Epidemiologists often use a retrospective study design to examine for associations between an exposure and the occurrence of adverse health effects. Several of these studies used this approach to examine for an association between elevated levels of nitrate in drinking water and related health effects such as methemoglobinemia, cancer, neural tube effects, or spontaneous abortions. Often, data on exposures that occurred before these health outcomes were not available. Consequently, researchers use measurements of exposures at the time of the study to represent exposures that occurred before people developed these conditions. An opportunity to examine the stability of nitrate in water occurred during a survey of private water wells in nine Midwestern states. In this survey, water samples from 853 homes with drilled wells were collected in May 1994 and in September 1995 and nitrate-nitrogen (nitrate-N) was measured by the colorimetric cadmium reduction method. Nitrate-N levels from the same well over time were assessed by a mixed-effects analysis of variance. Analysis showed no significant difference in between the initial level and those measured 16 months later. Furthermore, analysis showed that most of the variance in the nitrate concentrations in well water was due to well-to-well variation (89%) rather than to residual error (12%). This observation showed that a single measurement of nitrate in water from drilled wells could represent longer periods of exposure.
The level of nitrate in groundwater has been increasing over the past three decades (Mueller et al., 1995; Mueller and Helsel, 1996). Fertilizers and organic waste account for much of the increase in the nitrate in groundwater. Soil bacteria may also increase nitrate levels in groundwater by oxidizing ammonia into nitrite and then to nitrate ion in oxygen-rich water (Dorsch et al., 1984). Although the nitrate ion is not toxic, enzymes in saliva and gastrointestinal secretions can reduce it to the more toxic nitrite (Swann, 1975). Nitrite is not only directly toxic, it also reacts with secondary and tertiary amines and amides to form carcinogenic N-nitroso compounds.
Elevated nitrate levels in drinking water have been associated with methemoglobinemia, cancer, neural tube defects, and spontaneous abortions (Swann, 1975; Scragg et al., 1982; Maloney et al., 1983; Dorsch et al., 1984; Fan et al., 1987; CDC, 1994). Methemoglobinemia is the most widely recognized health outcome associated with ingestion of nitrate-contaminated water. It is an acute and potentially fatal illness that primarily affects infants. While N-nitroso compounds have been linked to cancer in animals (Swann, 1975), the evidence of stomach, brain, esophageal, and nasopharyngeal cancers in humans from these compounds remains inconclusive (Eichholzer and Gutzwiller, 1998). Studies by Scragg et al. (1982) and Dorsch et al. (1984) suggest a possible association between neural tube defects and elevated nitrate levels. An investigation of a cluster of women with spontaneous abortions in LaGrange County, Indiana, from 1991 to 1994 suggested that elevated nitrate levels in drinking water may be associated with spontaneous abortions (CDC, 1994).
To prevent adverse health effects from ingesting nitrate in drinking water, the Environmental Protection Agency (EPA) established a maximum contaminant level (MCL) of 10 mg/l for nitrate (nitrate-N) in public water systems (US EPA, 1994). Unlike other MCLs, the nitrate MCL does not include a 10–100-fold uncertainty factor (Fan and Steinberg, 1996). The nitrate MCL is based on studies of infant methemoglobinemia and nitrate levels in drinking water.
A key element when studying the relationship between nitrate-contaminated drinking water and adverse acute health outcomes is the exposure to nitrate levels before the disease occurrence. Since measurements of nitrate levels may not be available, information on the stability of nitrate levels in ground water would be helpful to determine whether historical data may be used or testing conducted on new samples. Several studies reported the stability of nitrate in groundwater. A study in Czechoslovakia showed that nitrate levels in farming communities doubled from 24 mg/l in 1961–1965 to 43 mg/l in 1986 as fertilizer use increased eight times (Benes et al., 1989). A smaller increase (21 mg/l in 1968 to 28 mg/l in 1986) was observed in wells surrounded by protection zones. The study concluded that climatic conditions were mainly responsible for short-term changes in nitrate levels in the upper parts of the aquifer whereas anthropogenic influences (especially farming) were mainly responsible for long-term increases.
A study in Minnesota showed that nitrate concentrations at the water table are closely linked to seasonal recharge and with nitrogen fertilizer applications (Landon et al., 2000). Nitrate levels in the 65-ha research area ranged from 0.05 to 40 mg/l over a 2-year period. A comparison of nitrate concentrations from the same irrigation wells in Kansas taken over 20 years showed a slight increasing nitrate concentration (Townsend and Young, 2000). This conclusion was based on results from 32 wells which had concentrations less than 5 mg/l in the 1970s and no data on rainfall and land use practices. Nitrate concentrations in 66 shallow groundwater sources in Kentucky showed seasonal variations. This was attributed to low flow rates in the summer and higher flow rates in the winter (Montgomery et al., 1997).
