Infertility treatment is associated with increased risk of postpartum hospitalization due to heart disease

Cardiovascular disease is a major cause of maternal mortality, but the extent to which infertility treatment is implicated in heart disease remains unclear.


Introduction
In the United States, approximately 13% of women aged 15-54 years have difficulty conceiving or carrying a pregnancy to live birth [1].As such, infertil-ity treatment is increasingly used to achieve pregnancy.It is estimated that 12.2% of reproductiveaged women accessed infertility services from 2015 to 2019 [1], and assisted reproductive technology contributed to 2% of births in the United States in 2018 [2].Unfortunately, infertility treatment has also been linked to adverse maternal and neonatal outcomes, including an increased risk of hypertensive disorders of pregnancy and gestational diabetes [3][4][5][6].These associations are important to consider, as they may have implications beyond the peripartum period, including the development of chronic disease along the life course, particularly cardiovascular disease (CVD) [7][8][9].However, the link between infertility treatment and the incidence of CVD is not well studied.
Research on infertility treatment and CVD is conflicting.A Swedish population-based study of women who delivered after in vitro fertilization (IVF) reported a higher rate of hypertension and more incident strokes compared with those who delivered after spontaneous conception [10].However, rates of incident heart disease were similar between the groups.In contrast, a populationbased study in Canada found that women who conceived via infertility treatment had fewer CVD events than those who did not undergo treatment [11].In a subsequent meta-analysis of six observational studies, the authors found no definitive increased risk of CVD associated with infertility treatment [12].However, they concluded that the small number of studies and substantial heterogeneity precluded reassurance about the cardiovascular safety of infertility treatment.In our study of patients in the United States, we found that infertility treatment was associated with a 66% increased risk of stroke hospitalizations in the postpartum period [13].On the other hand, a recent study of patients from four Nordic counties reported that infertility treatment was not associated with increased risks of ischemic heart disease, cerebrovascular disease, stroke, cardiomyopathy, heart failure, pulmonary embolism, or deep vein thrombosis over a median follow-up of 11 years [14].These observations underscore substantial knowledge gaps in this area.
As CVD is the leading cause of death globally [15], it is important to understand the relationship between prior infertility treatment and the later development of CVD.CVD is also a major cause of maternal mortality, contributing to an increased number of maternal deaths over the years and now accounting for 26% of pregnancy-related deaths in the United States [16,17].Outside of pregnancy, women still experience worse cardiovascular outcomes than similar-aged men [18,19].There has been an increased focus on sex-based differences in CVD, including risk factors that are sex-specific, such as pregnancy and infertility treatment.We performed this population-based study to examine the associations between exposure to infertility treatment and maternal cardiac complications within the first year postpartum.We hypothesized that infertility treatment is associated with an increased risk of cardiac complications and tested this hypothesis in a large population-based cohort of over 31 million patients who delivered in hospitals in the United States.

Methods
We designed a retrospective cohort study using the Nationwide Readmissions Database (NRD).The NRD is part of the Healthcare Cost and Utilization Project (HCUP) and contains nationally representative data on hospital readmissions for all ages and types of payers [20].The NRD is one of the largest publicly available, all-payer databases in the United States, comprising about 18 million unweighted (35 million weighted) hospital discharges each year.The database utilizes verified patient linkage numbers, enabling follow-up of a patient across hospitalizations within a given calendar year [20].It also contains diagnosis codes, which enable the identification of specific target populations for research as well as the identification of reasons for readmission.The NRD does not contain patient identifiers, so Institutional Review Board approval was not required.We structured the manuscript to follow the Strengthening the Reporting of Observational Studies in Epidemiology Reporting guidelines [21].

Cohort composition
We compiled a list of International Classification of Diseases, Ninth Revision (ICD-9) and ICD, Tenth Revision (ICD-10) diagnosis codes, as shown in Table S1, to obtain a cohort of patients who had a hospital delivery with or without infertility treatment.We identified infertility-related ICD-9 and ICD-10 codes by referencing published literature with similar study designs [22,23] as well as by searching through available diagnoses in our electronic medical records system.ICD-9 was used from 2010 up to the third quarter of 2015 and ICD-10 was used from the fourth quarter of 2015 to 2018.We included females 15-54-year old who had a hospital delivery from January to November in a given calendar year.The exclusion of delivery hospitalizations in December was necessary to ensure that all women had at least 30 days of follow-up in the postpartum period.The cohort included singleton and multiple births, irrespective of the delivery outcome (live births and stillbirths) or gestational age.
Of a total of 36,310,342 (weighted) hospitalizations from January to December, we sequentially excluded the following: 3011,462 ectopic pregnancies/molar pregnancies/abortive outcomes, 2809,622 deliveries that occurred in December, 144,496 admissions between January and November for patients with CVD prior to delivery, and 529,264 admissions for patients with CVD during delivery.After all exclusions, 31,339,991 (weighted) delivery hospitalizations from January to November remained for analysis.

