Trends in Adverse Pregnancy Outcomes Among Women with Systemic Sclerosis in the United States

Abstract

Objectives:

We sought to examine temporal trends in adverse pregnancy outcomes among SSc pregnancies in a large nationwide sample.

Methods:

We used the National Inpatient Sample (NIS) database from 2000 – 2017 to derive national estimates of delivery-associated hospitalizations in the United States among patients with SSc. Each SSc delivery was matched to 100 non-SSc deliveries by age, delivery year, and race. We evaluated adverse pregnancy outcomes (APOs) including maternal and fetal death, cesarean delivery, hospital length of stay, preterm delivery, intrauterine growth restriction, and hypertensive disorders of pregnancy. We used multivariable regression models with an interaction term between SSc and year and adjusting for race, advanced maternal age, diabetes mellitus, and pre-existing hypertension to evaluate temporal trends in APOs among SSc and non-SSc deliveries.

Results:

From 2000 to 2017, there were 3,740 delivery-associated hospitalizations for women with SSc. SSc was associated with an increased risk of all APOs compared to non-SSc deliveries. Fetal death declined in SSc deliveries from 49.0 per 1,000 delivery-related admissions in 2000 – 2005 to 16.2 per 1,000 in 2012 – 2017. There was a significant difference in trends for fetal death between SSc and non-SSc deliveries (p=0.043), but the trends for other APOs did not differ between the two groups.

Conclusions:

In this large nationwide sample, the risk of fetal death among women with SSc markedly improved over the past 18 years. The risk for other APOs remained high in SSc deliveries compared to non-SSc deliveries, and further studies are needed to determine what strategies can improve these outcomes.

INTRODUCTION

Autoimmune connective tissue diseases often disproportionately affect women of childbearing age. Systemic sclerosis (SSc) has a strong female predominance with an estimated female-to-male ratio ranging from 4:1 to 8:1 [1-3]. The peak age of onset of SSc tends to be in the 40s, but the combination of recent trends toward older maternal age nationally and earlier diagnosis of SSc has translated into more SSc patients being cared for during their childbearing years [4]. SSc has previously been shown to be associated with high risk of adverse pregnancy outcomes (APOs) including preterm delivery, prolonged hospital length of stay, hypertensive disorders of pregnancy, and intrauterine growth restriction (IUGR) [5-8]. Although rare, scleroderma renal crisis poses additional morbidity and mortality, and may be difficult to distinguish from preeclampsia [9, 10].

With increasing awareness and more sensitive and specific classification criteria introduced in 2013 by the American College of Rheumatology and European League Against Rheumatism (ACR/EULAR), the diagnosis of SSc can be made earlier and at milder stages of disease [11, 12]. Additionally, obstetric care has evolved over time and high-risk pregnancies are being managed with an emphasis on early identification and management of complicated pregnancies [13, 14]. Furthermore, there is increasing evidence to support the use of medications during pregnancy to reduce APOs in patients with connective tissue disease [15-17]. Taken together, we hypothesize that there has been an improvement in obstetric outcomes in women with SSc over time. Such a trend in improvement has been observed in a recent study of women with systemic lupus erythematosus [18].

We sought to investigate obstetric outcomes over time in patients with SSc in the United States compared to the general obstetric population. Specifically, we aimed to examine the risk of APOs among patients with SSc compared to the general obstetric population, and whether these complications have declined over the past 18 years in a large nationwide sample.

METHODS

Study design and data source

We performed a retrospective cohort study using data from the Healthcare Cost and Utilization Project’s (HCUP) National Inpatient Sample (NIS) for the years 2000 – 2017. The NIS is the largest publicly available inpatient database in the United States and is used to produce national estimates of healthcare utilization and outcomes [19]. Prior to 2012, the NIS included all discharge data from a stratified sample of U.S. hospitals; after a redesign in 2012, the NIS now includes a sample of discharge data from all participating hospitals. It contains data from an estimated 7 million hospitalizations per year, and when weighted it estimates about 35 million hospitalizations annually. The discharge records contain demographic and clinical data codified using International Classification of Diseases (ICD) diagnosis and Procedure Coding System (PCS) procedure codes. To preserve patient privacy, the data are anonymized and not associated with any patient identifiers; therefore, the unit of analysis is hospitalization rather than individual patients. Since NIS is a publicly available database with deidentified data, this study was considered exempt by the Stanford University institutional review board.

