Maternal exposure to nitrosamines in drinking water during pregnancy and birth outcomes in a Chinese cohort

Abstract

Maternal exposure to regulated disinfection by-products (DBPs) during pregnancy has been linked with adverse birth outcomes. However, no human studies have focused on drinking water nitrosamines, a group of emerging unregulated nitrogenous DBPs that exhibits genotoxicity and developmental toxicity in experimental studies. This cohort study included 2457 mother-infant pairs from a single drinking water supply system in central China, and maternal trimester-specific and entire pregnancy exposure of drinking water nitrosamines were evaluated. Multivariable linear and Poisson regression models were used to estimate the associations between maternal exposure to nitrosamines in drinking water and birth outcomes [birth weight (BW), low birth weight (LBW), small for gestational age (SGA) and preterm delivery (PTD)]. Elevated maternal N-nitrosodimethylamine (NDMA) exposure in the second trimester and N-nitrosopiperidine (NPIP) exposure during the entire pregnancy were associated with decreased BW (e.g., β = −88.6 g; 95% CI: −151.0, −26.1 for the highest vs. lowest tertile of NDMA; p for trend = 0.01) and increased risks of PTD [e.g., risk ratio (RR) = 2.16; 95% CI: 1.23, 3.79 for the highest vs. lowest tertile of NDMA; p for trend = 0.002]. Elevated maternal exposure of N-nitrosodiethylamine (NDEA) in the second trimester was associated with increased risk of SGA (RR = 1.80; 95% CI: 1.09, 2.98 for the highest vs. lowest tertile; p for trend = 0.01). Our study detected associations of maternal exposure to drinking water nitrosamines during pregnancy with decreased BW and increased risks of SGA and PTD. These findings are novel but require replication in other study populations.

Graphical Abstract

1. Introduction

Use of drinking water disinfectants is protective of human health as it reduces the likelihood of spread of water-borne pathogens, although harmful disinfection byproducts (DBPs) are produced simultaneously (Bond et al. 2015; Krasner 2009a; Zhang et al. 2015). More than 600 DBPs in drinking water have been reported (Richardson et al. 2007). Exposure to DBPs is associated with multiple adverse human health outcomes (Mashau et al. 2018; Richardson et al. 2007), of which reproductive and developmental toxicity is one of particular emerging concern (Kim et al. 2020; Narotsky et al. 2015; Säve-Söderbergh et al. 2020; Smith et al. 2016). Although the results are inconsistent, a number of studies have provided evidence of associations between maternal exposure to regulated DBPs including trihalomethanes (THMs) and haloacetic acids (HAAs) during pregnancy and adverse developmental outcomes, such as small for gestational age (SGA), reduced birth weight (BW), prematurity, congenital anomalies, and stillbirth (Cao et al. 2016; Hinckley et al. 2005; Nieuwenhuijsen et al. 2008; Smith et al. 2016; Toledano et al. 2005; Wright et al. 2004). However, no epidemiological studies of reproductive and developmental outcomes to date have focused on nitrosamines in drinking water, a group of emerging unregulated DBPs (Krasner et al. 2012).

Nitrosamines in drinking water are largely formed from disinfection using chloramination, as well as chlorine dioxide and ozonation (Chuang et al. 2013; Kristiana et al. 2013; Zhao et al. 2008). To date, several kinds of nitrosamines, including N-nitrosodiethylamine (NDEA), N-nitrosodiethylamine (NDEA), N-nitrosopiperidine (NPIP), N-nitrosomethylethylamine (NMEA), N-nitrosodi-n-butylamine (NDBA), N-nitrosopyrrolidine (NPYR), N-nitrosomorpholine (NMOR), N-nitrosodi-n-propylamine (NDPA) and N-nitrosodiphenylamine (NDPhA) have been detected in drinking water systems around the world (Asami et al. 2009; Bei et al. 2016; Goslan et al. 2009; Jurado-Sánchez et al. 2012; Krasner et al. 2013; US EPA 2016; Woods et al. 2015). The concentrations of nitrosamines in drinking water can vary from undetectable to hundreds of ng/L across different geographic regions and water systems and is related to many factors, such as nitrogenous organic matter as precursors in raw water, type, amount and timing/order of disinfectants application, and the use of anion exchange resins or amine-based cationic coagulation polymers (Bei et al. 2020; Kemper et al. 2009; McCurry et al. 2015; Uzun et al. 2015; Zhao et al. 2008).

Although the detected concentrations of nitrosamines in drinking water are relatively low compared to regulated DBPs such as THMs and HAAs, accumulating evidence from experimental and epidemiological studies raise concern that nitrosamines may pose greater health risks than those regulated DBPs such as THMs due to their high genotoxicity, cytotoxicity and carcinogenicity (Bond et al. 2011; Richardson et al. 2007; US EPA 2016). Moreover, human placental perfusion and animal experiments have demonstrated that nitrosamines, such as NDMA and NDEA, can cross through the placenta into the developing fetus (Annola et al. 2009; Donovan and Smith 2008; Pour 1986). Studies in fish embryos show nitrosamines can induce developmental toxicity including increases in the rate of embryo mortality and incidence of morphological abnormalities (Chaves et al. 2020; Park et al. 1992). The zebrafish study by Chaves et al. (2020) also reported more developmental toxicity for the nitrosamines compared to THMs. Nitrosamines have also been shown to induce disruption of thyroid and reproductive hormones in rabbits and rats (Sheweita et al. 2017; Somade et al. 2016), which may indirectly influence the developing fetus.

