Prior Oral Contraceptive Use and Longer Term Mortality Outcomes in Women with Suspected Ischemic Heart Disease

Abstract

Background: Previous Women's Ischemia Syndrome Evaluation (WISE) work demonstrated prior oral contraceptive (OC) use was associated with lower coronary artery disease (CAD) in women with suspected ischemia. The association of prior OC use with longer term all-cause and cardiovascular disease (CVD) mortality is unclear.

Materials and Methods: WISE women undergoing coronary angiography for suspected ischemia (enrolled 1996–2001) with prior OC use history and 10-year follow-up data were analyzed. A blinded core laboratory assessed atherosclerotic CAD severity. Kaplan–Meier analyses evaluated prior OC use relative to all-cause and CVD mortality. Cox regression analyses adjusted for baseline differences. Mediation, interaction, and multicollinearity were analyzed.

Results: Our 686 women had a mean age 62.5 ± 9.6 years, multiple cardiac risk factors, and 39% previously used OC. Prior OC users were younger, with less lipid-lowering medication use and lower atherosclerotic CAD severity scores (all p < 0.05). Prior OC use was associated with lower 10-year all-cause (p = 0.007) and CVD mortality (p = 0.019). After adjustment, this was no longer significant (p = 0.77 and p = 0.90, respectively). Atherosclerotic CAD severity score mediated one-third of the observed association. Prior OC use was associated with increased CVD mortality among women with very elevated menopausal systolic blood pressure (SBP).

Conclusions: Unadjusted prior OC use was associated with lower longer-term all-cause and CVD mortality. One-third of this observed effect appears mediated by the atherosclerotic CAD severity score. Prior OC was adversely associated with CVD mortality in women with very elevated menopausal SBP. Additional investigation is needed to understand the potential benefits and harms of prior OC use. Clinical Trial Number: NCT00000554, or https://www.clinicaltrials.gov/ct2/show/NCT00000554

Introduction

Oral contraceptives (OC) are used by 80% of American women, typically during their premenopausal years.¹ Indeed, the utility of OC for prevention of pregnancy is well-recognized, as well as for various other indications, including treatment of polycystic ovarian syndrome, endometriosis, adenomyosis, acne, hirsutism, amenorrhea, menstrual cramps, menstrual migraines, menorrhagia, menstruation-related or fibroid-related anemia, and dysmenorrhea.² However, the longer term benefits and risks associated with OC use, particularly among women with suspected ischemic heart disease (SIHD), are controversial and less understood.³

Current OC use in younger women increases venous thromboembolism¹ and new-onset hypertension (HTN) in previously normotensive women.^1,4 There is also a 1.6-fold relative increased risk of adverse cardiovascular disease (CVD) events, including nonfatal myocardial infarction (MI) and stroke, with current OC use.⁵ Although these short-term CVD risks of current OC use among younger women are well established, animal and human data have postulated that OC¹ use may in fact be preventive for coronary atherosclerosis. However, there have been no studies, to our knowledge, exploring the relationship between prior OC use and longer term outcomes, especially among women with SIHD, also specifically evaluating potential OC mediation of risk.

Analysis of prior OC use in menopausal women with SIHD provides an opportunity to further investigate longer term outcomes in this population. Accordingly, we evaluated the associations of prior OC use with longer term all-cause and CVD mortality, within a densely phenotyped cohort of menopausal women, referred for coronary angiography for SIHD.

Materials and Methods

Database query

As we were interested in exploring longer term associations with prior premenopausal use and mortality outcomes, we only included menopausal women in the original Women's Ischemia Syndrome Evaluation (WISE) cohort of 944 women with suspected SIHD (enrolled 1996–2001).⁶ IRB approval was given and informed consent obtained. Women who were not menopausal (n = 237), missing longer term follow-up and/or mortality data (n = 9), and missing data for reproductive questionnaire variables (n = 22) were excluded from the analysis. After exclusion, our cohort comprised n = 686 women. This represents a larger sample size than the n = 672 in our 2006 OC analysis,⁷ owing to the earlier study's additional exclusion of women missing specific variables of interest.

