Association of Domestic Water Hardness with All-Cause and Cause-Specific Cancers: Evidence from 447,996 UK Biobank Participants

Abstract

Background:

Accumulating evidence suggests that domestic water hardness is linked to health outcomes, but its association to all-cause and cause-specific cancers warrants investigation.

Objective:

The aim of this study was to investigate the association of domestic hard water with all-cause and cause-specific cancers.

Methods:

In the prospective cohort study, a total of 447,996 participants from UK Biobank who were free of cancer at baseline were included and followed up for 16 y. All-cause and 22 common cause-specific cancer diagnoses were ascertained using hospital inpatient records and self-reported data until 30 November 2022. Domestic water hardness, measured by ${\text{CaCO}}_3$ concentrations, was obtained from the local water supply companies across England, Scotland, and Wales in 2005. Data were analyzed using Cox proportional hazard models, with adjustments for known measured confounders, including demographic, socioeconomic, clinical, biochemical, lifestyle, and environmental factors.

Results:

Over a median follow-up of 13.6 y (range: 12.7–14.4 y), 58,028 all-cause cancer events were documented. A U-shaped relationship between domestic water hardness and all-cause cancers was observed ($p$ for nonlinearity $<0.001$). In comparison with individuals exposed to soft water ($0–60\mathrm{mg}/\mathrm{L}$), the hazard ratios (HRs) and 95% confidence intervals (CIs) of all-cause cancer were 1.00 (95% CI: 0.98, 1.02) for those exposed to moderate hard water ($>60–120\mathrm{mg}/\mathrm{L}$), 0.88 (95% CI: 0.84, 0.91) for those exposed to hard water ($>120–180\mathrm{mg}/\mathrm{L}$) and 1.06 (95% CI: 1.04, 1.08) for those exposed to very hard water ($>180\mathrm{mg}/\mathrm{L}$). Additionally, domestic water hardness was associated with 11 of 22 cause-specific cancers, including cancers of the esophagus, stomach, colorectal tract, lung, breast, prostate, and bladder, as well as non-Hodgkin lymphoma, multiple myeloma, malignant melanoma, and hematological malignancies. Moreover, we observed a positive linear relationship between water hardness and bladder cancer.

Discussion:

Our findings suggest that domestic water hardness was associated with all-cause and multiple cause-specific cancers. Findings from the UK Biobank support a potentially beneficial association between hard water and the incidence of all-cause cancer. However, very hard water may increase the risk of all-cause cancer. https://doi.org/10.1289/EHP13606

Introduction

Cancer is a leading cause of mortality worldwide in 2020, nearly 10 million deaths were due to cancer, accounting for approximately one-sixth of all deaths.^1–3 It is projected that there will be a total of 28.4 million new cancer cases worldwide by 2040, which is a 47% increase in comparison with the record in 2020.⁴ Cancer is a multifactorial disease with a complex web of etiological factors and the exact mechanisms underlying its pathogenesis and precise mechanism remain unknown.^5,6

Epidemiological studies have suggested that multiple risk factors are involved in the development of cancer, including alcohol consumption,^7,8 tobacco use,⁹ physical exercise,¹⁰ and environmental factors.^11,12 One proposed environmental risk factor for cancer is domestic hard water,^13–17 which contains higher levels of calcium and magnesium.^18,19 Previous studies have investigated the association between water hardness and cancer risk. Four studies from Taiwan found a negative association between water hardness and specific cancers, including esophageal, gastric, and colorectal cancers.^13–15,20 In addition, in a cross-sectional analysis, Cheng et al. reported that a high intake of magnesium from drinking water was linked to a lower lung cancer risk among women, whereas calcium levels did not show similar associations.²¹ Yang et al. found an inverse association between calcium and magnesium levels in drinking water and breast cancer mortality.²² Conversely, a case–control study observed a positive association between calcium intake and bladder cancer risk.²³ However, these studies were limited by their small sample size and cross-sectional design. Therefore, the relationship between hard water and cancer requires further investigation.

Hard water is recognized as a significant source of calcium and magnesium minerals.¹⁸ Calcium and magnesium are involved in the metabolic and physiological functions of the body that might affect the pathophysiology of carcinogenesis. Calcium is important for regulating proliferation and apoptosis in both malignant and immune cells.²⁴ Magnesium is vital for maintaining genomic stability, facilitating DNA replication and repair, regulating cell proliferation and apoptosis, and supporting immune responses.^18,25–28 Both calcium and magnesium are essential minerals and beneficial to human health. Therefore, we hypothesized that hard water may be associated with a lower risk of cancer because of its high calcium and magnesium content.

The British Drinking Water Inspectorate²⁹ indicates that drinking water in England is generally very hard, particularly in the east of England, where calcium carbonate (${\text{CaCO}}_3$) equivalent levels exceed $200\mathrm{mg}/\mathrm{L}$. Consequently, it is crucial to explore the correlation between water hardness and cancer within the context of the United Kingdom, where very hard water is prevalent. The UK Biobank has longitudinal follow-up of a large cohort for over a decade, with information on participants’ domestic water supply.³⁰ Thus, this cohort provides an opportunity to explore the correlation between hard water and cancer within a large-scale, population-based prospective study. Therefore, we aimed to assess the relationship between exposure to hard water and the incidence of all-cause and cause-specific cancers in participants enrolled in the UK Biobank.