In the present study, we examined whether maternal exposure to drinking water nitrosamines during pregnancy was associated with birth outcomes [BW, low birth weight (LBW), SGA and preterm delivery (PTD)] in a Chinese cohort. To the best of our knowledge, this is the first epidemiological study to investigate the effects of maternal exposure to drinking water nitrosamines during pregnancy on birth outcomes.

2. Methods

2.1. Study site and population

Our population-based cohort study was conducted in a city in central China (Hubei Province), where residents’ drinking water were supplied by a single water supply system, as described in our prior study (Luo et al. 2020). The source water of the city was taken from the Hanjiang River, one big branch of the Yangtze River, and the disinfection process for local water supply system is the combination of chlorine dioxide and chlorine. We included a total of 3,345 pregnant women who lived in the city and gave births at a local hospital during October 2015 and December 2016. Maternal and infant information including maternal age at delivery, race, education level, drinking and smoking status, hukou status (a population registration management system in China, including rural and urban hukou), gravidity, delivery mode, infant sex and birth size were obtained from the hospital records. No subjects reported that they had cigarette smoking or alcohol consumption during pregnancy. The inclusion criteria and excluded participants are detailed in Figure 1 with analyses restricted to singleton live births (n=66 exclusions). Mothers with missing information on residential address at time of delivery were not included (n=84). We also excluded infants who had birth defects (n=3). A total of 2,457 mother-infant pairs were included in the analysis. The available data provided by the hospital were de-identified and was approved by the Ethics Committee of Tongji Medical College (2016: IEC-S077).

Fig. 1.

The inclusion and exclusion of study population.

2.2. Birth outcomes

Birth outcomes including BW in grams and gestational age (GA) in weeks were extracted from the local hospital records. GA was calculated based on the interval between the date of infant delivery and the self-reported last menstrual period. An infant with BW at the lowest 10% for GA according to a national Chinese referent population was categorized as SGA (Chen and Jin 2011). An infant was identified as LBW if the BW was less than 2,500 g at the time of birth. PTD was defined as infants born prior to 37 completed weeks of gestation.

2.3. Exposure assessment

A monitoring survey of drinking water nitrosamine concentrations within this water distribution system in the city was carried out as described in detail elsewhere (Bei et al. 2016; Luo et al. 2020). Briefly, a total of 90 drinking water samples were collected monthly during October 2015 and December 2016 from six monitoring sites in the water supply system. A total of nine nitrosamines including NDMA, NDEA, NPIP, NMEA, NDBA, NPYR, NMOR, NDPA and NDPhA were analyzed by solid-phase extraction coupled with liquid chromatograph-mass spectrometer. The recoveries of matrix spikes and the method detection limits (MDLs) for the target compounds ranged from 64.2% to 115.6% and 0.5 ng/L to 1.0 ng/L, respectively. A value of MDL/2 was assigned to measurements below the MDL (Almberg et al. 2017). The monitoring results showed that NDMA, NDEA and NPIP were frequently detected (>65%), while the others were undetectable or detected in low frequency (<25%), as described in detail in our prior study (Luo et al. 2020). Thus, we included NDMA, NDEA and NPIP into the following analyses in this study. The total nitrosamines (TNA3) were defined as the sum of the concentrations of NDMA, NDEA and NPIP, which was highly correlated with NDMA but weekly correlated with NDEA and NPIP (data not shown).

Average nitrosamine exposures in the first (1–93 days), second (94–186 days) and third (187-the day proceeding delivery) trimesters, as well as entire pregnancy were estimated based on the detected nitrosamine concentrations and maternal residential address, following a previous method of time-weighted exposure assessment for drinking water DBPs (Ileka-Priouzeau et al. 2015; Lewis et al. 2007). In brief, we employed monthly data from the monitoring site nearest to maternal residential address to assign the daily nitrosamine exposures, assuming that the daily maternal exposure was constant during each month. The straight-line distance between the monitoring sites and the mother’s residential address was calculated by latitude and longitude. We calculated the exposure during pregnancy to nitrosamines for each mother by multiplying the monthly monitoring values of nitrosamines and the days of pregnancy falling in each month. We then calculated the average exposure of nitrosamines during trimester-specific and entire pregnancy by dividing the total nitrosamine exposures by the number of gestational days. The entire pregnancy length in days was estimated by subtracting the date of delivery from the date of last menstrual period.

2.4. Statistical analysis

Descriptive analyses for maternal and infant characteristics and averaged nitrosamine exposures during pregnancy were performed. Spearman correlation coefficients (ρ) were determined to measure the correlations between each nitrosamine and across critical time windows. T tests and chi-square tests were used to examine the differences of characteristics between included participants and excluded participants in this study. Statistical analyses were conducted by Statistical Package for Social Science (version 25.0, SPSS, Inc., USA) and R software (version 3.6.3, R Foundation for Statistical Computing, Austria). A p-value less than 0.05 was recognized as statistically significant.

For the continuous birth outcome such as BW, we estimated the associations with trimester-specific and entire pregnancy averaged nitrosamine exposures by using linear regression models. For the categorical outcomes such as SGA, LBW and PTD, we used Poisson regression models to estimate risk ratios (RRs) for the outcomes. In all analyses, tertiles of NDMA, NDEA, NPIP and TNA3 exposure variables were categorized based on the average exposures during the entire pregnancy. The regression coefficients (βs) or (RRs) and their 95% confidence intervals (CIs) for the higher tertiles of nitrosamine exposures compared to the first tertiles were estimated. To test the trend of exposure-response association, tertiles of nitrosamine variables were included as continuous variable using integer values (0, 1 and 2) in the regression models.

