Introduction

Nitrate is a water-soluble ion made up of nitrogen and oxygen with the chemical formula NO3−. It is a naturally occurring ion that is part of the nitrogen cycle involving the interchange of nitrogen between the atmosphere, land and living organisms1. In humans the main intake of nitrates is from food; vegetables constitute at least 85% of nitrate consumption2. Drinking water normally contributes only a small percentage of total nitrate intake based on consumption habits. For example, in New Zealand, less than 10% of total nitrate intake is from drinking water, with most of the remainder coming from the diet3. However, if the nitrate concentration in drinking water is high, it may contribute a much larger proportion of total nitrate intake.

Nitrogen is very important for plant nutrition and growth, being incorporated by plants into amino acid synthesis, and is therefore commonly used in inorganic fertilizers. However, because nitrate is highly water soluble, it leaches through soils and into groundwater very easily, particularly after heavy rainfall. About 80–90% of the world’s freshwater comes from groundwater4, and 50% of the total population relies on groundwater for daily drinking water5. The increasing use of artificial fertilizers, the disposal of wastes, particularly from animal farming, and changes in land use have become significant contributors to the progressive increase in nitrate levels in groundwater supplies1.

The current World Health Organization (WHO) guideline value for nitrate in drinking water is 50 mg/L as nitrate (NO3−) or 11.3 mg/L as nitrate-nitrogen (NO3-N) (multiply NO3− mg/L by 0.2259)1. This concentration is approximately equivalent to the current U.S. federal maximum contaminant level (MCL) for nitrate in public drinking water supply of 10 mg/L as NO3-N. This limit was established to protect against methemoglobinemia in infants, or blue baby syndrome, the most widely recognized health consequence of high nitrate exposure6.

While there is some evidence that nitrate in drinking water is associated with colorectal cancer7,8,9,10,11, the potential for adverse reproductive effects of chronic exposure to low levels of nitrate has also been raised recently12,13,14,15. Animal studies have indicated that nitrate from the mother can cross the placenta, affect the fetus in utero, and increase adverse outcomes, such as abortion, birth defects, gastroschisis, microphthalmia, anophthalmia, and craniofacial hypoplasia16,17,18,19,20.

The proposed mechanism for nitrate to cause adverse birth outcomes is via its reduction to nitrite, leading to the transformation of hemoglobin to methemoglobin, which cannot carry oxygen, thus reducing transfer of oxygen to body cells21,22. It has been suggested that fetal plasma nitrate levels may be higher than those of the mother because nitrate or nitrite can transfer to the fetus23 and fetal hemogolobin is especially vulnerable to oxidation. In addition, antioxidant defence is relatively deficient in newborns and, even when antioxidants are present, they do not fully mitigate oxidative/nitrosative stress or its consequences24. Further, since water does not contain antioxidants, nitrate intake from drinking water may be more damaging than that consumed from food which may contain antioxidants. Other proposed mechanisms for the potential effects of nitrate on reproductive health including the creation of N-nitroso compounds and thyroid and endocrine disturbance8,25.

Several epidemiological studies in humans have reported an association between prenatal nitrate exposure and adverse reproductive outcomes, including congenital abnormalities, preterm birth, low birth weight and small-for-gestational-age (SGA) infants13,26,27,28,29. Two previous systematic reviews have assessed the association between maternal nitrate intake and risk of neural tube defects30, birth defects, and preterm birth31, but the exposures included nitrate intake from food and drugs as well as drinking water.

The purpose of this study was to systematically review the published evidence to determine the association between human exposure to nitrate in drinking water and adverse reproductive and birth outcomes. This study was focused on the international context but uses New Zealand examples.

Results

Search results, study characteristics and quality

In total, 1005 records were identified from database searching. After removal of the duplicates, we completed title and abstract screening for 544 records and then full-text screening for 94 records, of which 65 did not meet our inclusion criteria. Sixteen studies (29 records) met our inclusion criteria (Fig. 1).

Figure 1
figure 1

Flow diagram of included studies.

