Air Pollution and Risk of Placental Abruption: A Study of Births in New York City, 2008–2014

Abstract

We evaluated the associations of exposure to fine particulate matter (particulate matter with an aerodynamic diameter ≤2.5 μm (PM_2.5) at concentrations of <12 μg/m³, 12–14 μg/m³, and ≥15 μg/m³) and nitrogen dioxide (at concentrations of <26 parts per billion (ppb), 26–29 ppb, and ≥30 ppb) with placental abruption in a prospective cohort study of 685,908 pregnancies in New York, New York (2008–2014). In copollutant analyses, these associations were examined using distributed-lag nonlinear models based on Cox models. The prevalence of abruption was 0.9% (n = 6,025). Compared with a PM_2.5 concentration less than 12 μg/m³, women exposed to PM_2.5 levels of ≥15 μg/m³ in the third trimester had a higher rate of abruption (hazard ratio (HR) = 1.68, 95% confidence interval (CI): 1.41, 2.00). Compared with a nitrogen dioxide concentration less than 26 ppb, women exposed to nitrogen dioxide levels of 26–29 ppb (HR = 1.11, 95% CI: 1.02, 1.20) and ≥30 ppb (HR = 1.06, 95% CI: 0.96, 1.24) in the first trimester had higher rates of abruption. Compared with both PM_2.5 and nitrogen dioxide levels less than the 95th percentile in the third trimester, rates of abruption were increased with both PM_2.5 and nitrogen dioxide ≥95th percentile (HR = 1.44, 95% CI: 1.15, 1.80) and PM_2.5 ≥95th percentile and nitrogen dioxide <95th percentile (HR = 1.43 95% CI: 1.23, 1.66). Increased levels of PM_2.5 exposure in the third trimester and nitrogen dioxide exposure in the first trimester are associated with elevated rates of placental abruption, suggesting that these exposures may be important triggers of premature placental separation through different pathways.

Abbreviations

CI

confidence interval

DLNM

distributed-lag nonlinear model

HR

hazard ratio

NYC

New York City

PM_2.5

particulate matter with an aerodynamic diameter ≤2.5 μm

PM₁₀

particulate matter with an aerodynamic diameter ≤10 μm

ppb

parts per billion

Accumulating evidence suggests that exposure to air pollution—especially particulate matter with an aerodynamic diameter ≤2.5 μm (PM_2.5), particulate matter with an aerodynamic diameter ≤10 μm (PM₁₀), nitrogen dioxide, ozone, and elemental carbon—is associated with increased risks of stillbirth, restricted fetal growth, preeclampsia, and gestational hypertension (1–8). These complications may originate during the stages of placental implantation or early in pregnancy (9, 10), with oxidative stress at the maternal-fetal interface and uteroplacental ischemia leading to inflammation as the chief causal mechanisms (11). Inhalation of fine particulate matter, particularly ultrafine matter, can cause oxidative stress (12–14), a proinflammatory response (15, 16), which may contribute to placental hypoperfusion. This pathway may be responsible for the association between air pollution and abruption and related obstetrical complications.

Placental abruption is the premature separation of the placenta from its implantation site on the uterine wall due to rupture of maternal vessels in the decidua basalis prior to delivery of the fetus; it complicates about 1% of pregnancies (17, 18). Multiple risk factors for this complication have been identified, including advanced maternal age, multiparity, multiple births, African-American race/ethnicity, smoking, cocaine use, a short time interval between pregnancies, chronic hypertension, preeclampsia, and prior abruption (19–22). Abruption is associated with increased risks of stillbirth and perinatal mortality (23), as well as preterm delivery and neurodevelopmental disorders in children (24, 25). Women with abruption also have increased risks of cardiovascular and cerebrovascular complications later in life (26–33).

We have previously hypothesized that abruption can be characterized clinically by 2 distinct pathways: 1) acute-onset abruption resulting from premature placental detachment and 2) chronic abruption secondary to defective placental implantation (17). In a case-crossover analysis, we showed that PM_2.5 and nitrogen dioxide exposures 4–6 days prior to delivery were associated with 15%–25% increased risks of abruption, suggesting that these pollutants may “trigger” acute-onset abruptions secondary to the premature placental detachment pathway (34).

In this study, we hypothesized that exposures incurred around the time of conception and placental implantation may be associated with abruption. If this hypothesis is true, observed associations may provide clues to the origins of defective implantation and an underlying chronic pathophysiological process. Our objectives in this study were to evaluate the associations of exposure to PM_2.5 and nitrogen dioxide with placental abruption. We further examined air pollution exposures during the month before conception, as well as by month and trimester of exposure during pregnancy, to identify critical windows of vulnerability.

