Pregnancy Outcomes in Women With Multiple Sclerosis

Abstract

Few studies have assessed the risk of adverse pregnancy outcomes in women with multiple sclerosis (MS). We used 2 large US administrative databases, the Truven Health MarketScan Database (2011–2015; Truven Health Analytics Inc., Ann Arbor, Michigan) and the Nationwide Inpatient Sample (2007–2011), to identify delivery cohorts. MS and pregnancy outcomes (infections, cesarean delivery, preterm delivery, poor fetal growth, preeclampsia, chorioamnionitis, postpartum hemorrhage, stillbirth, and infant malformations) were identified during pregnancy and at delivery. We calculated adjusted risk ratios according to MS status and relapse(s) in the year before delivery. Among over 5 million pregnancies, we identified 3,875 pregnancies in women with MS. Women with MS had an increased risk of infections during pregnancy (Truven Health: adjusted risk ratio (aRR) = 1.22, 95% confidence interval (CI): 1.16, 1.27) and preterm delivery (Truven Health: aRR = 1.19 (95% CI: 1.04, 1.35); Nationwide Inpatient Sample: aRR = 1.30 (95% CI: 1.16, 1.44)). The risks of other outcomes were similar for women with and without MS. In the Truven Health database, risk ratios for the pregnancy outcomes in women experiencing relapses versus those without relapses were between 0.9 and 1.4, and confidence intervals overlapped the null. Overall, women with MS had an increased risk of infections and preterm delivery; however, their risks for other adverse pregnancy outcomes were not elevated. Disease activity before delivery was not a strong predictor of outcomes.

administrative claims, multiple sclerosis, pregnancy

Women are 2–3 times more likely to be diagnosed with multiple sclerosis (MS) than men (1, 2), and onset commonly occurs between ages 20 and 40 years (3). As a result, many newly diagnosed patients are women in the process of building their families. It is therefore critical that the safety of pregnancies in women with MS be adequately studied.

Because only a small proportion of pregnant women have MS (4–7), large cohorts are needed to ascertain their risks for rare pregnancy outcomes. In previous studies, some researchers (5–8), but not all (4, 9–13), have found increased proportions of cesarean deliveries and preterm births among women with MS as compared with women without MS. However, more rare outcomes, like chorioamnionitis and postpartum hemorrhage, have received less attention (5). It is theoretically possible for women with MS to have an elevated risk of these disorders, given their increased risk for prolonged (9, 12) and induced (12, 14, 15) labor. In prior studies assessing infant malformations, investigators have generally reported no elevated risks in women with MS, but sample sizes have been small (8, 9, 11).

Finally, few investigators have studied whether risks of adverse pregnancy outcomes are elevated in women with more active disease. Indeed, a relatively low risk of adverse pregnancy outcomes in women with MS could be partially explained by a tendency for women with milder disease to be more likely to choose pregnancy than women with more severe MS. Understanding the risk of adverse pregnancy outcomes in women with moderate-to-severe disability has been stated as a key knowledge gap in the field (12).

Our objective in this study was to estimate the risks of adverse pregnancy outcomes in women with and without MS and, among women with MS, the risks of adverse pregnancy outcomes by relapse history in the year before delivery.

METHODS

Study population

We identified pregnancy cohorts within 2 large administrative health-care databases, the Truven Health MarketScan Commercial Claims and Encounters Database (Truven Health Analytics Inc., Ann Arbor, Michigan) and the Nationwide Inpatient Sample (NIS), both of which have been previously used in many pregnancy-related studies (6, 16–18). Both databases provided unique contributions, the former allowing for longitudinal exposure and outcome ascertainment and the latter providing estimates from a representative sample of hospital discharges in the United States.

The Truven Health database is a convenience sample of health-care claims from approximately 350 payers across the United States. Unique insurance enrollee identification numbers are available to link claims from the same individual, and a unique family identification number links families in the same insurance plan. We identified a cohort of pregnant women aged 12–55 years who delivered live or stillborn infants between September 1, 2011, and September 30, 2015. Codes indicating gestational age for each pregnancy were ascertained at delivery and in the 30 days postdelivery in maternal and infant (where available) claims. We calculated the date of the first day of the last menstrual period (LMP) by subtracting estimated gestational age from the delivery date. We required women to have been continuously enrolled in their insurance plan from 3 months before the LMP date to the delivery date.

