Chlorination by-products in drinking water and risk of bladder cancer – A population-based cohort study

Chlorination by-products have been consistently associated with risk of bladder cancer in case-control studies, but confirmation from large-scale cohort studies is lacking. We assessed the association of drinking water tri-halomethanes (THM), a proxy for chlorination by-products, with risk of bladder cancer in 58,672 men and women. Data came from two population-based cohorts, parts of the Swedish Infrastructure for Medical Population-Based Life-Course and Environmental Research (SIMPLER). Individual exposure to THM was assessed by combining residential information with tap water monitoring data. Participants were categorized into non-exposed, low ( < 15 µ g/L) or high ( ≥ 15 µ g/L) THM exposure. Incident cases were ascertained from 1998 through 2019 via register linkage. During 16 years of follow-up (965,590 person-years), 831 bladder cancer cases were ascertained. We observed no overall association of THM with risk of bladder cancer, hazard ratio for the highest exposed compared to the non-exposed 0.90 (95% confidence interval: 0.73 – 1.11). The null association remained after restricting the analysis to long-term residents and across strata of smoking status and cancer stage. Our results indicate that chlorination by-product exposure at THM concentrations representative of chlorinated drinking waters in most European countries, is not associated with an increased risk of bladder cancer.


Introduction
Drinking water chlorination is an important public health intervention for control of waterborne infectious disease, and the most common drinking water disinfectant globally (World Health Organization, 2017).Nevertheless, the concomitant introduction of potentially carcinogenic chlorination by-products to the drinking water, presents a drawback of the method.These by-products are formed when chlorine reacts with natural organic matter and other substances in the water, and today more than 600 different by-product substances have been identified in drinking water (Richardson, 2011).
Trihalomethanes (THMs) are together with haloacetic acids, the chlorination by-products generally found at the highest concentrations in chlorinated drinking water (Richardson et al., 2007).The four most common THMs, chloroform, chlorodibromomethane, bromoform and bromodichloromethane, are all rodent carcinogens (Richardson et al., 2007) and two of them, chloroform and bromodichloromethane, are classified as category 2B (possibly carcinogenic to humans) by IARC (IARC, 1991(IARC, , 1999)).Several case-control studies (Beane Freeman et al., 2017;Bove et al., 2007;Cantor et al., 1987;Cantor et al., 1998;Chevrier et al., 2004;King and Marrett, 1996;Koivusalo et al., 1998;McGeehin et al., 1993;Villanueva et al., 2007), a historical cohort (Koivusalo et al., 1997) and a small prospective cohort study (Wilkins and Comstock, 1981) have all consistently reported associations of chlorination by-products with increased risk of bladder cancer.In addition, the most recent meta-analysis of case-control studies estimated 47% increased odds of bladder cancer for men exposed to chlorination by-products corresponding to THM concentrations above 50 µg/L compared to those exposed to levels ≤ 5 µg/L (Costet et al., 2011).Although the combined existing evidence point towards a possible cancer link, there is still a lack of large cohort studies that confirm these indications.Moreover, in a recent study the average population weighted THM concentration in public drinking water of 28 European countries was estimated to 12 µg/L (Evlampidou et al., 2020), which suggest that most European populations are exposed to chlorination by-product levels lower than those for which association with increased cancer risks have been reported.
To corroborate previous indications of a potential cancer link with chlorination-byproducts, and to further explore this association at THM concentrations representative of large parts of the European population, we used data from two population-based prospective cohort studies to assess the association between exposure to THMs, as a proxy for chlorination by-products, in drinking water and incidence on bladder cancer among 50 000 middle aged to elderly men and women in Sweden.