METHODS

We designed a cohort analysis of pregnancies leading to births in New York City (NYC) between December 2008 and December 2014. Data were derived from all births taking place in NYC and linked to maternal and newborn hospitalizations. This linkage was undertaken and provided by the New York State Department of Health Statewide Planning and Research Cooperative System (35). Ethics approval was obtained from the institutional review board of Columbia University Irving Medical Center, as well as from the New York City Bureau of Vital Statistics and the Statewide Planning and Research Cooperative System.

Placental abruption

Cases of placental abruption were identified from both the birth certificates (diagnosed by the attendant at delivery) and maternal hospital delivery and discharge records (coded on the basis of International Classification of Diseases, Ninth Revision, Clinical Modification, code 641.2). We ascertained abruption from a diagnosis present in either of these 2 sources.

Air pollution exposure

We ascertained air monitoring data from 2 sources: temporally resolved regulatory monitors from the New York State Department of Environmental Conservation (36) and spatially resolved survey data from the New York City Community Air Survey (Sarah Johnson, New York City Department of Health and Mental Hygiene, unpublished data, 2017). We also collected a wide range of data on emissions sources and meteorological variables, including data on traffic, industrial land use, building area, green space, mixing depth (the vertical distance, measured from the surface of the earth, in which pollutants are vigorously mixed), precipitation, wind speed, and others. For spatial predictors that did not vary temporally, we linked the data to air quality monitoring sites by computing densities within a series of circular buffers of 100 m, 300 m, 500 m, and 1,000 m. For temporal variables, we computed citywide time series for the time period of the study and linked on day. In total, we computed more than 130 different variables used as candidate predictors in the air quality modeling. After splitting the data into training and testing sets, we used 10-fold cross-validation to identify the best modeling approach, tuning parameters, and the best model for predicting daily PM_2.5 and nitrogen dioxide levels. We found that a gradient boosting machine model with a quadratic kernel performed the best, with R² values in predicting the withheld test set of 0.74 (2.38 μg/m³) for PM_2.5 and 0.70 (4.73 parts per billion (ppb)) for nitrogen dioxide. All exposure assignments were at a high spatial resolution (25- × 25-m² grid), and specific details on the exposure modeling and assignment have been previously published (34).

Study population and cohort composition

Of the 840,916 births that took place in NYC between 2008 and 2014, we restricted the study to women who delivered in one of the 5 NYC boroughs. Furthermore, we included exposure data from 1 month before conception and across all 3 trimesters. After all exclusions, 685,908 pregnancies remained (Figure 1).

Figure 1.

Selection of the cohort for a study of air pollution and risk of placental abruption, New York, New York, 2008–2014. Data were derived from all births taking place in New York City (NYC) between December 2008 and December 2014 and linked to maternal and newborn hospitalizations. LMP, last menstrual period.

Potential confounders

We examined several potential confounders, all chosen on the basis of those reported in the literature. Sociodemographic confounders included year of conception, season of conception (spring (March–May), summer (June–August), autumn (September–November), or winter (December–February)), maternal age, parity (1, 2, or ≥3), education (≤8, 9–11, 12, 13–16, or ≥17 years of schooling), marital status, Medicaid enrollment, and plurality (singleton or twin/higher-order birth).

Behavioral factors considered as confounders included smoking (nonsmoker, smoking in the 3 months before pregnancy, or smoking during pregnancy) and prepregnancy body mass index (weight (kg)/height (m)²). We also considered socioeconomic characteristics at the census tract level, including the proportion of households below the federal poverty line and median annual family income (US dollars). For continuous covariates (maternal age, body mass index, proportion of households below the poverty level, and median family income), we also included the squared terms of these covariates in the models for adjustment to allow for a nonlinear relationship.

Statistical analysis

Average PM_2.5 and nitrogen dioxide exposures by gestation month and trimester were calculated from 1 month prior to the estimated conception date to the week of delivery. Associations between average air pollution exposures and time to abruption were examined in 2 ways. First, we examined the associations between abruption and monthly air pollution exposures as continuous variables, coupling distributed-lag nonlinear models (DLNMs) in the setting of Cox proportional hazards regression (37, 38). Second, we evaluated the association based on categorical variables for PM_2.5 concentration (<12 μg/m³, 12–14 μg/m³, and ≥15 μg/m³) and nitrogen dioxide concentration (<26 ppb, 26–29 ppb, and ≥30 ppb) in each trimester, using “standard” (non-DLNM) Cox proportional hazards models. These categorizations correspond approximately to the 75th and 95th percentiles of the respective distributions. Exposure data were simultaneously included in all models for all 4 periods (1 month before conception and 3 trimesters of pregnancy) to account for the potentially high correlations across trimesters, as well as negative control exposures for confounding (39, 40).