The NIS, sponsored by the Agency for Healthcare Research and Quality, is a 20% stratified sample of all US community hospitals as defined by the American Hospital Association. Approximately 7 million (unweighted) discharge abstracts from almost 1,000 hospitals are recorded each year (19). Discharge-level sampling weights based on the sampling scheme are available to obtain national estimates from all US community hospitals. Using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes from a previously validated method (20), we identified hospitalizations where a delivery occurred among women aged 12–55 years in 2007–2011.

MS was ascertained using ICD-9-CM code 340.xx. In the Truven Health database, we required this code to be recorded on at least 2 unique days from 90 days before LMP to the delivery date. In the NIS database, we required at least 1 code recorded at the delivery hospitalization, since this data set does not allow linkage across health-care encounters.

Baseline characteristics

We ascertained age, year, region, payer type, hospital location (available in NIS only), race (NIS only), and quartile of median household income by zip code (NIS only) at delivery using variables directly available in the data sets. Other chronic conditions and baseline characteristics (e.g., preexisting hypertension) were ascertained using ICD-9-CM codes (see Web Table 1, available at https://academic.oup.com/aje) at the delivery admission (NIS and Truven Health) and throughout pregnancy (Truven Health). In the Truven Health database, we calculated the mean number of health service dates (inpatient or outpatient) from LMP to (and excluding) delivery.

We present the estimated prevalence of these characteristics and conditions by MS status in both cohorts. To determine whether the cohort selected from the Truven Health database differed from persons who were excluded for noncontinuous enrollment, we compared the baseline characteristics ascertained at delivery in women included in and women excluded from the Truven Health cohort.

Outcomes

Because hospital encounters in the NIS database cannot be linked across pregnancy within individuals, infections during pregnancy were ascertained only in the Truven Health cohort (LMP to delivery). Infections during pregnancy were reported overall and in the following specific categories, using ICD-9-CM codes (21; Web Table 1): genitourinary infections, upper respiratory tract infections, lower respiratory tract infections, sexually transmitted infections, skin and soft tissue infections, otitis media, influenza, and gastroenteritis. An infection event was defined as 1 or more codes for the above types of infections in outpatient or inpatient data. To separately capture severe infections requiring hospitalization, we only included infection events with 1 or more infection codes as the primary diagnosis for a hospitalization from LMP up to, but excluding, the delivery date.

We identified cesarean delivery, preterm delivery, poor fetal growth, preeclampsia, chorioamnionitis, postpartum hemorrhage, and stillbirth in women with and without MS using ICD-9-CM codes (Web Table 1). In the NIS database, all pregnancy outcomes were ascertained at the delivery admission. However, in the Truven Health database, the ability to link data between health-care encounters allowed us improved ascertainment of some outcomes. Like the NIS, we ascertained cesarean delivery and postpartum hemorrhage in the delivery interval. However, since poor fetal growth, preeclampsia, and chorioamnionitis may be diagnosed before a woman gives birth, we ascertained these outcomes in the 30 days before and including the delivery interval. We only used inpatient claims before the delivery admission to ascertain preeclampsia, as previous work has suggested a high risk of misclassification for preeclampsia recorded in outpatient claims (22). Preterm delivery was defined as a gestational age less than 37 weeks and was based on the codes in Web Table 1 (see gestational age definition above). Stillbirth was a distinct pregnancy outcome in the cohort and was ascertained at delivery.

We ascertained major nonchromosomal structural malformations as described previously (23) using Centers for Disease Control and Prevention guidelines (24) in pregnancies with linked infants continuously enrolled for at least 90 days after delivery (defined as at least 1 claim in the first month and stated enrollment from the enrollment data set for remaining months) as well as in infants with evidence of a neonatal death (ascertained in the first 30 days of the maternal record and the first 90 days of the infant record (ICD-9-CM code(s) 768.0x, 768.1x, 798.0x–798.2x, 798.9x, or 656.4x or a discharge status of “died”)). Because of small numbers, we combined all types of malformations.

Relapse status

Relapses during pregnancy were assessed in the Truven Health database among women with prescription drug benefit coverage. We used the following validated algorithm (25, 26) to define a relapse: either 1) an inpatient admission (excluding the delivery admission) with MS (340.xx) as the primary diagnosis or 2) an outpatient admission with MS as the primary or secondary diagnosis, plus a corticosteroid prescription or procedure code for corticosteroids commonly used to treat acute relapses (methylprednisolone, dexamethasone, prednisone, prednisolone, or adrenocorticotropic hormone; Healthcare Common Procedure Coding System procedure codes J1020, J1030, J1040, J2920, J2930, J8540, J1094, J1100, and J7506) in the 7 days after the outpatient claim.