The present study examines the stability of nitrate levels in 853-drilled wells in nine upper mid-western states by comparing nitrate-N levels that were measured three times over a 17-month period.
Almost 2.5 million households in the upper Midwestern states depend on private well water as their source of drinking water. State health and environmental departments in nine Midwestern states (Illinois, Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota, and Wisconsin) and the Centers for Disease Control and Prevention (CDC) conducted a study to determine the presence of bacteria and chemicals in water drawn from private wells (CDC, 1998).
Households were originally selected for inclusion into the study using a systematic geographical sampling approach. A 10-mile sampling grid of the entire study area was constructed using ArcInfo, a geographic information system software package. County health department staff traveled to each grid intersection in the 10-mile sampling grid and asked a resident of the household closest to and within three miles of each intersection to participate in the study. If a resident consented, a water sample was collected at the point of use that most often supplied drinking water to household residents (generally the kitchen tap). Additionally, a short interview was conducted to obtain information on well characteristics, potential sources of contamination, the number of people who drink the well water, and the occurrence of diarrhea among residents of the household.
To be eligible to participate in the survey, at least one member of the household had to drink well water, and the well could not have been chlorinated in the previous 4 days. If the resident refused to participate or if an eligibility requirement was not met, the next closest household was contacted. From May to November 1994, samples were collected from 5520 domestic wells and tested for total coliforms, Escherichia coli, nitrate, and atrazine. Nitrate-nitrogen concentration (mg/l NO3-N and hereby called nitrate) was measured according to the colorimetric cadmium-reduction method (American Public Health Association, 1992).
In the following year (May–September 1995), a subset of household wells (n=1487) that tested positive in 1994 for E. coli or had elevated nitrate levels above the MCL were tested for bacteria and nitrate. A random sample of 10% of the uncontaminated wells was also re-sampled. In this survey, household respondents were re-interviewed with the same questionnaire as the 1994 survey. A third well water sample was collected approximately 2 weeks after the second sampling for quality assurance/quality control (QA/QC) purposes. This analysis in the study consists of the 853-drilled wells that were sampled in all three sampling rounds.
The sampling rounds were classified into three phases: Phase 1 are results of sampling of wells done in 1994; Phase 2(a) are results of the first sampling of wells in 1995, and Phase 2(b) are results of follow-up sampling conducted 2 weeks after Phase 2(a). In this manuscript, we present the (1) mean nitrate levels, (2) correlation coefficients between the phases, (3) a categorical analysis of the distribution of water-nitrate measurements in Phase 2(a) as a percentage of Phase 1 measurements, and (4) a mixed-effects analysis of variance procedure. Because nitrate levels of 2 mg/l or less are considered background levels (USGS, 2001) and the current standard is 10 mg/l, these levels provided the basis for the following categories: ≤2 mg/l (background level), >2 and <10 mg/l (low exposure), ≥10 and <20 mg/l (moderate exposure), and greater than or equal to 20 mg/l (high exposure). We include this categorical analysis because nitrate ranges like these might be used in an epidemiological study to assess health effects, and these analyses illustrate stability of this type of classification over time.
Nitrate levels were logarithmically transformed (base 10) to stabilize the variance of increasing nitrate levels. To include samples with a nitrate level that was below the level of detection, the number one was added to each nitrate measurement before the levels were transformed. The mixed-effects analysis of variance for all three phases (log-transformed) tested the model: Yij=α+Ai+βj+Eij where Yij is a measured value in well i at time j, α is a constant, Ai is a random effect for well i, βj is a fixed time effect, γ is a parameter for the fixed factor X, and Eij is an error term. Time j was treated categorically (to indicate Phase 1, Phase 2a or Phase 2b samples, respectively) and nearly the same results were obtained when time was treated as a continuous variable (t=0, 1, or 1.04; data not shown). The association among well age, depth, and diameter with nitrate levels was examined by adding these variables to the model as fixed effects (represented by X in the model equation above). The data were analyzed using SAS for Windows Release 6.12 (SAS Institute Inc., 1992).