Methods

Study Design and Population

The UK Biobank is a nationwide, prospective cohort study that recruited over 500,000 participants (37–73 y of age) between April 2006 and December 2010 and conducted follow-up until November 2022.³⁰ The participants were invited to attend one of 22 assessment centers across the United Kingdom (including England, Scotland, and Wales). Extensive information on demographic, clinical, biochemical, lifestyle, and environmental factors was collected from touch-screen questionnaires, computer-assisted interviews, physical and function examinations, sample assays, and electronic health records. Details of the UK Biobank’s data collection, assessment, and methodology have been previously reported.³¹

From the 502,412 initially enrolled participants, those with prevalent all-cause and cause-specific cancer (defined as cancer diagnosed prior to the date of baseline assessment), with missing data on domestic water hardness, and who dropped out during the follow-up period were excluded. Finally, 447,996 participants who were free of cancer and had available records of domestic water hardness were eligible for the analysis. Flowchart for the selection of the study sample was shown in Figure 1.

Figure 1.

Flowchart for the selection of the analyzed study sample ($n=\mathrm{447,996}$) from the UK Biobank Study.

Domestic Water Hardness

Data on domestic water hardness were collected in 2005. Each participant’s residential address was collected at baseline assessment (2006–2010) by the UK Biobank. Data on domestic water supply were collected from local water supply companies in England, Wales, and Scotland by the University of Melbourne,³² which then provided this category to UK Biobank (Field IDs: 21100; 21101; 21103; 21104; 21105). Data on domestic water, including ${\text{CaCO}}_3$ (mg/L), calcium (mg/L), and magnesium (mg/L) concentrations were collected. Domestic water ${\text{CaCO}}_3$ concentration, which related to the concentrations of calcium and magnesium in the water, were used to measure the water hardness. In our study domestic water hardness has been classified in two ways: a) categories into four groups according to US Geological Survey (USGS) classification of ${\text{CaCO}}_3$ concentration: soft water ($0–60\mathrm{mg}/\mathrm{L}$), moderately hard water ($>60–120\mathrm{mg}/\mathrm{L}$), hard water ($>120–180\mathrm{mg}/\mathrm{L}$), and very hard water ($>180\mathrm{mg}/\mathrm{L}$); b) categories into two groups following the World Health Organization (WHO) classification of ${\text{CaCO}}_3$ concentration: soft water ($<200\mathrm{mg}/\mathrm{L}$), and hard water ($\ge 200\mathrm{mg}/\mathrm{L}$).

Participants were assigned postcodes based on their residential address reported at baseline, and the researchers³² used these assigned postcodes to link to the water supply companies. A postcode, serving as an abbreviated form of address, is composed of a combination of five to seven alphanumeric characters. A postcode may cover between 1 and 100 addresses, the average number per postcode being 15 addresses. Information on water supplies in the United Kingdom is available by entering a postcode at: https://www.water.org.uk/customers/find-your-supplier. The range of hard water concentrations in the UK is wide and covers areas of soft, moderately hard, hard, and very hard water.

Assessment of All-Cause and Cause-Specific Cancer

Cases of all-cause and cause-specific cancer were ascertained through self-report of physician diagnosis, linkage to national cancer registries, hospital admissions, or diagnosed by a doctor. Cancer outcomes were coded according to International Classification of Diseases, ninth and 10th revision (ICD-9 and ICD-10), depending on date of diagnosis. The cancers included all-cause cancer, oropharyngeal cancer, esophageal cancer, stomach cancer, colorectal cancer, liver cancer, pancreatic cancer, lung cancer, malignant melanoma, breast cancer, cervical cancer, endometrial cancer, ovarian cancer, prostate cancer, kidney cancer, bladder cancer, brain cancer, thyroid cancer, cancer of the hepatobiliary tract, non-Hodgkin lymphoma, multiple myeloma, leukemia, and hematological malignancies. We set the cause-specific cancer inclusion criteria to require at least 100 incident cases to ensure statistical power for analyses. Details about the codes for outcomes are provided in Table S1.