Study population
The Swedish Mammography Cohort (SMC) and the Cohort of Swedish Men (COSM) are two large population-based prospective cohorts, parts of the Swedish Infrastructure for Medical Population-based Life-course and Environmental Research (SIMPLER; www.simpler4 health.se).The SMC was initiated in [1987][1988][1989][1990], when all women residing in two counties in central Sweden (Uppsala or adjacent Västmanland county), born 1914-1948, were invited to the study (n = 90,303) by receiving a diet and lifestyle questionnaire (74 % response rate).In 1997, an extended questionnaire was sent out to all women still alive and residing in the study area (n=56,030) to update information on diet and lifestyle (70 % response rate).After excluding those with incorrect or missing personal identification number and pre-baseline cancer, 37,193 women remained in the study.
At the time of the second SMC questionnaire, the male cohort, COSM, was initiated.All men residing in Västmanland and Örebro counties and who were born 1918-1952 (n = 100,303) received a questionnaire essentially identical to the SMC questionnaire.In total 48,850 men returned a completed questionnaire (response rate 49 %).After excluding those with incomplete or missing personal identification numbers or a cancer diagnosis prior to baseline, the cohort consisted of 45,872 men.The present study is based on the 1997 questionnaire for both cohorts.In total 83,065 men and women were eligible for further chlorination by-product exposure assessment (Fig. 1).The study was approved by the Regional Ethical Review Board in Stockholm, Sweden.

Assessment of chlorination by-product exposure
We only included participants served by municipal drinking water, because this water is closely monitored with respect to several important drinking water parameters.Moreover, by not including populations served by private wells, we minimized the risk of introducing unmeasured confounding.To identify participants receiving municipal drinking water, we collected information on individual annual residential history from the National Register for Regional Divisions Based on Real Estate (Statistics Sweden) for all years available in the database .We then included in the study only those residing in localities  (coherent and densely populated areas) with ≥1,000 inhabitants at baseline in 1997 (n=63,804).The reason for not including smaller localities was that some inhabitants in those localities may not yet be connected to the municipal drinking water system (i.e., to avoid exposure misclassification).For each locality, we then identified the drinking water treatment plant(s) responsible for the drinking water distribution in each area and collected data on their method(s) for disinfection and THM monitoring data from Swedish Water analytical reports and Vattentäktsarkivet, Geological Survey of Sweden.If no monitoring data were available, or if changes were made in water source or chlorination procedure during follow-up, participants in these localities were excluded (n=5,132).This resulted in a final study population of 58,672 men and women (Fig. 1).
We used the sum of THMs (sum of chloroform, bromoform, dibromochloromethane and bromodichloromethane, n=406 sampled in 2008-2017) as a proxy of chlorination by-product exposure.These are the by-products that are formed at the highest concentrations when hypochlorite is used as the main disinfectant (Richardson et al., 2007), and are routinely monitored according to the European drinking water regulation (98/83/EG).In our study, more than 99% of the study population in the exposed areas was served by drinking water treatment plants using hypochlorite as their main disinfectant (only two plants serving in total only 543 participants used chloramine).To account for seasonal variability in by-product formation and municipal monitoring strategies, we first calculated the average THM concentration in drinking water for each water treatment plant (n=52) and season (December-February; March-May; June-August; September-November), and then averaged these treatment plant-specific seasonal averages (Fig. S1).The rationale for using this approach is that season is the most important determinant of temporal variations in THM concentrations in Sweden (Andersson et al., 2019), displaying bimodal peaks during spring and autumn due to concomitant increase of organic matter in the water.Conditioning on that no changes had occurred in the drinking water production, we extrapolated these averages back to baseline (1997).In two cases, when localities were served by two water treatment plants, we used the average THM concentrations of the two plants.
The non-exposed reference group, denoting zero THM exposure, were those participants residing in localities served by water treatment plants that used no chlorination in their production.For participants living in areas where chlorination was used, THM exposure was separated into two exposure categories that were predefined based on prior knowledge on the THM exposure distribution in Sweden, representing low (<15 μg/L) and high (≥15 μg/L) THM exposure, respectively.

Assessment of covariates
Data on arsenic concentrations in drinking water was based on municipal monitoring data.Information on household income and level of education was obtained from the Longitudinal integrated database for health insurance and labor market studies (LISA, Statistics Sweden).Self-reported information on smoking status, alcohol consumption, body mass index (BMI; calculated from body weight (in kg) divided by height (in m) squared) and physical activity was obtained from the questionnaire completed at baseline.Information on intake of cultured milk (Larsson et al., 2008), processed meat (Crippa et al., 2018), fruits and vegetables (World Cancer Research Fund/American Institute for Cancer Research, 2018), tea (World Cancer Research Fund/American Institute for Cancer Research, 2018) and coffee, all of which are potentially associated with bladder cancer risk, was derived from the validated 96-item food frequency questionnaire (FFQ) (Harris et al., 2013).Information on prevalent diabetes was based on self-reports and on linkage of the cohorts to the National Diabetes and Patient Registers (the National Board of Health and Welfare).