The determination of covariates in the final models were based on a priori and statistical considerations. Covariates (maternal age, infant sex, education level and hukou status) in relation to both the exposures and the outcomes based on previously reported literature were included in the final models (Ebisu and Bell 2012; Kogevinas et al. 2016; Smith et al. 2016; Villanueva et al. 2011). Moreover, the gravidity which resulted in changes of the effect estimates (RRs or regression coefficients) for the associations between the exposures and the outcomes by ≥ 10% were also retained in the final models. The retained covariates included maternal age (age squared) as continuous variables and infant sex (girls vs. boys), education level (college and above vs. high school or less), hukou status (rural vs. urban) and gravidity (≥2 vs. 1) as dichotomous variables in the models.

3. Results

Table 1 shows the characteristics of mothers and infants included in this study. Mothers were nearly all ethnic Han and had a mean age (± standard deviation, SD) of 28.4 (± 4.5) years old at delivery. Most mothers reported their education levels as high school or less (71.1%) and hukou status as rural (63.0%). Mothers who had never conceived before (45.1%) and who experienced spontaneous labor (43.8%) in the current birth accounted for less than half of the study population. Of the 2457 single gestation live infants, 100 (4.1%) were classified as SGA and PTD, and 64 (2.6%) were classified as LBW. The mean (± SD) GA and BW were 38.8 (± 1.3) weeks and 3311.5 (± 419.5) g, respectively. There were no statistically significant differences between included participants and excluded participants in the characteristics except for education and urbanicity (Table S1).

Table 1.

Characteristics of study subjects (n = 2457).

Characteristics	Mean ± SD or n (%)
Maternal characteristics
Maternal age at delivery (years)	28.4 ± 4.5
Race
Han	2412 (98.2%)
Other	12 (0.5%)
Missing observations	33 (1.3%)
Education level
High school or less	1747 (71.1%)
College and above	626 (25.5%)
Missing observations	84 (3.4%)
Hukou status
Rural	1549 (63.0%)
Urban	862 (35.1%)
Missing observations	46 (1.9%)
Gravidity
1	1108 (45.1%)
≥2	1349 (54.9%)
Delivery mode
Spontaneous labor	1077 (43.8%)
Cesarean delivery	1380 (56.2%)
Infant characteristics
Sex
Boys	1382 (56.2%)
Girls	1075 (43.8%)
Gestational age (weeks)	38.8 ± 1.3
<37	100 (4.1%)
≥37	2357 (95.9%)
Birth weight (g)	3311.5 ± 419.5
< 2500	64 (2.6%)
≥ 2500	2393 (97.4%)
SGA
Yes	100 (4.1%)
No	2357 (95.9%)

Note: SGA, small for gestational age.

The distribution of maternal exposure to drinking water nitrosamines during pregnancy is presented in Table 2. Among the analyzed nitrosamines, NDMA was the highest in terms of maternal average exposure, followed by NDEA and NPIP. Average TNA3 exposures in the first, second, third trimesters and entire pregnancy were 29.5 ng/L, 19.7 ng/L, 24.6 ng/L and 26.1 ng/L, respectively. Statistically significant correlations among the maternal exposures of drinking water most nitrosamines were observed in each period of gestation (Table S2, ρ=−0.63 to 0.98) and across different periods of gestation (Table S3, ρ=−0.65 to 0.85).

Table 2.

Distribution of maternal exposure to drinking water nitrosamines (ng/L) during pregnancy.

	Min	25tha	Median	75tha	Max	Mean ± SD
First-trimester
NDMA	1.9	8.0	16.7	35.3	51.5	21.8 ± 15.4
NDEA	0.2	2.2	3.4	5.6	9.6	4.0 ± 2.4
NPIP	1.1	3.1	3.4	4.1	7.3	3.8 ± 1.1
TNA3b	4.1	15.4	24.8	44.0	58.7	29.5 ± 15.9
Second-trimester
NDMA	1.9	8.6	12.4	15.9	48.5	13.0 ± 6.5
NDEA	0.3	2.7	4.1	5.2	9.5	4.0 ± 1.7
NPIP	0.4	1.4	3.3	4.6	7.3	3.2 ± 1.9
TNA3b	5.7	16.1	19.1	21.9	52.2	19.7 ± 6.5
Third-trimester
NDMA	4.2	13.5	18.6	28.4	43.2	21.1 ± 8.4
NDEA	0.5	1.3	2.1	4.4	7.6	2.8 ± 1.7
NPIP	0.3	0.7	1.3	1.7	10.5	1.7 ± 1.4
TNA3b	9.3	19.7	22.6	29.7	45.4	24.6 ± 7.0
Entire pregnancy
NDMA	7.8	15.9	18.4	21.4	32.0	18.6 ± 4.1
NDEA	1.9	2.9	3.3	4.1	7.5	3.6 ± 1.1
NPIP	1.8	2.5	2.9	3.3	4.9	2.9 ± 0.6
TNA3b	14.8	22.4	25.8	30.7	39.2	26.1 ± 5.0

^aPercentile.

^bTNA3 is the sum of NDMA, NDEA and NPIP.