Covariates

Data on sex, age, educational attainment, socioeconomic status, employment status, family history of cancer, smoking status, alcohol consumption, and physical activity were available from a touch-screen questionnaire, and diet was derived from the Food Frequency Questionnaire. Previous studies have found that the incidence of cancer differs by race.^33–35 Therefore, we included race as a confounder and effect modifier. Data on race were self-reported and collected through the touch-screen questionnaire, including White, mixed (White and Black Caribbean, White and Black African, White and Asian, and any other mixed background), Asian or Asian British, Black or Black British, Chinese, and other ethnic group. Socioeconomic status was defined based on the Townsend deprivation index³⁶ (encompassing information on social class, employment, car availability, and housing) and categorized as low (highest quintile), middle (quintiles 2–4), or high (lowest quintile).³⁷ Physical activity was divided into three levels according to metabolic equivalent (MET) of tasks min/wk: low ($<600$ MET min/wk), middle (600–3,000 MET min/wk), and high ($>3000$ MET min/wk).³⁸ Employment status was categories as employed, retired, and others. Smoking status was defined as never, former, and current smokers. Alcohol consumption was calculated based on the self-reported intake of white wine, red wine, beer, spirits, and fortified wine. Excessive alcohol consumption was defined as $>14\mathrm{g}/\mathrm{d}$ alcohol for women and $>28\mathrm{g}/\mathrm{d}$ alcohol for men.³⁹ A healthy diet score was generated based on the seven commonly eaten food groups (vegetables $\ge 3$ servings/d; fruits $\ge 3$ servings/d; unprocessed red meats $\le 1.5$ servings/wk; processed meats $\le 1$ serving/wk; fish $\ge 2$ servings/wk, whole grains $\ge 3$ servings/d; refined grains $\le 1.5$ servings/d), and a healthy diet was based on intake of at least four of these seven commonly eaten food groups.⁴⁰ Body mass index (BMI) was calculated as weight (kg) divided by height squared (${\mathrm{m}}^2$) and categorized as $<25{\mathrm{kg}/\mathrm{m}}^2$ (normal weight), $25–29.9{\mathrm{kg}/\mathrm{m}}^2$ (overweight), and $\ge 30{\mathrm{kg}/\mathrm{m}}^2$ (obese). Data on the C-reactive protein (CRP) levels were obtained from fasting blood samples collected from participants at the initial assessment center visit. Data on residential duration was self-reported and collected through the touchscreen questionnaire. Percentages of residential natural environment was the proportion of natural environment land-use type in a $\mathrm{1,000}\text{-m}$ buffer from each participant’s dwelling coordinates, taken from the 2007 Land Cover Map (LCM) data of the Center for Ecology and Hydrology (CEH).⁴¹ Use of sun/ultraviolet (UV) protection was categorized as never/rarely, sometimes, most of the time, always, and others. Data on sex-specific factors, including prostate-specific antigen (PSA) test, hormone replacement therapy (HRT), menopause, parity, oral contraceptive pill or minipill, age when periods started (menarche), and hysterectomy were collected via a touch-screen questionnaire. PSA test, HRT, menopause, and hysterectomy were dichotomized as “yes” and “no.” Normal menarche was defined as the age of the first period (menarche) at 12–14 y of age. Information on heart disease and stroke was derived from medical records (codes I20–I25 for heart disease and codes I60–I64 for stroke). Diabetes was ascertained on the basis of medical records (codes E10–E14), self-reports, glycated hemoglobin $\ge 6.5\%$, and use of antidiabetic drugs. Hypertension was defined as systolic blood pressure (SBP) $\ge 140\text{\hspace{0.17em}mmHg}$ or diastolic blood pressure (DBP) $\ge 90\text{\hspace{0.17em}mmHg}$, use of antihypertension agents, or medical records (codes I10–I13, I15). Comorbidities were defined as having any of the following: heart disease, stroke, hypertension, and diabetes; and comorbidities were categorized as none, one, two or more comorbidities. Data on death was obtained through linkage to national death registries from May 2006 to November 2022.

Statistical Analyses

Baseline characteristics of the study population were summarized across domestic water hardness groups as means and standard deviations (SDs) for continuous variables and as number and percentages for categorical variables and compared by domestic water hardness using one-way analysis of variance (ANOVA) or Mann–Whitney U–test for continuous variables and chi-square tests for categorical variables.

Restricted cubic spline models were used to evaluate the does-response relationship between domestic water hardness and incident all-cause and cause-specific cancers, with four knots at the 25th, 50th, 75th, and 95th centiles. Cox-proportional hazard models were conducted to estimate the prospective association of domestic water hardness with the risk of incident developing all-cause and cause-specific cancer. The years of follow-up were determined as the time scale. Hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) of incident all-cause and cause-specific cancer were calculated to report the results. The proportional hazards assumptions for the Cox model were tested using the Schoenfeld residuals method, and no violation of the assumption was observed. All analyses were adjusted for sex, age, race, educational attainment, employment status, socioeconomic status, family history of cancer, alcohol consumption, smoking status, physical activity, diet pattern, CRP, hypertension, heart disease, stroke, and diabetes. Use of sun/UV protection was further included as covariate in malignant melanoma model, and gastro-esophageal reflux disease was included in the esophageal cancer model. For sex-specific cancers (e.g., breast cancer, cervical cancer, endometrial cancer, ovarian cancer, and prostate cancer), separate models were run for males and females and additionally adjusted for PSA in males for prostate cancer and HRT, menopause, parity, oral contraceptive pill or minipill, age of menarche, and hysterectomy in females for breast cancer, cervical cancer, endometrial cancer, ovarian cancer. Multiple imputations for the missing covariates were performed using the R package “mice.”⁴² Specifically, predictive mean matching was used for continuous variables (socioeconomic level, alcohol consumption, physical activities, diet score, BMI, and CRP), whereas logistic regression was used for categorical variables (race, education level, current employment status, smoking status, use of sun/UV protection, menopause, parity, age of menarche, HRT, PSA test, oral contraceptive pill, and hysterectomy). The process was repeated five times, and five imputed datasets were obtained. These five imputed datasets were entered into the model individually, and the results were then pooled to one estimate. Detailed information on missing covariates was shown in Table S2.