Full size image

Among the 65 records that did not meet the study design criteria for this review, we identified six ecological, three cross-sectional and one record-based prevalence study exploring the association between nitrate exposure in drinking water and adverse reproductive and birth outcomes. These ten studies were included in the qualitative analysis and are presented as additional evidence.

The 16 studies included in the analysis were published between 1982 and 2022 and included 7,268,991 participants (range from 394 to 4,160,998). Five were cohort studies carried out in Denmark(2), France(1), US(1) and Sweden(1). The other eleven were case–control studies in US(7), Canada(2), Australia(1) and Sweden(1). The study characteristics are described in Table 1.

Table 1 Study characteristics.
Full size table

According to the Newcastle–Ottawa Scale (NOS) (Table 2), all five cohort studies and four of the eleven case–control studies were considered of high quality; seven of the eleven case–control studies were assessed as moderate study quality.

Table 2 Quality of the included studies using the Newcastle–Ottawa scale.
Full size table

Primary outcome

Four studies reported the odds ratios (ORs) for preterm birth, one study reported the ORs for SGA and two studies reported the ORs for low birth weight, but it was not possible to pool data about the composite of any of these outcomes.

Secondary outcomes

Preterm birth

When analyzed as a linear relationship, there was an association between nitrate in drinking water and preterm birth (3 cohort studies, 5,014,487 participants; OR for 1 mg/L (OR1) = 1.01, 95% CI 1.00, 1.01; I2 = 23.9%, Fig. 2a). There also was an association when comparing the highest nitrate exposure to the lowest exposure (3 cohort studies, 4,152,348 participants; pooled OR (ORp) = 1.05, 95% CI 1.01, 1.10, I2 = 0%, Fig. 2b). When comparing the combined higher nitrate exposures to the lowest exposure did not show a significant association (3 cohort studies, 5,023,271 participants; ORp = 1.08, 95% CI 0.99, 1.17; I2 = 47.2%, Fig. 2c). However, one case–control study48 reported that there was no association between nitrate in drinking water and preterm birth.

Figure 2
figure 2

Forest plot for the association between nitrate in drinking water and preterm birth. (a) Linear relationship; (b) Highest versus lowest exposure; (c) All combined higher versus the lowest exposure.

Full size image

Any birth defects

One cohort study36 reported a weak inverse association between nitrate in drinking water and any birth defect when treating nitrate as a continuous variable (1 cohort study, 1,018,914 participants; OR1 = 0.98, 95% CI 0.97, 1.00). When analyzed as a linear relationship, findings from two case–control studies27,46 did not show an association between nitrate in drinking water and any birth defect (2 case–control studies, 2676 participants; OR1 = 1.13, 95% CI 0.92, 1.38; I2 = 69.7%; Fig. 3a); however, significant heterogeneity was present. Comparing the highest to the lowest nitrate exposure groups did not show a significant association between nitrate in drinking water and any birth defects (2 case–control studies, 1150 participants, ORp = 2.32, 95% CI 0.99, 5.42; I2 = 41.0%; Fig. 3b), but comparing all combined higher exposures to the lowest exposure group showed evidence of an association between the risk of any birth defect and nitrate in drinking water (2 case–control studies, 2676 participants; ORp = 2.17, 95% CI 1.31, 3.60; I2 = 57.7%, Fig. 3c). Aschengrau43 reported there was no evidence of an association between detectable nitrate concentration in drinking water and any birth defects.

Figure 3
figure 3

Forest plot for the association between nitrate in drinking water and any birth defects. (a) Linear relationship; (b) Highest versus the lowest exposure; (c) All combined higher versus the lowest exposure.