Outcome assessment
Incident cases of bladder cancer, including bladder cancer in situ, were ascertained through a linkage of the cohort to the National Cancer Register, with over 90% completeness for common cancers (Barlow et al., 2009), and the National Patient Register (National Board of Health and Welfare).The International classification of disease 10th revision (ICD-10) codes used were C67 (bladder cancer) and D09.0 (bladder cancer in situ).The rationale for including bladder cancer in situ is that it is a well-defined form of bladder cancer, with a high risk of progressing into muscle invasive bladder cancer if left untreated (Tang and Chang, 2015).To examine any potential differences in the association of THM with bladder cancer by disease stage, incident cases were further classified by tumor-node-metastasis (TNM) classification stage, into non-muscle invasive (stage Ta, Tcis and T1) and muscle-invasive bladder cancer (≥ T2), with the latter signaling a more aggressive disease with worse prognosis (Witjes et al., 2021).Information TNM-stage was obtained from National Cancer Register.

Statistical analysis
We used Cox proportional hazards regression analysis with attained age as underlying time scale, to estimate hazard ratio (HR) and the 95% confidence interval (CI).Participants contributed to person-time from January 1 st , 1998 until the date of bladder cancer diagnosis, death, emigration, moving from the baseline locality (+2 years) or end of follow-up (December 31 st , 2019), whichever occurred first.The rationale for censoring when participants moved was to ensure unceasing exposure to THMs, and because bladder cancer has a long induction time, the date was set to 2 years after the moving date.
All multivariable-adjusted models were adjusted for the following potential confounders: age (as time scale), cohort (as a stratum variable), level of education, household income, smoking status and drinking water arsenic concentration.Further adjustment for prevalent diabetes, BMI, physical activity, alcohol intake, intake of cultured milk, processed meat, fruits and vegetables, tea and coffee consumption did not have any impact on the association and estimates from these models are therefore not reported for subsequent analyses.Missing information on categorical covariates was handled using a missing indicator category, while missing data on drinking water arsenic (9%) was replaced by the median.The proportional hazard assumption was tested using the Schoenfeld's residuals and no department from the assumption was observed.
In addition to the main analysis, we performed several sensitivity analyses.First, to explore the impact of sex and smoking status (both strong risk factors for bladder cancer) on the THMbladder cancer association, we conducted a likelihood ratio test for interaction between THM and sex/smoking status and performed stratified analyses for men and women as well as never and ever smokers.Second, to assess potential influence of long-term exposures, we repeated the main analyses after restricting to individuals with constant long-term THM exposure (since 1982).Lastly, we explored potential differences in the association with bladder cancer of different stages, by performing separate analyses for non-muscle invasive bladder cancer and muscle invasive bladder cancer.All tests were two sided with the level of significance set at 0.05.The software used for statistical analysis was STATA/SE version 16.0 (Stata Corporation, Inc., Collage Station, TX, USA).

