Hormonal Contraception and Risk of Cardiometabolic Disease in Women with HIV

Abstract

Objective:

We sought to determine the association of hormonal contraception (HC) and cardiometabolic outcomes among women with human immunodeficiency virus (HIV).

Methods:

We included women with HIV aged 18–45 years in clinical care in the Southeastern United States between 1998 and 2018. Oral and injectable HC use was captured from medication records. Our outcomes included incident cardiovascular/thrombotic disease (CVD) (atherosclerosis, hypertension, cerebrovascular disease, thrombosis, and heart failure) and incident metabolic disorders (diabetes, dyslipidemia, obesity, and non-alcoholic steatohepatitis). We excluded women with prevalent conditions. We used multivariable marginal structural models to examine time-varying current and cumulative HC use and cardiometabolic outcomes in separate analyses, adjusting for age, race, smoking, time-varying comorbidities, CD4 cell count, HIV RNA, and antiretroviral use. Women with HC exposure were compared with women without HC exposure.

Results:

Among the 710 women included, 201 women (28%) used HC. CVD analyses included 603 women without prevalent CVD and 93 incident events; metabolic analyses included 365 women without prevalent metabolic disease and 150 incident events. Current and cumulative oral HC use was associated with increased odds of CVD, though this was not statistically significant (adjusted odds ratio [aOR] = 2.08, [95% confidence interval (CI): 0.80–5.43] and aOR = 1.24 [95% CI: 0.96–1.60] per year of use, respectively). Oral HC was not associated with risk of incident metabolic disorders. Depot medroxyprogesterone acetate (DMPA) was not associated with risk of incident CVD. Current and cumulative DMPA use was significantly associated with decreased odds of incident metabolic disorders (aOR = 0.48 [95% CI: 0.23, 1.00] and aOR = 0.65 [95% CI: 0.42–1.00] per year of use, respectively).

Conclusion:

Our results suggest that cardiovascular risk should be considered when selecting contraception for women with HIV.

Introduction

Women make up nearly one-quarter of all people living with human immunodeficiency virus (HIV) in the United States and >50% of people with HIV worldwide.¹ Globally, 15% of all women with HIV are young women of reproductive age between 15 and 24 years.² Due to effective antiretroviral therapy (ART), the burden of noncommunicable disease morbidity and mortality has increased among people living with HIV (PLWH), making prevention and management of comorbidities extremely important, including among women with HIV.^3,4 By 2030, it is estimated that 73% of PLWH will be over the age of 50.⁵ In the United States, older women are more likely to be diagnosed with HIV than men at the same age, increasing the risk of comorbidities among aging women with HIV.⁶

However, studies have also shown that women with HIV are more likely to experience comorbidities at an earlier age.⁷ Women with HIV are particularly affected by cardiovascular disease compared with women without HIV and younger women with HIV, particularly those under the age of 50, are especially susceptible to cardiovascular disease.^8

–11 Studies have shown that women with HIV are more likely to be diagnosed with hypertension, dyslipidemia, diabetes, and obesity compared with women without HIV and that these comorbidities occur at an earlier age than HIV negative women.^12

–15 Women with HIV remain underrepresented in research and it is important to identify unique female-specific risk factors for the development of comorbidities among women with HIV.^16,17

Hormonal contraception (HC) has been associated with adverse cardiometabolic effects among people without HIV.^18,19 When oral hormonal contraceptives were first brought to the market, they contained high levels of estrogen and progestin, leading to reports of arterial and venous thrombosis occurring among young women.²⁰ Over time, the dose of estrogen and progestin in oral contraceptives has decreased, which has also led to a reduction in side effects.²¹ Nevertheless, some adverse cardiovascular effects persist, including thrombotic stroke and myocardial infarction.²² Studies have also shown that HC affects the lipid profile and may increase the risk of diabetes mellitus; however, the data are conflicting and depend on the type and dose of HC.^23

–26

A few studies have examined whether HC is associated with cardiometabolic disease in women with HIV.²⁷ Prior studies are limited by small sample size and older time period.^24,28 In this retrospective cohort study of women with HIV, we hypothesized that current or prior exposure to HC is associated with an increased incidence of cardiometabolic disease compared with no exposure to HC.

Methods

Study setting

We conducted a retrospective observational cohort study among women with HIV receiving care at the Vanderbilt Comprehensive Care Clinic (VCCC), an outpatient clinic affiliated with Vanderbilt University Medical Center in Nashville, Tennessee that serves as a medical home for PLWH. The VCCC serves nearly 4000 patients annually, of whom 21% are female. Routine follow-up care for patients occurs every 3–6 months. The VCCC serves as the primary care clinic for the majority of its patients. It provides women's health services, including cervical cancer screening, contraception, prenatal services, and obstetrics care. All laboratory and clinical data were systematically extracted and validated from the electronic medical record by research staff. Demographic, clinical, medication (including HC use), height and weight, and laboratory data were included.

Clinical diagnoses including cardiovascular and metabolic diseases were abstracted from diagnoses recorded in the chart and validated by medical research staff. HC exposure data were extracted and validated from medication lists in the electronic medical records by research staff, which undergo reconciliation by physicians at each clinic visit. When a medication was removed from the medication list, listed as inactive, or if the duration of exposure time expired (e.g., for injectable contraception), the exposure was considered complete.