Full size image

Neural tube defects

One cohort study36 reported there was no evidence of an association between nitrate in drinking water and any neural tube defects. However, when analyzed as a linear relationship, findings from case–control studies showed a significant relationship between nitrate in drinking water and neural tube defects (2 studies, 2,196 participants; OR1 = 1.06, 95% CI 1.02, 1.10; I2 = 0%; Fig. 4a). There also was an association between nitrate in drinking water and neural tube defects when comparing the highest versus the lowest nitrate exposure groups (3 studies, 1,501 participants; ORp = 1.51, 95% CI 1.12, 2.05; I2 = 0%; Fig. 4b) and when comparing all combined higher exposure versus the lowest exposure group (3 studies, 2,306 participants; ORp = 1.40, 95% 1.14, 1.71; I2 = 0%; Fig. 4c). However, there was no evidence of an association between nitrate in drinking water and spina bifida when comparing the highest versus the lowest exposure groups (2 case–control studies, numbers of participants unknown; ORp = 2.84, 95% CI 0.90, 8.97; I2 = 44.3%) or when comparing all combined higher versus the lowest nitrate exposure group (2 case–control studies, numbers of participants unknown; ORp = 2.69, 95% CI (0.69, 10.49); I2 = 56.9%). There was also no association between nitrate in drinking water and anencephaly when comparing the highest to the lowest exposure groups (2 case–control studies, numbers of participants unknown; ORp = 0.82, 95% CI 0.50, 1.36; I2 = 0.0%), or when comparing all combined higher versus the lowest nitrate exposure group (2 case–control studies, numbers of participants unknown; ORp = 0.71, 95% CI 0.48, 1.05; I2 = 0.0%).

Figure 4
figure 4

Forest plot for the association between nitrate in drinking water and neural tube defects. (a) Linear relationship; (b) Highest versus the lowest exposure; (c) All combined higher versus the lowest exposure.

Full size image

Any congenital heart defects

There were insufficient data to allow the linear relationship analysis. There was no evidence of an association between nitrate in drinking water and any congenital heart defects when comparing the highest versus the lowest nitrate exposure group (2 cohort studies, numbers of participants unknown; ORp = 1.08, 95% CI 0.96, 1.20; I2 = 19.7%; Fig. 5a). However, when comparing all combined higher versus the lowest nitrate exposure group, there was significant between study heterogeneity (2 cohort studies, numbers of participants unknown; ORp = 1.05, 95% CI 0.89, 1.24; I2 = 63.9%; Fig. 5b).

Figure 5
figure 5

Forest plot for the association between nitrate in drinking water and any heart defects. (a) Highest versus the lowest exposure; (b) All combined higher versus the lowest exposure.

Full size image

There were insufficient data to allow meta-analysis for any of the other outcomes. The direction of findings from the included studies are summarised in Table 3, with the detailed individual study results provided in Table S1.

Table 3 Direction of findings.
Full size table

Aschengrau43 did not find an association between nitrate in drinking water and neonatal death. Thomsen37 and Aschengrau42 reported no association between nitrate in drinking water and stillbirth. Aschengrau43 found a decrease in the frequency of spontaneous abortion with any detectable level of nitrate. However, Ebdrup33, in a large cohort study, reported there was no association between drinking water nitrate and the risk of pregnancy loss. The continuous analysis indicated a risk of pregnancy loss in the lower nitrate exposure groups for the first trimester.

Migeot38 reported exposure to the second tertile (3.19–6.10 mg/L NO3-N) of nitrate compared with exposure to the lowest tertile (< 3.19 mg/L NO3-N) was associated with a possible increased risk of SGA birth (OR = 1.74, 95% CI 1.10, 2.75), but there was no association between exposure to the highest (> 6.10 mg/L NO3-N) and lowest tertiles of nitrate with SGA birth (OR = 1.51, 95% CI 0.96, 2.40)38.

Coffman13 and Liu48 reported no association between nitrate in drinking water and low birth weight, although compared to the lowest exposure group (≤ 0.23 mg/L NO3-N), Coffman13 reported that all the other exposure groups were associated with a small decrease in birth weight. They also reported that mean body length at birth decreased with increased nitrate in drinking water, but only in the second highest exposure group (1.13 to ≤ 5.65 mg/L NO3-N) not the highest exposure group (> 5.65 mg/L NO3-N). They reported no association between nitrate in drinking water and head circumference at birth.