We calculated relapses in 4 main intervals: 1) from 90 days before LMP to 1 day before LMP (prepregnancy); 2) from LMP to 90 days after LMP (trimester 1); 3) from 91 days after LMP to 180 days after LMP (trimester 2); and 4) from 181 days after LMP to delivery (trimester 3). Women with at least 1 relapse during any of these intervals were categorized in the relapse group, and their risks of the pregnancy outcomes defined above were compared with those of women without a relapse in any of the intervals. As a secondary analysis, we also calculated relapse risk in 2 postpartum trimesters—1) from 1 day after delivery to 90 days after delivery (postpartum trimester 1) and 2) from 91 days after delivery to 180 days after delivery (postpartum trimester 2)—in the subgroup of women with continuous medical and prescription coverage during these postdelivery intervals.

Statistical analyses

Risk ratios for pregnancy outcomes were estimated using standardized probabilities from logistic regression (27), with 95% confidence intervals obtained from 500 bootstrap samples. In the NIS database, we employed survey procedures and discharge-level sampling weights in the logistic regression to account for the complex sampling design and to obtain national estimates of discharges from all US community hospitals.

Risk ratios were calculated in 4 ways: 1) unadjusted; 2) adjusted for maternal age, delivery region, year of delivery, and insurance type (private vs. public; NIS only; model 1); 3) model 1 with additional adjustment for preexisting diabetes, chronic renal disease, preexisting hypertension, asthma, thyroid disorders, depression, and alcohol or substance abuse (model 2); and 4) in the NIS, model 2 with additional adjustment for race, quartile of median household income by patient zip code, and hospital location (rural or urban; model 3). Because model 1 represents an interpretable measure of the risk for a woman with MS compared with a woman without MS of the same age, delivery year, and region of treatment, we present results from this model as the primary measures; data from additional models are available in the Web tables. Risk ratios comparing women with and without a relapse are presented unadjusted, since the sample size was small.

Variables with missing values were assigned a missing indicator level. In accordance with the data-use agreement, we could not present data for any cell sizes less than 11. All statistical analyses were conducted using SAS Enterprise Guide, version 6.1 (SAS Institute, Inc., Cary, North Carolina). This study received ethics board approval from the Harvard T.H. Chan School of Public Health (Boston, Massachusetts).

RESULTS

In the Truven Health database, there were 1,830,941 live or stillborn pregnancies from September 1, 2011, to September 30, 2015, in women aged 12–55 years (Figure 1). Of these, 728,337 pregnancies (39.8%) were excluded for noncontinuous enrollment during the period from 90 days before LMP to the delivery date, leaving 1,102,604 pregnancies in the Truven Health cohort. Besides year of delivery, covariates were similar between women included and women excluded for noncontinuous enrollment (Web Table 2).

Figure 1.

Sample sizes in a study of pregnancy outcomes among women with multiple sclerosis (MS), United States, 2007–2015. Data were obtained from 2 databases: the Truven Health MarketScan Commercial Claims and Encounters Database (Truven Health Analytics Inc., Ann Arbor, Michigan) (2011–2015) and the Nationwide Inpatient Sample (2007–2011). Pregnancy outcomes included cesarean delivery, preterm delivery, poor fetal growth, preeclampsia, chorioamnionitis, postpartum hemorrhage, and stillbirth. Black dashed arrows indicate the analysis conducted for each sample.

There were 4,186,816 live and stillborn deliveries in the NIS cohort from 2007 to 2011 among women aged 12–55 years.

The estimated prevalence of MS was 0.13% (n = 1,439) in the Truven Health cohort and 0.06% (n = 2,436) in the NIS cohort. Because MS was ascertained throughout the full pregnancy period in Truven Health and only for the delivery admission in the NIS, a lower estimated prevalence of MS in the NIS data set was expected.

Baseline characteristics and conditions are presented in Table 1 (Truven Health) and Table 2 (NIS). In the NIS cohort, women with MS were more likely to be white, less likely to be Hispanic, Asian, or Pacific Islander, and more likely to be in the highest quartile of patient income. In both cohorts, pregnant women with MS were older than those without MS; were more likely to be from the Northeast than from the South; and had a higher prevalence of alcohol or substance abuse, depression, anxiety disorders, chronic lower gastrointestinal conditions, preexisting hypertension, thyroid disorders, bipolar disorder, and asthma. The estimated prevalence of recorded chronic conditions was lower overall in the NIS data set because of differences in the ascertainment window; however, the direction of the association with MS was similar to that in the Truven Health data set.