Identification of the patient sample

We used ICD codes to identify adult women with SSc who had delivery-associated hospitalizations in the United States between 2000 and 2017 in the NIS database. ICD-9 codes were used from 2000 to September 2015, and ICD-10 codes from October 2015 to the end of 2017. SSc patients were identified by ICD-9 code 710.1 and ICD-10 code M34* and those with concomitant diagnoses for systemic lupus erythematosus (SLE; ICD codes 710.0, M32*) or rheumatoid arthritis (RA; ICD codes 714.0, M05* and M06*) were excluded. ICD codes have been previously validated in identifying patients with SSc who meet the 2013 ACR/EULAR criteria [20]. Delivery-associated hospitalizations were identified by delivery-related ICD-9/10 diagnosis or procedure codes (Supplemental Table 1). Each SSc delivery was matched in a 1:100 ratio to randomly selected obstetric patients (without SSc, SLE or RA) by age, race/ethnicity, and delivery year. Since the unit of analysis is hospitalization rather than individual patients, we restricted our analysis to delivery-associated hospitalizations to minimize biases introduced by multiple antenatal hospitalizations of sicker patients.

Patient characteristics obtained for SSc and non-SSc delivery-associated hospitalizations included: maternal age at delivery, race/ethnicity (white, African American, Hispanic, Asian or Pacific Islander, other, or missing), and hospital characteristics. Comorbidities identified by discharge ICD codes included diabetes, pre-existing hypertension, renal disease, interstitial lung disease, antiphospholipid syndrome, pulmonary hypertension, and smoking status.

Outcome variables

We assessed the proportion of SSc and non-SSc deliveries with the following pregnancy-related outcomes noted at the delivery hospitalization: cesarean section, preterm delivery (defined as delivery before 37 weeks of gestation), hypertensive disorders of pregnancy (including gestational hypertension, preeclampsia, eclampsia, and preeclampsia superimposed on preexisting hypertension), and intrauterine growth restriction (IUGR). We determined in-hospital maternal mortality by the discharge status field in NIS. Fetal death was determined based on ICD codes and reported per 1,000 delivery admissions. Number of inpatient days was determined for each delivery-associated hospitalization.

Statistical analysis

We applied sampling weights to calculate national estimates and 95% confidence intervals (CI) for SSc and non-SSc delivery-associated hospitalizations for each year. Because SSc is a rare condition and the NIS restricts reporting of small cell sizes to protect patient privacy, we collapsed the data into three 6-year intervals: 2000 – 2005, 2006 – 2011, and 2012 – 2017. To account for NIS redesign of the survey sampling in 2012, we used revised “trend weights” instead of the original discharge weights for the years 2000 – 2011 per HCUP recommendations [21]. All analyses accounted for the complex survey design and clustering, and sampling and stratification weights were applied.

Using the overall sample from 2000 – 2017, we determined the proportion of SSc and non-SSc deliveries affected by each of the APOs. We also determined the average hospital length of stay in each group. Rao-Scott Chi-squared test and Student’s t test were used to compare the two groups. We used multivariable logistic regression to estimate the odds ratios (OR) for each of the dichotomous outcomes; linear regression was used for hospital length of stay. In our multivariable models, we included race, advanced maternal age (≥ 35 years), diabetes mellitus, and pre-existing hypertension as covariates. We selected these covariates as they are known risk factors for APOs but not expected to be on the causal pathway between the exposure (SSc) and outcome (APOs).

We evaluated whether temporal trends in APOs differed between SSc and non-SSc deliveries by using logistic or linear regression with an interaction term between SSc (yes vs. no) and year interval. Two-sided p value < 0.05 was considered statistically significant. All analyses were performed on SAS version 9.4 (SAS Institute, Inc., Cary, NC).

