In-Hospital Complications in Pregnancies Conceived by Assisted Reproductive Technology.

The data that support the findings of this study are available from the corresponding author upon reasonable request. Data from the US National Inpatient Sample (NIS) database containing hospital discharges between January 1, 2008, and December 31, 2016, was used in this study. The NIS is the largest, all‐payer inpatient health care database in the United States, developed by the Agency for Healthcare Research and Quality within their Healthcare Cost and Utilization Project. The NIS contains data on ≈20% stratified samples of all discharges from US hospitals, which is equivalent to 7 to 8 million hospital discharges per annum. This study involved the analysis of deidentified data and was exempt from institutional review board approval. We included all women admitted for delivery using a validated algorithm. 11 Briefly, International Classification of Diseases, Ninth Revision and Tenth Revision, Clinical Modification ( ICD‐9‐CM and ICD‐10‐CM ) obstetric diagnosis codes were used to identify delivery hospitalization episodes (Table  S1A ). For the exposure of ART, we extracted records with an ART code (Table  S1A ). For outcomes, we extracted information on selected vascular complications (acute kidney injury, arrhythmia, ischemic stroke, peripartum cardiomyopathy, venous thromboembolism) and obstetric outcomes (cesarean delivery, placental abruption, preterm birth; Table  S1B ), as well as cost outcomes (length of stay and total billed hospitalization charge). In addition, we extracted covariate information on demographics and comorbidities (Table  S1C ). The ICD‐9‐CM and ICD‐10‐CM codes are included in Table  S1 . In the temporal analyses, we grouped the years as follows: 2008 to 2010, 2011 to 2013, and 2014 to 2016. We performed 2 stratified analyses in the ART and non‐ART groups: presence versus absence of CVD risk factors, and singleton versus multifetal pregnancies. The NIS includes sampling weights that can be used to calculate national estimates and correct variances. 12 During the study period, there was a change in the sampling design in 2012, from using all discharges from a sample of hospitals, to using a sample of discharges from all hospitals participating in the Healthcare Cost and Utilization Project. To ensure compatible data across the study period, we applied a discharge trend weight provided by the Agency for Healthcare Research and Quality in all our analyses. We applied the NIS population survey weights ( svy prefix in Stata; StataCorp, College Station, TX) to all analyses as per Agency for Healthcare Research and Quality recommendations to decrease error margin of the national estimates and provide more stable estimates. We created a missing category for missing data in the race and ethnicity, and median ZIP code income variables because of the large number of missing data, so we could perform sensitivity analyses to assess the effect of excluding observations with missing data. For variables with <2% missing data overall, hospitalization episodes with missing data were removed and assumed to be missing at random. Stata/MP version 14.0 statistical package was used to conduct all analyses. Continuous variables are shown as median and interquartile range, while categorical data are shown as numbers and percentages. To determine statistical difference between the groups for categorical and continuous variables, chi‐square and t tests were used, respectively. The “nptrend” package was used for trend across ordered groups. Binary logistic regression analyses were used to assess the association of ART with maternal and obstetric outcomes. We adjusted for the following potential confounders: age, median ZIP code income quartile, primary payer, race and ethnicity, weekday admission, year of admission, and comorbidities associated with CVD (chronic kidney disease, congenital heart disease, congestive heart failure, depression, diabetes, dyslipidemia, hypertension, multifetal pregnancy, obesity, preeclampsia/eclampsia, previous myocardial infarction, previous transient ischemic attack or stroke, smoker, valvular disease). The odds ratios are shown with the corresponding 95% CIs. In addition, we performed nearest neighbor propensity score matching on race and ethnicity, age, and cardiovascular risk factors (congenital heart disease, smoking, previous myocardial infarction, previous stroke, dyslipidemia, valvular disease, depression, diabetes, chronic kidney disease, obesity, congestive heart failure) using the psmatch2 command in Stata. The effect estimate was generated from logistic regression analysis in the matched cohort. An extension of the strengthening the reporting of observational studies in epidemiology checklist, the reporting of studies conducted using observational routinely‐collected health data checklist, is shown in Table  S2 to summarize our study. 2 , 13