Table 1.

Demographic Characteristics of Women With and Without Multiple Sclerosis in the Truven Health MarketScan Commercial Claims and Encounters Database^a (n = 1,102,604), United States, 2011–2015

Demographic Characteristicb	Women With MS (n = 1,439)		Women Without MS (n = 1,101,165)
	No.	%	No.	%
Mean age at delivery, yearsc	32.6 (4.9)		30.3 (5.6)
Year of delivery
2011	<11d	—	5,312	0.5
2012	488	33.9	374,157	34.0
2013	347	24.1	283,672	25.8
2014	420	29.2	286,330	26.0
2015	175	12.2	151,694	13.8
Region
Northeast	363	25.2	198,926	18.1
Midwest	286	19.9	242,981	22.1
South	480	33.4	404,097	36.7
West	286	19.9	232,329	21.1
Unknown	24	1.7	22,832	2.1
Previous cesarean delivery	309	21.5	202,851	18.4
Multifetal pregnancy	44	3.1	36,390	3.3
Preexisting diabetes mellitus	78	5.4	48,711	4.4
Chronic renal disease	22	1.5	13,880	1.3
Preexisting hypertension	141	9.8	74,066	6.7
Asthma	88	6.1	53,081	4.8
Thyroid disorder	175	12.2	95,820	8.7
Chronic lower gastrointestinal condition	20	1.4	9,389	0.9
Alcohol/substance abuse	97	6.7	21,721	2.0
Depression	140	9.7	49,522	4.5
Bipolar disorder	11	0.8	6,090	0.6
Anxiety disorder	135	9.4	53,215	4.8
No. of health service dates during pregnancyc	22 (11)		16 (10)

Abbreviation: MS, multiple sclerosis.

^a Truven Health Analytics Inc., Ann Arbor, Michigan.

^b Outcomes are not mutually exclusive.

^c Values are expressed as mean (standard deviation).

^d Cell sizes less than 11 have been suppressed in accordance with the data-use agreement.

Table 2.

Demographic Characteristics of Women With and Without Multiple Sclerosis in the Nationwide Inpatient Sample (n = 4,186,816), United States, 2007–2011

Demographic Characteristica,b	Women With MS (n = 2,436)		Women Without MS (n = 4,184,380)
	No.	%	No.	%
Mean age at delivery, yearsc	31.5 (0.1)		27.6 (0.05)
Year of delivery
2007	525	21.6	927,517	22.2
2008	484	19.9	865,887	20.7
2009	491	20.2	819,992	19.6
2010	430	17.7	776,249	18.6
2011	506	20.8	794,735	19.0
Region
Northeast	559	24.0	653,544	16.3
Midwest	593	24.6	883,917	21.5
South	779	31.5	1,599,387	38.1
West	505	19.9	1,047,532	24.2
Primary health insurance payer
Private	1,728	71.0	2,091,057	50.0
Public	605	24.8	1,825,933	43.6
Other	103	4.2	267,390	6.4
Hospital location
Urban area	2,223	91.3	3,671,351	87.7
Rural area	200	8.1	468,938	11.3
Missing data	13	0.5	44,091	1.1
Race/ethnicity
White	1,454	59.7	1,792,817	43.0
Black	268	10.9	484,094	11.6
Hispanic	149	6.1	811,798	19.2
Asian or Pacific Islander	30	1.2	181,345	4.3
Native American or other	79	3.3	198,354	4.8
Missing data	456	18.9	715,972	17.1
Quartile of median household income for patient’s zip coded
Quartile 1	439	17.9	1,109,720	26.6
Quartile 2	477	19.5	1,041,599	24.8
Quartile 3	677	27.7	1,013,480	24.1
Quartile 4	816	33.6	939,334	22.5
Missing data	27	1.2	80,247	2.0
Previous cesarean delivery	458	18.9	675,697	16.2
Multifetal pregnancy	66	2.7	76,343	1.8
Preexisting diabetes mellitus	34	1.4	38,737	0.9
Chronic renal disease	11	0.4	10,424	0.3
Preexisting hypertension	87	3.6	81,618	2.0
Asthma	158	6.5	131,669	3.2
Thyroid disorder	165	6.7	98,237	2.4
Chronic lower gastrointestinal condition	13	0.6	9,050	0.2
Alcohol/substance abuse	40	1.7	56,373	1.4
Depression	134	5.6	79,384	1.9
Bipolar disorder	32	1.3	22,102	0.5
Anxiety disorder	50	2.0	31,821	0.8

Abbreviation: MS, multiple sclerosis.