Results
The average THM concentration in all areas using chlorination was 8.8 ± 6.5 µg/L, and more than 99% of the participants in the study were exposed to THM concentrations below 20 µg/L (Supplemental Table S1, Supplemental Fig. S1).While THMs were always below the limit of detection in the non-chlorinated areas, the average THM concentrations were 6.6 ± 2.2 µg/L and 17.1 ± 2.1 µg/L in the low and high exposure category, respectively.Baseline age-standardized main characteristics of the study population by drinking water THM exposure categories are presented in Table 1.Participants in the non-chlorinated area were less E. Helte et al. likely to be former heavy smokers (>10 cigarettes/day), somewhat less educated and to a greater degree served by water treatment plants with a slightly higher arsenic concentration in drinking water.
During a mean follow-up of 16 years (in total 965,590 person-years), 831 out of the 58,672 participants developed bladder cancer (699 men and 132 women).TNM status was obtained for 70% of the cases resulting in 474 and 112 non-muscle invasive and muscle invasive bladder cancer cases, respectively.In the multivariable-adjusted model we observed overall no association of exposure to THM with bladder cancer risk, HR: 0.90 (95% CI: 0.73 -1.11) for the highest exposed category compared to the non-chlorinated area (Table 2).We found no evidence of effect modification by sex, and results are therefore presented for the entire study population as a whole (analyses stratified by sex are shown in Table S2).Likewise, we observed overall no association of THM with risk of bladder cancer when stratifying the analyses by smoking status (ever/never smokers, p interaction = 0.88, Table 2) or restricting it to only long-term residents (Table 3).The association did not vary by cancer stage, HRs for the high exposure category versus nonchlorinated area were 0.86 (95% CI: 0.65 -1.13) and 0.65 (95% CI: 0.37 -1.14) for non-muscle invasive and muscle invasive bladder cancer, respectively (Table 4).

Discussion
In this population-based cohort study of upper middle-aged to elderly men and women, we observed no association of drinking water THMs with increased risk of bladder cancer.The results did not vary with smoking status or when the analysis was restricted to participants with long-term constant exposure.Including more than 800 cases, this is the first larger cohort study to evaluate the association of chlorination by-products with bladder cancer risk.Our results suggest that at least up to THM concentrations around 20 µg/L, which is representative of chlorinated drinking waters in most European countries, chlorination by-products seem not to be associated with increased risk of bladder cancer.
Chlorination by-products is a complex mixture of substances that is formed when chlorine oxidizes naturally occurring anthropogenic substances, iodine and bromide drinking water (Richardson, 2011).Since the first discoveries of these by-products were made in the 70´s, numerous studies have been performed to increase our understanding of the potentially associated health risks.THMs, which often is the most  c Further adjusted for level of education (< 9 yrs, 9 yrs, 10-11 yrs, 11-12 yrs, >12 yrs), household income (quartiles), smoking status (never, former < 10 cig/ day, former ≥ 10 cig/day current < 10 cig/day, current ≥ 10 cig/day) and drinking water arsenic (continuous, µg/L).
E. Helte et al. prevalent class of chlorination by-products in chlorinated drinking water, has been mostly studied.In toxicological in vitro assays, three out of the four most common THMs bromodichloromethane, chlorodibromomethane, and bromoform test positive for genotoxicity after glutathione S-transferase-theta (GSTT1) activation, while the fourth one -chloroform, generally test negative (Richardson et al., 2007).In rodent bioassays, all four THMs induce tumours, the latter likely through a non-genotoxic mode of action, suggested to involve cytotoxicity and regenerative cell proliferation (Richardson et al., 2007).Also, other classes of chlorination by-products potentially present, such as haloacetic acids, display similar patterns of genotoxicity and/or rodent carcinogenicity (Richardson et al., 2007).
The epidemiological evidence linking chlorination by-products to overall cancer is somewhat less clear.The strongest evidence for a carcinogenic potential in humans exists for bladder cancer, with several case-control studies (Beane Freeman et al., 2017;Bove et al., 2007;Cantor et al., 1987Cantor et al., , 1998;;Chevrier et al., 2004;King and Marrett, 1996;Koivusalo et al., 1998;McGeehin et al., 1993;Villanueva et al., 2007), a historical cohort (Koivusalo et al., 1997), and a small cohort study with only 52 cases (Wilkins and Comstock, 1981), repeatedly reporting associations with increased risks, although some inconsistencies exist (Doyle et al., 1997;Freedman et al., 1997).In addition, in one of the most comprehensive studies reporting direct associations, the authors found that a polymorphism in GSTT1 -a key metabolizing enzyme of THMs, modified the THMbladder cancer association (Cantor et al., 2010).Stronger associations were observed among individuals with one or two copies of the functioning allele than among those without, supporting a possible causal relationship.Further, different case-control meta-and pooled analyses have been performed, indicating a possible link at THM exposures above approximately 25 µg/L, that was