Exposure
Any infertility treatment, as identified by ICD-9 and ICD-10 codes (Table S1), was the exposure.

Outcome measures
The primary outcome was hospitalization within 12-month post-delivery due to heart disease, including ischemic heart disease, atherosclerotic heart disease, cardiomyopathy, hypertensive disease, heart failure, and cardiac dysrhythmias.The secondary outcome was the cumulative risk of heart disease hospitalization at <30, <60, <90, and <180 days after delivery.The comparison group consisted of patients without any heart disease hospitalizations.

Statistical analysis
The rate of heart disease hospitalizations per 100,000 patients was calculated for those who conceived spontaneously and for those who conceived with infertility treatment.We estimated the rate difference (RD) in heart disease hospitalizations between patients who underwent infertility treatment and spontaneous conceptions.We derived the number needed to be exposed to infertility treatment to result in one heart disease hospitalization, adjusted for confounders (discussed below) [24].
We fit Cox proportional hazards regression models to estimate the associations between infertility treatment and heart disease hospitalizations.From these models, we derived hazards ratio (HR) and 95% confidence intervals (CI) before and after adjustments for confounders as the effect mea-sure.We adjusted for the following confounders: maternal age (grouped in 5-year categories as <15, 15-19, …, and 45-54 years), hospital bed size (small, medium, large, and unknown), hospital type (government/nonfederal, private/nonprofit, private/investor-owned, and unknown), hospital teaching status (metropolitan nonteaching, metropolitan teaching, and nonmetropolitan hospital), income quartile (≤25th percentile as low, 26th-50th percentiles as medium-low, 51st-75th percentiles as medium-high, 76th-100th percentiles as high, and unknown), insurance (Medicare, Medicaid, private, self-pay, other, and unknown), and year of hospital discharge.Hospital bed size and median household income varied by each year in the database [25,26].

Secondary outcomes
As secondary outcomes, we fit discrete-time survival models to estimate the risk of heart disease hospitalization within 30-, 60-, 90-, and 180day post-deliveries for those who conceived with infertility treatment compared with those who conceived spontaneously.The discharge date of the first heart disease hospitalization encounter was considered the event time for analysis.As the NRD database does not include the day of hospitalization (although the month and year are available), we simulated the day based on a uniform distribution, considering the variable number of days in each month and leap years.

Probabilistic bias analysis
We undertook a quantitative bias analysis [26] to simultaneously address biases due to exposure misclassification (infertility treatment), selection, and unmeasured confounding.Misclassification was assumed non-differential with respect to the outcomes, and, under a uniform distribution, the sensitivity and specificity of infertility treatment in the HCUP data were assumed to range between 0.5-0.8 and >0.99, respectively [27,28].For selection bias, we allowed the bias parameter to be 0.56 (28 of 50 states that contributed data to HCUP).Unmeasured confounding was addressed by setting the prevalence of the confounder(s) among those with and without the infertility treatment groups to range between 0.05 and 0.25 (nondifferential) and the confounder-outcome rate ratio to range between 0.1 and 10.0.Based on these assumptions, we generated 500,000 simulation patterns and reported the median bias-corrected rate ratio with 95% CI.

E-values
To account for residual unmeasured confounders, such as maternal obesity and smoking status, we conducted a sensitivity analysis using the E-value.The E-value is defined as the minimum strength of association that an unmeasured confounder would need to have, over and above the adjustment for observed confounders, with both the exposure and the outcome to nullify a specific exposure-outcome association [29,30].Thus, a large E-value relative to the confounder-adjusted effect measure suggests that considerable unmeasured confounding would be needed to nullify the reported association.