For both approaches, analyses were based on fitting a marginal Cox model with time to abruption defined by gestation week, with the start time being the date of conception and the end time being the date of delivery. We accounted for repeat pregnancies in a given woman using the method of generalized estimating equations. We allowed baseline abruption rates to differ according to maternal race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, or other) and hospital. Using the approach of time-to-event analysis, we aligned the data within each gestational-month risk set (41, 42). DLNMs using forward lags were employed to fit all monthly exposures to PM_2.5 and nitrogen dioxide into a single 2-pollutant model (37, 43, 44). The objective of using the 2-pollutant model was to account for potential confounding by copollutant and to obtain surrogates for traffic-related (nitrogen dioxide) and non–traffic-related (PM_2.5) associations with abruption. We used the weighted Akaike’s Information Criterion (45) to evaluate deviations from linearity in the air pollution–abruption association and, if deviations were present, to select the appropriate number of degrees of freedom (37). There was no evidence for a nonlinear exposure-response curve for nitrogen dioxide, but the association between PM_2.5 and abruption was nonlinear with 3 degrees of freedom (see Web Table 1, available online at https://doi.org/10.1093/aje/kwaa259). We estimated associations between air pollution and abruption for an interquartile-range scaled increase in nitrogen dioxide concentration (6 ppb). For the lag constraints in the DLNMs, we assigned 3 degrees of freedom for both lag months and PM_2.5.

A “traditional” marginal Cox proportional hazards model (i.e., non-DLNM) was also used to examine the joint association between the 2 pollutants and abruption. For this analysis, PM_2.5 and nitrogen dioxide levels were categorized as ≥95th percentiles. Joint effects of PM_2.5 and nitrogen dioxide on abruption were assessed on the basis of the relative excess risk due to interaction with 95% confidence intervals on an additive scale (46). We used SAS, version 9.4 (SAS Institute Inc., Cary, North Carolina), and RStudio, version 3.4.2 (R Foundation for Statistical Computing, Vienna, Austria), with the “dlnm” package (version 2.1.3) (47) for statistical analysis.

Sensitivity analyses

We undertook 3 sets of sensitivity analyses. First, we examined the association between persistently high exposure (≥95th percentiles) to both PM_2.5 and nitrogen dioxide from 1 month preconception through the entire pregnancy and rates of abruption based on copollutant models. Second, we repeated the main analyses including only those cases with an abruption diagnosis recorded in both the birth records and the hospitalization data (1,102 abruption cases). Third, since abruption rates and risk factors for abruption differ between singletons and twins/higher-order births, we examined the associations between PM_2.5 and nitrogen dioxide and abruption with the data restricted to singleton births (the numbers of twin and higher-order births were too low for stable analysis).

RESULTS

The proportions of pregnancies to women of advanced maternal age (35–39 years and ≥40 years), multiparous women (parity ≥3), and non-Hispanic Black women were higher in the abruption groups than in the nonabruption group (Table 1). Similarly, single marital status, multiple births, smoking, chronic hypertension, and preeclampsia proportion were higher among abruption groups than in the nonabruption group. We also compared maternal characteristics among abruption births with diagnosis based on both the birth and hospitalization records versus those with diagnosis from neither source. These analyses showed that the distributions of maternal characteristics were similar for the 2 groups.

Table 1.

Distribution of Maternal Characteristics in Relation to Placental Abruption, New York, New York, 2008–2014