We further performed several additional analyses to examine the robustness of the results. They are as follows: a) We conducted stratification analysis to explore whether the association of domestic water hardness with incident all-cause and cause-specific cancer varied by sex, age, race, educational attainment, socioeconomic status, employment status, family history of cancer, smoking status, alcohol consumption, physical activity, diet pattern, and comorbidities. In addition, we examined the interaction between domestic water hardness and potential confounders. b) We excluded participants with missing values on covariates (before imputation) to conduct the analysis (included $n=\mathrm{349,321}$ participants with complete information). c) We applied a subdistribution method⁴³ to investigate the competing risk of noncancer death on the association of domestic water hardness with all-cause cancer. d) We restricted analyses to individuals living at the same residential address for $\ge 10$ y to minimize exposure misclassification. e) We further adjusted for residential natural environment percentage. Furthermore, we evaluated the association of domestic water hardness, as classified by WHO terms, with the risk of incident developing all-cause and cause-specific cancer.

R software (version 4.2.1; R Development Core Team) and STATA 15 (StataCorp LLC) were used to perform the statistical analyses. A $p$-value $<0.05$ was set as statistical significance.

Results

A total of 447,996 UK Biobank participants who were free of cancer were included for analysis. Among those, the median age was 57 y [interquartile range (IQR), 50 to 63 y], and 46.5% individuals were men. Over the follow-up (median, 13.6; IQR: 12.7–14.4 y), 58,028 all-cause cancer events were documented. The characteristics of the participants according to domestic water hardness, as classified by USGS, are shown in Table 1. According to the USGS classification, 198,208 participants (44.2%) were exposed to hard or very hard water. The distribution of domestic water hardness was depicted in Figure 2. Participants exposed to hard water were likely to be younger, women, have a normal weight, and have higher education and socioeconomic status in comparison with those exposed to soft water, who were less likely to be White, excessive alcohol drinkers, physically active, and have a higher prevalence of comorbidities (e.g., heart disease, stroke, and diabetes).

Table 1.

Baseline characteristics of UK Biobank participants ($n=\mathrm{447,996}$) across domestic water hardness categories.