Pharmacy refill data were not available. Data regarding non-hormonal forms of birth control including condoms were not available for this study. The study was approved by the Institutional Review Board of Vanderbilt University Medical Center.

Study population

Adult (age ≥18 years) women with HIV <45 years old at clinic entry who enrolled in care between January 1, 1998, and December 31, 2018 and who had at least two clinic visits in their first year of care were included in this study. Transgender people, women with breast or ovarian cancer, hysterectomy, or bilateral tubal ligation at or before clinic entry, and women with no lab data were excluded. These exclusion criteria were chosen to identify women that were likely to be prescribed HC (i.e., not receiving hormonal therapy for another reason or did not have a contraindication to HC).

Observation time started at the first non-pregnant visit (baseline), and women were followed until cardiovascular/thrombotic disease (CVD) or metabolic outcome, last clinic visit, age 45 years, death, or December 31, 2018, whichever occurred first. Person-time during pregnancy and gaps in care >12 months were censored as they were not exposed to HC during these periods. There was no minimum follow-up time required for inclusion. Baseline demographic and clinical characteristics were compared across HC status (any exposure during follow-up vs. no exposure) using rank sum and chi-square tests for continuous and categorical variables, respectively.

Variables

The primary exposures of interest were time-varying and cumulative use of HC during clinic follow-up, which included oral, transdermal, implant, and injectable contraceptives (depot-medroxyprogesterone acetate or DMPA). Intrauterine devices were not considered HC given their lack of systemic effects.²⁹ Data on HC use before clinic entry are not routinely collected in clinical care and were not available. Given their predominance as HC used in the clinic and differing estrogen and progesterone components, we focused on oral contraception and DMPA exposures.

All observation time was divided into 28-day months with time-varying current hormonal contraceptive exposure in a given month defined by any use during the month. Time-updated cumulative exposure reflected total duration of HC exposure during clinic follow-up, up to a given month.

The primary outcomes included incident clinical diagnoses of CVD (including diagnoses of hypertension, atherosclerotic disease [including coronary artery disease, myocardial infarction, and peripheral vascular disease], heart failure, and deep venous thrombosis as defined by clinical diagnoses included in patient's medical chart problem list or medical history) and metabolic disorders (defined by clinical diagnoses or laboratory parameters for diabetes, dyslipidemia, obesity, and non-alcoholic steatohepatitis).

Diabetes was indicated by either a clinical diagnosis or a measured hemoglobin A1c >6.5%; dyslipidemia was indicated by either a clinical diagnosis or a measured low-density lipoprotein >160 mg/dL or a total cholesterol >200 mg/dL; obesity was defined as a body-mass index (BMI) ≥30 kg/m²; hypertension was indicated by a clinical diagnosis; and non-alcoholic steatohepatitis was indicated by a clinical diagnosis in the medical record.

Analysis

We used multivariable marginal structural models with inverse probability weights (IPW) to estimate the effect of HC use on cardiometabolic outcomes.³⁰ Separate analyses were performed for each outcome (cardiovascular disorders and metabolic disorders), and only the time to first outcome in each class was considered (i.e., follow-up for the cardiovascular disorders analysis ended at the occurrence of first cardiovascular disorder). Women with a prevalent condition of any of the cardiovascular/thrombotic disorders listed earlier were excluded from the analyses looking at incident cardiovascular/thrombotic disorders, and women with a prevalent condition of any of the metabolic disorders listed earlier were excluded from analyses looking at incident metabolic disorders.

For each outcome, we fit two marginal structural models: one where the time to the first occurrence of the potential outcome was modeled based on current hormonal contraceptive use and a second where the time to the first occurrence of the potential outcome was modeled using cumulative (since baseline) hormonal contraceptive use. Both marginal structural models were discrete failure time models using a logit link function; hence, results were summarized as odds ratios (ORs) (i.e., given the event has not yet occurred, the ORs is the relative odds of the event occurring in the next month).

When the frequency of events is low (<10%), these ORs approximate hazard ratios from Cox proportional hazards model; but since in our setting event rates are >10%, we interpret results as ORs. Marginal structural models with IPWs are used to address time-varying confounders that may be on the causal pathway between use of hormonal contraceptives and the outcomes of interest; for cardiovascular events, metabolic disorders are potential time-varying confounders, and for the metabolic events, cardiovascular events are potential time-varying confounders. Time-varying confounding is addressed by including potential confounders in time-varying IPW.

To fit the marginal structural models, IPW were first calculated as the cumulative products of the estimated inverse probabilities of women following the HC exposure pattern that they followed and not being lost to follow-up (LTFU). HC exposure in a given month was defined as either no exposure, oral HC exposure, or DMPA exposure and was modeled using multinomial logistic regression to obtain predicted probabilities of each HC exposure.

LTFU was defined as no visit within 1 year of December 31, 2018, and was modeled using pooled logistic regression to obtain predicted probabilities of not being LTFU.³¹ The models predicting HC and LTFU included the potential confounding variables time since baseline, 1-month lagged HC exposure status, baseline age, race (white vs. non-white), smoking status (ever vs. never), and time-varying CD4 cell count (cell/μL, square-root transformed), time-varying HIV RNA (copies/mL, log-10 transformed), time-varying cardiometabolic comorbidity status (any prior cardiovascular/thrombotic disorder for the metabolic analyses and any metabolic disorder for the cardiovascular analyses), and time-varying ART use by class (including no ART, protease inhibitor-based ART use, integrase strand transferase inhibitor-based ART use, or other ART use).