Table 3 presents the associations between maternal nitrosamine exposures during pregnancy and BW. We found inverse exposure-response associations of maternal exposures to NDMA in the second trimester and exposures to NPIP during entire pregnancy with infant BW (both p for trends ≤ 0.05). The highest versus lowest tertile of maternal NDMA and NPIP exposures had lower mean infant BW of 88.6 g (95% CI: −151.0, −26.1) and 41.7 g (95% CI: −83.2, −0.3), respectively. We did not find exposure-response associations between maternal nitrosamine exposures during pregnancy and the risk of LBW (Table 4).

Table 3.

Regression coefficients^a [β (95% CI)] for BW associated with maternal tertile exposure to drinking water nitrosamines during pregnancy (n = 2457).

	First-trimester	Second-trimester	Third-trimester	Entire pregnancy
	β (95% CI)	β (95% CI)	β (95% CI)	β (95% CI)
NDMA (ng/L)
T1 (< 16.7)	0.0 (reference)	0.0 (reference)	0.0 (reference)	0.0 (reference)
T2 (16.7 – 20.2)	15.8 (−64.9, 96.6)	−20.0 (−69.5, 29.6)	69.8 (23.1, 116.4)	−5.9 (−47.3, 35.5)
T3 (> 20.2)	−2.3 (−37.3, 32.9)	−88.6 (−151.0, −26.1)	22.3 (−16.6, 61.2)	−40.7 (−82.5, 1.1)
P for trend	0.90	0.01	0.34	0.06
NDEA (ng/L)
T1 (< 3.0)	0.0 (reference)	0.0 (reference)	0.0 (reference)	0.0 (reference)
T2 (3.0 – 3.8)	−15.2 (−71.1, 40.7)	30.5 (−23.3, 84.3)	−21.3 (−88.5, 45.8)	−37.8 (−79.3, 3.6)
T3 (> 3.8)	−13.6 (−49.9, 22.7)	17.0 (−21.5, 55.6)	−5.7 (−42.8, 31.3)	3.5 (−38.2, 45.2)
P for trend	0.46	0.45	0.73	0.90
NPIP (ng/L)
T1 (< 2.6)	0.0 (reference)	0.0 (reference)	0.0 (reference)	0.0 (reference)
T2 (2.6 – 3.2)	−33.8 (−109.4, 41.8)	35.7 (−23.7, 95.1)	−106.9 (−224.7, 10.9)	−5.1 (−46.7, 36.5)
T3 (> 3.2)	−49.3 (−116.5, 17.8)	26.9 (−9.6, 63.5)	−8.2 (−56.0, 39.6)	−41.7 (−83.2, −0.3)
P for trend	0.14	0.16	0.57	0.05
TNA3 (ng/L)
T1 (< 21.4)	0.0 (reference)	0.0 (reference)	0.0 (reference)	0.0 (reference)
T2 (21.4 – 27.1)	−13.8 (−77.0, 49.3)	−15.1 (−56.7, 26.4)	−2.9 (−45.7, 39.9)	−0.9 (−42.3, 40.4)
T3 (> 27.1)	3.8 (−32.0, 39.6)	−16.9 (−82.5, 48.6)	−7.8 (−47.7, 32.1)	−22.1 (−63.9, 19.6)
P for trend	0.83	0.44	0.70	0.30

Note: BW, birth weight, NDMA, N-nitrosodimethylamine; NDEA, N-nitrosodiethylamine; NPIP, N-nitrosopiperidine; TNA3, total nitrosamines, the sum of NDMA, NDEA and NPIP.

^aAdjusted for maternal age at delivery (squared, years²), hukou status (rural vs. urban), education level (college and above vs. high school or less), gravidity (≥2 vs. 1) and infant sex (girls vs. boys).

Table 4.

Risk ratios^a [RRs (95% CI)] for LBW associated with maternal tertile exposure to drinking water nitrosamines during pregnancy (n = 2457).

	First-trimester		Second-trimester		Third-trimester		Entire pregnancy
	RR (95% CI)	n/N	RR (95% CI)	n/N	RR (95% CI)	n/N	RR (95% CI)	n/N
NDMA (ng/L)
T1 (< 16.7)	1.0 (reference)	36/1229	1.0 (reference)	45/1923	1.0 (reference)	21/844	1.0 (reference)	23/823
T2 (16.7 – 20.2)	1.05 (0.32, 3.44)	3/123	1.40 (0.72, 2.72)	12/333	0.89 (0.44, 1.81)	12/527	0.99 (0.55, 1.78)	23/815
T3 (> 20.2)	0.92 (0.54, 1.55)	25/1105	1.62 (0.73, 3.60)	7/201	0.94 (0.53, 1.66)	31/1086	0.83 (0.44, 1.58)	18/819
P for trend	0.76		0.16		0.84		0.59
NDEA (ng/L)
T1 (< 3.0)	1.0 (reference)	32/1120	1.0 (reference)	18/739	1.0 (reference)	43/1472	1.0 (reference)	17/835
T2 (3.0 – 3.8)	0.89 (0.39, 2.04)	7/288	0.90 (0.39, 2.08)	9/371	1.31 (0.55, 3.09)	6/179	1.91 (1.02, 3.57)	29/809
T3 (> 3.8)	0.90 (0.52, 1.54)	25/1049	1.10 (0.62, 1.96)	37/1347	0.73 (0.40, 1.32)	15/806	1.32 (0.66, 2.63)	18/813
P for trend	0.70		0.70		0.34		0.44
NPIP (ng/L)
T1 (< 2.6)	1.0 (reference)	7/172	1.0 (reference)	31/906	1.0 (reference)	55/2035	1.0 (reference)	28/822
T2 (2.6 – 3.2)	0.53 (0.18, 1.52)	8/449	0.39 (0.12, 1.27)	3/255	1.76 (0.42, 7.27)	2/56	0.63 (0.33, 1.20)	15/815
T3 (> 3.2)	0.72 (0.31, 1.70)	49/1836	0.75 (0.44, 1.25)	30/1296	0.79 (0.36, 1.75)	7/366	0.86 (0.48, 1.54)	21/820
P for trend	0.87		0.28		0.65		0.60
TNA3 (ng/L)
T1 (< 21.4)	1.0 (reference)	33/1076	1.0 (reference)	44/1732	1.0 (reference)	26/1084	1.0 (reference)	17/817
T2 (21.4 – 27.1)	0.52 (0.16, 1.71)	4/218	1.38 (0.79, 2.44)	17/543	1.02 (0.54, 1.93)	16/597	1.76 (0.95, 3.26)	30/823
T3 (> 27.1)	0.89 (0.53, 1.50)	27/1163	0.74 (0.23, 2.39)	3/182	0.99 (0.55, 1.78)	22/776	1.16 (0.58, 2.30)	17/817
P for trend	0.65		0.78		0.98		0.66