Characteristics	Water hardness				$p$-Valuea
	$0–60\mathrm{mg}/\mathrm{L}$	$>60–120\mathrm{mg}/\mathrm{L}$	$>120–180\mathrm{mg}/\mathrm{L}$	$>180\mathrm{mg}/\mathrm{L}$
[$n$ (%)]	161,359 (36.0)	88,429 (19.7)	26,361 (5.9)	171,847 (38.4)	—
Sex [$n$ (%)]	—	—	—	—	$<0.001$
Female	86,074 (53.3)	46,465 (52.5)	14,263 (54.1)	93,110 (54.2)	—
Male	75,285 (46.7)	41,964 (47.5)	12,098 (45.9)	78,737 (45.8)	—
Age [$\text{mean}\pm \text{SD}$ (y)]	$56.46\pm 8.08$	$56.61\pm 8.01$	$56.00\pm 7.96$	$55.94\pm 8.20$	—
Race [$n$ (%)]	—	—	—	—	$<0.001$
White	156,215 (96.8)	85,451 (96.6)	25,363 (96.2)	155,625 (90.6)	—
Others	5,144 (3.2)	2,978 (3.4)	998 (3.8)	16,223 (9.4)	—
Education level [$n$ (%)]	—	—	—	—	$<0.001$
College or university	49,741 (30.8)	23,350 (26.4)	8,989 (34.1)	64,179 (37.3)	—
Other	111,618 (69.2)	65,079 (73.6)	17,372 (65.9)	107,668 (62.7)	—
Socioeconomic status [$n$ (%)]	—	—	—	—	$<0.001$
High	32,862 (20.4)	18,141 (20.5)	6,545 (24.8)	32,887 (19.1)	—
Middle	95,701 (59.3)	56,091 (63.4)	15,980 (60.6)	100,329 (58.4)	—
Low	32,796 (20.3)	14,197 (16.1)	3,836 (14.6)	38,631 (22.5)	—
Current employment [$n$ (%)]	—	—	—	—	$<0.001$
Employed	91,660 (56.8)	50,215 (56.8)	15,863 (60.2)	105,318 (61.3)	—
Retired	54,994 (34.1)	30,518 (34.5)	8,294 (31.5)	50,517 (29.4)	—
Others	14,705 (9.1)	7,696 (8.7)	2,204 (8.4)	16,012 (9.3)	—
Had family history of cancer [$n$ (%)]	—	—	—	—	$<0.001$
No	104,623 (64.8)	57,354 (64.9)	17,462 (66.2)	114,649 (66.7)	—
Yes	56,736 (35.2)	31,075 (35.1)	8,899 (33.8)	57,198 (33.3)	—
Smoking status [$n$ (%)]	—	—	—	—	$<0.001$
Never smoker	88,539 (54.9)	49,270 (55.7)	14,636 (55.5)	94,648 (55.1)	—
Former smoker	54,739 (33.9)	30,250 (34.2)	8,890 (33.7)	59,333 (34.5)	—
Current smoker	18,081 (11.2)	8,909 (10.1)	2,835 (10.8)	17,866 (10.4)	—
Alcohol consumption [$n$ (%)]	—	—	—	—	$<0.001$
Never or moderate	72,238 (44.8)	39,112 (44.2)	11,772 (44.7)	73,278 (42.6)	—
Excessive	89,121 (55.2)	49,317 (55.8)	14,589 (55.3)	98,569 (57.4)	—
Physical activities category [$n$ (%)]	—	—	—	—	$<0.001$
Low	30,267 (18.8)	16,553 (18.7)	5,428 (20.6)	31,251 (18.2)	—
Middle	79,724 (49.4)	43,315 (49.0)	13,176 (50.0)	89,217 (51.9)	—
High	51,368 (31.8)	28,561 (32.3)	7,757 (29.4)	51,379 (29.9)	—
Diet [$n$ (%)]	—	—	—	—	—
Unhealthy	87,175 (54.0)	48,183 (54.5)	15,087 (57.2)	96,854 (56.4)	—
Healthy	74,184 (46.0)	40,246 (45.5)	11,274 (42.8)	74,993 (43.6)	—
Comorbidities	—	—	—	—	—
Heart disease [$n$ (%)]	—	—	—	—	$<0.001$
No	153,685 (95.2)	84,193 (95.2)	25,285 (95.9)	165,636 (96.4)	—
Yes	7,674 (4.8)	4,236 (4.8)	1,076 (4.1)	6,211 (3.6)	—
Stroke [$n$ (%)]	—	—	—	—	$<0.001$
No	158,384 (98.2)	86,928 (98.3)	25,926 (98.3)	169,426 (98.6)	—
Yes	2,975 (1.8)	1,501 (1.7)	435 (1.7)	2,421 (1.4)	—
Diabetes [$n$ (%)]	—	—	—	—	$<0.001$
No	152,859 (94.7)	83,675 (94.6)	25,152 (95.4)	162,956 (94.8)	—
Yes	8,500 (5.3)	4,754 (5.4)	1,209 (4.6)	8,891 (5.2)	—
Hypertension [$n$ (%)]	—	—	—	—	$<0.001$
No	101,553 (62.9)	52,193 (59.0)	15,669 (59.4)	105,928 (61.6)	—
Yes	59,806 (37.1)	36,236 (41.0)	10,692 (40.6)	65,919 (38.4)	—
Comorbidities categories [$n$ (%)]	—	—	—	—	$<0.001$
0	93,900 (58.2)	48,301 (54.6)	14,680 (55.7)	99,118 (57.7)	—
1	59,060 (36.6)	35,329 (40.0)	10,410 (39.5)	64,709 (37.7)	—
2 or more	8,399 (5.2)	4,799 (5.4)	1,271 (4.8)	8,020 (4.7)	—
BMI [$\text{mean}\pm \text{SD}$ (${\mathrm{kg}/\mathrm{m}}^2$)]	$27.59\pm 4.82$	$27.71\pm 4.78$	$27.58\pm 4.80$	$27.16\pm 4.78$	—
BMI category [$n$ (%)]	—	—	—	—	$<0.001$
Normal weight	50,860 (31.5)	26,575 (30.1)	8,341 (31.6)	61,103 (35.6)	—
Overweight	69,521 (43.1)	38,700 (43.8)	11,302 (42.9)	71,572 (41.6)	—
Obesity	40,978 (25.4)	23,154 (26.2)	6,718 (25.5)	39,172 (22.8)	—
CRP [$\text{mean}\pm \text{SD}$ (mg/L)]	$2.64\pm 4.36$	$2.64\pm 4.36$	$2.60\pm 4.36$	$2.47\pm 4.19$	—
Use of sun/UV protection [$n$ (%)]	—	—	—	—	$<0.001$
Never/rarely	15,462 (9.6)	8,587 (9.7)	2,445 (9.3)	19,894 (11.6)	—
Sometimes	52,071 (32.3)	29,484 (33.3)	8,402 (31.9)	59,059 (34.4)	—
Most of the time	55,223 (34.2)	30,992 (35.0)	9,499 (36.0)	60,103 (35.0)	—
Always	33,929 (21.0)	18,690 (21.1)	5,833 (22.1)	30,348 (17.7)	—
Other	4,674 (2.9)	676 (0.8)	182 (0.7)	2,443 (1.4)	—
GERD [$n$ (%)]	—	—	—	—	$<0.001$
No	146,155 (90.6)	79,340 (89.7)	24,371 (92.5)	157,329 (91.6)	—
Yes	15,204 (9.4)	9,089 (10.3)	1,990 (7.5)	14,518 (8.4)	—
Ever had PSA test [$n$ (%)]	—	—	—	—	$<0.001$
No	142,441 (88.3)	78,035 (88.2)	22,613 (85.8)	148,430 (86.4)	—
Yes	18,918 (25.1)	10,394 (24.8)	3,748 (31.0)	23,417 (29.7)	—
Ever use HRT [$n$ (%)]	—	—	—	—	$<0.001$
No	127,904 (79.3)	70,439 (79.7)	20,936 (79.4)	138,941 (80.9)	—
Yes	33,455 (38.9)	17,990 (38.7)	5,425 (38.0)	32,906 (35.3)	—
Had menopause [$n$ (%)]	—	—	—	—	$<0.001$
No	109,349 (67.8)	60,252 (68.1)	18,127 (68.8)	117,294 (68.3)	—
Yes	52,010 (60.4)	28,177 (60.6)	8,234 (57.7)	54,553 (58.6)	—
Parity [$n$ (%)]	—	—	—	—	$<0.001$
0	92,069 (57.1)	50,956 (57.6)	14,590 (55.3)	96,415 (56.1)	—
$\ge 1$	59,055 (68.6)	32,209 (69.3)	9,904 (69.4)	61,676 (66.2)	—
Ever taken oral contraceptive pill [$n$ (%)]	—	—	—	—	$<0.001$
No	92,069 (57.1)	50,956 (57.6)	14,590 (55.3)	96,415 (56.1)	—
Yes	69,290 (80.5)	37,473 (80.6)	11,771 (82.5)	75,432 (81.0)	—
Normal menarche [$n$ (%)]	—	—	—	—	$<0.001$
No	109,027 (67.6)	60,352 (68.2)	17,669 (67.0)	114,432 (66.6)	—
Yes	52,332 (60.8)	28,077 (60.4)	8,692 (60.9)	57,415 (61.7)	—
Ever had hysterectomy [$n$ (%)]	—	—	—	—	$<0.001$
No	155,536 (96.4)	85,238 (96.4)	25,477 (96.6)	166,324 (96.8)	—
Yes	5,823 (6.8)	3,191 (6.9)	884 (6.2)	5,523 (5.9)	—