If a laboratory measurement was not obtained in a given month, the most recent previous lab measurement was carried forward up until at most 12 months at which time patient follow-up was censored. ART exposure was defined as ≥15 days of exposure in a given month. If multiple classes of ART were used during the month, the class having the longest exposure was assigned to that month. Time since baseline and age were included in models with restricted cubic splines using four knots to relax assumptions of linearity.

Weights were stabilized (i.e., prevented from becoming extremely large) by including in the numerator of the weights the product of the cumulative predicted probabilities of hormonal contraceptive exposure and not being LTFU at each month estimated with models that included a subset of baseline and time-invariant covariates (i.e., baseline age, baseline CD4 cell count, race, smoking status, and prevalent metabolic disorder for the cardiovascular/thrombosis diseases outcome and prevalent CVD for the metabolic disorder outcome) and the previous month's HC exposure.

Baseline CD4 cell count was the CD4 count closest to the date of the first non-pregnant visit within a window of −180 to 7 days. Prevalent cardiometabolic disease was any diagnosis that occurred before or within 30 days after the date of first non-pregnant visit. The stabilized IPW were trimmed at the 5th and 95th percentiles.

Next, weighted pooled logistic regression models with robust standard errors were fit to estimate the ORs for the outcomes of interest in separate analyses (incident CVDs or metabolic disorders). Models included time since baseline (expanded using restricted cubic splines with four knots), HC exposure (examined separately as either time-varying current use or time-varying cumulative time on oral HC and DMPA), and the same baseline covariates that were included in the numerator of the stabilized weights.

A secondary set of analyses was done in which HC exposure was dichotomized as any exposure versus none. Weighting models for HC included the same variables as before and modeled the probability of exposure to any HC using binary logistic regression.

All analyses were conducted in R version 3.5.2, and analysis code is available online at https://biostat.app.vumc.org/ArchivedAnalyses. All p-values are two sided.

Results

Our study included 710 women between the ages of 18–44 years who met inclusion criteria, of whom 509 did not use any HC and 201 (28%) used some form of HC during follow-up. The most common forms of HC were oral contraception (103/201, 51.2%) and DMPA (134/201, 66.7%). The median follow-up time, excluding time during pregnancies, was 3.6 years (interquartile range [IQR]: 1.2, 6.9 years). The median total time on HC was 1.25 years (IQR: 0.56, 2.82 years) among women who used HC.

Baseline demographic and clinical characteristics of women with and without any HC use are shown in Table 1. Women with any contraception use were younger at baseline, had a higher baseline CD4 cell count, and were less likely to have co-infection with hepatitis C virus at any point during follow-up. Women who received HC had a higher median duration of clinic follow-up and were less likely to be LTFU. Prevalent metabolic disorders were present at clinic entry in close to half of all women with and without HC exposure during follow-up. CVD was prevalent in 15% of women, regardless of HC use status.

Table 1.

Baseline Characteristics of Women with None or Any Use of Hormonal Contraception During Follow-Up

	Women with no HC use (n = 509)	Women with any HC use (n = 201)	p ^a
Age at baseline^b in years, median (IQR)	32 (27–39)	27 (24–32)	<0.001
Race, n (%)			0.67
White	181 (36)	71 (35)
Black	294 (58)	118 (59)
Other	34 (7)	12 (6)
Hispanic, n (%)	27 (5)	14 (7)	0.39
Year of clinic entry, median (IQR)	2007 (2002–2012)	2006 (2001–2011)	0.12
CD4 cell count (cells/μL) at baseline^b, median (IQR)	432 (197–653)	532 (300–749)	<0.001
HIV-1 RNA log₁₀ at baseline,^b median (IQR)	3.6 (2.6–4.7)	3.1 (1.7–4.5)	0.02
Hepatitis C virus infection ever	63 (12)	13 (6)	0.02
Ever smoker	203 (40)	96 (48)	0.055
Prevalent CVD^c at clinic entry	76 (15)	31 (15)	0.87
Prevalent metabolic disorder^d at clinic entry	242 (48)	103 (51)	0.37
Any PI exposure during follow-up^e	304 (60)	148 (74)	<0.001
Any INSTI exposure during follow-up^e	142 (28)	79 (39)	0.003
Any other ART exposure during follow-up^e	152 (30)	80 (40)	0.001
Months on any contraception	0	15 (6.7–33.8)	<0.001
Months on oral contraception	0	0.8 (0–10.1)	<0.001
Months on DMPA	0	5.6 (0–23.9)	<0.001
Duration of follow-up in years, median (IQR)	3.0 (1.1–6.0)	7.7 (4.0–12.0)	<0.001
Censored follow-up time in years, median (IQR)^f	0.0 (0.0–0.9)	0.2 (0.0–1.5)	0.003
Lost to follow-up, n (%)	210 (42)	61 (30)	0.007

Women with and without any hormonal contraception use compared using Wilcoxon and Pearson tests for continuous and categorical variables, respectively.