Note: LBW, low birth weight, NDMA, N-nitrosodimethylamine; NDEA, N-nitrosodiethylamine; NPIP, N-nitrosopiperidine; TNA3, the sum of NDMA, NDEA and NPIP; n, number of cases in each exposure category; N, total number of participants in each exposure category; RR, risk ratio.

^aAdjusted for maternal age at delivery (squared, years²), hukou status (rural vs. urban), education level (college and above vs. high school or less), gravidity (≥2 vs. 1) and infant sex (girls vs. boys).

As shown in Table 5, we detected an exposure-response association between maternal exposure to NDEA during the second trimester and increased risk of SGA (p for trend = 0.01) with an elevated RR of 1.80 (95% CI:1.09, 2.98) for the highest versus lowest tertile of maternal NDEA exposure. We also observed an inverse exposure-response between maternal NPIP exposure during entire pregnancy and risk of SGA (p for trend = 0.03) with an decreased RR of 0.55 (95% CI: 0.33, 0.94) for the highest versus lowest tertile of maternal NPIP exposure.

Table 5.

Risk ratios^a [RRs (95% CI)] for SGA associated with maternal tertile exposure to drinking water nitrosamines during pregnancy (n = 2457).

	First-trimester		Second-trimester		Third-trimester		Entire pregnancy
	RR (95% CI)	n/N	RR (95% CI)	n/N	RR (95% CI)	n/N	RR (95% CI)	n/N
NDMA (ng/L)
T1 (< 16.7)	1.0 (reference)	52/1229	1.0 (reference)	77/1923	1.0 (reference)	28/844	1.0 (reference)	28/823
T2 (16.7 – 20.2)	0.82 (0.29, 2.28)	4/123	1.09 (0.61, 1.93)	15/333	1.54 (0.90, 2.64)	26/527	1.44 (0.88, 2.36)	39/815
T3 (> 20.2)	0.98 (0.65, 1.48)	44/1105	1.12 (0.54, 2.33)	8/201	1.24 (0.76, 2.03)	46/1086	1.19 (0.71, 2.00)	33/819
P for trend	0.92		0.70		0.44		0.53
NDEA (ng/L)
T1 (< 3.0)	1.0 (reference)	42/1120	1.0 (reference)	21/739	1.0 (reference)	64/1472	1.0 (reference)	31/835
T2 (3.0 – 3.8)	1.61 (0.90, 2.90)	16/288	1.06 (0.50, 2.28)	11/371	0.78 (0.34, 1.82)	6/179	1.08 (0.65, 1.79)	33/809
T3 (> 3.8)	1.13 (0.72, 1.75)	42/1049	1.80 (1.09, 2.98)	68/1347	0.85 (0.54, 1.32)	30/806	1.24 (0.76, 2.04)	36/813
P for trend	0.60		0.01		0.45		0.38
NPIP (ng/L)
T1 (< 2.6)	1.0 (reference)	10/172	1.0 (reference)	45/906	1.0 (reference)	85/2035	1.0 (reference)	42/822
T2 (2.6 – 3.2)	0.67 (0.30, 1.53)	16/449	0.33 (0.12, 0.91)	4/255	2.67 (1.16, 6.17)	6/56	0.89 (0.56, 1.41)	36/815
T3 (> 3.2)	0.74 (0.37, 1.48)	74/1836	0.78 (0.51, 1.18)	51/1296	0.64 (0.32, 1.27)	9/366	0.55 (0.33, 0.94)	22/820
P for trend	0.62		0.27		0.38		0.03
TNA3 (ng/L)
T1 (< 21.4)	1.0 (reference)	46/1076	1.0 (reference)	66/1732	1.0 (reference)	42/1084	1.0 (reference)	23/817
T2 (21.4 – 27.1)	0.92 (0.43, 1.96)	9/218	1.35 (0.85, 2.13)	27/543	0.93 (0.54, 1.58)	21/597	1.88 (1.12, 3.16)	44/823
T3 (> 27.1)	0.93 (0.61, 1.42)	45/1163	1.09 (0.50, 2.38)	7/182	1.21 (0.77, 1.92)	37/776	1.49 (0.86, 2.57)	33/817
P for trend	0.74		0.39		0.43		0.18

Note: SGA, small for gestational age, NDMA, N-nitrosodimethylamine; NDEA, N-nitrosodiethylamine; NPIP, N-nitrosopiperidine; TNA3, the sum of NDMA, NDEA and NPIP; n, number of cases in each exposure category; N, total number of participants in each exposure category; RR, risk ratio.