					Placental Abruption
Maternal Characteristic	All Births (n = 685,908)		No Abruption (n = 679,813)		Diagnosis From Either Birth Records or Hospital Discharge (n = 6,095)			Diagnosis From Both Birth Records and Hospital Discharge (n = 1,212)
	No. of Births	Column %	No. of Births	Column %	No. of Cases	Column %	Row %	No. of Cases	Column %	Row %
Year of conception
2007	35,360	5.2	35,023	5.2	337	5.5	1.0	70	5.8	0.20
2008	103,193	15.0	102,289	15.0	904	14.8	0.9	196	16.2	0.19
2009	103,010	15.0	102,163	15.0	847	13.9	0.8	189	15.6	0.18
2010	105,562	15.4	104,581	15.4	981	16.1	0.9	187	15.4	0.18
2011	104,153	15.2	103,246	15.2	907	14.9	0.9	169	13.9	0.16
2012	104,502	15.2	103,628	15.2	874	14.3	0.8	155	12.8	0.15
2013	103,955	15.2	103,020	15.2	935	15.3	0.9	187	15.4	0.18
2014	26,173	3.8	25,863	3.8	310	5.1	1.2	59	4.9	0.23
Season of conception
Spring (March–May)	160,490	23.4	159,072	23.4	1,418	23.3	0.9	270	22.3	0.17
Summer (June–August)	153,364	22.4	152,013	22.4	1,351	22.2	0.9	299	24.7	0.20
Autumn (September–November)	185,606	27.1	183,986	27.1	1,620	26.6	0.9	319	26.3	0.20
Winter (December–February)	186,448	27.2	184,742	27.2	1,706	28.0	0.9	324	26.7	0.20
Maternal age, yearsa	29.3 (6.2)		29.2 (6.2)		29.8 (6.4)			30.1 (6.6)
<20	37,043	5.4	36,706	5.5	337	5.5	0.9	66	5.4	0.18
20–24	133,364	19.4	132,285	19.5	1,079	17.7	0.8	211	17.4	0.16
25–29	180,013	26.2	178,579	26.3	1,434	23.5	0.8	286	23.6	0.16
30–34	189,681	27.7	187,956	27.6	1,725	28.3	0.9	316	26.1	0.17
35–39	112,735	16.4	111,605	16.4	1,130	18.5	1.0	240	19.8	0.21
≥40	33,072	4.8	32,682	4.8	390	6.4	1.2	93	7.7	0.28
Parity
1	310,552	45.3	307,847	45.3	2,705	44.4	0.9	551	45.5	0.18
2	204,694	29.8	203,050	29.9	1,644	27.4	0.8	324	26.7	0.16
≥3	170,662	24.8	168,916	24.8	1,746	28.7	1.0	337	27.8	0.20
Race/ethnicity
Non-Hispanic White	206,003	30.0	204,463	30.1	1,540	25.3	0.7	327	27.0	0.16
Non-Hispanic Black	144,042	21.0	141,951	20.9	2,091	34.3	1.5	409	33.7	0.29
Hispanic	217,808	31.8	216,069	31.8	1,739	28.5	0.8	337	27.8	0.16
Other	118,055	17.2	117,330	17.3	725	11.9	0.6	139	11.5	0.12
Education, years
≤8	47,653	6.9	47,284	7.0	369	6.1	0.8	77	6.4	0.16
9–11	112,564	16.4	111,491	16.4	1,073	17.6	1.0	208	17.2	0.19
12	160,119	23.3	158,604	23.3	1,515	24.9	0.9	299	24.7	0.19
13–16	271,703	39.6	269,374	39.6	2,329	38.2	0.9	458	37.8	0.17
≥17	91,506	13.3	90,748	13.3	758	12.4	0.8	155	12.8	0.17
Unknown	2,363	0.3	2,312	0.3	51	0.8	2.2	15	1.2	0.64
Single marital status	299,654	43.7	296,561	43.6	3,093	50.7	1.0	619	51.1	0.21
Medicaid enrollee	419,474	61.2	415,628	61.1	3,846	63.1	0.9	717	59.2	0.17
Multiple births	11,603	1.7	11,365	1.7	238	3.9	2.1	45	3.7	0.39
Stillbirth	9,504	1.4	9,331	1.4	173	2.8	1.8	39	3.2	0.41
Prepregnancy body mass indexa,b	25.1 (5.7)		25.1 (5.7)		25.3 (5.7)			25.4 (5.9)
<18.5	38,023	5.5	37,693	5.5	330	5.4	0.9	62	5.1	0.16
18.5–24.9	364,595	53.2	361,536	53.2	3,059	50.2	0.8	594	49.0	0.16
25.0–29.9	162,838	23.7	161,308	23.7	1,530	25.1	0.9	313	25.8	0.19
30.0–34.9	71,810	10.5	71,114	10.5	696	11.4	1.0	136	11.2	0.19
≥35.0	42,053	6.1	41,677	6.1	376	6.2	0.9	81	6.7	0.19
Unknown	6,589	1.0	6,485	1.0	104	1.7	1.6	26	2.1	0.40
Maternal smokingc	16,233	2.4	16,011	2.4	222	3.6	1.4	55	4.5	0.34
Chronic hypertension	21,308	3.1	20,987	3.1	321	5.3	1.5	66	5.4	0.31
Gestational hypertension	18,062	2.6	17,912	2.6	150	2.5	0.8	19	1.6	0.11
Preeclampsia	28,383	4.1	27,633	4.1	750	12.3	2.6	207	17.1	0.73
Pregestational diabetes	6,095	1.0	6,279	1.0	103	1.7	1.7	21	1.7	0.34
Gestational diabetes	37,861	5.5	37,447	5.5	414	6.8	1.1	87	7.2	0.23
Census tract datad
% of households below federal poverty level	18.2 (8.6–30.2)		18.2 (8.6–30.2)		19.1 (9.5–29.6)			18.2 (8.9–29.5)
Annual income, US dollars	45,417 (32,150–65,595)		45,417 (32,150–65,595)		44,375 (31,750–62,070)			45,714 (32,195–65,313)

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

^b Weight (kg)/height (m)².