Our study population included 7 236 075, which were weighted to represent 34 273 494 delivery hospitalization episodes between 2008 and 2016 (Figure  1 ). The percentage of delivery episodes from ART increased over time from 0.05% in 2008 to 0.51% in 2016 ( P trend <0.0001; Figure  S1 ). We stratified our study population according to use of ART for conception, which consisted of 0.31% (weighted n=106 248) with ART and 99.69% (weighted n=34 167 246) without ART (Table  1 ). Women who conceived by ART were older (median age 35 versus 28 years; P <0.0001), had higher proportion of White women (66.13% versus 47.82%; P <0.0001), and had more women residing in the wealthiest quartile of household income (54.63% versus 21.42%; P <0.0001), compared with non‐ART (Table  1 ). Patient Characteristics 11 983 (7937–18 341) 18 705 (11 428–29 968) ART indicates assisted reproductive technology; IQR, interquartile range; MI, myocardial infarction; and TIA, transient ischemic attack. Denotes statistical significance. Women who conceived by ART had a higher prevalence of multifetal pregnancy (28.70% versus 1.95%; P <0.0001) and current preeclampsia/eclampsia (10.33% versus 4.42%; P <0.0001; Table  1 ). Over the 9‐year study period, the prevalence of current preeclampsia/eclampsia increased ( P trend =0.005), while multifetal pregnancy decreased ( P trend <0.0001), in both ART and non‐ART groups (Table  S3 ). Women who conceived with ART had a higher prevalence of cesarean delivery, preterm birth, and placental abruption in comparison with women who conceived without ART (Table  2 ). Between 2008 and 2016, the prevalence of preterm birth and cesarean delivery decreased ( P trend <0.05) in both ART and non‐ART groups over time (Table  S3 ). In the univariable and multivariable regression models, women who conceived with ART had increased odds of placental abruption (adjusted odds ratio [aOR], 1.57; 95% CI, 1.41–1.74; P <0.0001), cesarean delivery (aOR, 1.38; 95% CI, 1.33–1.43; P <0.0001), and preterm birth (aOR, 1.26; 95% CI, 1.20–1.32; P <0.0001), compared with women who conceived without ART (Table  3 ). In‐Hospital Obstetric Outcomes and Vascular Complications (per 10 000 Deliveries) Non‐ART n=34 167 246 ART n=106 248 ART indicates assisted reproductive technology. Denotes statistical significance. Association Between Pregnancies Conceived by Assisted Reproductive Techniques and In‐Hospital Obstetric Outcomes and Vascular Complications Data expressed as odds ratios and 95% CIs, reference group is no assisted reproductive technology. Adjustment includes age, median ZIP code income quartile, primary payer, race and ethnicity, weekend admission, year of admission, chronic kidney disease, congenital heart disease, congestive heart failure, depression, diabetes, dyslipidemia, hypertension, multifetal pregnancy, obesity, preeclampsia/eclampsia, previous myocardial infarction, previous transient ischemic attack/stroke, smoker, valvular disease. Denotes statistical significance. We further stratified the groups according to CVD risk (Table  S4 ) and found that the prevalence of adverse obstetric outcomes was higher in women with CVD risk factors compared with their counterparts without CVD risk factors (Table  S5 ). For example, cesarean delivery was prevalent in 68% of women with ART pregnancies and CVD risk factors. In the regression models, after adjusting for demographic data, women with CVD risk factors and conceived with ART had highest risks of adverse obstetric outcomes out of all groups (Table  S6 ). For example, women with ART pregnancies and CVD risks have an 8‐fold and 4‐fold increased risk for preterm birth and cesarean delivery, respectively, in comparison with women without ART pregnancies or CVD risks. The ART and non‐ART groups were also stratified according to whether women had singleton or multifetal pregnancies (Table  S7 ). For singleton pregnancies, the prevalence of adverse obstetric outcomes remained more common in ART pregnancies. Furthermore, the increased risks of these outcomes also persisted in singleton pregnancies conceived using ART compared with those that were spontaneously conceived (placental abruption: aOR, 1.86; 95% CI, 1.64–2.12; P <0.0001; preterm birth: aOR, 1.57; 95% CI, 1.47–1.68; P <0.0001; and cesarean delivery: aOR, 1.44; 95% CI, 1.39–1.50; P <0.0001). However, multifetal pregnancies conceived using ART did not have increased risk of placental abruption compared with spontaneously conceived multifetal pregnancies (Figure 2 ). Women who conceived by