Note:—, no data; ANOVA, analysis of variance; BMI, body mass index; CRP, C-reactive protein; GERD, gastro-esophageal reflux disease; HRT, hormone replacement therapy; PSA, prostate-specific antigen; SD, standard deviation; UV, ultraviolet.

^a $p$-Value was calculated by comparing the baseline characteristics across the domestic water hardness groups, using $t$-test or one-way ANOVA or Mann–Whitney U–test for continuous variables and chi-square tests for categorical variables.

Figure 2.

The distribution of domestic water hardness among UK Biobank participants ($n=\mathrm{447,996}$).

Association of All-Cause and Cause-Specific Cancer with Domestic Water Hardness

A nonlinear (U-shaped) relationship between domestic water hardness and all-cause cancer was observed ($p$ for nonlinearity $<0.001$). Similarly, a U-shaped relationship between calcium concentrations in domestic water and all-cause cancer was observed ($p$ for nonlinearity $<0.001$). The relationship between magnesium concentration and all-cause cancer was not statistically significant. For cause-specific cancers, a U-shaped relationship was observed between ${\text{CaCO}}_3$ concentrations and cancers of the esophagus, colorectal tract, and breast. Likewise, a U-shaped relationship was observed between calcium concentrations in domestic water and cancers of the oropharyngeal organs, colorectal tract, and breast. Furthermore, we observed a U-shaped relationship between magnesium concentration and lung cancer. In addition, a positive linear relationship between ${\text{CaCO}}_3$ and calcium concentrations with bladder cancer was observed, while a nonlinear relationship between magnesium concentrations and bladder was observed. The dose-response relationship between concentrations of ${\text{CaCO}}_3$, calcium, and magnesium in domestic water with all-cause cancer, gastrointestinal cancers, and bladder cancer were shown in Figure 3. Relationships with other cause-specific cancers were shown in Figure S1.

Figure 3.

Restricted cubic spline models for the relationship between domestic water hardness and the risk of all-cause cancer, gastrointestinal cancers, and bladder cancer among UK Biobank participants ($n=\mathrm{447,996}$).

In multi-adjusted Cox regression models, in comparison with individuals exposed to soft water ($0–60\mathrm{mg}/\mathrm{L}$, USGS classification), the HRs and 95% CIs of all-cause cancer were 1.00 (95% CI: 0.98, 1.02) for those exposed to moderate hard water ($>60–120\mathrm{mg}/\mathrm{L}$), 0.88 (95% CI: 0.84, 0.91) for those exposed to hard water ($>120–180\mathrm{mg}/\mathrm{L}$), and 1.06 (95% CI: 1.04, 1.08) for those exposed to very hard water ($>180\mathrm{mg}/\mathrm{L}$), as shown in Figure 4.

Figure 4.