RESULTS

National estimates and baseline characteristics

From 2000 to 2017, there were 3,740 delivery-associated hospitalizations for women with SSc (95% CI 3,446–4,034). There were 374,035 matched non-SSc delivery-associated hospitalizations. Table 1 shows the demographics, hospital characteristics, and clinical comorbidities in the cohort. Mean maternal age at delivery was 30 years old. 49% were white, 13% black, 3% Asian or Pacific Islander, 14% Hispanic ethnicity, and 22% had missing race/ethnicity. Most patients were hospitalized at large hospitals and had private insurance. SSc deliveries had more frequent comorbid diagnoses of pre-existing hypertension, renal disease, and pulmonary hypertension (Table 1) compared to the non-SSc cohort. There were very few patients with concomitant diagnoses of interstitial lung disease and antiphospholipid antibody syndrome, so these results were suppressed per the NIS’s data use agreement.

Table 1:

Baseline demographic and clinical characteristics of SSc and non-SSc deliveries

	SSc (n = 3,740)	Non-SSc (n = 374,035)
Age, mean (SD)	30.2 (0.2)	30.2 (0.04)
Race, %
White	48.7	48.9
Black	12.5	12.5
Hispanic	14.0	13.8
Asian or Pacific Islander	2.8	2.7
Other or missing	22.0	22.0
Primary Payer, %
Medicare	1.9	0.7
Medicaid	29.5	35.2
Private insurance	61.9	58.3
Other or missing	6.7	5.8
Hospital characteristics, %
Small	8.0	12.1
Medium	24.6	27.3
Large	67.2	60.3
Missing	0.2	0.3
Clinical comorbidities, %
Diabetes*	9.2	7.7
Pre-existing Hypertension	7.0†	2.5
Renal Disease	0.9‡	0.1
Pulmonary Hypertension	1.6†	0.06
Current smoking	5.7	4.5

Abbreviations: SSc, systemic sclerosis; SD, standard deviation

All values are weighted estimates

^*Includes pre-existing and gestational diabetes

^†p < 0.001 compared to non-SSc

^‡p < 0.05 compared to non-SSc

Overall outcomes

Maternal death was very rare in the SSc group. Since the NIS prohibits reporting of cells with few events to avoid individual patient identification, we are not able to report this result. The results of other pregnancy outcomes are shown in Table 2. Compared to non-SSc deliveries, SSc deliveries were more likely to be associated with APOs including fetal death (28.9 per 1,000 vs. 7.0 per 1,000), cesarean section (44% vs. 32%), preterm delivery (22% vs. 9%), hypertensive disorders of pregnancy (15% vs. 8%), and IUGR (6% vs. 2%), and had a longer inpatient length of stay (4.1 days vs. 2.7 days), all p < 0.001.

Table 2:

Adverse pregnancy outcomes in SSc and Non-SSc delivery-related hospitalizations over entire study period

Pregnancy Outcome	SSc (n = 3,740)	Non-SSc (n= 374,035)	OR* (95% CI)
Fetal death per 1,000 admissions, n	28.9	7.0	4.00 (2.55–6.27)
Mean LOS, days	4.1	2.7	1.35 (0.98–1.71) †
Cesarean section, %	44.4	32.0	1.66 (1.44–1.92)
Preterm delivery, %	21.9	9.4	2.65 (2.23–3.14)
Hypertensive disorders of pregnancy‡, %	15.1	7.9	1.97 (1.61–2.41)
IUGR, %	5.5	2.2	2.45 (1.74–3.45)

Abbreviations: SSc, systemic sclerosis; OR, odds ratio; CI, confidence interval; LOS, length of stay; IUGR, intrauterine growth restriction

^*OR expressed as adjusted odds ratios (95% CI) from multivariable logistic regression analyses except as noted; covariates included maternal age, race, diabetes mellitus and preexisting hypertension

^†Linear regression expressed as β coefficients (95% CI)

^‡Includes gestational hypertension, preeclampsia, eclampsia, superimposed preeclampsia

In our multivariable models we adjusted for potential confounders including race, advanced maternal age (≥ 35 years), diabetes mellitus, and pre-existing hypertension. We found that SSc deliveries were independently associated with higher odds of fetal death (OR 4.0, 95% CI 2.55 – 6.27), cesarean section (OR 1.66, 95% CI 1.44 – 1.92), preterm delivery (OR 2.65, 95% CI 2.23 – 3.14), hypertensive disorders of pregnancy (OR 1.97, 95% CI 1.61 – 2.41), and IUGR (OR 2.45, 95% CI 1.74 – 3.45). In our multivariable linear regression model adjusting for the same confounders, we found that SSc deliveries were also associated with a longer hospital length of stay by 1.35 days compared to non-SSc deliveries (β = 1.35, 95% CI 0.98 – 1.71) (Table 2).