Stayner36 reported there was no association between nitrate in drinking water and limb deficiencies or oral cleft defects. However, Brender28 reported that compared to the lowest tertile (< 0.71 mg/L), infants exposed to the highest tertile (> 3.5 mg/L NO3-N) of nitrate in drinking water had a potential increased risk of limb deficiencies and oral cleft defects.

Both a cohort study36 and a case–control study49 reported there was no association between nitrate in drinking water and abdominal wall defects/ gastroschisis.

Evidence from other study types

Blake50 used a spatial analysis to explore the relationship between nitrate exposure level within ZIP codes and low birth weight. The authors reported there was no correlation between low birth weight and unsafe nitrate levels (> 10 mg/L as NO3-N). Two ecological studies15,26 of the same cohort explored the association between average county-level nitrate concentrations in drinking water and adverse birth outcomes. Antenatal exposure to nitrate in drinking water was associated with a possible increased risk of limb deficiencies, but no association was found between antenatal exposure to nitrate in drinking water and preterm birth, low birth weight, neural tube defects or oral cleft defects. However, Bukowski51 explored the association between residential postcode-level nitrate concentration in drinking water and low birth weight and preterm birth and reported that antenatal exposure to > 3.1 mg/L median nitrate concentration was associated with preterm birth and low birth weight. Mattix52 linked the monthly abdominal wall defect rates to monthly surface water nitrate concentration, and reported no association between nitrate levels in surface water and monthly abdominal wall defects rate. Winchester53 linked mean monthly nitrate concentration to birth defects and found there was no association between increased nitrate level (1.31 ± 0.20 vs. 0.16 ± 0.02 mg/L in log) in April–July (annual peaks in nitrates) and spina bifida, oral cleft, circulation system defects, Down syndrome, gastroschisis, urogenital defects and clubfoot or oral cleft. Ouattara54 assessed the occurrence of birth defects and nitrate concentrations collected from selected Nebraska watershed boundaries, and observed a positive association between high levels of nitrate (> 6.94 mg/L) in drinking water and the prevalence of birth defects with incidence rate ratio 1.44 (1.40–1.50). Two cross-sectional studies55,56 linked birth records, maternal and infant hospital discharge records to the CalEnviro Screen 3.0 dataset from California Communities Environmental Health Screening Tool to explore the relationship between preterm birth, gestational hypertension, eclampsia and environmental factors including nitrate in drinking water. The investigators reported that nitrate in drinking water is potentially associated with preterm birth in California. There was insufficient evidence suggesting that nitrate in drinking water was associated with hypertensive disorders or eclampsia in pregnancy. However, Super57 found there were no associations between well-water with high nitrate regions (> 4.518 mg/L as NO3-N) and preterm birth or size of infant at birth.

Joyce58 conducted a record-based prevalence study assigning water contaminant measurements to the maternal residential address. The authors found increasing exposure to nitrate in drinking water was associated with an increased risk of prelabor rupture of membranes (Table S2).

There were insufficient data to allow the sensitivity analyses or assessment of publication bias funnel plots.

Discussion

In this systematic review of evidence from nine high-quality and seven moderate-quality observational studies, involving 7,268,991 participants, we found nitrate in drinking water may be associated with some adverse reproductive and birth outcomes. Studies included in the meta-analysis found an association between nitrate in drinking water and increased risk of preterm birth. Three high quality cohort studies contribute to this result. Coffman35 linked 1,009,189 liveborn resident in Denmark to household level of nitrate in drinking water. They observed an increasing risk of preterm birth with increases in nitrate in household tap water at levels below current regulatory levels. Sherris34, linked over 6 million birth certificate records to public water system monitoring records in US and suggested a robust association between nitrate in tap water and risk of preterm birth in within-mother analyses. They also found modestly increased odds of preterm birth within nitrate levels below the regulatory limit. Albouy-Llaty40 linked 13,654 mother/neonate pairs in France to mean nitrate concentrations in maternal place of residence and did not find a relationship between exposure to nitrate in drinking-water during the second trimester of pregnancy and preterm birth.