ART had a higher prevalence of maternal congenital heart disease (0.30% versus 0.12%; P 0.001), valvular disease (0.75% versus 0.28%; P <0.001), and obesity (6.36% versus 5.52%; P =0.0002). However, there were more current smokers in the non‐ART group compared with the ART group (7.68% versus 3.66%; P <0.0001). Over the 9‐year study period, the prevalence of maternal dyslipidemia, obesity, and current smokers increased ( P trend <0.02), while the prevalence of valvular disease decreased ( P trend =0.003) in both ART and non‐ART groups (Table  S3 ). Overall, compared with women who conceived without ART, the prevalence of all complications was higher in women who conceived with ART, except for peripartum cardiomyopathy (Table  2 ). Between 2008 and 2016, the prevalence of acute kidney injury, arrhythmia, and venous thromboembolism increased in the non‐ART group but not the ART group over time ( P trend <0.001; Table  S3 ). Arrhythmia consisted mainly of sinus node dysfunction and supraventricular tachycardia and did not differ between ART and non‐ART groups. Our univariable and multivariable regression models showed that women who conceived with ART had increased odds of acute kidney injury (aOR, 2.52; 95% CI, 1.99–3.19; P <0.0001), and arrhythmia (aOR, 1.65; 95% CI, 1.46–1.86; P <0.0001), in comparison with women who conceived without ART (Table  3 ). There was non–statistically significant association, after adjustment, for ischemic stroke, peripartum cardiomyopathy, or venous thromboembolism. Similar to obstetric outcomes, the prevalence of vascular complications was also found to be higher in women with CVD risk factors, compared with their counterparts without CVD risk factors in both non‐ART and ART groups in the stratified analyses (Table  S5 ). These were particularly high in women with ART pregnancies and CVD risk factors. For example, the prevalence of arrhythmia was 174 per 10 000 deliveries. Multivariable modeling showed that women with ART pregnancies and CVD risk factors had the highest risks for acute kidney injury, arrythmia, ischemic stroke, and venous thromboembolism, followed by women with non‐ART pregnancies and CVD risk factors, in comparison with women with non‐ART pregnancies and no CVD risk factors (Table  S6 ). For women with CVD risk factors, the risk of peripartum cardiomyopathy was greater in spontaneous compared with ART‐conceived pregnancies. For singleton pregnancies, the prevalence of vascular complications (acute kidney injury, arrhythmia, ischemic stroke, and venous thromboembolism) remained higher in ART compared with the non‐ART group (Table  S7 ). Increased risks of acute kidney injury (singleton: aOR, 2.82; 95% CI, 2.06–3.87; P <0.0001; multifetal: aOR, 1.49; 95% CI, 1.05–2.12; P <0.0001), and arrhythmia (singleton: aOR, 1.89; 95% CI, 1.62–2.19; P <0.0001; multifetal: aOR, 1.32; 95% CI, 1.08–1.62; P <0.0001) were also present in ART pregnancies compared with spontaneous conception (Table  S8 ). Additional sensitivity analyses on complications and outcomes were conducted to examine for the effects of excluding records with missing data (data not shown) and revealed no clinically significant changes in the odds ratios. To account for the baseline differences between the non‐ART and ART groups, we repeated our analyses using propensity score matched cohorts based on age, race and ethnicity, and cardiovascular risk factors (Tables  S9 and S10 ). The matched cohort consisted of 6 050 692 (after removal of 993 706 unmatched delivery hospitalizations), which were weighted to represent 29 459 351 delivery hospitalizations, with a good balance of matched variables (Table  S11 ). These supplementary analyses showed similar results to the main analyses. Women with ART had more weekday admissions (82.94% versus 80.67%; P <0.0001), which were mainly paid for by private insurance (91.86% versus 49.17%; P <0.0001). For the cost outcomes, the ART group experienced longer peridelivery length of stay (median 3 versus 2 days; P <0.0001) and higher hospital charges (median $18 705, interquartile range $11 428–$29 968 versus $11 983, interquartile range $7937‐$18 341; P <0.0001) in comparison with women without ART (Table  1 ). These charges increased between 2008 and 2016 in both groups ( P trend <0.0001), after adjusting for inflation (Table  S3 ). 14

Dr Wu is funded by a National Institute for Health Research Transitional Research Fellowship (TRF‐2017‐10‐005). Drs Chew‐Graham and Jordan are partially funded by West Midlands NIHR Applied Research Collaborations. Dr Chappell is funded by a National Institute for Health Research Professorship (RP‐2014‐05‐019). Dr Sharma is supported by the Blumenthal Scholarship in Preventive Cardiology at Johns Hopkins.