Baseline is first non-pregnant visit.

CVDs included myocardial infarction, coronary artery disease, ischemic heart disease, peripheral vascular disease, transient ischemic attack, cerebrovascular disease, hypertension, heart failure, deep vein thrombosis and pulmonary embolism.

Metabolic disorders included diabetes mellitus, dyslipidemia, non-alcoholic fatty liver disease, obesity.

ART use during follow-up includes person-time of exposure to ART regimens including a PI, ART regimens including an INSTI, or other regimens not containing either a PI or INSTI. ART use descriptions by class exposure are not mutually exclusive as women may have received multiple regimens over follow-up.

Censored time during follow-up includes time censored for pregnancies or gaps in care >12 months during which time a woman would not be at risk for or could have confirmed exposure to hormonal contraception.

ART, antiretroviral therapy; CVD, cardiovascular/thrombotic disease; DMPA, depot medroxyprogesterone acetate; HC, hormonal contraception; HIV, human immunodeficiency virus; INSTI, integrase strand transfer inhibitor; IQR, interquartile range; PI, protease inhibitor.

Table 2 shows the specific prevalent and incident cardiometabolic disease events in the study. There were 93 incident CVD events that occurred, the most common of which was hypertension. There were 150 incident metabolic events, the most common of which were dyslipidemia and obesity.

Table 2.

Prevalent and Incident Cardiovascular/Thrombotic Disease and Metabolic Disorder Diagnoses^a Among Women With and Without Any Hormonal Contraception Use During Follow-Up

	Prevalent diagnoses		Incident diagnoses
	Women with no HC use (n = 509)	Women with any HC use (n = 201)	Women with no HC use (n = 509)	Women with any HC use (n = 201)
All cardiovascular disease and thrombosis^b events, n	76	31	54	39
Hypertension, n (%)	70 (92)	28 (90)	49 (91)	33 (85)
Thrombosis,^a n (%)	2 (3)	3 (10)	4 (7)	4 (10)
Cerebrovascular disease, n (%)	3 (4)	1 (3)	0	2 (5)
Coronary artery disease, n (%)	3 (4)	0	1 (2)	0 (0)
All metabolic disorder events, n	242	103	97	53
Obesity, n (%)	155 (64)	66 (64)	38 (39)	20 (38)
Dyslipidemia, n (%)	78 (32)	38 (37)	57 (59)	32 (60)
Diabetes, n (%)	16 (7)	2 (2)	2 (2)	1 (2)
NAFLD, n (%)	3 (1)	0 (0)	1 (1)	0 (0)

Diagnoses are not mutually exclusive. Incident diagnoses include only the first of the potential cardiovascular/thrombotic or metabolic disorder diagnoses.

Thrombosis included deep vein thrombosis and pulmonary embolism. There were no prevalent nor incident diagnoses of myocardial infarction, peripheral vascular disease, nor heart failure among all women with or without hormonal contraception exposure.

NAFLD, non-alcoholic fatty liver disease.

Of the 710 women included, 107 had prevalent CVD at baseline and were excluded from analyses for incident CVD events. Of the 603 included women, 93 had incident cardiovascular/thrombotic events. Compared with no HC use, current use of oral HC was associated with a twofold increase in the odds of developing CVD during the next month in adjusted analyses (adjusted OR [aOR] = 2.08, 95% confidence interval [CI]: 0.80–5.43; Fig. 1a), though this association did not meet statistical significance.

FIG. 1.

aORs for time-varying current (a) and cumulative (b) HC use and incident CVD events (n = 117 events). ^aMultivariable marginal structural model comparing oral contraception and DMPA use to no hormonal contraception use. Model directly adjusted for race, age at first non-pregnant visit, smoking status, CD4 cell count at first non-pregnant visit, and prevalent metabolic disorder; analyses also adjusted for these variables as well as time-varying CD4 cell count, HIV RNA, ART, and incident metabolic disorders through the use of inverse-probability weights. ^bMarginal structural logistic regression model comparing any hormonal contraception use to no hormonal contraception use. Model directly adjusted for race, age at first non-pregnant visit, smoking status, CD4 cell count at first non-pregnant visit, and prevalent metabolic disorder; analyses also adjusted for these variables as well as time-varying CD4 cell count, HIV RNA, ART, and incident metabolic disorders through the use of inverse-probability weights. aOR, adjusted odds ratio; CI, confidence interval; CVD, cardiovascular/thrombotic disease; DMPA, depot-medroxyprogesterone acetate; HC, hormonal contraception.

Similarly, cumulative use of oral HC was associated with increased odds of CVD (aOR for every additional month of oral HC 1.24, 95% CI: 0.96–1.60, Fig. 1b). Current and cumulative use of DMPA was not associated with increased odds of CVD compared to no contraception. Results were similar in secondary analyses in which HC exposure was dichotomized to any versus none.

There were 345 women who had a prevalent metabolic disorder at baseline who were excluded from the metabolic disorder analyses. The metabolic analyses included 365 women and 150 incident events. In adjusted analyses, neither current nor cumulative use of oral HC showed increased odds of a metabolic disorder (Fig. 2). Conversely, current and cumulative use of DMPA showed decreased odds of metabolic disease (Fig. 2). Results showed no statistically significant association in secondary analyses in which HC exposure was dichotomized to any versus none.