Reproductive questionnaire

The WISE reproductive questionnaire is an assessment of menopause status that ascertains history of menarche, date of last menstrual period, prior menstrual cycling patterns, prior reproductive events (pregnancy, hysterectomy, and oophorectomy), current and prior perimenopausal symptoms, and active/prior OC or menopausal hormone therapy (HT) use.⁷ Non-OC users were defined as those women who have never used OC. Menopausal HT use was defined as those currently on HT at time of enrollment.

Measurement of obstructive coronary artery disease and atherosclerotic coronary artery disease severity score

Coronary angiograms were analyzed by a WISE core laboratory masked to all other patient data. The analysis included quantitative and qualitative assessments of the presence, severity, and complexity of epicardial coronary artery disease (CAD), as previously published.⁸ Obstructive CAD was defined as ≥50% luminal diameter stenosis in at least one epicardial coronary artery. In addition, we calculated a continuous atherosclerotic CAD severity score to estimate the severity of coronary artery atherosclerosis.⁹ This severity score assigns points according to percent diameter stenosis, which is adjusted for any partial or complete collaterals. More proximal lesions receive a higher weighting factor. The final score is then calculated by summing the individual lesion point totals. The score can range from 5 (no detectable stenosis) to 100 (multiple severe lesions).

Follow-up for all-cause and CVD mortality

All-cause mortality was obtained by a search of the National Death Index. CVD deaths were adjudicated based on death certificates and/or statements of hospital staff or family members, surrounding these deaths. The investigators performing these adjudications had access to selected medical records. Women were followed for CVD outcomes over 6.94 ± 3.39 years (mean ± SD); all-cause and CVD mortality data were available over 10 years.

Survival analyses

Survival analyses, including log-rank testing and Kaplan–Meier estimates, were performed, to compare the outcomes of all-cause and CVD mortality rates among prior OC users versus non-OC users. A base Cox regression model was constructed utilizing prior OC use status and known predictors of mortality as described in existing literature, including self-reported age, history of HTN, history of dyslipidemia, history of diabetes mellitus (DM), tobacco use status, measured blood pressure, lipid panel, glucose, body mass index (BMI), and calculated atherosclerotic CAD severity score. Socioeconomic status (SES) variables of annual income, level of education, and marital status were also included. Variables were defined as given in Table 1.

Table 1.

Demographic and Clinical Attributes of Prior Oral Contraceptive and Nonoral Contraceptive Users

	Prior OC users (n = 269)	Non-OC users (n = 417)	Total (n = 686)	Age-adjusted p-value
Age, years				<0.001
Median (IQR)	57.1 (10.4)	66.6 (12.8)	62.5 (13.6)
Ethnicity, White	236 (87.7%)	329 (78.9%)	565 (82.4%)	<0.001
Marital status				0.07
Never married	4 (1.5%)	23 (5.5%)	27 (3.9%)
Divorced or separated	52 (19.3%)	50 (12.0%)	102 (14.9%)
Widowed	34 (12.6%)	121 (29.1%)	155 (22.6%)
Presently married	169 (62.8%)	216 (51.9%)	385 (56.2%)
Living in a marriage like relationship	10 (3.7%)	6 (1.4%)	16 (2.3%)
Education				0.50
<High school	50 (18.6%)	104 (25.1%)	154 (22.5%)
High school diploma or GED	100 (37.2%)	174 (42.0%)	274 (40.1%)
>High school	119 (44.2%)	136 (32.9%)	255 (37.3%)
Income				0.98
<$20,000	99 (36.9%)	143 (34.8%)	242 (35.6%)
$20,000–$34,000	51 (19.0%)	103 (25.1%)	154 (22.7%)
$35,000–$49,999	47 (17.5%)	66 (16.1%)	113 (16.6%)
$50,000–$99,000	37 (13.8%)	52 (12.7%)	89 (13.1%)
$100,000+	14 (5.2%)	10 (2.4%)	24 (3.5%)
Unknown	20 (7.5%)	37 (9.0%)	57 (8.4%)
Body mass index, kg/me				0.53
Median (IQR)	28.7 (8.0)	28.3 (7.9)	28.5 (7.8)
History of hypertension	155 (57.8%)	267 (64.2%)	422 (61.7%)	0.55
History of diabetes mellitus	58 (21.6%)	119 (28.6%)	177 (25.9%)	0.27
Ever smokers	170 (63.2%)	205 (49.3%)	375 (54.7%)	0.15
History of dyslipidemia	154 (61.1%)	220 (57.0%)	374 (58.6%)	0.035
Lipid-lowering medication use	65 (24.3%)	157 (37.7%)	222 (32.5%)	0.046
Aspirin use	159 (59.3%)	281 (67.5%)	440 (64.3%)	0.60
Beta blocker use	90 (33.6%)	176 (42.4%)	266 (38.9%)	0.19
Diuretic use	80 (29.9%)	136 (32.7%)	216 (31.6%)	0.6
Current hormone therapy	148 (55.4%)	152 (36.6%)	300 (44.0%)	0.026
Duration of OC use, years
Median (IQR)	1.3 (6.5)		1.3 (6.5)
CAD status				0.13
<20% stenosis	101 (37.5%)	114 (27.3%)	215 (31.3%)
20%–49% stenosis	80 (29.7%)	99 (23.7%)	179 (26.1%)
≥50% stenosis	88 (32.7%)	204 (48.9%)	292 (42.6%)
Atherosclerotic CAD Severity Score				0.003
Median (IQR)	7.5 (9.3)	11.5 (20.6)	9.3 (15.5)