^a Outcomes are not mutually exclusive.

^b Proportions were weighted using sampling weights to achieve nationally representative estimates.

^c Values are expressed as mean (standard error).

^d The exact amount varied by year. In 2011, the first quartile was $1–$37,999, the second quartile was $38,000–$47,999, the third quartile was $48,000–$63,999, and the fourth quartile was $64,000 or more.

The estimated cumulative incidence of the measured infections during pregnancy was 51.7% in women with MS and 43.4% in women without MS (risk ratio (RR) = 1.22, 95% confidence interval (CI): 1.16, 1.27 (Table 3)), with genitourinary infections being the most common type of infection (30.4% in women with MS vs. 24.4% in women without MS). Risk was higher in women with MS than in women without MS for all infections except influenza, with risk ratios ranging from 1.18 to 1.58. Demographic characteristics and concomitant chronic conditions did not account for the elevated risk of infections overall (model 2).

Table 3.

Prevalences of and Risk Ratios for Infections During Pregnancy Among Women With and Without Multiple Sclerosis in the Truven Health MarketScan Commercial Claims and Encounters Database^a (n = 1,102,604), United States, 2011–2015

Infection Type	Women With MS (n = 1,439)		Women Without MS (n = 1,101,165)		Unadjusted Results		Model 1b		Model 2c
	No.	%	No.	%	RR	95% CI	RR	95% CI	RR	95% CI
Any infection	744	51.7	478,219	43.4	1.19	1.13, 1.25	1.22	1.16, 1.27	1.18	1.12, 1.24
Genitourinary infection	437	30.4	268,472	24.4	1.25	1.16, 1.34	1.31	1.20, 1.40	1.26	1.17, 1.36
Upper respiratory tract infection	352	24.5	197,164	17.9	1.37	1.23, 1.49	1.38	1.25, 1.51	1.33	1.20, 1.45
Lower respiratory tract infection	75	5.2	42,675	3.9	1.35	1.08, 1.66	1.33	1.05, 1.63	1.22	0.94, 1.50
Sexually transmitted infection	61	4.2	44,796	4.1	1.04	0.80, 1.31	1.18	0.90, 1.47	1.12	0.85, 1.41
Infection of skin and soft tissues	66	4.6	41,388	3.8	1.22	0.94, 1.51	1.23	0.94, 1.53	1.17	0.89, 1.45
Otitis media	24	1.7	16,662	1.5	1.11	0.66, 1.52	1.18	0.76, 1.64	1.13	0.68, 1.60
Influenza	11	0.8	11,355	1.0	0.74	0.34, 1.20	0.77	0.36, 1.27	0.74	0.34, 1.21
Gastroenteritis	13	0.9	6,855	0.6	1.46	0.70, 2.22	1.58	0.82, 2.53	1.46	0.69, 2.29
Infection requiring hospitalization	<11d	—	5,044	0.5	NA	NA	1.52	0.57, 2.56	1.28	0.34, 2.21

Abbreviations: CI, confidence interval; MS, multiple sclerosis; NA, not available; RR, risk ratio.

^a Truven Health Analytics Inc., Ann Arbor, Michigan.

^b Results were adjusted for year, age, and region at delivery.

^c Results were adjusted for year, age, region at delivery, preexisting diabetes, chronic renal disease, preexisting hypertension, asthma, thyroid disorders, depression, and substance abuse.

^d Cell sizes less than 11 have been suppressed in accordance with the data-use agreement.

Estimated risks of pregnancy outcomes in women with and without MS are presented in Table 4, with adjusted risk ratios shown in Figure 2 (all risk ratios and 95% confidence intervals are listed in Web Table 3). The proportion of women with cesarean delivery was higher in women with MS than in those without MS, both in Truven Health (39.5% vs. 35.3%) and in the NIS (42.5% vs. 33.1%). After adjustment for basic demographic information, the risk ratio was 1.16 in the NIS (95% CI: 1.09, 1.22) and 1.04 in Truven Health (95% CI: 0.97, 1.11). To see whether the different risk ratios in Truven Health and the NIS could be explained by effect modification by insurance type, we restricted the NIS cohort to only privately insured women but obtained a similar risk ratio (RR = 1.15, 95% CI: 1.04, 1.27).