Association of domestic water hardness and all-cause and cause-specific cancers among UK Biobank participants ($n=\mathrm{447,996}$) according to the US Geological Survey classification. Note: The reference group comprised participants who were exposed to domestic water with a ${\text{CaCO}}_3$ concentration $0–60\mathrm{mg}/\mathrm{L}$. Cox regression models were adjusted for sex, age, race, education attainment, employment status, socioeconomic status, family history of cancer, alcohol consumption, smoking status, physical activity, diet pattern, CRP, hypertension, heart disease, stroke, and diabetes. Use of sun/UV protection was further included as covariate in malignant melanoma model, and gastro-esophageal reflux disease was included in the esophageal cancer model. For sex-specific cancers (breast cancer, cervical cancer, endometrial cancer, ovarian cancer, and prostate cancer), separate models were run for males and females. We additionally adjusted for PSA test for prostate cancer and adjusted for HRT, menopause, parity, oral contraceptive pill or minipill, age of menarche, and hysterectomy for female specific cancers (breast cancer, cervical cancer, endometrial cancer, ovarian cancer). Note: CRP, C-reactive protein; HRT, hormone replacement therapy; PSA, prostate-specific antigen; UV, ultraviolet.

Domestic water hardness was associated with 11 of 22 cause-specific cancers, including cancers of the esophagus, stomach, colorectal, lung, breast, prostate, and bladder, as well as non-Hodgkin lymphoma, multiple myeloma, malignant melanoma, and hematological malignancies. Participants exposed to domestic water with a ${\text{CaCO}}_3$ concentration of $120–180\mathrm{mg}/\mathrm{L}$ had a lower risk of esophageal cancer ($\text{HR}=$ 0.79; 95% CI: 0.62, 1.00), colorectal cancer ($\text{HR}=$ 0.87; 95% CI: 0.78, 0.97), lung cancer ($\text{HR}=$ 0.85; 95% CI: 0.75, 0.96), female breast cancer ($\text{HR}=$ 0.81; 95% CI: 0.72, 0.90), prostate cancer ($\text{HR}=$ 0.91; 95% CI: 0.83, 0.99), multiple myeloma ($\text{HR}=$ 0.72; 95% CI: 0.54, 0.97), and hematological malignancies cancer ($\text{HR}=$ 0.81; 95% CI: 0.72, 0.90) in comparison with those exposed to domestic water with ${\text{CaCO}}_3$ concentration $<60\mathrm{mg}/\mathrm{L}$.

Association of All-Cause and Cause-Specific Cancer with Domestic Water Hardness Stratified by Potential Risk Factors

We investigated whether the associations of all-cause cancer with domestic hard water varied according to participant characteristics. Descriptive characteristics of the risk factors and all-cause cancer were provided in the Table 2. The association between domestic water hardness and the risk of incident all-cause cancer was not modified by sex, age, race, education level, employment status, socioeconomic status, family history of cancer, alcohol consumption, physical activity, diet pattern, hypertension, heart disease, stroke, and diabetes (all $p$ for interaction $>0.05$).

Table 2.

Association of water hardness with all-cause cancer among UK Biobank participants ($n=\mathrm{447,996}$) stratified by participant characteristics.