Our analysis of over 34 million delivery hospitalization episodes is the first population‐based study and the largest analysis to consider both obstetric outcomes and vascular complications at time of delivery in women who conceived with ART. We show that the prevalence of pregnancies conceived by ART increased over time, in women who are older and with more comorbidities including dyslipidemia and congenital heart disease. After adjusting for baseline risk profile, ART‐conceived pregnancies are independently associated with 1.7‐ and 2.5‐fold risks for arrhythmia and acute kidney injury, respectively. These risks persisted after further stratification into women with and without CVD risk factors, as well as within‐subgroup analysis on singleton pregnancies. Although the absolute risk remains low, our study highlights the need for close monitoring of both obstetric and vascular complications during admission for delivery in women who conceived with ART, particularly in those with CVD risk factors. In keeping with previous smaller studies, our analysis showed increased risks of obstetric outcomes, such as cesarean deliveries and preterm births. 15 , 16 , 17 , 18 , 19 A single‐center cohort study on 650 ART‐conceived pregnancies showed that ART was associated with 3.6‐fold increased risk of intra‐ and postpartum serious and potentially life‐threatening conditions, including preeclampsia/eclampsia, placental abruption, and maternal cardiovascular dysfunction. 20 An older study using claims data from one health insurance company studied 1 million deliveries, and showed that the odds of severe maternal morbidity, including conditions such as eclampsia, puerperal cerebrovascular disorders, and acute kidney injury, were 1.8 times higher among singleton ART pregnancies compared with non‐ART pregnancies during delivery or postpartum readmissions. 21 We found that there was a large difference in multifetal pregnancies between ART and non‐ART groups. This is likely attributable to the practice of placing ≥2 embryos at the time of embryo transfer during the ART procedure. In 2016, the national elective single‐embryo transfer rate was only 42.7%. 22 The literature also supports the marked difference we found in the multifetal pregnancy rate between the ART and non‐ART groups. The US Centers for Disease Control and Prevention reported that 31.1% of multiple‐birth infants were among infants conceived with ART in 2016. 22 The rate of multifetal pregnancy in the United States remained stable at around 2% until the 1970s, with the advent of ART. 23 , 24 , 25 In our stratified analysis, women with ART pregnancies and CVD risk factors had increased risks of acute kidney injury, arrhythmia, ischemic stroke, peripartum cardiomyopathy, and venous thromboembolism, compared with women without ART or CVD risk factors. Derangements in the renin‐angiotensin‐aldosterone axis and endothelial dysfunction in ART have been suggested to contribute to increased CVD risk. 26 This may be a contributory cause for our finding of increased risk of vascular complications in the ART group. Our study also showed that there is a disproportionate number of women with congenital heart disease seeking ART, who are already at an increased risk of cardiovascular events. Acute changes in maternal hemodynamics attributable to changes in the endogenous hormone levels, including increases in heart rate and decreases in blood pressure, occurs during an IVF cycle, with cardiac functional changes reported in agonist IVF protocols. 27 Furthermore, ovarian hyperstimulation syndrome, a complication of ART, causes an increasingly procoagulant state, severe multiorgan dysfunction, and dramatic fluid shifts within the body. 26 As ART is associated with increased risks of preeclampsia and venous thromboembolism, together with the increased complications of ovarian hyperstimulation syndrome and multifetal gestations arising from ART causing prothrombotic environment and hemodynamic shifts, women who conceived using ART undergo additional circulatory burden. Although multifetal pregnancy acts as a mediator, with ART increasing the risk of multifetal pregnancy that in turn increases the risk of adverse vascular complications, the impact of ART itself is still observed in singleton pregnancy. Infertility may also act as a confounding factor and contribute to increased cardiovascular risk. For example, thrombophilia and placental disease may cause both infertility and increased cardiovascular risks. Polycystic ovary syndrome, a common cause of female infertility, 28 has been associated with a 7‐fold increased risk of myocardial infarction. 29 Similarly, irregular menstrual cycle has been associated with both infertility 30 and increased CVD risk. 31 Furthermore, women with infertility are more likely to have cardiometabolic risk factors, such as atherogenic lipid profile and obesity, compared with fertile women. 32 Nevertheless, it remains challenging to determine whether the increased prevalence of adverse outcomes is attributable to the ART procedure itself 33 , 34 , 35 , 36 or maternal factors associated with infertility 37 or is multifactorial. 38 , 39 , 40 , 41 , 42 , 43 , 44 Despite CVD being the leading cause of maternal death, the perinatal confidential inquiry report from the United Kingdom showed that many women who died from heart disease during or after pregnancy following ART had preexisting cardiovascular risk factors that were not formally assessed before ART treatment. 