^aAdjusted for maternal age at delivery (squared, years²), hukou status (rural vs. urban), education level (college and above vs. high school or less), gravidity (≥2 vs. 1) and infant sex (girls vs. boys).

As shown in Table 6, we found an exposure-response relationship for risks of PTD for mothers with higher second trimester exposure of NDMA (RR = 1.73, 95% CI: 1.05, 2.86 and RR =2.16, 95% CI: 1.23, 3.79 for the second and third versus first tertile, respectively; p for trend = 0.002). We also detected an exposure-response association between maternal exposure to NPIP during entire pregnancy and elevated risk of PTD (RR = 1.66, 95% CI: 1.03, 2.68 for the third versus first tertile; p for trend = 0.03). The above-mentioned associations between maternal nitrosamine exposures during pregnancy and infant birth outcomes (BW, LBW, SGA and PTD) in the adjusted models were similar to the crude models (Table S4–S7).

Table 6.

Risk ratios^a [RRs (95% CI)] for PTD associated with maternal tertile exposure to drinking water nitrosamines during pregnancy (n = 2457).

	First-trimester		Second-trimester		Third-trimester		Entire pregnancy
	RR (95% CI)	n/N	RR (95% CI)	n/N	RR (95% CI)	n/N	RR (95% CI)	n/N
NDMA (ng/L)
T1 (< 16.7)	1.0 (reference)	55/1229	1.0 (reference)	65/1923	1.0 (reference)	41/844	1.0 (reference)	33/823
T2 (16.7 – 20.2)	0.40 (0.10, 1.64)	2/123	1.73 (1.05, 2.86)	20/333	0.55 (0.30, 1.01)	14/527	0.92 (0.56, 1.50)	31/815
T3 (> 20.2)	0.91 (0.61, 1.36)	43/1105	2.16 (1.23, 3.79)	15/201	0.82 (0.53, 1.26)	45/1086	1.12 (0.70, 1.81)	36/819
P for trend	0.63		0.002		0.40		0.63
NDEA (ng/L)
T1 (< 3.0)	1.0 (reference)	42/1120	1.0 (reference)	29/739	1.0 (reference)	59/1472	1.0 (reference)	28/835
T2 (3.0 – 3.8)	0.79 (0.38, 1.63)	9/288	1.28 (0.72, 2.29)	19/371	2.11 (1.18, 3.78)	14/179	1.47 (0.91, 2.37)	41/809
T3 (> 3.8)	1.25 (0.83, 1.89)	49/1049	0.94 (0.60, 1.48)	52/1347	0.87 (0.55, 1.37)	27/806	1.14 (0.68, 1.91)	31/813
P for trend	0.29		0.69		0.73		0.62
NPIP (ng/L)
T1 (< 2.6)	1.0 (reference)	4/172	1.0 (reference)	46/906	1.0 (reference)	83/2035	1.0 (reference)	27/822
T2 (2.6 – 3.2)	1.72 (0.58, 5.12)	17/449	0.74 (0.36, 1.52)	9/255	-	0/56	1.08 (0.64, 1.83)	28/815
T3 (> 3.2)	1.77 (0.65, 4.85)	79/1836	0.71 (0.47, 1.08)	45/1296	1.17 (0.69, 1.97)	17/366	1.66 (1.03, 2.68)	45/820
P for trend	0.35		0.11		-		0.03
TNA3 (ng/L)
T1 (< 21.4)	1.0 (reference)	49/1076	1.0 (reference)	64/1732	1.0 (reference)	46/1084	1.0 (reference)	31/817
T2 (21.4 – 27.1)	0.73 (0.33, 1.61)	7/218	1.48 (0.95, 2.31)	28/543	0.85 (0.51, 1.41)	22/597	1.13 (0.69, 1.83)	35/823
T3 (> 27.1)	0.87 (0.57, 1.31)	44/1163	1.20 (0.58, 2.52)	8/182	0.94 (0.60, 1.48)	32/776	1.14 (0.70, 1.85)	34/817
P for trend	0.49		0.19		0.75		0.61

Note: PTD, preterm delivery, NDMA, N-nitrosodimethylamine; NDEA, N-nitrosodiethylamine; NPIP, N-nitrosopiperidine; TNA3, the sum of NDMA, NDEA and NPIP; n, number of cases in each exposure category; N, total number of participants in each exposure category; RR, risk ratio.

^aAdjusted for maternal age at delivery (squared, years2), hukou status (rural vs. urban), education level (college and above vs. high school or less), gravidity (≥2 vs. 1) and infant sex (girls vs. boys).

4. Discussion

In this Chinese birth cohort, we are the first study to examine the effects of maternal exposure to drinking water nitrosamines during pregnancy on birth outcomes. We found evidence of inverse exposure-response associations between both elevated maternal NDMA exposure in the second trimester and NPIP exposure during entire pregnancy and infant BW. Moreover, we observed that maternal exposure of NDEA in the second trimester was positively associated with risk of SGA and that maternal NDMA exposure in the second trimester and NPIP exposure during entire pregnancy were positively associated with risk of PTD.