Our analysis from case–control studies also showed nitrate in drinking water may be associated with increased risk of neural tube defects, with a pooled odds ratio of 1.06 (1.02, 1.10) per 1 mg/L NO3-N increase in drinking water, although this finding is in contrast to a cohort study showing no linear association36. However, our analyses did not find a relationship between nitrate and any birth defects, or any congenital heart defects, although visual inspection of forest plots from meta-analyses shows mostly positive associations between nitrate and adverse reproductive and birth outcomes but with wide confidence intervals. It may therefore be possible that there is small impact of nitrate in drinking water on these outcomes, but the available studies together are insufficiently powered to detect a clinically significant effect.

We observed moderate heterogeneity in some of our meta-analyses. Several factors may contribute to the inconsistent findings amongst studies. The first factor is the different study characteristics. In the analysis of preterm birth, for example, Albouy-Llaty 2016 was conducted in France with a reported preterm birth rate of 7.5% of all live births in 201659, while Sherris 2021 was conducted in the US with a reported preterm birth rate of 10.1% of all live births in 202060, and Coffman 2022 was conducted in Demark with a reported preterm birth rate of 6.29% per year between 2016 and 202061. In addition, the three studies monitored different exposure periods: Albouy-Llaty 2016 estimated the nitrate exposure during the second trimester of pregnancy, while Coffman 2022 and Sherris 2021 measured the nitrate exposure throughout pregnancy. Moreover, while Albouy-Llaty 2016 studied 13,481 participants, only 4625 participants were included in their adjusted analytical model, so that statistical power was very limited. In comparison, Coffman 2022 analyzed data from 1,009,189 participants and Sherris 2021 analyzed data from 3,832,090 participants. Another factor may be the different statistical models used and adjustment for different confounding factors. Albouy-Llaty 2016 used a multivariable logistic regression model adjusted for rural area, season, maternal age, mother’s occupation, smoking during pregnancy, single-parent family, history of preterm birth, primiparity and quality of follow-up; Coffman 2022 used generalized logistic regression models adjusted for the non-independence of births from the same mother and year, sex, gravidity, urbanicity, and maternal age, smoking, education, income, and employment status; Sherris 2021 used mixed-effects logistic regression adjusted for maternal age, parity, education, race, payer for delivery, and timing of initiation of care. Therefore, without individual participant data we could not further investigate nonlinear effects.

Further, the limits of exposure categories were different between studies. To calculate the exposure–response relationship for nitrate in drinking water, our study used several transformations. For the study by Albouy-Llaty 2016, when adjusting for confounders, both relationships between preterm birth and second tertile nitrate exposure group, and preterm birth and third tertile nitrate exposure group were non-significant, but the direction of the findings was towards a protective effect. Therefore, the result of the generalized least squares regression analysis yielded negative study-specific slopes and the odds ratio for 1 mg/L increase in nitrate indicated a protective effect. For the study by Sherris 2021, the odds ratio for 1 mg/L was reported by the authors but for different gestational age subgroups (20–31 gestational weeks and 32–36 gestational weeks). The study by Coffman 2022 reported adjusted odds ratio for per 10 mg/L NO3−. We pooled the odds ratios for all these subgroups to obtain the results for 1 mg/L increase in nitrate for overall preterm birth. This pooled estimate indicates nitrate in drinking could be a risk factor for preterm birth, but the several steps of conversion may reduce the accuracy of the estimation.

Periconceptional intake of folic acid reduces the risk of women having an infant affected by neural tube defects62. The inconsistent findings for neural tube defects from cohort36 and case–control28,44,45 studies may result from the different years of outcome ascertainment as it is likely more pregnant women received folic acid in the later studies. However, Stayner 2022 included all singleton infants born in Denmark between 1991 and 2013, where periconceptional folic acid has been recommended since 199763. The three case–control studies included infants born in U.S. between 1989 and 1991, 1995–2000 or 1997–2005. The U.S. recommended periconceptional intake of folic acid in 199264 and implemented mandatory fortification of cereal grains with folic acid in 199865.