FIG. 2.

aORs for time-varying current (a) and cumulative (b) HC use and incident metabolic events (n = 172 events). ^aMultivariable marginal structural model comparing oral contraception and DMPA use with no hormonal contraception use. Model directly adjusted for race, age at first non-pregnant visit, smoking status, CD4 cell count at first non-pregnant visit, and prevalent cardiovascular disorder; analyses also adjusted for these variables as well as time-varying CD4 cell count, HIV RNA, ART, and incident cardiovascular disorders through the use of inverse-probability weights. ^bMarginal structural logistic regression model comparing any hormonal contraception use with no hormonal contraception use. Model directly adjusted for race, age at first non-pregnant visit, smoking status, CD4 cell count at first non-pregnant visit, and prevalent cardiovascular disorder; analyses also adjusted for these variables as well as time-varying CD4 cell count, HIV RNA, ART, and incident cardiovascular disorders through the use of inverse-probability weights.

Discussion

Our study suggests that current and cumulative use of oral HC is associated with a trend in increased odds of incident cardiovascular disease in women with HIV, though this did not meet statistical significance. Current and cumulative DMPA use was associated with a decreased risk of metabolic disease. These findings support the need to take an individual's risk of cardiometabolic disease into consideration when selecting a hormonal contraceptive method for women with HIV.

Morbidity and mortality from cardiovascular disease is a serious health threat among people with HIV and particularly women with HIV. Previous studies have shown that women with HIV are at an increased risk of cardiovascular disease compared with women without HIV. When evaluating the incidence of cardiovascular disease among women with HIV, studies have shown that women <50 years of age are particularly prone to cardiovascular disease compared with women without HIV <50 years of age and compared with men living with HIV <50 years of age.^9,32,33

Importantly, women under the age of 50 years are also likely to be prescribed HC. Use of oral HC appeared to be associated with an increased incidence of cardiovascular disease, particularly hypertension, in our cohort of women with HIV; however, it was not statistically significant. Combined oral contraceptives have been associated with an increase in blood pressure and may increase the systolic blood pressure by as much as 8 mmHg.³⁴ However, this increase is small and the development of hypertension attributed to oral HC use is generally uncommon.³⁵

Our prior analysis found that the incidence of cardiovascular disease was increased threefold in women with HIV on HC compared with women with HIV not on HC; however, our earlier result was not statistically significant due to smaller sample size.²⁷ Individual cardiovascular risk should be considered when prescribing HC to women living with HIV.

The interaction between HIV, HC, and metabolic disorders is complex as both HIV and HC have been associated with the development of metabolic disorders.³⁶ However, the interaction is complicated due to various formulations of HC. For instance, prior studies have demonstrated that oral hormonal contraceptives have an effect on lipids and lipoprotein metabolism; however, these effects can vary due to the different doses and formulations of estrogen and progestin used in combined oral contraceptives and the use of progestin-only oral contraceptives.^37,38

The effects of oral HC on glucose metabolism are also conflicting and difficult to study; however, several studies have shown that oral contraceptives affect glucose metabolism, leading to pre-diabetes and diabetes mellitus.^39
–41 Among women with HIV, progestin-only hormonal contraceptives have been associated with lower high-density lipoprotein (HDL) levels as well as increased insulin resistance whereas combined hormonal contraceptives were associated with higher HDL levels.²⁴

Similarly, data examining metabolic outcomes and DMPA use are also complex. Our study showed that current and cumulative DMPA use was associated with a decreased risk of incident metabolic disorder. While many women often report weight gain with DMPA use, data on this are conflicting.^42

–45 One study evaluated weight gain among obese and non-obese women who used DMPA and found that non-obese women's BMI did not change while on DMPA whereas there was an increase in BMI among obese women.⁴⁴

In our study, there was a high baseline prevalence of obesity (>30%) and women with obesity at baseline were excluded from analyses of incident metabolic disorder. With regards to other metabolic diseases, studies of the effects of DMPA on insulin resistance have also shown conflicting results, likely due to small sample size, short follow-up, as well as not taking into consideration the effect of obesity.⁴⁶ A recent study showed that DMPA use among women with normal BMI resulted in an increase in insulin resistance though not incident diabetes; however, DMPA use in obese women was associated with both insulin resistance and incident diabetes.⁴⁵

It is possible that our findings of decreased metabolic disorders among users of DMPA are because we excluded women with baseline metabolic disorders, including obesity and diabetes mellitus. Given that both HIV and HC have effects on metabolic dysregulation, it is difficult to determine how HC may potentiate the development of metabolic disorders in women with HIV, which likely accounts for our results, as we were unable to look at individual metabolic events and specific formulations and dosages of hormonal contraceptive methods.

Based on prior studies, uptake of HC is low among women with HIV. There have been several theories for this, including concern regarding drug-drug interactions between HC and ART, lack of provider comfort in prescribing HC, and lack of access to reliable contraception. Our study shows that while current and cumulative use of oral HC was associated with a trend in increased odds of incident cardiovascular disease in women with HIV, this did not meet statistical significance and the overall risk is low among women with HIV. Therefore, use of HC should not be limited among women with HIV. Our study supports offering all women with HIV effective HC and that individual cardiometabolic risk should be considered in provider-patient discussions.