Values in bold indicate significant p-values.

CAD, coronary artery disease; GED, graduate equivalency degree; IQR, interquartile range; OC, oral contraceptive.

The final Cox model was constructed based on the variables included in the base model (Table 1) and any additional variables determined by stepwise selection. Stepwise selection was conducted using SAS software (version 9.2; SAS, Cary, NC). As the result of this procedure, menopausal HT and diuretic use were added to both models. Systolic blood pressure (SBP) was also added to the CVD mortality model.

The data were found to be non-normally distributed. However, to evaluate for a linear relationship, the Pearson's correlation coefficient test was applied.

Correlation, interactions, and mediation analyses

Correlation was assessed between each of the quantitative variables in the final Cox models to evaluate for any confounding variables exhibiting multicollinearity or nonlinear relationships with each other. We also explored for any interactions between all variables included in the final regression models. Finally, we conducted mediation analyses as proposed by Valeri and VanderWeele¹⁰ to determine the degree to which the statistical relationship between prior OC use (the predictor) and outcomes is a direct one, or may be accounted for by an intervening variable (mediator). Each variable in the final model was evaluated as a potential mediator. There were no significant interactions between mediator and exposure on either of the mortality outcomes, and thus a simple mediation model without a mediator × exposure interaction was used.

Results

Demographic attributes

Within the cohort, 39% (269) were prior OC users. Table 1 summarizes the demographic attributes, adjusted for age. Compared with non-OC users, prior OC users were younger, more likely to be Caucasian, with higher frequency of dyslipidemia but lower atherosclerotic CAD severity scores than non-OC users (all p < 0.05). The majority had nonobstructive CAD. Prior OC users were less likely to be receiving lipid-lowering therapy but more likely to be receiving current menopausal HT (p < 0.05).

Outcomes

Unadjusted prior OC use was associated with lower 10-year all-cause (p = 0.007) and CVD mortality (p = 0.019), versus the non-OC user group (Fig. 1A, B, respectively). Following adjusted Cox regression modeling, these relationships were no longer statistically significant (p = 0.77 and p = 0.90, respectively). The proportional hazards assumption was met for both the final models. There were no statistically significant relationships between duration of prior OC use and the mortality outcomes. The dimensions of SES analyzed, including annual income, education, and marital status, also did not statistically influence either outcome model, nor did HT.

FIG. 1.

(A) Unadjusted prior OC use was associated with reduced all-cause mortality when compared with no history of OC use (log rank p = 0.007). (B) Unadjusted prior OC use was associated with reduced CVD mortality when compared with no history of OC use (log rank p = 0.019). CVD, cardiovascular disease; OC, oral contraceptive.