Of the 5520 wells sampled in 1994, 1471 were sampled in all three sampling rounds and were tested for nitrate. For this analysis, all the dug, bored, and driven wells were excluded to reduce the effect of contamination. Thus the analysis is based upon the 853-drilled wells. Of the drilled wells sampled in Phase 1, there were 425 wells (49.8%) testing positive for total coliforms, and 199 wells (23.3%) testing positive for E. coli (concentrations of at least 1 per 100 ml of sample), and 394 wells (46.2%) had nitrate levels above the MCL of 10 mg/l NO3-N. In Phase 2(a), 381 wells (44.7%) tested positive for total coliforms, 109 wells (12.8%) tested positive for E. coli, and 316 wells (37.1%) had nitrate levels above the MCL of 10 mg/l NO3-N. Most of the drilled wells had a diameter <6 inches (80.9%), were <25 years (64.5%), and <100 ft deep (55.1%). Statistically significantly higher nitrate levels were measured in wells that had diameters ≥6 inches (P<0.004), were at least 25 years of age (P<0.001), or were less than 100 ft deep (P<0.0001). There were 57.5% of the well owners reported using fertilizer within 100 ft of the well.
The mean nitrate levels in Phases 2(a) and 2(b) was 12.0 mg/l and similar to the Phase 1 level of 13.1 mg/l (Table 1). For the initial Phase 1 sampling, the nitrate levels ranged from non-detectable levels to 266.0 mg/l, with a median of 7.1 mg/l; most (99%) of the values were below 103.0 mg/l. In Phase 2(a), the levels ranged from non-detectable levels to 190 mg/l, with a median of 4.2 mg/l; most (99%) of the values were below 109 mg/l.
The nitrate levels measured in Phases 2(a) and 2(b) were highly positively correlated in both transformed and non-transformed states (r=0.92 and 0.95, respectively, P=0.0001 in both cases; Table 2). Levels in Phase 1 were highly positively correlated with levels measured in both Phase 2(a) and Phase 2(b). The association became stronger when the levels were logarithmically transformed (r=0.88, P=0.0001 and r=0.87, P=0.0001, respectively).
When the nitrate levels of wells from Phase 1 were categorized as described previously into background, low, moderate, and high, most of the wells (75.4%, n=643) remained in the same category in Phase 2(a) (Table 3). Of the 345 wells that initially had background nitrate levels, 92.2% (n=318) remained in the background category in Phase 2(a); of the 114 wells that were initially low, 74.6.2% (n=85) remained low; of the 226 wells that were initially moderate, 55.8% (n=126) remained high; and of the 168 wells that were initially high, 67.9% (n=114) remained high. When the nitrate levels did not remain in the same categories, most changes involved a shift to the adjacent lower exposure category.
The mixed-effects analysis of variance showed that most (89%) of the variance in the nitrate measurements was due mainly to well-to-well variation (s2=0.33). Eleven percent of the variance was due to the within-subject residual error (s2=0.040). Similar results were obtained when we compared Phase 1 to Phase 2(a). The association of higher nitrate levels with well age (P=0.02), depth (P<0.0001) and larger diameter (P<0.0001) was apparent in the mixed model. However, well-to-well variation remained substantial (s2=0.30) and residual error small (s2=0.04), even after accounting for these factors.
In this study of 853-drilled private water wells in nine mid-western states, we found that intra-well nitrate levels were not statistically significantly different over a 17-month period. Most of the variance in the nitrate measurements in the well water was due to well-to-well variation. Variability owing to short-term changes, sample collection and laboratory analysis were found to be small because nitrate levels in samples taken from the same well 2 weeks apart were extremely similar. The residual error for intra-well measurements is consistent with the precision in the laboratory/analytical method used to measure nitrate further indicating a small degree of short-term variability (American Public Health Association, 1992).
The information included here about variability and stability of nitrate levels is important for design, analysis, and interpretation of epidemiologic studies intended to address possible health effects of nitrates. Design and analysis of such studies depend on the variability of the true exposure. Devine and Smith (1998), for example, have shown that the sample size is inversely proportional to the variance of the exposure . In other words, if other factors (average disease risk in the population, and the dose-response) remain constant, a smaller sample size will be required if the population variance of the exposure is larger.