Our study is strengthened by a large and diverse cohort as well as 20 years of longitudinal data. However, there are limitations to consider. First, this was a single-site study in Nashville, Tennessee, and may not be generalizable to other populations of women with HIV. Further, the single clinic source limited the total number of women included in our cohort, affecting our power and ability to measure statistically significant differences.

Second, due to the observational nature of the study, there could have been unmeasured confounders; as women who use HC differ from women who do not use HC, and among those who choose HC there are differences among women who choose DMPA versus oral contraceptives. Further, there may be selection bias among providers as they may prescribe HC more frequently to women at lower risk of cardiometabolic disease. To account for this, marginal structural models included time-varying covariates such as CD4 count, HIV-1 RNA, ART use, and other cardiometabolic diseases, which were both potential confounders and suspected to be on the causal pathway.

Marginal structural models use inverse probability weighting to create a pseudo-population that is balanced on available covariates. The duration of follow-up was longer for women with any HC use compared with no HC use, which could be explained by the fact that women with a longer follow-up time are more likely to receive HC and women receiving HC may be more likely to undergo follow-up in clinic. Women engaged in follow-up would also be more likely to experience an observed cardiometabolic event.

Our primary analysis limited potential bias by unequal follow-up time, as it did not compare ever versus never use of hormonal contraceptives, but rather allowed person-time during a given month to be specific for HC exposure and permitted study subjects to contribute follow-up time in both the exposed and unexposed states. Third, there was no HIV-negative control group and so we were unable to determine whether the association between HC use and cardiometabolic disorders would also have been seen in HIV-negative women.

Fourth, the medication history was determined by prescriptions in the electronic health record, and we were unable to verify medication adherence. DMPA users were less likely to be incorrectly classified as active users since DMPA administration requires a clinic visit. However, it is possible that some women using oral contraceptives were not adherent and were mislabeled as active users. Similarly, clinical data before clinic entry were not available and thus we were unable to determine the full history of women's HC use or other potential contributors to cardiometabolic risk, including cumulative HIV RNA.

Fifth, due to low numbers of individual events, outcomes were grouped together in larger categories instead of looking at individual diagnoses, such as coronary artery disease or diabetes mellitus. This could affect findings since we included metabolic disorders as a whole rather than individual metabolic disorders. Further, due to the low numbers of individual events, we were unable to categorize the cardiovascular events, particularly hypertension, by severity.

Lastly, as this study only included eligible to receive HC, we excluded follow-up time during pregnancy and gaps in medical care. The impact of pregnancy and care interruptions on cardiometabolic outcomes among women with HIV could not be assessed and is deserving of future, dedicated investigations.

In conclusion, HC, particularly oral contraception, appeared to be associated with an increased incidence of cardiovascular disease in women with HIV. When discussing contraceptive options for women with HIV, cardiovascular risk factors should be considered. Similarly, providers may consider current and previous HC use when identifying women living with HIV for interventions to reduce cardiometabolic risk, including smoking cessation, exercise, or statin therapies. Further studies evaluating the risk of metabolic disease with DMPA use are needed, as well as studies looking at incidence of cardiometabolic disorders with other forms of contraceptives.

Footnotes

Acknowledgments

The authors recognize the efforts of Daniel Rasbach, Paul No, Henry Heaton, James Logan, Asghar Kheshti, Carmen Bofill, William Ben Rogers, and Fernanda Maruri for their contributions toward the collection and validation of cardiometabolic disease diagnoses from medical records. They also recognize the tireless dedication to women's health at the VCCC by Ms. Beverly Byram, FNP, and other providers over the years.

Authors' Contributions

J.N.: conceptualization, methodology, funding acquisition, and writing—original draft. C.A.J.: methodology, formal analysis, and writing—review and editing. M.B.: conceptualization, investigation. M.T.: conceptualization, investigation, and data curation. A.B.: methodology, formal analysis. J.N.: investigation, data curation. I.R.: investigation, data curation. D.H.: investigation, supervision. J.R.K.: conceptualization, methodology, and writing—review and editing. T.R.S.: conceptualization, methodology, supervision, funding acquisition, resources, and writing—review and editing. B.E.S.: conceptualization, methodology, resources, supervision, and writing—review and editing. J.L.C.: conceptualization, methodology, funding acquisition, supervision, project administration, and writing—review and editing.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This study was supported by grant funding by the National Institutes of Health: K23 AI120875 (J.L.C., C.A.J.), P30AI110527 (T.R.S., B.E.S., M.T.); UL1TR002243 (A.B.).

References

CDC. HIV and Women. 2020. Available from: https://www.cdc.gov/hiv/group/gender/women/index.html [Last accessed: August 27, 2020 ].

UNAIDS. The Gap Report; 2014; p. 135.