^c Smoking in the 3 months before pregnancy or during pregnancy.

^d Values are expressed as median (interquartile range).

Web Table 2 shows the distributions of PM_2.5 and nitrogen dioxide exposure at 1 month prepregnancy and across the trimesters for abruption and nonabruption groups. The association between monthly PM_2.5 exposure and rate of abruption is shown in Figure 2. For pregnancies with exposure data covering up to 10 months during pregnancy, the adjusted copollutant association between PM_2.5 and abruption was higher at months 7–9. For instance, for exposure to PM_2.5 at a concentration of 20 μg/m³ in months 7, 8, and 9, the hazard ratios for abruption were 1.39 (95% confidence interval (CI): 0.96, 2.03), 1.71 (95% CI: 1.01, 2.90), and 2.13 (95% CI: 0.98, 4.61), respectively (Figure 2, Web Table 3). These hazard ratios were considerably stronger for higher levels of exposure to PM_2.5. PM_2.5 was not associated with abruption for deliveries taking place in earlier months. No associations between nitrogen dioxide exposure and abruption were seen (Figure 3). These associations are consistent with the analysis of average trimester exposure (Table 2).

Figure 2.

Association between exposure to particulate matter with an aerodynamic diameter ≤2.5 μm (PM_2.5) from 1 month before pregnancy to 10 months of pregnancy and rate of placental abruption, by month of delivery, New York, New York, 2008–2014. Adjusted hazard ratios were based on copollutant analysis adjusting for nitrogen dioxide exposure and confounders. A) Pregnancies lasting 10 months (n = 630,808); B) pregnancies lasting 9 months (n = 43,746); C) pregnancies lasting 8 months (n = 6,938); D) pregnancies lasting 7 months (n = 893). Distributed-lag nonlinear models based on marginal Cox proportional hazards regression were developed to account for repeat pregnancies in a given woman based on generalized estimating equations, where we allowed baseline abruption rates to differ according to maternal race/ethnicity and hospital at study entry. PM_2.5 exposure was modeled based on 3 degrees of freedom. Hazard ratios were adjusted for year of conception, season of conception, maternal age, maternal age squared, parity, education, smoking 3 months prior to conception or during pregnancy, prepregnancy body mass index (weight (kg)/height (m)²), body mass index squared, marital status, Medicaid enrollment, multiple births, percentage of households below the federal poverty level, percentage of households below the poverty level squared, median annual income, and median income squared.

Figure 3.

Association between nitrogen dioxide exposure from 1 month before pregnancy to 10 months of pregnancy and rate of placental abruption, by month of delivery, New York, New York, 2008–2014. Adjusted hazard ratios (circles) and 95% confidence intervals (bars) for placental abruption were based on a 6–parts per billion increase in nitrogen dioxide exposure, derived from a copollutant analysis adjusting for PM_2.5 exposure and confounders. A) Pregnancies lasting 10 months (n = 630,808); B) pregnancies lasting 9 months (n = 43,746); C) pregnancies lasting 8 months (n = 6,938); D) pregnancies lasting 7 months (n = 893). Distributed-lag nonlinear models based on marginal Cox proportional hazards regression were developed to account for repeat pregnancies in a given woman based on generalized estimating equations, where we allowed baseline abruption rates to differ according to maternal race/ethnicity and hospital at study entry. Hazard ratios were adjusted for year of conception, season of conception, maternal age, maternal age squared, parity, education, smoking 3 months prior to conception or during pregnancy, prepregnancy body mass index (weight (kg)/height (m)²), body mass index squared, marital status, Medicaid enrollment, multiple births, percentage of households below the federal poverty level, percentage of households below the poverty level squared, median annual income, and median income squared.

Table 2.