Temporal trends in outcomes

Temporal trends in APOs in SSc and non-SSc deliveries are shown in Table 3. Fetal deaths notably declined in SSc deliveries from 49.0 per 1,000 delivery-related admissions in 2000–2005 to 16.2 per 1,000 in 2012–2017. Difference in trends over time for fetal death was significant (p for difference in trends = 0.043) between SSc and non-SSc deliveries. Hospital length of stay also tended to decrease among SSc deliveries from 4.8 days in 2000 – 2005 to 3.9 days in 2012—2017, but the difference in trends was not statistically significant (p = 0.087). There were no differences in trends over time between SSc and non-SSc deliveries for cesarean section, preterm delivery, hypertensive disorders of pregnancy, or IUGR (Table 3; Figure 1).

Table 3:

Temporal trends in adverse pregnancy outcomes in SSc and non-SSc deliveries^*

Outcome	2000 – 2005	Time period 2006 – 2011	2012 – 2017	P value†
Fetal death per 1,000 admissions, n
SSc	49.0	22.3	16.2	0.043
Non-SSc	7.1	6.7	7.4
Mean LOS, days (SD)
SSc	4.8 (0.4)	3.7 (0.3)	3.9 (0.3)	0.087
Non-SSc	2.7 (0.02)	2.8 (0.02)	2.7 (0.02)
Cesarean section, n (%)
SSc	510 (42.4)	580 (44.6)	570 (46.2)	0.43
Non-SSc	33,669 (27.6)	43,766 (34.0)	42,060 (34.1)
Preterm delivery, n (%)
SSc	326 (27.1)	249 (19.1)	245 (19.8)	0.17
Non-SSc	11,815 (9.7)	12,771 (9.9)	10,670 (8.6)
Hypertensive disorders of pregnancy‡, n (%)
SSc	184 (15.3)	177 (13.6)	205 (16.6)	0.17
Non-SSc	7,739 (6.4)	9,754 (7.6)	11,905 (9.6)
IUGR, n (%)
SSc	51 (4.2)	94 (7.2)	60 (4.9)	0.15
Non-SSc	1,910 (1.6)	2,470 (1.9)	3,660 (3.0)

Abbreviations: SSc, systemic sclerosis; LOS, hospital length of stay; SD, standard deviation; IUGR, intrauterine growth restriction

^*AII n's are weighted values; % given as % of deliveries

^†P value for difference in trends

^‡Includes gestational hypertension, preeclampsia, eclampsia, superimposed preeclampsia

Figure 1: Temporal trends in adverse pregnancy outcomes in SSc and non-SSc deliveries.

A: Fetal deaths per 1,000 admissions declined among SSc deliveries from 2000 – 2005 to 2012 – 2017, with significant difference in trends between SSc and non-SSc deliveries (p=0.043). There was no significant difference in trends between SSc and non-SSc deliveries in average hospital LOS (B), or in the percentages of cesarean section (C), preterm delivery (D), hypertensive disorders of pregnancy (E), or IUGR (F).

P value is for difference in trends.

SSc, systemic sclerosis; LOS, length of stay; IUGR, intrauterine growth restriction

DISCUSSION

In this large nationwide sample, the risk of fetal death among SSc patients markedly declined over the 18-year period from 2000 – 2017. Maternal deaths among SSc deliveries were too few to report, which is reassuring. Nevertheless, similar to what has been reported previously, SSc deliveries still had significantly higher risk of APOs. Despite the improvement over time in rates of fetal death, the overall risk of serious APOs including fetal death, IUGR, and hypertensive disorders of pregnancy remain substantially elevated among SSc patients compared to the general obstetric population.