Palella

, Baker

, Moorman

, et al. Mortality in the highly active antiretroviral therapy era: Changing causes of death and disease in the HIV outpatient study. J Acquir Immune Defic Syndr, 2006; 43(1):27–34; doi: 10.1097/01.qai.0000233310.90484.16

Smith

, Ryom

, Weber

, et al. Trends in underlying causes of death in people with HIV from 1999 to 2011 (D:A:D): A multicohort collaboration. Lancet, 2014; 384(9939):241–248; doi: 10.1016/S0140-6736(14)60604-8

Hsue

, Waters

. Time to recognize HIV infection as a major cardiovascular risk factor. Circulation, 2018; 138(11):1113–1115; doi: 10.1161/CIRCULATIONAHA.118.036211

CDC. HIV and Women: HIV Diagnoses. 2021. Available from: https://www.cdc.gov/hiv/group/gender/women/diagnoses.html [Last accessed: June 21, 2021 ].

Guaraldi

, Orlando

, Zona

, et al. Premature age-related comorbidities among HIV-infected persons compared with the general population. Clin Infect Dis, 2011; 53(11):1120–1126; doi: 10.1093/cid/cir627

Chow

, Regan

, Feske

, et al. Comparison of ischemic stroke incidence in HIV-infected and non-HIV-infected patients in a U.S. health care system. J Acquir Immune Defic Syndr, 2012; 60(4):351–358; doi: 10.1097/QAI.0b013e31825c7f24

Triant

, Lee

, Hadigan

, et al. Increased acute myocardial infarction rates and cardiovascular risk factors among patients with human immunodeficiency virus disease. J Clin Endocrinol Metab, 2007; 92(7):2506–2512; doi: 10.1210/jc.2006-2190

10.

Tate

, Willig

, et al. HIV infection and obesity: Where did all the wasting go?. Antivir Ther, 2012; 17(7):1281–1289; doi: 10.3851/IMP2348

11.

Crum-Cianflone

, Krause

, Wessman

, et al. Fatty liver disease is associated with underlying cardiovascular disease in HIV-infected persons(*). HIV Med, 2011; 12(8):463–471; doi: 10.1111/j.1468-1293.2010.00904.x

12.

Birabaharan

, Strunk

, Martin

TCS

. Burden of hypertension, diabetes, cardiovascular disease, and lung disease among women living with human immunodeficiency virus (HIV) in the United States. Clin Infect Dis, 2021; 73(1):169–170; doi: 10.1093/cid/ciaa1240

13.

Collins

, Sheth

, Mehta

, et al. The prevalence and burden of non-AIDS comorbidities among women living with or at risk for human immunodeficiency virus infection in the United States. Clin Infect Dis, 2021; 72(8):1301–1311; doi: 10.1093/cid/ciaa204

14.

Tien

, Schneider

, Cox

, et al. Association of HIV infection with incident diabetes mellitus: Impact of using hemoglobin A1C as a criterion for diabetes. J Acquir Immune Defic Syndr, 2012; 61(3):334–340; doi: 10.1097/QAI.0b013e31826bfc32

15.

Bares

, Smeaton

, Xu

, et al. HIV-infected women gain more weight than HIV-infected men following the initiation of antiretroviral therapy. J Womens Health (Larchmt), 2018; 27(9):1162–1169; doi: 10.1089/jwh.2017.6717

16.

Campbell

, Cowlings

, Tholanah

, et al. A community call to action to prioritize inclusion and enrollment of women in HIV cure-related research. J Acquir Immune Defic Syndr, 2022; 91(5):e12–e14; doi: 10.1097/QAI.0000000000003084

17.

Grewe

, Ma

, Gilbertson

, et al. Women in HIV cure research: Multilevel interventions to improve sex equity in recruitment. J Virus Erad, 2016; 2(1):49–51.

18.

Sun

, Peipert

, Zhao

, et al. Trends in contraceptive use among women with human immunodeficiency virus. Obst Gynecol, 2012; 120(4):783–790; doi: 10.1097/AOG.0b013e318269c8bb

19.

Ferreira

JRD

, Aleluia

, Figueiredo

CVB

, et al. Evaluation of cardiometabolic parameters among obese women using oral contraceptives. Front Endocrinol, 2017; 8; doi: 10.3389/fendo.2017.00256

20.

Vessey

. Oral contraceptives and thromboembolic disease. Am Heart J, 1969; 77(2):153–157; doi: 10.1016/0002-8703(69)90344-5

21.

Dhont

. History of oral contraception. Eur J Contracept Reprod Health Care, 2010; 15(sup2):S12–S18; doi: 10.3109/13625187.2010.513071

22.

Lidegaard Ø, Løkkegaard

, Svendsen

, et al. Hormonal contraception and risk of venous thromboembolism: National follow-up study. Br Med J, 2009; 339(7720):557–560.

23.

Kahn

, Curtis

, Marchbanks

. Effects of injectable or implantable progestin-only contraceptives on insulin-glucose metabolism and diabetes risk. Diabetes Care, 2003; 26(1):216–225.

24.

Womack

, Scherzer

, Cole

, et al. Hormonal contraception and metabolic outcomes in women with or at risk for HIV infection. J Acquir Immune Defic Syndr, 2009; 52(5):581–587.

25.

International Collaborative Post-Marketing Surveillance of Norplant. Post-marketing surveillance of Norplant((R)) contraceptive implants: II. Non-reproductive health(1). Contraception, 2001; 63(4):187–209; doi: 10.1016/s0010-7824(01)00187-1

26.