Associations of Exposure to PM_2.5 and Nitrogen Dioxide With Risk of Placental Abruption During the Month Before Conception and in Each Trimester of Pregnancy^a Among Singleton Pregnancies (Copollutant Model With Exposure Data Categorized as <75th, 75th–94th, and ≥95th Percentiles), New York, New York, 2008–2014

PM 2.5 , μg/m 3						Nitrogen Dioxide, ppb
Level of Exposure	Total No. of Births	Placental Abruption b		Adjusted HR c , d	95% CI	Level of Exposure	Total No. of Births	Placental Abruption b		Adjusted HR c , d	95% CI
		No. of Cases	%					No. of Cases	%
1 Month Before Conception
<12	477,103	3,935	0.82	1.00	Referent	<26	469,832	3,910	0.83	1.00	Referent
12–14	161,512	1,341	0.83	0.99	0.91, 1.07	26–29	145,094	1,176	0.81	0.90	0.83, 0.98
≥15	43,770	371	0.85	0.98	0.86, 1.13	≥30	67,459	561	0.83	0.95	0.83, 1.08
First Trimester
<12	499,974	4,167	0.83	1.00	Referent	<26	478,899	3,922	0.82	1.00	Referent
12–14	165,033	1,349	0.82	0.92	0.85, 1.00	26–29	147,930	1,296	0.88	1.11	1.02, 1.20
≥15	17,378	131	0.76	0.77	0.63, 0.94	≥30	55,556	429	0.77	1.06	0.96, 1.24
Second Trimester
<12	530,618	4,372	0.82	1.00	Referent	<26	504,402	4,207	0.83	1.00	Referent
12–14	142,711	1,201	0.84	1.05	0.97, 1.15	26–29	132,320	1,089	0.82	0.97	0.89, 1.06
≥15	9,056	74	0.82	1.11	0.86, 1.44	≥30	45,663	351	0.77	0.86	0.74, 1.01
Third Trimester
<12	534,646	4,407	0.82	1.00	Referent	<26	517,825	4,298	0.83	1.00	Referent
12–14	128,589	1,024	0.80	1.03	0.94, 1.13	26–29	121,508	985	0.81	1.00	0.92, 1.09
≥15	19,150	216	1.11	1.68	1.41, 2.00	≥30	43,052	364	0.85	1.05	0.91, 1.22

Abbreviations: CI, confidence interval; HR, hazard ratio; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm; ppb, parts per billion.

^a Analyses pertained to women with data on exposure in all 3 trimesters (682,385 total births; placental abruption rate = 0.83%).

^b Placental abruption was based on recorded diagnoses in birth records or hospitalizationdata.

^c A marginal Cox proportional hazards model was fitted to account for repeat pregnancies to a given woman using generalized estimating equations where we allowed baseline abruption rates to differ according to maternal race/ethnicity and hospital at study entry.

^d Hazard ratios were adjusted for year of conception, season of conception, maternal age, maternal age squared, parity, education, smoking 3 months prior to conception or during pregnancy, prepregnancy body mass index (weight (kg)/height (m)²), body mass index squared, marital status, Medicaid enrollment, plurality, percentage of households below the federal poverty level, percentage of households below the poverty level squared, median annual income, and median income squared.

We examined the associations of trimester-specific PM_2.5 and nitrogen dioxide exposure and abruption (Table 2). With available exposure data ranging between 1 month prepregnancy and all 3 trimesters of pregnancy, compared with PM_2.5 less than 12 μg/m³, exposure to PM_2.5 at levels of 12–14 μg/m³ and ≥15 μg/m³ in the first trimester were associated with decreased rates of abruption, whereas exposure to PM_2.5 ≥15 μg/m³ in the third trimester was associated with an increased rate of abruption. In the second trimester, compared with women with PM_2.5 exposure less than 12 μg/m³, abruption rates were increased with PM_2.5 exposures of 12–14 μg/m³ (hazard ratio (HR) = 1.05, 95% CI: 0.97, 1.15) and ≥15 μg/m³ (HR = 1.11, 95% CI: 0.86, 1.44). In contrast, compared with nitrogen dioxide exposure less than 26 ppb, abruption rates were increased with nitrogen dioxide exposures of 26–29 ppb (HR = 1.11, 95% CI: 1.02, 1.20) and ≥30 ppb (HR = 1.06, 95% CI: 0.96, 1.24) in the first trimester.

We examined the joint association of exposure to both PM_2.5 and nitrogen dioxide at concentrations ≥95th percentile and abruption (Table 3). Compared with exposure to both PM_2.5 and nitrogen dioxide at <95th percentile (referent), the rate of abruption was decreased with both PM_2.5 and nitrogen dioxide exposures ≥95th percentile in the first trimester and was increased with PM_2.5 exposure ≥95th percentile with nitrogen dioxide either <95th percentile or ≥95th percentile in the third trimester. However, these associations showed no evidence of interactions on an additive scale.

Table 3.