The nitrate levels in drinking water reported in the different studies vary, but most of the high-quality studies investigated levels that were within the WHO guideline nitrate level (11.3 mg/L NO3-N). However, Sherris, who reported a relationship between high nitrate exposure and preterm birth14 selected the high exposure cut-off as > 10 mg/L NO3-N and median exposure cut-off as 5 mg/L NO3-N to correspond to MCL level and half MCL level respectively for nitrate in drinking water in US, although only 0.6% of the population were exposed to this level. Further, the ecological study by Blake et al.50, which did not find an association between preterm birth and unsafe nitrate level in drinking water, also assessed exposure to the unsafe nitrate level in drinking water of 10 mg/L NO3-N. The maximum nitrate level in drinking water (19.3 mg/L NO3-N) reported in the two cross-sectional studies55,56 was almost equivalent to twice the US MCL of 10 mg/L NO3-N, but 75% of the population in the study region were exposed to < 2.44 mg/L NO3-N. Croen45 found exposure to the highest nitrate level, which was above the US MCL, was associated with increased risk for anencephaly, but only 3% of the included population were exposed to this level.

Some drinking water is pumped from the ground, other drinking water originates as surface water in streams and rivers. About half of New Zealand’s drinking water is pumped from the ground, with the remainder coming from surface sources66. In the US, more water systems use groundwater (78%) rather than surface water as a source, but more people (68%) receive their water from a system supplied by surface water67.

Groundwater is recharged from the surface, predominantly from rainfall, but can also receive leakage from rivers and lakes4. Drinking water suppliers in New Zealand are not required to routinely monitor or report on nitrate levels if levels have been previously found to be below 25 mg/L as NO3−68. Richard69 estimated the variability of nitrate levels in drinking water in New Zealand, and found the nitrate levels in drinking water from registered supplies ranged from less than detection (< 0.01 mg/L) to 41.8 mg/L as NO3−. More than 60% of the population were exposed to less than 2 mg/L as NO3−, 8.2% of the population were exposed to more than 5 mg/L as NO3−, 2.2% were exposed to more than 10 mg/L as NO3−, and 0.1% of the population were exposed to more than 25 mg/L as NO3−69. Although the nitrate level in groundwater in New Zealand is lower than the WHO guideline level, long-term trends (10 years, 2009–2018) showed 28–35% of sites had increasing levels of nitrate over time70. Further, nitrate contamination present in the groundwater would likely stay there for years or decades, so exposures identified are likely to continue to increase if nitrate removal technologies are not utilized71.

A recent report estimated that while New Zealanders have similar nitrate exposure from drinking-water to that in most other countries, total nitrate intake from drinking water is less than 10%3. While this is of note, the conclusions drawn in the report should be interpreted with caution as the data analyzed were from more than ten years ago and the outcome of interest was limited to the risk of colorectal cancer rather than adverse perinatal outcomes. As nitrate concentrations in drinking water increase, it will contribute a larger proportion of total nitrate intake, potentially making this exposure of greater importance for overall public and reproductive health.

High quality, large epidemiology studies are needed to further assess any associations with perinatal outcomes and nitrate exposure from drinking water. In the US, all public water systems are required to be monitored at least annually to determine compliance with the nitrate MCL of 10 mg/L as NO3-N72. However, nitrate concentrations in New Zealand are not regularly monitored if below 50% of MAV68, and there is no national repository of nitrate exposure data for the New Zealand population69. In the recently published high quality studies, Sherris 2021 included around 6 million participants and Coffman 2021 included 852,348 participants. In New Zealand, there are around 58,000 births annually and the estimated preterm birth rate is 7.4%73. Thus, it will be difficult to reach adequate sample size to draw reliable conclusions about nitrate exposure and reproductive outcomes, since this would require matching at least 15 years of birth data with individual or regional nitrate exposure in drinking water. This issue of very limited power to detect small effect sizes is a problem in many other smaller regions, making it difficult to investigate the effects of nitrates in local drinking water supplies.