Interpretation of epidemiologic studies also depends on variability of the exposure, particularly as it relates to measurement error. For example, the magnitude of measurement error relative to overall variability is a key determinant of bias owing to measurement error in logistic regression. With a classical error structure, the regression coefficient estimated in the presence of measurement error is biased towards 0 by a factor that depends on the ratio s12/(s12+s22), where s12 is the variance of the exposure and s22 is the variance due to error. In this study, we emphasize that we do not have true repeat measurements so that we do not have a true estimate of s22. However, our residual error estimate is likely a conservative overestimate of the variance owing to measurement error, since it should include components owing to changes in nitrate levels over time, variability between wells in the pattern of change over time, as well as measurement error. Nevertheless, our results are somewhat reassuring for the design of epidemiologic studies since our overestimate of measurement error is small relative to the overall variability of nitrate levels between wells. The variability from well-to-well may reflect differences in land and fertilizer use near the well head, individual characteristics of the well such as depth and aquifer geology, and perhaps maintenance of the well.
This study is unique because it evaluated the change in nitrate levels in the representative wells across a large and diverse geographic area. In addition, the same well was measured over time and we used a statistical approach that evaluated changes within and between wells. Previous studies were conducted with fewer wells, in smaller areas, and compared aggregate results (McDonald and Splinter, 1982; Hallberg et al., 1984; Benes et al., 1989).
Our study observed that nitrate levels were stable during the study period and that shallower wells were more likely to have higher nitrate levels. A study in Iowa examined long-term trends in nitrate concentrations in water supplies (McDonald and Splinter, 1982). In that study, the average nitrate concentrations in 4597 samples from municipal wells of all depths grouped together were stable over a 30-year period. However, the authors concluded that nitrate concentrations in well water were dependent on time, well depth, climate, and surface runoff (McDonald and Splinter, 1982). In a second study in Iowa, nitrate levels in 24 of 40 wells increased an average of 14 mg/l from 1975 to 1983 (Hallberg et al., 1984). The authors concluded that the increase in nitrates in unprotected shallow groundwater supplies was related to increased application of chemical fertilizers. Our study could not assess the impact of fertilizer application on nitrate levels because only qualitative information (such as commercial fertilizer ever used) was obtained and information on quantity and change over time was not collected.
About half of the wells were <100 ft deep and more that one-third were older than 25 years and a fifth had a diameter greater than 6 inches. These well features were associated with elevated nitrate levels in our water samples, a finding which is consistent with findings of other studies that have shown that well depth, diameter, age, and construction type are important indicators of well water contamination (Johnson and Kross, 1990; Kross et al., 1993; US EPA, 1994).
Although most nitrate levels remained stable (76% overall agreement between water nitrate categories in Phases 1 and 2(a); Table 3), some nitrate levels may have decreased by about 5%. This small decline could have been owing to regression to the mean, a statistical phenomenon, or it could represent an actual gradual decline in nitrate levels. It is possible that in the absence of continued sources of nitrate contamination, nitrate levels may naturally decay, and groundwater may be diluted with uncontaminated water. However, further work may be warranted to determine which of these factors explain the slight decline.
The limitations of this study include the use of self-reported data on well characteristics, land use, and fertilizer use. However, there was good agreement of reported well type, well depth, and well diameter between Phases 1 and 2(a).
In this study area, nitrate levels appear to be stable over a period of 17 months. This stability, a change of <0.5 mg/l, shows that a nitrate level may be representative of the nitrate concentration in well water for at least 17 months. This information is important in assessing exposure to nitrates and useful for conducting epidemiologic studies on the acute effects of elevated nitrate levels on adverse health outcomes.
American Public Health Association. Cadmium Reduction Method. In: Greenberg AE, Clesceri LS, Eaton AD (ed.) Standard Methods for the Examination of Water and Wastewater. American Public Health Association: Washington, DC, 1992 Section 4-87-88.
Benes V., Pekny V., Skorepa J., and Vrba J. Impact of diffuse nitrate pollution sources on groundwater quality-some examples from Czechoslovakia. Environ Health Perspect 1989: 83: 5–24.
Centers for Disease Control and Prevention (CDC). Spontaneous abortions possibly related to ingestion of nitrate-contaminated well water - LaGrange County, Indiana, 1991–1994. MMWR 1994: 45 (26): 569–572.
Centers for Disease Control and Prevention (CDC). A Survey of the quality of water drawn from domestic wells in nine midwestern states. Centers for Disease Control and Prevention, Atlanta, GA, 1998.
Devine O.J., and Smith J.M. Estimating sample size for epidemiologic studies: the impact of ignoring exposure measurement uncertainty. Stat Med 1998: 17: 1375–1389.
Dorsch M.M., Scragg R.K.R., McMichael A.J., Baghurst P.A., and Dyer K.F. Congenital malformations and maternal drinking water supply in rural South Australia: a case control study. J Epidemiol 1984: 119: 473–485.