There are numerous epidemiological studies investigating the associations between maternal exposure of drinking water regulated DBPs such as THMs and HAAs during pregnancy and birth outcomes (Grellier et al. 2010; Ileka-Priouzeau et al. 2015; Säve-Söderbergh et al. 2020; Smith et al. 2016; Summerhayes et al. 2020). Few epidemiological studies have focused on some of the emerging DBPs such as haloacetaldehydes (HAs) and haloacetonitriles (HANs) (Colman et al. 2011; Ileka-Priouzeau et al. 2015). To our knowledge, no epidemiological studies to date have been conducted to explore the effects of maternal exposure to nitrosamines in drinking water on birth outcomes, despite experimental evidence showing that nitrosamines can cause adverse effects on embryo development with an increasing rate of embryo mortality and incidence of morphological abnormalities (Chaves et al. 2020; Park et al. 1992).

BW has been suggested to be an important predictor of neonatal morbidity and mortality and related with chronic disease risks in later life (Saigal and Doyle 2008; Whincup et al. 2008). In the present study, we observed evidence of maternal NDMA exposure in the second trimester and NPIP exposure during entire pregnancy in associations with reduced infant BW. Similar to our results, studies from Lithuania, America, and England observed dose-response relationships between exposure to individual THMs in drinking water during the entire pregnancy and reduction in BW (Grazuleviciene et al. 2011; Smith et al. 2016; Wright et al. 2003). Several studies also found that exposures to total THMs (Lewis et al. 2006; Wright et al. 2003) and HAAs (Rivera-Nunez and Wright, 2013) during the second trimester were associated with decreased BW in exposure-response manners. However, some studies reported null associations of maternal DBPs during pregnancy with birth outcomes (Botton et al. 2015; Villanueva et al. 2011; Yang et al. 2000). In the present study, we observed evidence of maternal NDMA exposure in the second trimester and NPIP exposure during entire pregnancy in associations with reduced infant BW. Given the potential for chance findings in the first study of its kind, more experimental and epidemiological studies are required to confirm our findings. For example, developmental toxicity studies that included direct comparisons across and within DBPs would allow for more relevant DBP mixture metrics weighted by toxicity.

Because LBW is affected by both gestational duration and fetal growth rate, SGA has been generally considered to be a better marker for fetal growth (Colman et al. 2011). In this study, we did not observe evidence of associations between maternal exposure to drinking water nitrosamines and increased risks of LBW. However, increased risks of SGA were found for elevated maternal exposure of NDEA in the second trimester. Our results are in accordance with the findings that second trimester and third trimester are the critical period of fetal growth (Matheus and Sala 1980; Vorherr 1982). This is largely reflected in the DBP literature as well, with the associations between the regulated DBP exposures being reported in the second trimester and third trimester and increased risks of SGA or intrauterine growth retardation (Grazuleviciene et al. 2011; Porter et al. 2005; Summerhayes et al. 2020; Wright et al. 2003). Although not statistically significant, we found that pregnancy average NPIP exposure was associated with reduced risk of SGA. It is not clear whether the inverse association of NPIP and SGA is a chance finding due to the various outcomes and exposure combinations that were examined. In addition, misclassification of GA due to the use of self-reported last menstrual period may occur to some degree and decrease our overall study sensitivity. When feasible, future studies should also target more pathologically small cases (compared to constitutionally small) by examining smaller percentiles of SGA like SGA5% or SGA1%.

Previous studies reported inconclusive results on the associations between maternal exposure to regulated DBPs such as THMs and HAAs during the specific trimesters and entire pregnancy and PTD, with some null (Kogevinas et al. 2016; Patelarou et al. 2011; Villanueva et al. 2011) and negative associations (Hoffman et al. 2008; Lewis et al. 2007; Säve-Söderbergh et al. 2020; Wright et al. 2004). The lack of evidence of an association between maternal THM exposure during the third trimester and PTD was also observed in a meta-analysis (Grellier et al. 2010). In contrast, we found evidence of exposure-response associations of maternal NDMA exposure in the second trimester and NPIP exposure during entire pregnancy with increased risks of PTD. This may be indicative of a more relevant earlier window being examined for PTD or may reflect increased sensitivity if more participants were included with a longer averaging time since not all PTD cases will have third trimester exposures.

The mechanisms by which nitrosamines adversely affect birth outcomes are largely not well understood. Experimental studies have shown that nitrosamines induce oxidative stress via increasing production of lipid peroxidation and decreasing antioxidants (Ahmad and Ahmad 2014; Dong et al. 2022; Hebels et al. 2010). Excessive oxidative stress has been linked to abnormal fetal growth triggered by disrupting the development of the placenta, impairing the placental tissue, and promoting premature senescence in fetal membranes (Burton and Jauniaux 2018; Hu et al. 2020; Polettini et al. 2015). Another potential biological mechanism is genotoxicity. An in vivo study in HepG2 cells revealed that exposure to mixture of nitrosamines induced DNA and chromosome damage (Dong et al. 2022), which may lead to adverse fetal growth (Gluckman et al. 2008).

