the anterior pituitary are the most commonly occurring pituitary adenoma subtype; incidence of this subtype decreases after menopause, and perhaps higher levels of circulating estrogen stimulate the growth of these lesions, increasing the likelihood that they would come to clinical attention (40, 41). The difference in association between OCs and MHT and pituitary adenoma is less easily explained, particularly given that MHT generally contains lower levels of estrogen than OCs. Nevertheless, the possibility of a threshold effect remains, given that the lifetime cumulative estrogen exposure for women taking MHT is substantially higher than that of OC users, particularly since MHT users may have previously used OCs.

Importantly, the effect of calendar time may also explain the minor discrepancies between the 2 separate analyses. The case-control analysis, in general, included cases diagnosed more recently than the prospective cohort analysis. If anything, we would expect that the active hormonal content of OCs may have decreased over time (42), which would not explain why the finding of increased risk of pituitary adenoma with OC use was stronger in the case-control analysis than in the prospective cohort analysis. It is possible, however, that the case-control analysis included more long-duration users, or this difference could simply be due to the higher statistical power of the later analysis.

The 2 separate analyses include complementary strengths and limitations. The prospective cohort analysis included repeat, prospective, highly accurate assessment of OC and MHT use, allowing the calculation of duration and assessment of specific types of MHT, but was limited to self-reported pituitary adenomas. The case-control analysis, on the other hand, included an extremely large number of clinician-diagnosed pituitary adenomas (n > 5000), but had fewer granular data on OC/MHT use and may have included some use that occurred after initial diagnosis, even in the analysis intended to account for this. The fact that we generated overall similar results from 2 completely separate data sets with 2 different methodologies lends credence to our overall findings. Limitations include the lack of data on specific tumor types as well as difficulty identifying the exact date of diagnosis, which were limitations in both data sets. In the case-control analysis, we attempted to identify patients with prolactinoma by using pituitary adenoma cases codiagnosed with hyperprolactinemia, but it remains possible that this may include patients with nonfunctioning (or other functioning) pituitary adenomas who have elevated prolactin due to the stalk effect, in addition to the possibility of undercoding or overcoding in an administrative data set (43). For example, physicians may not be inclined to code hyperprolactinemia systematically for patients with prolactinoma, and may elect to use only the nonspecific code for adenoma. Notably, although the pituitary adenoma cases in the prospective cohorts were self-reported, all participants were nurses, and thus able to provide accurate information regarding medical conditions; the accuracy of recall has previously been confirmed in these cohorts by comparing self-report with medical records for a wide range of medical conditions (14, 44-49).

Conclusion

Compared to never use, current and past MHT use and longer duration of MHT use were positively associated with incidence of pituitary adenoma in 2 independent data sets. OC use was not associated with risk in the prospective cohort analysis and was associated with only mildly increased risk in the case-control analysis. These findings warrant further investigation, including corroboration in past and ongoing randomized trials of MHT, especially those with brain imaging studies.

Abbreviations

Abbreviations
 
• BMI

  body mass index

 
• OC

  oral contraceptive

 
• ICD

  International Classification of Diseases;

 
• MHT

  menopausal hormone therapy

 
• MVHR

  multivariable-adjusted hazard ratio

 
• MVOR

  multivariable-adjusted odds ratio

 
• NHS

  Nurses’ Health Study

 
• NHSII

  Nurses’ Health Study II

 
• OR

  odds ratio

 
• RPDR

  Mass General Brigham Research Patient Data Registry

Acknowledgments

We would like to thank the participants and staff of the Nurses’ Health Study, Nurses’ Health Study II, and Health Professionals Follow-Up Study for their valuable contributions as well as the following state cancer registries for their help: Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Mississippi, Nevada, New Hampshire, New Jersey, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Virginia, Washington, and Wyoming. The authors assume full responsibility for the analyses and interpretation of these data.

Financial Support: This work was supported by the National Institutes of Health (grant Nos. NIH PO1 CA87969, U01 CA167552, UM1 CA186107, U01 CA176726, UM1 CA167552, F30 CA235791 to D.J.C., and R37 HD019938 to U.B.K.).

Additional Information

Disclosures: The authors have nothing to disclose.

Data Availability

All supplementary tables are available in a data repository (21) and all data are available from the investigators on reasonable request, as noted on the NHS website, www.nurseshealthstudy.org.