Two previous systematic reviews have assessed the association between maternal nitrate intake and risk of neural tube defects30, and between maternal intake of nitrate and risk of birth defects and preterm birth31. They reported no association between maternal nitrate intake and the risk of preterm birth, limb deficiency, cleft lip, and neural tube defects through high versus low meta-analysis. However, non-linear analysis showed the risk of neural tube defects increased with increasing intake of maternal dietary nitrate > 3 mg/day, and there was positive correlation between nitrate intake and heart defects through high versus low meta-analysis, with each additional 0.5 mg/day of maternal nitrate intake increasing the risk of heart defects. These two systematic reviews assessed maternal dietary nitrate intake, including nitrate from drinking water, food, or drugs. However, there was some overlapping of participants between the included studies that used the same cohort, with some participants counted more than once, potentially resulting in overestimation of the effects.

This review has some limitations. First, most of the studies were carried out in US and Europe. Given that nitrate levels in drinking water vary widely among different regions and countries, findings from these studies should be interpreted with caution when extrapolated to other regions. Second, the studies included in this systematic review do not consistently account for other potential confounding factors such as maternal diet, nitrosatable drug use, and antioxidant intake. Third, the concentration of nitrate in water often fluctuates with the season1. Not all studies included in this review took seasonal variation of nitrate into account in the measurement of exposure or as a factor in the adjusted model. Nitrogen shows significant seasonal relationships with high-intensity agriculture, with the difference between summer and winter water quality increasing as the proportion of high intensity agriculture in a catchment increases. For example, spatial modelling in New Zealand showed that regions dominated by high-intensity agriculture typically have poorer clarity, turbidity and nutrient concentrations in winter than in summer74. Fourth, in meta-analysis of observational studies, it is challenging or impossible to identify any unpublished studies, as pre-registration of a protocol is not mandatory75. Finally, of all the pre-specified outcomes, only a few outcomes could be incorporated into a meta-analysis to help determine the overall association. Sixteen studies have evaluated the nitrate in drinking water and adverse reproductive and birth outcomes, but the number of studies of any individual outcome was limited. Although some outcomes were reported by more than one study, studies using different designs cannot be combined.

We conclude that currently there is sufficient evidence of a possible association between nitrate in drinking water and preterm birth and specific congenital anomalies, to warrant nitrate exposure monitoring and reporting, and regular review as new evidence becomes available.

Methods

The study was reported according to the PRISMA guidelines (Note S1)76. The review protocol was not registered but was prepared before the review was conducted (Note S2).

Criteria for considering studies for this review

Type of studies: randomized trials, cohort and case–control studies published in English that reported the relationship between nitrate intake from drinking water and the risk of perinatal outcomes were eligible.

Type of participants: pregnant women and their infants.

Type of intervention: the exposure of interest is nitrate intake from drinking water during the antenatal period.

Type of outcome measure:

Primary outcome: a composite of any of the following outcomes: preterm birth; small-for-gestational-age (SGA) infant; low birth weight infant; miscarriage; stillbirth; and neonatal death.

Secondary outcomes: For infants: preterm birth, SGA, low birth weight, stillbirth, neonatal death, perinatal death, hypoglycemia, need for respiratory support after birth, infection, congenital abnormality, necrotizing enterocolitis, bronchopulmonary dysplasia, intraventricular hemorrhage, neonatal lung disease, neonatal intensive care unit (NICU) admission, jaundice, methemoglobinemia (as defined by the authors).

For women: any pregnancy complications (miscarriage, high blood pressure, preeclampsia, gestational diabetes, infection, obstetric hemorrhage; as defined by the authors).

Search strategy

We conducted a comprehensive search of databases from inception to 30 November 2022, including: Ovid MEDLINE via PubMed, Embase, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL, current issue) in the Cochrane Library, Web of Science, Scopus, GEOBASE and ProQuest Agricultural and Environmental Science Database, using search terms unique to the review topic (Note S3). We searched using both English and American spelling. We did not apply language restrictions, but only full text articles in English were included. Additionally, we reviewed the reference lists of all identified articles for relevant articles not identified in the primary search.

Study selection

Two authors (LL and SSC) independently evaluated and appraised the retrieved studies using COVIDENCE, extracted data and assessed risk of bias. Any disagreements were resolved by discussion and, if necessary, by discussion with a third review author (JH).