Correlation analysis demonstrated significant Pearson correlation coefficients for all quantitative variables included in the final Cox models, including age, BMI, and SBP. However, none of the coefficients were >0.23, and thus, there was no distinguishable multicollinearity. Furthermore, these variables did not appear to exhibit any nonlinear relationships with one another.

Interactions and mediation

Interaction analyses for the all-cause mortality model were not informative. In terms of the CVD mortality model, there was a statistically significant interaction (p = 0.013) between prior OC use and SBP (Fig. 2). For the purposes of acquiring representative hazard ratios for this continuous SBP variable, specific SBP values of 100 and 175 were selected to represent very low and very elevated SBP, respectively. For women with very elevated SBP (i.e., with SBP = 175, hazard ratio (HR) = 2.36, 95% confidence interval [CI, 1.1–5.3], p = 0.036), prior OC use had a harmful association, whereas for women with very low SBP (i.e., with SBP = 100, HR = 0.26 (95% CI [0.07–0.92], p = 0.037), prior OC use showed a protective association with CVD mortality.

FIG. 2.

Predicted probability of CVD mortality based on the final model, including interaction between prior OC use status and SBP. For women with very high SBP (i.e., with SBP = 175, HR = 2.36 [95% CI, 1.1–5.3], p = 0.036), prior OC use had a harmful association, but for women with very low SBP (i.e., with SBP = 100, HR = 0.26 [95% CI, 0.07–0.92], p = 0.037), prior OC use had a protective association (overall p = 0.013 for this interaction analysis). CI, confidence interval; HR, hazard ratio; SBP, systolic blood pressure.

Mediation analyses revealed that the atherosclerotic CAD severity score was a significant mediator of the observed effects of prior OC use on all-cause and CVD mortality. As detailed in Table 2, both the direct effect of prior OC use on each mortality outcome absent, the mediator and the indirect effect of prior OC use by way of the mediator were analyzed. The atherosclerotic CAD severity score was found to mediate 34% of the effect of prior OC use on all-cause mortality and 35% of its effect on CVD mortality.

Table 2.

Mediation Analyses

Mediator	Outcome	Direct effect		Indirect effect		Total effect		Proportion mediated
		HR (95% CI)	p-Value	HR (95% CI)	p-Value	HR (95% CI)	p-Value
CAD	All-cause mortality	0.73 (0.52–1.04)	0.08	0.83 (0.75–0.92)	<0.001	0.61 (0.42–0.87)	0.006	32%
CAD	CVD mortality	0.70 (0.45–1.09)	0.12	0.82 (0.74–0.92)	0.001	0.57 (0.36–0.90)	0.017	29%
CAD severity score	All-cause mortality	0.77 (0.53–1.10)	0.15	0.84 (0.77–0.91)	<0.001	0.64 (0.45–0.92)	0.016	34%
CAD severity score	CVD mortality	0.75 (0.47–1.20)	0.23	0.82 (0.74–0.90)	<0.001	0.61 (0.39–0.97)	0.037	35%

CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio.

Discussion

In this WISE cohort, unadjusted prior OC use is associated with lower longer term CVD and all-cause mortality in women with suspected ischemia, and one-third of this observed effect appears to be mediated by the atherosclerotic CAD severity score. Our findings support the hypothesis that prior OC use may contribute to lower longer term mortality by way of lower coronary atherosclerosis burden. Of note, menopausal HT did not appear to contribute to the longer term outcome adjusted models. Taken together, these new analyses suggest further support for an earlier timing hypothesis,^11–13 in that premenopausal exogenous OC use may have a beneficial influence on vascular health and subsequent CVD mortality, as opposed to later menopausal HT use.