Table 4.

Pregnancy Outcomes Among US Women With and Without Multiple Sclerosis in the Truven Health MarketScan Commercial Claims and Encounters Database^a (2011–2015; n = 1,102,604) and the Nationwide Inpatient Sample (2007–2011; n = 4,186,816)

Pregnancy Outcomeb	Truven Health MarketScan Commercial Claims and Encounters Database				Nationwide Inpatient Samplec
	Women With MS (n = 1,439)		Women Without MS (n = 1,101,165)		Women With MS (n = 2,436)		Women Without MS (n = 4,184,380)
	No.	%	No.	%	No.	%	No.	%
Cesarean delivery	568	39.5	389,112	35.3	1,036	42.5	1,383,636	33.1
Preterm delivery	189	13.1	114,176	10.4	265	10.9	348,253	8.3
Poor fetal growth	128	8.9	96,665	8.8	59	2.4	88,814	2.1
Preeclampsia	73	5.1	56,828	5.2	102	4.2	174,008	4.2
Chorioamnionitis	46	3.2	37,604	3.4	35	1.5	76,008	1.8
Postpartum hemorrhage	21	1.5	24,236	2.2	78	3.2	114,766	2.7
Stillbirth	<11d	—	6,256	0.6	15	0.6	27,442	0.7
Infant malformatione	35	3.9	26,896	4.4	NA	NA	NA	NA

Abbreviations: MS, multiple sclerosis; NA, not available.

^a Truven Health Analytics Inc., Ann Arbor, Michigan.

^b Outcomes are not mutually exclusive.

^c Proportions were weighted using sampling weights to achieve nationally representative estimates.

^d Cell sizes less than 11 have been suppressed in accordance with the data-use agreement.

^e In a subcohort of women with continuous enrollment during the 30 days after delivery and a linked infant with continuous enrollment during the 90 days after delivery or evidence of neonatal death (n = 902 women with MS; n = 604,808 women without MS).

Figure 2.

Risk ratios (RRs) for pregnancy outcomes in women with and without multiple sclerosis, United States, 2007–2015. Data were obtained from 2 databases: the Truven Health MarketScan Commercial Claims and Encounters Database (Truven Health Analytics Inc., Ann Arbor, Michigan) (2011–2015; n = 1,102,604) and the Nationwide Inpatient Sample (NIS) (2007–2011; n = 4,186,816). RRs were adjusted for year, age, region at delivery, and insurance type (NIS only). Infant malformations were assessed in a subcohort of women with continuous enrollment during the 30 days after delivery and a linked infant with continuous enrollment during the 90 days after delivery or evidence of neonatal death (n = 902 women with multiple sclerosis; n = 604,808 women without multiple sclerosis). Bars, 95% confidence intervals (CIs).

Women with MS had a 20%–30% increased risk of preterm delivery in both cohorts (Truven Health: RR = 1.19 (95% CI: 1.04, 1.35); NIS: RR = 1.30 (95% CI: 1.16, 1.44)). Notably, risks were increased for women with MS among both vaginal deliveries and cesarean deliveries (Web Table 3). Risk ratios for poor fetal growth, preeclampsia, chorioamnionitis, postpartum hemorrhage, and stillbirth were close to the null in both cohorts. Varying the definition of poor fetal growth to at least 2 codes before delivery did not substantially alter the effect estimates. Overall, adjustment for chronic conditions (model 2) and additional socioeconomic factors (model 3) did little to alter the risk ratios beyond the effect estimates generated in model 1.

We assessed 605,710 pregnancies for major malformations (n = 6,263 stillbirths; n = 598,370 livebirths continuously enrolled; and n = 1,077 livebirths with evidence of neonatal death). The risk of malformation was 4% in both women with MS and women without MS (RR = 0.85, 95% CI: 0.59, 1.12).

To assess relapses in the Truven Health database, we excluded 271 (18.8%) women with MS from the original cohort for not having prescription drug benefit coverage (Web Table 4), thus leaving 1,168 (81.2%) women (Figure 1). In total, 210 (18.0%) of those women had at least 1 relapse during or before pregnancy (Figure 3): 10% during the prepregnancy interval, 5% in the first trimester, 4% in the second trimester, and 3% in the third trimester. There were 844 (72%) women with MS with continuous enrollment and prescription drug coverage in the 180 days postdelivery. In this subgroup, the risk of relapse was 11% in the first postpartum trimester and 9% in the second. Prevalences of pregnancy outcomes in women with and without a relapse before or during pregnancy are presented in Table 5. Overall, no meaningful differences were found, and risk ratios ranged from 0.9 to 1.4.