Selection of studies followed the steps below:

  1. 1.

    Import all the records from the database into COVIDENCE (https://www.covidence.org/).

  2. 2.

    Screen titles and abstracts to select relevant reports and exclude studies not relevant for this review.

  3. 3.

    Examine full-text studies for compliance with the eligibility criteria for this review.

  4. 4.

    Make final decisions on study inclusion and proceeded to data collection.

Data extraction

We recorded the selection process in sufficient detail to complete a PRISMA flow diagram.

We developed a data form (Note S4) to extract data for eligible studies. Information extracted included: source details, eligibility assessment, methodological details, characteristics of participants, details of intervention and outcomes reported.

Study quality

We assessed the quality of case–control and cohort studies according to the Newcastle–Ottawa Scale (NOS)77. The NOS evaluates nine methodological items and their reporting (participant selection, comparability of groups, and ascertainment of exposure/outcome), with values ≥ 7 compatible with good study quality (least bias, results are considered valid), between 2 and 7 with moderate study quality (susceptible to some bias but probably not enough to invalidate the results), and ≤ 2 with poor study quality (significant bias that may invalidate results).

We planned to assess the quality of randomized trials using the methods specified in the Cochrane Handbook for Systematic Reviews of Interventions78: (1) random sequence generation (selection bias); (2) allocation concealment (selection bias); (3) blinding of participants, personnel and outcome assessment (performance and detection bias); (4) incomplete outcome data (attrition bias); (5) selective reporting (reporting bias); (6) other bias (checking for bias due to problems not covered by (1) to (5) above).

Statistical analysis

The nitrate values in the analyses are expressed as nitrate-nitrogen (NO3-N). We converted the nitrate (NO3− mg/L) values of the included studies to nitrate-nitrogen (NO3-N mg/L) by multiplying 0.22591.

The relationships between nitrate intake from drinking water and the risk of adverse birth outcomes were examined based on the effect size. Nitrate intake from drinking water, odds ratios (ORs), risk ratios (RRs), hazard ratios (HRs), with 95% confidence intervals (CIs) were extracted (both crude and adjusted).

Because different studies used different exposure categories and have presented data in a variety of ways, we pooled the study-specific risk per mg/L increase in nitrate for each outcome, using the mid-point of each reported exposure category. If the lower and upper limits of the category were given, the midpoint intake of nitrate in drinking water was calculated as: midpoint intake = (lower limit + upper limit) divided by 2. If the midpoint intake was given, the data were used directly. If the interval for any category of nitrate intake was not provided, we assigned a value following the algorithms suggested by Il'yasova et al.79. For the upper open-ended category, we assigned the value of its lower limit plus the width of the previous (second-to-highest) interval. For the lower open-ended category, we assigned the value of its upper limit minus half the width of the next (second-to-lowest) interval. If the range of lower open-ended category was smaller than the half width of the next (second-to-lowest) interval, we assigned the value of half of the upper limit.

Generalized least squares regression analysis was used to generate study-specific slopes representing the estimated increase in log odds ratio (OR) per mg/L increase in drinking water nitrate concentration and standard errors for these slopes. Study-specific slopes and their standard errors were then used to calculated ORs and 95% confidence intervals (CIs) per mg/L increase in nitrate for each outcome. When the OR per mg/L was given, the data were used directly. We then incorporated the OR per mg/L into meta-analysis using a random effects model to derive a weighted pooled estimate with 95% CIs based on the DerSimonian and Laird method80.

We also calculate the pooled ORs for the highest compared with the lowest exposure level, and for all combined exposure levels above the lowest compared with the lowest exposure level using random effects models.

A random effects model was used instead of a fixed effects model in order to account for both within-study and inter-study variation. Heterogeneity tests were performed using the I-square and Q-statistic, and significant heterogeneity was defined as I2 > 50% or p < 0.1078. We planned to assess potential bias due to small study effects by visual inspection of funnel plots when there were more than 10 studies. We planned to conduct sensitivity analyses by examining only studies considered to be of good quality.

Statistical analyses were performed using Stata Statistical Software (version 14, STATA).