Results
We analyzed a total of 31,339,991 patients with a hospital delivery in the United States from 2010 to 2018.The distribution of patients who delivered after spontaneous conception versus infertility treatment is shown in Table 1.Any infertility treatment was identified in 0.9% (n = 287,813) of patients.The number of delivery hospitalizations in the infertility treatment group increased from 2010 to 2016 and decreased in 2017 and 2018.Compared with patients who conceived spontaneously, those who used infertility treatment were older, came from high-income households (39.6% vs. 21.2%), and had private insurance (83.2% vs. 51.3%).
The rates of hospitalization for heart disease complications within 12 months following delivery for patients with and without infertility treatment are shown in Table 2. Patients who conceived with infertility treatment had a higher rate of heart disease hospitalizations compared with patients who conceived spontaneously (550 vs. 355 per 100,000; RD 195, 95% CI: 143-247).Further classification of heart disease showed the greatest RD for hypertensive disease (380 vs. 209 per 100,000; RD 171, 95% CI: 127-214), with much lower absolute rates of hospitalizations, as well as RDs, for other types of heart diseases.The numbers needed to be exposed to infertility treatment were 518 (95% CI: 443-624) and 580 (95% CI: 497-695) per 100,000 for any heart disease and for hypertensive disease, respectively.
The associations between infertility treatment and the risk of heart disease complications (before and after adjusting for confounders) are shown in Table 3. Patients who conceived with infertility treatment had a doubling in the risk of hospitalization due to any heart disease within 12 months after delivery compared with those who conceived spontaneously.The corresponding E-value for this effect measure was 3.39 (95% CI closest to null, 3.00).Further classification of heart disease showed the greatest increase in risk for hypertensive disease, followed by cardiac dysrhythmias.Corrections of the risk estimates for exposure misclassification, selection, and unmeasured confounding confirmed the associations.
Finally, the risk of heart disease at different time points post-delivery is shown in Table 4.The increased risk with infertility treatment was apparent as early as 30-day post-delivery for any heart disease hospitalization (adjusted HR 1.61, 95% CI: 1.39-1.86)and continued to be observed at an even higher risk at 180-day post-delivery (adjusted HR 1.92, 95% CI: 1.73-2.13).When these risk estimates were corrected for various biases, the associations became stronger (Fig. 1).

Discussion
In this population-based retrospective cohort study of women who underwent a hospital delivery in the United States (2010-2018), infertility treatment was associated with an increased risk of hospitalization due to heart disease within 12-month postpartum.Notably, patients who conceived with infertility treatment had a doubling in the risk of hospitalization due to hypertensive disease compared with those who conceived spontaneously.Our findings also suggest that the increased risk of heart disease associated with infertility treatment is present in the weeks after delivery and continues to rise beyond the postpartum period.Although these findings should be interpreted cautiously given the low absolute rates of hospitalization, this study highlights the importance of timely postpartum follow-up as well as referral to primary care physicians for long-term followup.
Observed increased heart disease risk associated with infertility treatment suggests that providers should consider infertility treatment as a potential sex-specific risk factor for heart disease, especially hypertensive disease.These results are contrary to those of a systematic review and meta-analysis [12], as well as those of others.In a meta-analysis of six observational studies with 41,910 patients who conceived with infertility treatment and Note: The person-months of follow-up among those who delivered with spontaneous conception and with infertility treatment were 31,100,181 and 288,321, respectively.a The number needed to be exposed is not reported since the estimate was negative.
1400,202 patients who did not, Dayan et al. [12] concluded there was no definitive increased risk of cardiac events with infertility treatment.Similarly, in a retrospective cohort study of 33,520 women in Australia who underwent infertility treatment, the authors reported that infertility treatment was associated with lower all-cause and CVD mortality compared with the general population [31].Our study with national estimates and a large sample size-including 287,813 patients who conceived via infertility treatment and over 31 million patients who did not-strengthens the generalizability of the association between infertility treatment and increased risk of heart disease.
Results from the discrete-time analysis highlight the importance of scheduled and timely postpartum follow-up visits and the likely need for ongoing care of women who receive infertility treatment.This supports the American College of Obstetricians and Gynecologists (ACOG) recommendation for all women to have contact with their obstetric care providers within the first 3-week postpartum [32].ACOG also recommends that women with pregnancies complicated by preterm birth, gestational diabetes, or hypertensive disor-ders of pregnancy should be counseled about the associated higher lifetime risk of maternal cardiometabolic disease.We argue that women who conceive with infertility treatment should also be included with these higher risk patients.In addition, ACOG recommends that women with chronic medical conditions should receive timely followup and ongoing care.Our results suggest that women who undergo infertility treatment should also be included in this group, requiring greater surveillance.
Infertility treatment has been associated with an increased risk of adverse maternal outcomes, such as hypertensive disorders of pregnancy and gestational diabetes [3][4][5][6], which in turn may contribute to the development of heart disease.
Another hypothesis for heightened heart disease risk with infertility treatment, based on current knowledge of ovarian hyperstimulation syndrome, is that exposure to infertility therapies produces non-physiologic ovarian stimulation, resulting in supraphysiologic production of hormones and other substances that affect vascular function [33].In a systemic review, Fujitake et al. [34] found that IVF leads to acute maternal hemodynamic changes, including an increase in heart rate and a decrease in blood pressure, but it remains unclear as to whether these changes are associated with cardiovascular outcomes.Other studies have suggested derangements in the renin-angiotensinaldosterone axis, endothelial dysfunction, and the induction of prothrombotic state with infertility treatment that may confer elevated heart disease risk [33,35,36].Although the exact mechanism remains unknown, it is likely multifactorial.
It is also possible that the causes and diagnosis of infertility itself, and not infertility treatment, per se, may contribute to the increased CVD risk.For example, endometriosis, which is present in 25%-50% of women with infertility, has been associated with an increased risk of CVD [37,38].Menstrual irregularity, one of the most com-mon causes of infertility, as well as premature menopause, has also been associated with an increased risk of CVD [39,40].Further, polycystic ovarian syndrome is strongly linked with metabolic syndrome [41], a known risk factor for CVD.Although these infertility-related conditions may obscure the association between infertility treatment and CVD, results from existing literature remain mixed.In a cross-sectional analysis of 744 US women, Gleason et al. [42] found that women with self-reported infertility had 71% higher odds of reporting a cardiovascular event after controlling for metabolic syndrome.Similarly, in a study of 863,324 Swedish women who gave birth, Parikh et al. [43] found that women reporting ≥5 years of subfertility had 19% higher odds of CVD after controlling for cardiovascular risk factors and adverse pregnancy outcomes.On the other hand, in an analysis of a 10-year cohort study of 2990 women, Cairncross et al. [44] found no association between infertility and the development of metabolic syndrome or CVD events.In a prospective study of 38,528 women from the Women's Health Initiative, Lau et al. [45] found that although infertility was associated with the development of heart failure, this association was independent of traditional cardiovascular factors and other infertilityrelated conditions.Notably, these studies did not differentiate between patients with infertility who underwent treatment and those who did not, and thus, did not examine the impact of treatment on CVD risk.Overall, discerning the impact of infertility diagnosis itself versus infertility treatment on CVD risk remains a challenge.