There have been several prior studies evaluating APOs among women with SSc, [6, 7, 22, 23]. In one of the larger prospective studies of SSc pregnancies by Steen et al in 1999, 15% of the women who became pregnant after SSc diagnosis suffered a miscarriage, though this was similar to rates among RA and general obstetric comparators and there was no report of stillbirths later in the pregnancy. In Taraborelli et al’s study in 2012, miscarriages, fetal deaths, and voluntary/therapeutic abortions occurred in 10% of SSc patients. In their cohort, 25% of SSc pregnancies were affected by preterm delivery and 6% with IUGR, similar to the findings in our study (22% and 6%, respectively). In a recent retrospective cohort study of SSc and systemic lupus erythematosus (SLE) patients and healthy controls, Barilaro et al found that SSc patients had higher risk of IUGR, preterm delivery, and preeclampsia compared to the general obstetric comparators, and higher risk of preterm birth compared to SLE patients.

Our current study builds on these prior studies and provides an update on pregnancy complications among SSc patients in a large nationwide sample and trends in APOs over time. While the risks of APOs remain high in SSc patients, we found encouraging trends that fetal death has declined over time for SSc patients. This is likely due to a combination of factors, including earlier diagnosis of SSc at younger age and with milder disease manifestations, improvements in rheumatologic and obstetric care, and close monitoring by interdisciplinary teams during pregnancy. We did not see a significant difference in trends for rates of cesarean section; however, the decision to pursue a cesarean delivery involves a complex interplay of patient and provider preferences in addition to medical indications [24], thus this finding needs to be interpreted with caution. The frequency of other APOs including preterm birth, IUGR, and hypertensive disorders of pregnancy have remained steady over time; this may be because these outcomes are closely linked to placental insufficiency that results from SSc vasculopathy [25, 26]. Currently we do not have any pharmacologic interventions that are known to improve outcomes related to placental insufficiency. A randomized controlled trial of sildenafil in a high-risk obstetric population with early severe IUGR unfortunately failed to show any reduction in perinatal mortality and led to higher incidence of neonatal pulmonary hypertension [27]. However, such trials have not been undertaken in an SSc cohort. Future studies are needed to determine whether immunomodulatory agents or medications targeting SSc vasculopathy have the potential to improve pregnancy and neonatal outcomes in SSc.

Our study has several limitations. First, we used ICD codes to identify cases of SSc delivery-associated hospitalizations, comorbidities, and APOs, which carries the risk of misclassification. However, since we are using discharge codes, the inclusion of SSc in the discharge diagnosis tends to bias toward a sicker or more complicated SSc patient and may overestimate the proportion of SSc pregnancies with complications. That said, there is also likely an element of self-selection as patients who have substantial comorbidities or severe SSc may choose not to or be advised not to pursue a pregnancy. Furthermore, since the introduction of the 2013 ACR/EULAR classification criteria for SSc [12], the diagnosis of SSc can be made earlier and in patients with milder disease; this may mean that women who delivered in 2012-2017 have milder forms of SSc compared to prior years and could contribute to the lower rate of fetal deaths that we observed. In addition, since our study focused on delivery-associated hospitalizations, fetal death should represent stillbirths and not early miscarriages prior to 20 weeks of gestation, though misclassification from ICD-coding is possible. Since the NIS does not include patient identifiers, we are not able to link the maternal records to infant records to track neonatal outcomes; therefore, early neonatal deaths may have been missed. Due to the nature of the NIS dataset, we do not have access to detailed patient-level data including SSc subtype (diffuse vs. limited), autoantibody status, disease duration, disease severity, medication use, or antiphospholipid antibody status. We are also not able to track parity or other obstetric factors that may affect pregnancy outcomes. Finally, many patients had missing data on race and ethnicity, partly due to differences in reporting practices among the different states, including a few states which did not report race at all or collect information regarding Hispanic ethnicity [28]. Therefore, there may be some residual confounding.

In conclusion, we found that fetal death has declined among SSc deliveries in recent years in a large nationwide sample. The risks of other APOs remain high, and close monitoring by a multi-disciplinary team of rheumatologists and high-risk obstetricians is paramount. Further studies are needed to determine what strategies can be implemented to improve these outcomes. We anticipate that the results of the current study will help guide women and clinicians in pre-conception counseling and obstetric care.

Data availability

This study includes deidentified patient data from a publicly available database. The data that support the findings of this study may be obtained from the Healthcare Cost and Utilization Project (https://www.hcup-us.ahrq.gov/tech_assist/centdist.jsp).