Association of Joint Exposure to PM_2.5 and Nitrogen Dioxide (Based on 95th Percentile Cutoffs^a) With Risk of Placental Abruption Among Women With Exposure Data on 3 Trimesters of Pregnancy^b, New York, New York, 2008–2014

		Placental Abruption c		Adjusted Hazard Ratio d , e	95% Confidence Interval
Joint Exposure to PM 2.5 and Nitrogen Dioxide	Total No. of Births	No. of Cases	%
1 Month Before Conception
PM2.5 <95th and NO2 <95th	618,099	5,161	0.83	1.00	Referent
PM2.5 ≥95th and NO2 <95th	24,762	209	0.84	0.93	0.80, 1.09
PM2.5 <95th and NO2 ≥95th	24,733	200	0.80	1.04	0.89, 1.23
PM2.5 ≥95th and NO2 ≥95th	9,144	77	0.84	1.11	0.87, 1.41
Excess risk due to interactionf				0.13	−0.41, 0.67
First Trimester
PM2.5 <95th and NO2 <95th	619,738	5,212	0.83	1.00	Referent
PM2.5 ≥95th and NO2 <95th	23,204	186	0.80	0.86	0.73, 1.01
PM2.5 <95th and NO2 ≥95th	23,070	177	0.76	0.90	0.76, 1.07
PM2.5 ≥95th and NO2 ≥95th	10,726	72	0.67	0.73	0.57, 0.94
Excess risk due to interactionf				−0.03	−0.28, 0.23
Second Trimester
PM2.5 <95th and NO2 <95th	619,805	5,179	0.83	1.00	Referent
PM2.5 ≥95th and NO2 <95th	23,015	218	0.94	1.17	1.01, 1.35
PM2.5 <95th and NO2 ≥95th	23,110	178	0.76	0.95	0.80, 1.12
PM2.5 ≥95th and NO2 ≥95th	10,808	72	0.66	0.82	0.63, 1.05
Excess risk due to interactionf				−0.30	−0.59, 0.01
Third Trimester
PM2.5 <95th and NO2 <95th	618,624	5,112	0.82	1.00	Referent
PM2.5 ≥95th and NO2 <95th	24,297	253	1.03	1.43	1.23, 1.66
PM2.5 <95th and NO2 ≥95th	24,396	188	0.76	1.00	0.85, 1.18
PM2.5 ≥95th and NO2 ≥95th	9,421	94	0.99	1.44	1.15, 1.80
Excess risk due to interactionf				0.01	−0.39, 0.42

Abbreviations: NO₂, nitrogen dioxide; PM_2.5, particulate matter with an aerodynamic diameter ≤2.5 μm.

^a The 95th percentile cutoffs for PM_2.5 and NO₂ were derived from the nonabruption births shown in Web Table 2.

^b Analyses pertained to women with exposure data in 3 trimesters (682,385 total births; placental abruption rate, 0.83%).

^c Abruption was based on recorded diagnoses in birth records or hospitalizationdata.

^d A marginal Cox proportional hazards model was fitted to account for repeat pregnancies to a given woman using generalized estimating equations where we allowed baseline abruption rates to differ according to race and individual hospital at study entry.

^e Hazard ratios were adjusted for year of conception, season of conception, maternal age, maternal age squared, parity, education, smoking 3 months prior to conception or during pregnancy, prepregnancy body mass index (weight (kg)/height (m)²), body mass index squared, marital status, Medicaid enrollment, multiple births, percentage of households below the federal poverty level, percentage of households below the poverty level squared, median annual income, and median income squared.

^f Values are expressed as relative excess risk due to interaction.

Sensitivity analyses

We conducted a sensitivity analysis by identifying the abruption cases based on both birth certificates and hospitalization records in order to improve the specificity of abruption diagnosis (Web Table 4). We found that restricting the analysis to highly confirmative abruption cases yielded stronger associations between exposure to high levels of PM_2.5 and nitrogen dioxide and placental abruption.

The rates of abruption among women with consistently high exposures to PM_2.5 between 1 month preconception and through all 3 trimesters at the 95th percentile or higher versus less than the 95th percentile were 1.26% and 0.83%, respectively (HR = 1.74, 95% CI 1.06, 2.84; Web Table 5). No such associations were evident for consistently high exposure to nitrogen dioxide. In an analysis restricted to singleton births, the associations of PM_2.5 and nitrogen dioxide with abruption (Web Table 6) were very similar to those in the overall analyses.

DISCUSSION

This large cohort study of births in NYC showed 3 important findings. First, exposures to high levels of PM_2.5 in the third trimester and high levels of nitrogen dioxide in the first trimester were associated with increased rates of abruption (Table 2). Second, exposure to high levels of PM_2.5 in the first trimester was associated with slightly reduced rates of abruption. Third, a higher risk of abruption was seen among women with consistently high exposure to PM_2.5 ≥95th percentile (but not nitrogen dioxide) between 1 month preconception and the third trimester in comparison with women with consistently low exposure to PM_2.5 (all exposure windows <95th percentile; Web Table 4). These findings, coupled with those of previous studies (34, 48, 49), suggest that air pollution exposures incurred several months before delivery may be associated with an increased rate of placental abruption.