restricted to men (Costet et al., 2011;Villanueva et al., 2004).Nevertheless, although the combined existing evidence point towards a possible association, there is still a lack of large-scale cohort studies that confirms these indications, and it is also uncertain if this association persist at lower exposure levels, more representative of large parts of the population today (Evlampidou et al., 2020).In our study, we observed no overall association of drinking water THM up to 20 µg/L with risk of bladder cancer, confirming the lack of an association at low exposure levels observed in pooled analysis of case-control studies (Costet et al., 2011).This is also in line with the results of a small population-based cohort study of postmenopausal women in Iowa, US (Doyle et al., 1997), and a more recent population-based case-control study from New England, US (Beane Freeman et al., 2017), in which the investigators found overall no association of individual lifetime THM concentration with bladder cancer risk below the 95 th percentile of THM exposure (46 µg/L).
Strengths of this study include the population-based cohort design, the linkage to high quality registers for outcome ascertainment and covariate information, the detailed information on important risk factors for bladder cancer, the use of a representative non-chlorinated reference area and the access to a large database on municipal THM monitoring data.
The complexity of THM formation requires careful consideration, since THM levels vary with season, temperature, pH and retention time in the distribution system.In our study, we tried to minimize the impact of seasonal variations, which are larger than the annual variations in Northern European countries (Andersson et al., 2019), by using multiannual seasonal averages.Moreover, while spatial differences in THM levels in large distribution systems is a common issue, most water treatment plants in the present study were small and had a rapid turnover time, thus relatively low spatial variations can also be assumed.
Another point to consider is that we used THMs as a proxy for general chlorination by-product exposure.Although these are frequently used as such, THMs are poor indicators for some other by-products that are formed when disinfectants other than hypochlorite is used.Nevertheless, in our study, more than 99% of the exposed study population were served by drinking water treatment plants employing hypochlorite treatment.Moreover, we had limited data on each of the four components of the THMs -which all display slightly different toxicological patterns in vitro and in vivo (Richardson et al., 2007), and thus, we were not able to assess the associations of each of these separately with risk of bladder cancer.On the other hand, chlorination by-products does not only consist of THMs, but is a complex mixture of several hundreds of substances which may have different toxicological patterns and may or may not interact with each other.This complexity must be taken into consideration when interpreting results of studies, including ours, that are relying on data on a single group or substance of by-products, for exposure assessment.One limitation of our study is that we did not have access to individual information on drinking water consumption.However, given that THM exposure occurs not only through ingestion but also to a large extent via inhalation and dermal absorption (Villanueva et al., 2007), this might not be an issue of major concern.By contrast, only considering ingested water could potentially lead to misclassification of individuals highly exposed through routes other than consumption.The risk for unmeasured or residual confounding is unavoidable in observational studies, and in our study, we did not have access to information on employment in risk occupations.However, we had detailed information on other, more important potential confounders such as smoking, which has been estimated to contribute to 45 % of all cases compared to the estimated 6 % caused by occupational exposure (Cancer Research UK, 2021).Finally, the THM concentrations in our study were low in comparison to those in earlier studies reporting direct associations with bladder cancer, thus, we cannot rule out that increased risks may occur at higher concentrations.

Conclusion
In this population-based cohort study of upper middle-aged to elderly men and women, we observed no association of THMs, as a proxy for chlorination by-products, in drinking water with increased bladder cancer risk.Our results suggest that at least at THM concentrations up to 20 µg/L, which are typically found in chlorinated drinking water in major parts of Europe, chlorination by-products seem not to be associated with increased risk of bladder cancer.

Fig. 1 .
Fig. 1.Flow chart of SIMPLER study population *This number refers to women who had cancer between the first (1987) and the second (1997) questionnaire.Women who had a cancer diagnosis before the first questionnaire were excluded from the cohort already in 1987.

Table 1
Baseline age-standardized main characteristics of the study population (n=58,672) by THM exposure categories.

Table 2
Hazard ratios of bladder cancer by THM exposure categories in the 58,672 men and women in SIMPLER and stratified by smoking status (ever or never smokers).

Table 3
Hazard ratios of bladder cancer by THM exposure categories in the 48,508 men and women in SIMPLER with long-term constant exposure to THM since 1982.