Exposure assessment plays an important role in studying health effects associated with drinking water DBPs (Morgenstern and Thomas 1993; Symanski et al. 2004). On the whole, previous studies estimated the maternal exposure of DBPs mainly based on the limited data from national or local database or sampling campaigns (Grellier et al. 2010; Mashau et al. 2018; Nieuwenhuijsen et al. 2000). Quarterly DBPs concentrations from limited sampling sites in a water supply system are the most commonly applied approach used in previous studies; these may not sufficiently capture the temporal and spatial variability of DBPs in drinking water, and further lead to the inaccuracy of exposure assessment (Parvez et al. 2011; Symanski et al. 2004). To improve the precision and accuracy of exposure estimation, we utilized monthly water sampling strategy and selected six monitoring points in the studied water supply distribution system. We found significant monthly variation for drinking water nitrosamines (Luo et al. 2020), indicating that the use of more frequent nitrosamine sampling may be more critical to exposure assessment especially to address smaller critical windows that occur during pregnancy. A strength of our study is that the NDMA and NDEA levels shown here are higher than those reported in initial sampling occurrence studies in the US, this should increase our overall study sensitivity for examination of maternal DBP estimates (US EPA, 2016).

An additional study strength was the use of a time-weighted method to maximize the frequent nitrosamine monthly sampling data and to better estimate the individual-specific pregnancy and trimester-specific average exposures. Nonetheless, exposure misclassification can be introduced from different sources including the possibility of residential mobility during pregnancy or exposure from other sources and pathways. The assignment of the nearest measurement concentration may result in inaccurate exposure assessment, because it is uncertain that the levels of nitrosamines in the sampling sites are identical to the tap water concentrations in residents’ home. Moreover, some measurement error was inevitably introduced due to the omission of individual variation in tap water consumption and other water use activities. Given that the nitrosamines are semi-volatile, the use of indirect water concentration-based exposure measures might not fully characterize the impact of dermal and inhalation exposures due to lack of individual water use activity data. Therefore, similar to other studies, we would largely anticipate that random measurement error and misclassification of exposure may have biased the effect estimates toward the null (Nieuwenhuijsen et al. 2000).

5. Limitations

As with other observational studies, a source of uncertainty is the potential impact of unmeasured or uncontrolled confounding variables. Although we adjusted for some potential confounders, many other factors such as pre-pregnancy weight, weight gain during pregnancy, certain maternal diseases and dietary habits associated with fetal growth (Goldstein et al. 2017) were not available. Although many of these covariates are strong risk factors for fetal growth restriction and prematurity, residual confounding may be minimal since there is limited evidence that any of the non-dietary covariates are related to nitrosamine exposures. Additionally, we only focused on the exposure of nitrosamines in drinking water, and it is well-known that foods and beverages have much higher concentrations of nitrosamines than drinking water. To the extent that food may contribute to overall nitrosamine exposures, future studies may be interested in further delineating relative source contributions. These studies may also further clarify whether any dietary nitrosamine sources may be related to underlying water concentrations. Also, it is not negligible that nitrosamines contamination in bottled water (De Mey et al. 2017; Gushgari and Halden 2018; Li et al. 2021; Zhang et al. 2020). It was a pity that we did not collect the information on the amount or frequency of use of bottled water. Two studies in China have investigated nitrosamine concentrations in bottled water, with lower detectable rates (≤10%) and the detectable concentrations of 3.16–11.67 ng/L (Li et al., 2015; Zhang et al., 2020). Furthermore, maternal nutrition varied by socioeconomic factors that are independently associated with birth outcomes (Drewnowski et al. 2013). Other unmeasured parameters including maternal medical risk factors (e.g., gestational diabetes and hypertension) potentially in relation to birth outcomes and water quality factors (e.g., temperature, disinfection dosing, and water flow speed) associated with significant variation of nitrosamines in water variation were not investigated, which may bias the results (Almberg et al. 2017).

Given that mothers are generally exposed to a mixture of DBPs in drinking water, a challenge in epidemiological studies is determining which DBP metric best represents the most relevant toxic combinations of interest. For example, our summary TNA3 measure was based on the predominant nitrosamines in these water systems and may not be reflective of total exposure to all nitrosamines. Other regulated DBPs such as THMs and HAAs have been reported to be linked with adverse effects on fetal growth (Almberg et al. 2017; Cao et al. 2016; Porpora et al. 2019; Säve-Söderbergh et al. 2020). The monitoring data of the regulated DBPs, such as THMs and HAAs, were also not available in this study. In contrast to nitrosamines which are more prevalent in chloraminated or ammonia-rich systems, THMs and HAAs will predominate in chlorinated water systems and have been shown to be poorly correlated with THMs in systems with a chlorine residual (Krasner et al. 2009b). Since THMs and HAAs not likely co-occurring with nitrosamines in our water distribution system, the observed nitrosamine findings would not likely be completely confounded by these DBPs. Moreover, given that a recent meta-analysis has suggested that if there is any risk for SGA births and related outcomes with water concentration-based THM and HAA metrics, it is likely small in magnitude (Summerhayes et al. 2020) and would not likely fully explain the effect estimates for nitrosamines detected here.

6. Conclusions

This study is the first investigation of relationships between maternal exposure to nitrosamines in drinking water during pregnancy and birth outcomes. We found evidence of associations between elevated nitrosamine exposures and reduction in BW and some increased risks of SGA and PTD. However, more epidemiological research with rigorous study designs and refined exposure assessment, as well as mechanistic investigations are needed to confirm our findings.