45 There is evidence suggesting that lifestyle intervention in women who are obese before infertility treatment may improve cardiometabolic health at 6 months. 46 Therefore, women with cardiac risk factors may benefit from cardiac assessments and screening before ART, in line with recommendations from the perinatal confidential inquiry report. 45 , 47 The important clinical implication of our study is how to counsel patients desiring ART, especially those with existing cardiovascular risk factors and regarding long‐term cardiovascular implications from ART. While our study did not explore the relationship between ART and long‐term CVD, there is growing understanding that additive adverse effects of infertility may have implications on long‐term cardiovascular risks through shared pathogenesis and vascular dysfunction. 48 Women should be counseled that ART carries a higher risk of pregnancy complications, which will require close monitoring, particularly during delivery. This study has several strengths. First, we examined a large delivery cohort using nationwide data. Second, our study design allowed us to simultaneously examine several important clinical outcomes in a cohort that included both singleton and multifetal pregnancies. Moreover, we also provide type‐specific information for women who are already pregnant, in our subgroup analysis of singleton versus multifetal pregnancy. Third, we are the first to evaluate in‐hospital cardiovascular complications based on a population cohort. These short‐term complications may provide a basis for risk stratification of women at risk of cardiovascular events in the long term. Limitations of our study include inherent errors of retrospective database studies, such as exposure misclassification, ascertainment bias, and underreporting of secondary diagnoses. The recording of ART may have improved over time, as a previous study on delivery admissions in the NIS between 2008 and 2011 showed 0.17% prevalence of IVF compared with 0.31% in our study. 49 We were not able to assess all possible confounding factors, for example, the type and number of ART used before conception. Moreover, the NIS does not capture timing of diagnosis or follow‐up period over 12 months. As such, we were unable to perform analyses on time to events and duration of infertility or comorbidities. Because of the design of the NIS database, we could assess only hospitalization episodes rather than individual women. Therefore, 1 woman may have multiple delivery hospitalizations during our study period. As our study considers only the delivery population, our results are not generalizable to women who had ART but did not conceive. Because of the lack of drug information, we could not consider the effect of current or previous pharmacotherapy. Finally, some of the statistically significant results may have been a chance finding since we did not adjust for multiple testing. In conclusion, women with pregnancies conceived by ART pregnancies are at increased risk of adverse obstetric outcomes and vascular complications, compared with women who conceived spontaneously. Primary and specialist care clinicians should ensure that they communicate these risks and how to mitigate them. Future research should examine the impact of optimization of cardiovascular risk factors before ART on pregnancy complications and long‐term cardiovascular health.

None.

Infertility, defined as the inability to conceive within 1 year of unprotected intercourse, 1 affects an estimated 48 million couples worldwide 2 and ≈15% of women of reproductive age in the United States. 3 The common causes of female infertility include ovulatory dysfunction, fallopian tubal disease, pelvic adhesions, and endometriosis. 4 Assisted reproductive technology (ART) is a group of medical procedures for treating infertility, which includes medication to control timing of ovulation, as well as procedures such as in vitro fertilization (IVF) and intracytoplasmic sperm injection (ICSI). To date, ART has contributed to the birth of more than 5 million infants worldwide. 5 In 2019, there were ≈275 million women diagnosed with cardiovascular disease (CVD) worldwide, 6 which caused 35% of total female deaths. 7 While CVD is the leading cause of maternal mortality in the United States, 8 studies regarding the association between ART and future maternal CVD are limited with inconsistent findings. Although a meta‐analysis concluded that ART treatment does not increase risk of overall cardiac events, only 6 studies were included. 9 Advancing maternal age with increasing prepregnancy cardiovascular risk factors, such as chronic hypertension, obesity, and diabetes, has raised concerns about the pregnancy complications associated with ART. Adverse pregnancy outcomes, such as preeclampsia, have now been established as risk factors for future CVD. 10 Given the multimorbidity associated with women undergoing ART, understanding their risks of adverse obstetric and vascular outcomes may help to guide postpartum vascular risk reduction strategies. Because of the paucity of population‐based data on vascular complications such as arrhythmia and ischemic stroke, we aimed to assess the patient characteristics, obstetric outcomes, vascular complications, and temporal trends in pregnancies conceived by ART using a nationally representative database.

Tables S1–S11 Figure S1 Click here for additional data file.