Our results align with those of our previous study, which revealed that prior OC use was an independent predictor of a lower atherosclerotic CAD severity score, even after adjusting for age, DM, triglycerides, low-density lipoprotein cholesterol, smoking, aspirin use, and lipid-lowering medication therapy.⁷ Our findings also support and propose a mechanistic explanation for other observational reports,^14,15 demonstrating prior OC use associated with less CVD events. However, our results conflict with several negative studies of prior OC use and longer term CVD risk.^16–18 Our contrasting results likely reflect the variation in outcomes measured and populations studied. Whereas we focused specifically on mortality among prior OC users in menopause with SIHD, these previous analyses examined MI and ischemic stroke among younger healthier OC users. Our study also had a comparatively longer term follow-up interval.

Of note, inclusion of menopausal HT did not alter our results. Although this finding refutes earlier observational data supporting atherosclerosis and CVD mortality benefits of HT among menopausal women,¹⁹ it is consistent with more recent and rigorous studies, including a large meta-analysis,²⁰ as well as the many randomized menopause HT trials²¹ that have failed to find benefit. Over half of our current menopausal HT users previously used OC, as compared with one-third of the non-OC users. Taken in totality, these data suggest the novel concept that prior observational HT analyses may be confounded by prior OC use contributing to the observed CVD benefit.

There are a variety of pathophysiologic explanations for the influence of exogenous estrogen on vascular health, predominantly studied in animal models, as human data are limited. Estrogen mediates release of heat shock protein 27, which binds scavenger receptors, reducing uptake of cholesterol into macrophages and counteracts mediators of vascular inflammation.²² Estrogen also potentiates endothelium-dependent relaxation by enhancing nitric oxide bioavailability.²³ Among the relatively few human studies, estrogen appears to increase expression of prostacyclin receptors in aortic smooth muscle cells²⁴ and prevent acetylcholine-induced vasospasm in women with nonobstructive CAD.²⁵

Second, our interaction analysis revealed an association of menopausal SBP between prior OC use and CVD mortality. Namely, for our menopausal women with very high SBP, prior OC use had an adverse association with CVD mortality, whereas the presence of very low SBP was associated with a beneficial relationship. These results expand upon existing knowledge regarding the relationship between OC use and HTN. Previous studies have demonstrated an elevated risk of OC-induced HTN among active users, which increases with the potency of the progestin but diminishes after OC cessation.^1,4 In contrast, a recent large study suggested that neither prior OC use nor duration was associated with menopausal blood pressure. This difference likely stemmed from the variation in populations studied. Whereas we focused on postmenopausal women with SIHD, this large cohort incorporated primarily healthy postmenopausal women.²⁶ These investigators also excluded women who had been diagnosed with HTN prior to age 30, as it was unlikely that OC contributed to HTN in these women. Our findings suggest the hypothesis that prior OC-related HTN may identify an “at risk” group of women for future HTN-related CVD mortality. Further work is needed to follow cohorts of younger women, subsequent to diagnosis of OC use and HTN. These results further support the need for more human research regarding the influence of OC use on long-term vascular health.

Strengths and limitations

To our knowledge, this study is the first to explore the relationship between prior OC use and longer term outcomes, especially among women with SIHD. We have also conducted a novel detailed evaluation of potential OC mediation of risk. Some limitations deserve mention. We recognize that owing to the potential bias and confounding introduced by the observational nature of the study, reliable causal inferences cannot be made. As such, our results offer hypotheses to be tested in future investigations. In terms of analysis, although the models were created with and without current HT, we acknowledge that there may be collinearity between current HT and prior OC use, which may be misleading when evaluating for a prior OC use association. Self-report bias associated with the questionnaire data collection method is described. In addition, we did not acquire data regarding specific type of OC used. Our results also may not account for various socioeconomic variables impairing access to OC that could be confounding the reported relationships. Whereas we explored annual income, level of education, and marital status, these data were self-reported. Thus, they may not reflect the subjects' actual SES at the time they were previously using OC.

Conclusions

Unadjusted prior OC use was associated with lower longer term CVD and all-cause mortality, in a prospectively followed densely phenotyped cohort of women with suspected SIHD. One-third of this observed effect appeared to be mediated by atherosclerotic CAD severity. Elevated menopausal SBP was associated with a harmful relationship between prior OC use and CVD mortality. Our findings support the hypothesis that prior OC use may positively influence longer term mortality outcomes, mediated by lower coronary atherosclerosis burden. Elevated menopausal SBP among prior OC users may also identify individuals at risk of future CVD mortality. Additional prospective investigation is needed to understand the potential presence and pathways of these putative benefits and harms of prior OC use.