Eichholzer M., and Gutzwiller F. Dietary nitrates, nitrites, and N-nitroso compounds and cancer risk: a review of the epidemiologic evidence. Nutr Rev 1998: 56: 95–105.
Fan A.M., Willhite C.C., and Book S.A. Evaluation of the nitrate drinking water standard with reference to infant methemoglobinemia and potential reproductive toxicity. Regul Toxicol Pharmacol 1987: 7: 135–148.
Fan A.M., and Steinberg V.E. Health implications of nitrate and nitrite in drinking water: an update on methemoglobinemia occurrence and reproductive and developmental toxicity. Regul Toxicol Pharmacol 1996: 23: 35–43.
Hallberg G.R., Libra R.D., Ressmeyer G.G., Bettis E.A., and Hoyer B.E. Temporal changes in nitrates in groundwater in Northeastern Iowa. Iowa Geological Survey Open-File Report 1984, 81–84.
Johnson C.J., and Kross B.C. Continuing importance of nitrate contamination of groundwater and wells in rural areas. Am J Ind Med 1990: 18 (4): 449–456.
Kross B.C., Hallberg G.R., Bruner D.R., Cherryholmes K., and Johnson J.K. The nitrate concentration of private well water in Iowa. Am J Public Health 1993: 83 (2): 270–272.
Landon M.K., Delin G.N., Komor S.C., and Regan C.P. Relation of pathways and transit times of recharge water to nitrate concentration using stable isotopes. Ground Water 2000: 38: 381–395.
Maloney W.J., MacFarlane M.T., and Rimsza M.E. Methemoglobinemia in an infant. Arizona Med 1983: 40 (10): 700–702.
McDonald D.B., and Splinter R.C. Long-term trends in nitrate concentration in Iowa water supplies. J AWWA 1982: 74 (8): 437–440.
Montgomery E., Coyne M.S., and Thomas G.W. Denitrification can cause variable nitrate concentrations in shallow groundwater. Soil Sci 1997: 162: 148–156.
Mueller D.K., Hamilton P.A., Helsel D.R., Hitt K.J., and Ruddy B.C. Nutrients in ground water and surface water of the United States - an analysis of data through 1992, Water-Resources Investigations Report 95-4031. United States Geological Survey, Denver, 1995.
Mueller D.K., and Helsel D.R. Nutrients in the nation's waters - too much of a good thing?, US Geological Survey Circular 1136. United States Geological Survey, Denver, 1996.
SAS Institute Inc. SAS Technical Report P-229, SAS/STAT Software: Changes and enhancements, Release 6.0. SAS Institute Inc.: Cary, NC, USA, 1992 p 620.
Scragg R.R., Dorsch M.M., McMichael A.J., and Bayhurst P.A. Birth defects and household water supply. Epidemiologic studies in the Mount Zambia region of South Australia. Med J Australia 1982: 2: 577–579.
Swann P.F. The toxicology of nitrate, nitrite and N-nitroso compounds. J Sci Fd Agric 1975: 26: 1761–1770.
Townsend M.A., and Young D.P. Assessment of nitrate-nitrogen distribution in Kansas groundwater. Natural Resources Res 2000: 9: 125–134.
US EPA. Drinking Water Regulations and Health Advisories, EPA 822-R-94-001 Environmental Protection Agency, Washington, DC, 1994.
USGS. 2001. Nutrients in the nation's ground waters – too much of a good thing?, http://water.usgs.gov/nawqa/circ-1136/h4.html).
About this article
Cite this article
Ruckart, P., Henderson, A., Black, M. et al. Are nitrate levels in groundwater stable over time?. J Expo Sci Environ Epidemiol 18, 129–133 (2008) doi:10.1038/sj.jes.7500561
Understanding nitrate contamination based on the relationship between changes in groundwater levels and changes in water quality with precipitation fluctuations
Science of The Total Environment (2019)
International Journal of Environmental Research and Public Health (2017)
Spectroscopic and electrochemical behavior of a supramolecular tetrapyridylporphyrin encompassing four terpyridine(oxalate)chloridoruthenium(II) complexes and its use in nitrite sensors
Inorganica Chimica Acta (2015)
A Population-Based Case-Control Study of Drinking-Water Nitrate and Congenital Anomalies Using Geographic Information Systems (GIS) to Develop Individual-Level Exposure Estimates
International Journal of Environmental Research and Public Health (2014)
Private Well Testing in Oregon from Real Estate Transactions: An Innovative Approach toward a State-Based Surveillance System
Public Health Reports (2011)