Skouby

, Endrikat

, Düsterberg

, et al. A 1-year randomized study to evaluate the effects of a dose reduction in oral contraceptives on lipids and carbohydrate metabolism: 20 microg ethinyl estradiol combined with 100 microg levonorgestrel. Contraception, 2005; 71(2):111–117; doi: 10.1016/j.contraception.2004.08.017

27.

Castilho

, Jenkins

, Shepherd

, et al. Hormonal contraception and risk of psychiatric and other noncommunicable diseases in HIV-infected women. J Womens Health (Larchmt), 2015; 24(6):481–488; doi: 10.1089/jwh.2014.5003

28.

Kancheva Landolt

, Bunupuradah

, Achalapong

, et al. Carbohydrate, lipid, bone and inflammatory markers in HIV-positive adolescents on antiretroviral therapy and hormonal contraception. J Virus Erad, 2017; 3(1):56–60.

29.

Lähteenmäki

, Rauramo

, Backman

. The levonorgestrel intrauterine system in contraception. Steroids, 2000; 65(10–11):693–697; doi: 10.1016/s0039-128x(00)00176-8

30.

Robins

, Hernán

, Brumback

. Marginal structural models and causal inference in epidemiology. Epidemiology, 2000; 11(5):550–560; doi: 10.1097/00001648-200009000-00011

31.

Hernán

, Brumback

, Robins

. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology, 2000; 11(5):561–570; doi: 10.1097/00001648-200009000-00012

32.

Triant

, Regan

, Grinspoon

. MACE incidence among HIV and non-HIV-infected patients in a clinical care cohort. Abstract 738. In: Paper presented at: 21st Conference on Retroviruses and Opportunistic Infections (CROI); March 3–6, 2014; Boston, MA.

33.

Womack

, Chang

, So-Armah

, et al. HIV infection and cardiovascular disease in women. J Am Heart Assoc, 2014; 3(5); doi: 10.1161/JAHA.114.001035

34.

ACOG practice bulletin no. 206: Use of hormonal contraception in women with coexisting medical conditions. Obstet Gynecol, 2019; 133(2):e128–e150; doi: 10.1097/AOG.0000000000003072

35.

Chasan-Taber

, Willett

, Manson

, et al. Prospective study of oral contraceptives and hypertension among women in the United States. Circulation, 1996; 94(3):483–489; doi: 10.1161/01.cir.94.3.483

36.

Womack

, Richman

, Tien

, et al. Hormonal contraception and HIV-positive women: Metabolic concerns and management strategies. J Midwifery Womens Health, 2008; 53(4):362–375; doi: 10.1016/j.jmwh.2008.01.006

37.

Godsland

, Crook

. Update on the metabolic effects of steroidal contraceptives and their relationship to cardiovascular disease risk. Am J Obstet Gynecol, 1994; 170(5 Pt 2):1528–1536.

38.

Silva-Bermudez

, Toloza

FJK

, Perez-Matos

, et al. Effects of oral contraceptives on metabolic parameters in adult premenopausal women: A meta-analysis. Endocr Connect, 2020; 9(10):978–998; doi: 10.1530/EC-20-0423

39.

Mosorin

, Haverinen

, Ollila

, et al. Current use of combined hormonal contraception is associated with glucose metabolism disorders in perimenopausal women. Eur J Endocrinol, 2020; 183(6); doi: 10.1530/EJE-20-0406

40.

Rosenthal

, Shu

X-O

, Jin

, et al. Oral contraceptive use and risk of diabetes among Chinese women. Contraception, 2004; 69(3):251–257; doi: 10.1016/j.contraception.2003.10.015

41.

Deleskog

, Hilding

, Östenson

C-G

. Oral contraceptive use and abnormal glucose regulation in Swedish middle aged women. Diabetes Res Clin Pract, 2011; 92(2):288–292; doi: 10.1016/j.diabres.2011.02.014

42.

Jirakittidul

, Somyaprasert

, Angsuwathana

. Prevalence of documented excessive weight gain among adolescent girls and young women using depot medroxyprogesterone acetate. J Clin Med Res, 2019; 11(5):326–331; doi: 10.14740/jocmr3792

43.

Lange

HLH

, Manos

, Gothard

, et al. Bone mineral density and weight changes in adolescents randomized to 3 doses of depot medroxyprogesterone acetate. J Pediatr Adolesc Gynecol, 2017; 30(2):169–175; doi: 10.1016/j.jpag.2016.10.011

44.

Silva Dos Santos P de

, de Souza

, Batista

, et al. Binge eating and biochemical markers of appetite in new users of the contraceptive depot medroxyprogesterone acetate. Arch Gynecol Obstet, 2016; 294(6):1331–1336; doi: 10.1007/s00404-016-4192-9

45.

Segall-Gutierrez

, Xiang

, Watanabe

, et al. Deterioration in cardiometabolic risk markers in obese women during depot medroxyprogesterone acetate use. Contraception, 2012; 85(1):36–41; doi: 10.1016/j.contraception.2011.04.016

46.

Berenson

, van den Berg

, Williams

, et al. Effect of injectable and oral contraceptives on glucose and insulin levels. Obstet Gynecol, 2011; 117(1):41–47; doi: 10.1097/AOG.0b013e318202ac23