Authorship Confirmation Statements

L.B. contributed to the conception and design of the work, analysis and interpretation of the work, drafting and critical revision of the work for important intellectual content, gave final approval of the version to be published and agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

C.S., M.L., B.D.J., S.L.B., G.B., V.B., L.S., and E.H. contributed to the analysis and interpretation of the work, critical revision of the work for important intellectual content, gave final approval of the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

S.R. and C.J.P. contributed to the acquisition of data, critical revision of the work for important intellectual content, gave final approval of the version to be published and agrees to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

C.N.B.M. contributed to the acquisition of data, conception and design of the work, analysis and interpretation of the work, critical revision of the work for important intellectual content, gave final approval of the version to be published and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Authors' Disclosure Statement

N.B.M. reports personal fees from iRhythm, other from Sanofi, other from Abbott Diagnostics, during the conduct of the study. E.H. reports grants from NIH/NHLBI, during the conduct of the study; grants from Aastom Biosciences, Amgen, Amorcyte, AstraZeneca, Biocardia, Boehringer Ingelheim, Brigham and Women's Hospital, Capricor, Cytori Therapeutics, Department of Defense, Direct Flow Medical, Duke Clinical Research Institute, East Carolina University, Everyfit, Inc., Gilead, Ionis, Medtronic, Merck & Co., Mesoblast, PCORI, Relypsa, Sanofi Aventis, outside the submitted work. C.J.P. reports grants from NIH/NHLBI, during the conduct of the study; grants from NIH/NCATS, grants from BioCardia BC-14-001-02; Mesoblast, Inc., MSB-MPC-CHF001; Ventrix, Inc.; Athersys, Inc., AMI MultiStem; Verily Life Sciences LLC-Project Baseline OSMB; Ironwood MSB-MPC-CHF00-DMC, Imbria Pharmaceuticals, Inc.; Milestone Pharmaceuticals, Inc.; Caladrius Biosciences, Inc.; Gatorade Trust; and McJunkin Family Foundation, outside the submitted work. V.B. serves on the Executive Steering Committee of the ODYSSEY OUTCOMES trial (Sanofi), as National Coordinator for STRENGTH (Astra Zeneca), DalGene (Dalcor), CLEAR (Esperion), as local site investigator for ORION IV (The Medicines Company) and was a local site investigator for ARTEMIS (Astra Zeneca) and COMPASS (Bayer Healthcare), both completed. All activities are contracted through the University of Alabama at Birmingham. She has served as a consultant for Sanofi. All other authors report no disclosures.

Funding Information

This work was supported by contracts from the National Heart, Lung and Blood Institutes nos. N01-HV-68161, N01-HV-68162, N01-HV-68163, N01-HV-68164, grants U0164829, U01 HL649141, U01 HL649241, K23HL105787, K23HL127262, K23HL125941, T32HL69751, R01 HL090957, 1R03AG032631 from the National Institute on Aging, GCRC grant MO1-RR00425 from the National Center for Research Resources, the National Center for Advancing Translational Sciences Grant UL1TR000124, and grants from the Gustavus and Louis Pfeiffer Research Foundation, Danville, NJ, The Women's Guild of Cedars-Sinai Medical Center, Los Angeles, CA, The Ladies Hospital Aid Society of Western Pennsylvania, Pittsburgh, PA, and QMED, Inc., Laurence Harbor, NJ, the Edythe L. Broad and the Constance Austin Women's Heart Research Fellowships, Cedars-Sinai Medical Center, Los Angeles, California, the Barbra Streisand Women's Cardiovascular Research and Education Program, Cedars-Sinai Medical Center, Los Angeles, The Society for Women's Health Research (SWHR), Washington, DC, the Linda Joy Pollin Women's Heart Health Program, the Erika Glazer Women's Heart Health Project, and the Adelson Family Foundation, Cedars-Sinai Medical Center, Los Angeles, CA.