Strengths of the study

The large cohort of NYC births allowed precise estimation of associations. Second, women who deliver in NYC are very socioeconomically and racially diverse, and this enhances the generalizability of our findings. Third, exposure assignment was based on state-of-the-art prediction models with high spatial resolution. Fourth, the replication of findings based on 2 measures of abruption diagnosis (birth records or hospital discharge records vs. both sources) underscores that the associations were robust to the source of information used to define abruption.

Limitations of the data

First, we recognize the potential for error in the measurement of exposures to both PM_2.5 and nitrogen dioxide (50), since exposure assignment was based on the location of residence recorded at the time of delivery; we did not have data on daily activity patterns or on time spent away from the home. We speculate that any bias was probably nondifferential, since we do not believe that this error would have differed for abruption births and nonabruption births. Second, potential exposure misclassification due to maternal residential mobility during pregnancy (residing in NYC but working outside the city or vice versa) or the birth address differing from the early-pregnancy address is likely to have affected the associations to some extent. Only residential address was available at delivery, which may have introduced additional exposure measurement error (51). Third, there is a potential for misclassification of abruption diagnosis on birth records. However, since we based the abruption diagnosis on either birth records or the linked hospitalization records (which results in improved sensitivity when validated against medical record abstraction (52)), we anticipate that the misclassification would be small. The sensitivity analysis restricted to abruptions recorded in both birth and hospitalization records (to improve the specificity) yielded stronger associations between exposure to high levels of PM_2.5 and abruption. Fourth, the potential for the associations being affected by unmeasured confounding, particularly measures of socioeconomic status not considered in this study (53), cannot be overlooked.

Interpretation

Studies have documented increased risks of hypertensive disorders (6–8), gestational diabetes (54, 55), stillbirth (56, 57), small-for-gestational-age birth (58), and preterm delivery (59–61) among women exposed to increased levels of air pollutants, particularly PM_2.5 and nitrogen dioxide (62–64). We initially hypothesized that increased exposure to PM_2.5 and nitrogen dioxide during the month before conception or early in pregnancy would predispose women to abruption. This hypothesis was posited on the premise that abruption can be characterized as an obstetrical complication with defective placental implantation, a process whose origins extend to very early stages of pregnancy. The pathophysiological mechanisms for abruption include uteroplacental ischemia, decidual vasculopathy, inflammation, and oxidative stress (11, 65), all of which are also associated with increased exposure to air pollution (60).

Women exposed to PM_2.5 at concentrations ≥15 μg/m³ in the third trimester and those exposed to PM_2.5 ≥95th percentile were at increased risk of abruption, independent of nitrogen dioxide levels. The only other study that examined PM_2.5 in relation to abruption found no evidence of an association (49). There are some data to support a biologically plausible association between particulate matter and abruption. In a study on the Generation R cohort, van den Hooven et al. (2) reported that PM₁₀ (and nitrogen dioxide) exposures averaged over total pregnancy were associated with higher soluble fms-like tyrosine kinase 1 and lower placental growth factor levels in fetal umbilical cord blood, which is consistent with an anti-angiogenic state and is a phenomenon associated with increased risk of abruption (66).

The increased rate of abruption among women exposed to high levels of PM_2.5 in the third trimester is biologically plausible (67, 68). There is some evidence that exposure to particulate matter is associated with increased risk of premature rupture of the membranes (69, 70), which, in turn, is a strong risk factor for placental abruption (67, 68). The association between nitrogen dioxide in the first and third trimesters and abruption is weak, but it may suggest that nitrogen dioxide exposure may be important for both defective placental implantation and premature placental separation. The associations between third-trimester exposures to PM_2.5 and nitrogen dioxide and abruption are also supported by similar findings in case-crossover studies (34, 48). The observed protective effect of first-trimester PM_2.5 exposure against abruption in our study may have been a consequence of apparent left-truncation bias (71–73). Entry into the cohort was restricted to deliveries occurring at ≥20 weeks’ gestation, so earlier losses, arguably attributable to air pollution (71, 74) and/or abruption (23, 75–78), may have been selectively excluded.

Conclusions

This study of over 685,000 births in NYC showed that exposure to high levels of PM_2.5 in the third trimester and nitrogen dioxide in the first trimester were associated with increased rates of abruption. These findings, coupled with prior research on air pollution exposures a few days prior to delivery (34, 48), underscore that these exposures may be implicated as important environmental triggers of premature placental separation through different pathways (10).