Strengths and limitations
This study has several strengths.First, we utilized a large, nationally representative database that included hospitalizations from 28 states over a 9year period [20], which allows for generalizability.Second, we adjusted for multiple known confounders and assessed for unmeasured confounding biases based on the E-value, which strengthened the study's findings.We also corrected for potential biases due to exposure misclassification, selection, and unmeasured confounding.
This large population-based retrospective cohort study is not without its limitations.First, we could not consistently differentiate the types of infertility treatments between the ICD-9 and ICD-10 versions.Thus, an area for future research is assessing heart disease risk based on types of infertility treatments, including IVF, intrauterine insemination, and fertility-enhancing medications.Second, the cohort was restricted to hospitalizations 1 year postpartum, and thus, we did not assess for longterm heart disease outcomes.Finally, we caution that the associations between infertility treatment and heart disease cannot be regarded as being causal.

Conclusions
In this cohort of over 31 million patients who delivered in the United States from 2010 to 2018, we found that infertility treatment was associated with a higher risk of hospitalization due to heart disease, particularly hypertensive disease.Further, this increased risk was observed within weeks after delivery and continued beyond the postpartum period.The risks reported here should be interpreted cautiously, considering the low absolute rates of heart disease in the population, including hypertensive disease.However, the findings suggest that providers should consider counseling patients that infertility treatment is a potential sex-specific risk factor for CVD.Importantly, these patients should have timely postpartum follow-up and transition to primary care.Further studies are needed to understand the mechanism through which infertility treatment impacts cardiovascular outcomes.

Table 1 .
Distribution of patients with infertility treatment and spontaneous conceptions: Nationwide Readmissions Database, 2010-2018.

Table 2 .
Rates of heart disease hospitalizations and number needed to be exposed: Nationwide Readmissions Database, 2010-2018.

Table 3 .
Associations between infertility treatment and heart disease: Nationwide Readmissions Database, 2010-2018.Note: The median follow-up was 6.3 months (95% CI, 6.3-6.3) and 6.2 months (95% CI, 6.2-6.3)among those who delivered with spontaneous conception and with infertility treatment, respectively.Hazard ratios were adjusted for the confounding effects of maternal age, hospital bed size, hospital type, hospital teaching status, income quartile, insurance, and year of hospital discharge.
a Bias-correction resulted in over 50% missing data.
Note: Hazard ratios were adjusted for the confounding effects of maternal age, hospital bed size, hospital type, hospital teaching status, income quartile, insurance, and year of hospital discharge through discrete-time Cox proportional hazards regression models.