Figure 3.

Risk of relapse before and during pregnancy among women with multiple sclerosis in the Truven Health MarketScan Commercial Claims and Encounters Database (Truven Health Analytics Inc., Ann Arbor, Michigan) (2011–2015; n = 1,168). The postpartum trimesters were in the subcohort with continuous enrollment and prescription drug coverage from delivery to 180 days postdelivery (n = 844). Bars, 95% confidence intervals.

Table 5.

Infections and Pregnancy Outcomes Among Women With and Without Multiple Sclerosis Relapses During Pregnancy in the Truven Health MarketScan Commercial Claims and Encounters Database^a (n = 1,168), United States, 2011–2015

Pregnancy Outcomeb	Women With a Relapse (n = 210)		Women Without a Relapse (n = 958)		Unadjusted Risk Ratio	95% Confidence Interval
	No.	%	No.	%
Infection during pregnancy	105	50.0	483	50.4	0.99	0.85, 1.15
Cesarean delivery	91	43.3	367	38.3	1.13	0.95, 1.35
Preterm delivery	27	12.9	131	13.7	0.94	0.63, 1.32
Poor fetal growth	19	9.0	86	9.0	1.01	0.57, 1.63
Preeclampsia	14	6.7	47	4.9	1.36	0.59, 2.24

^a Truven Health Analytics Inc., Ann Arbor, Michigan.

^b Data for outcomes with cell sizes less than 11 (chorioamnionitis, postpartum hemorrhage, stillbirth, and infant malformation) have been suppressed in accordance with the data-use agreement and are therefore not presented.

DISCUSSION

Using 2 large administrative databases, we found that women with MS had increased risks of both infections (overall) in pregnancy and preterm delivery. Women with MS were also more likely to give birth via cesarean delivery, although this association was largely explained by a more advanced maternal age in women with MS. Relapses during the year prior to delivery were not strongly associated with adverse pregnancy outcomes.

Previous studies have found increased risk of urinary tract infections in women with MS (5, 14, 15); however, less was known about other infections. Here, women with MS had increased risks of all infection subtypes considered except influenza, although the confidence intervals were wide and included the null for all subtypes except genitourinary and respiratory infection subtypes. It is possible that women with MS have a higher likelihood of having infections detected or recorded due to their more frequent contact with the health-care system (surveillance bias). However, the mean number of health service visits during pregnancy for women without MS was quite high (16 visits), and thus this explanation is unlikely to completely account for the results. Further, surveillance bias is less likely to affect infections requiring hospitalization, and the risk ratio for hospitalized infections remained increased (although confidence intervals overlapped the null).

While, in the NIS cohort, women with MS had a higher risk of cesarean delivery than women without MS, in the Truven Health cohort the association was weaker. A preferential recording of MS at cesarean deliveries in the NIS could explain the larger risk ratio through the introduction of recording bias. When identification of MS was restricted to delivery hospitalization within Truven Health as well, the risk was slightly higher (risk ratio = 1.07 (28)). An alternative explanation is that the Truven Health database does not include a publicly insured population as the NIS database does. Restriction of the NIS data set to just the commercially insured still resulted in a similar risk ratio (RR = 1.15). Finally, confounding by center-level differences in the completeness of MS and pregnancy outcome recording could have been more prominent in the NIS, since ascertainment occurred at a single visit (i.e., delivery hospitalization) rather than throughout the full pregnancy period (as in the Truven Health database). Other investigators who have studied cesarean delivery in women with MS have generally found no increased risk (4, 10, 29, 30), and this is consistent with contemporary obstetrical practice, which does not indicate operative delivery for MS itself, and with current results suggesting no increased risk of conditions that might result in cesarean delivery (e.g., preeclampsia). Taken altogether, the difference in the risk ratios for the current study may likely be explained by some combination of random variability, differences in the population, and, potentially, bias induced by ascertainment of MS and outcomes at the same single visit (delivery hospitalization) in the NIS.

We found a modestly increased risk of preterm delivery, supporting findings from some previous studies (5, 7, 8) but not all (4, 10, 11, 14, 15). In at least 1 previous study (5), the estimate was further reduced after adjustment for chronic conditions, suggesting that at least part of the association may be explained by the tendency for women with MS to have a higher prevalence of other conditions known to increase the risk of preterm delivery (e.g., diabetes, hypertension). Most other previous studies assessing preterm delivery have had fewer than 500 cases (4, 7, 8, 10, 11, 15) or have not adjusted for chronic conditions (4, 7, 8, 10, 11, 14, 15), making conclusions about the nature of the association between MS and preterm delivery uncertain.

As in previous work, women with MS did not have an increased risk of stillbirth (12, 15, 31) or infant malformations (8, 9, 11, 14, 15, 32). Fewer studies have addressed the risks for poor fetal growth, postpartum hemorrhage, and chorioamnionitis. On the basis of the current findings, a substantially increased risk is unlikely.

Our study replicated the well-known pattern of declining relapses during pregnancy followed by an increase after delivery, with a gradual decline over time (12, 31). We found women with more active disease did not have substantially increased risks of the adverse pregnancy outcomes evaluated, which is reassuring. We are unaware of any other study that has examined the relationship between disease activity and risk of pregnancy outcomes. However, 2 studies found that women with more severe disability were at increased risk for cesarean delivery (29), assisted vaginal delivery (29), and labor induction (33).

The current study had limitations. First, with claims data, misclassification is inevitable. However, 2 studies carried out using other administrative databases (34, 35) assessed misclassification of the MS code plus other markers of the disease (e.g., prescriptions) and found high sensitivity (93%) and a high positive predictive value (81%–92%). In the NIS database, ascertainment of both MS and the pregnancy outcomes occurred at the delivery hospitalization, and thus differential misclassification is possible. However, differential misclassification of MS was minimized in the Truven Health database, since information on disease status and activity was recorded during pregnancy, before the delivery outcomes had occurred. In prior validation work carried out in the Truven Health pregnancy cohort, MacDonald et al. (28) found that 60% of women with at least 2 MS codes in pregnancy had at least 1 MS code at delivery and, conversely, 78% of women with 1 MS code at delivery had 2 MS codes overall in pregnancy. Additionally, the investigators found that recording at delivery is not more likely for deliveries with complications such as cesarean section, preterm birth, and preeclampsia (28). Second, since ICD-9-CM codes do not divide MS diagnosis by subtype, we were unable to restrict our analysis to only relapsing-remitting cases. However, with 85% of new patients being diagnosed as relapsing-remitting (13), we do not expect a large proportion with other subtypes, and thus the impact is probably small.

The greatest strength of this study was the use of 2 large complimentary data sets. The NIS database was a representative sample of discharges from community hospitals across the United States, allowing us to generate estimates of pregnancy outcome risks that are generalizable to the wider US population. As well, the NIS provided rich demographic data, enabling us to assess confounding by sociodemographic factors beyond the factors controlled for in Truven Health. On the other end, the Truven Health longitudinal database allowed improved disease and outcome ascertainment with inpatient and outpatient linkage across patients. Together, with over 5 million pregnancies, we examined the risks for rare outcomes (e.g., chorioamnionitis, postpartum hemorrhage) that previous studies have been unable to assess.

Altogether, our study suggests that women with MS may be at moderately increased risk of infections during pregnancy and preterm birth as compared with coetaneous women without MS. However, their risk for other adverse pregnancy outcomes was not meaningfully elevated.

ACKNOWLEDGMENTS

Author affiliations: Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, Massachusetts (Sarah C. MacDonald, Sonia Hernández-Díaz); and Division of Maternal-Fetal Medicine, Department of Obstetrics and Gynecology, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts (Thomas F. McElrath).

S.C.M. was supported by the Canadian Institutes for Health Research, and S.H-.D. was supported by the National Institute of Mental Health (grant R01 MH100216).

We thank Dr. Miguel Hernán for his contributions to the development of the study.

This work was presented as a poster at the 50th Annual Meeting of the Society for Epidemiologic Research, Seattle, Washington, June 20–23, 2017.

Conflict of interest: none declared.

Abbreviations

 
• CI

  confidence interval

 
• ICD-9-CM

  International Statistical Classification of Diseases, Ninth Revision, Clinical Modification

 
• LMP

  last menstrual period

 
• MS

  multiple sclerosis

 
• NIS

  Nationwide Inpatient Sample

 
• RR

  risk ratio