	Water hardness				$p$ For interactiona
	$0–60\mathrm{mg}/\mathrm{L}$	$>60–120\mathrm{mg}/\mathrm{L}$ [HR (95% CI)]	$>120–180\mathrm{mg}/\mathrm{L}$ [HR (95% CI)]	$>180\mathrm{mg}/\mathrm{L}$ [HR (95% CI)]
All participants	1.00 (Ref)	1.00 (0.98, 1.02)	0.88 (0.84, 0.91)	1.06 (1.04, 1.08)	—
Sex	—	—	—	—	0.129
Female	1.00 (Ref)	0.99 (0.96, 1.03)	0.85 (0.81, 0.90)	1.06 (1.03, 1.09)	—
Male	1.00 (Ref)	1.01 (0.98, 1.04)	0.90 (0.85, 0.95)	1.06 (1.03, 1.08)	—
Age (y)	—	—	—	—	0.620
$<50$	1.00 (Ref)	0.97 (0.90, 1.04)	0.85 (0.76, 0.95)	1.04 (0.98, 1.10)	—
50–60	1.00 (Ref)	0.99 (0.95, 1.03)	0.85 (0.79, 0.91)	1.07 (1.03, 1.11)	—
$>60$	1.00 (Ref)	1.01 (0.98, 1.04)	0.90 (0.85, 0.94)	1.05 (1.03, 1.08)	—
Race	—	—	—	—	0.958
White	1.00 (Ref)	1.00 (0.98, 1.03)	0.88 (0.84, 0.91)	1.06 (1.04, 1.08)	—
Mixed	1.00 (Ref)	1.28 (0.85, 1.93)	0.66 (0.33, 1.30)	1.05 (0.76, 1.44)	—
Asian or Asian British	1.00 (Ref)	0.79 (0.60, 1.05)	0.94 (0.62, 1.43)	0.94 (0.77, 1.15)	—
Black or Black British	1.00 (Ref)	1.06 (0.78, 1.45)	1.02 (0.65, 1.60)	1.01 (0.82, 1.23)	—
Chinese	1.00 (Ref)	0.95 (0.49, 1.86)	0.81 (0.28, 2.33)	0.95 (0.60, 1.52)	—
Others	1.00 (Ref)	0.79 (0.55, 1.12)	0.60 (0.34, 1.08)	0.87 (0.69, 1.10)	—
Education level	—	—	—	—	0.165
College or University	1.00 (Ref)	0.99 (0.94, 1.03)	0.90 (0.84, 0.96)	1.07 (1.04, 1.11)	—
Other	1.00 (Ref)	1.00 (0.98, 1.03)	0.87 (0.83, 0.91)	1.05 (1.03, 1.07)	—
Socioeconomic status	—	—	—	—	0.339
High	1.00 (Ref)	1.02 (0.97, 1.07)	0.83 (0.77, 0.90)	1.07 (1.03, 1.12)	—
Middle	1.00 (Ref)	1.00 (0.97, 1.03)	0.88 (0.83, 0.92)	1.06 (1.03, 1.08)	—
Low	1.00 (Ref)	0.97 (0.92, 1.03)	0.95 (0.87, 1.05)	1.04 (1.00, 1.08)	—
Current employment	—	—	—	—	0.123
Employed	1.00 (Ref)	0.99 (0.95, 1.02)	0.89 (0.84, 0.94)	1.07 (1.04, 1.10)	—
Retired	1.00 (Ref)	1.01 (0.98, 1.05)	0.87 (0.83, 0.93)	1.04 (1.02, 1.08)	—
Other	1.00 (Ref)	1.00 (0.92, 1.08)	0.81 (0.71, 0.93)	1.02 (0.95, 1.09)	—
Family history of cancer	—	—	—	—	0.584
No	1.00 (Ref)	0.99 (0.96, 1.02)	0.88 (0.84, 0.92)	1.05 (1.02, 1.07)	—
Yes	1.00 (Ref)	1.01 (0.97, 1.05)	0.88 (0.82, 0.93)	1.07 (1.03, 1.10)	—
Smoking status	—	—	—	—	$<0.001$
Never smoker	1.00 (Ref)	1.00 (0.97, 1.04)	0.86 (0.81, 0.91)	1.07 (1.05, 1.10)	—
Former smoker	1.00 (Ref)	1.00 (0.96, 1.03)	0.92 (0.87, 0.98)	1.07 (1.03, 1.10)	—
Current smoker	1.00 (Ref)	1.01 (0.95, 1.07)	0.81 (0.73, 0.90)	0.98 (0.93, 1.03)	—
Alcohol consumption	—	—	—	—	0.721
Excessive	1.00 (Ref)	0.99 (0.96, 1.03)	0.87 (0.82, 0.92)	1.05 (1.02, 1.08)	—
Never or moderate	1.00 (Ref)	1.00 (0.97, 1.03)	0.88 (0.84, 0.93)	1.06 (1.03, 1.09)	—
Physical activity	—	—	—	—	0.397
Low	1.00 (Ref)	0.98 (0.93, 1.03)	0.85 (0.78, 0.92)	1.02 (0.98, 1.07)	—
Middle	1.00 (Ref)	1.00 (0.96, 1.03)	0.88 (0.83, 0.93)	1.05 (1.02, 1.08)	—
High	1.00 (Ref)	1.02 (0.98, 1.06)	0.89 (0.83, 0.96)	1.09 (1.05, 1.13)	—
Diet	—	—	—	—	0.757
Unhealthy	1.00 (Ref)	1.00 (0.97, 1.03)	0.87 (0.83, 0.92)	1.06 (1.03, 1.08)	—
Healthy	1.00 (Ref)	1.00 (0.96, 1.03)	0.89 (0.84, 0.94)	1.06 (1.03, 1.09)	—
BMI category	—	—	—	—	0.834
Normal weight	1.00 (Ref)	1.00 (0.96, 1.05)	0.89 (0.83, 0.96)	1.06 (1.02, 1.09)	—
Overweight	1.00 (Ref)	1.01 (0.97, 1.04)	0.88 (0.83, 0.93)	1.07 (1.04, 1.10)	—
Obesity	1.00 (Ref)	0.99 (0.95, 1.03)	0.86 (0.80, 0.93)	1.04 (1.00, 1.08)	—
Comorbidities	—	—	—	—	0.779
0	1.00 (Ref)	1.01 (0.98, 1.05)	0.90 (0.85, 0.95)	1.05 (1.02, 1.08)	—
1	1.00 (Ref)	0.98 (0.94, 1.01)	0.83 (0.78, 0.88)	1.05 (1.02, 1.08)	—
2 or more	1.00 (Ref)	1.03 (0.95, 1.12)	1.00 (0.87, 1.15)	1.10 (1.02, 1.18)	—

Note: Cox regression models were adjusted for sex, age, race, education attainment, employment status, socioeconomic status, family history of cancer, alcohol consumption, smoking status, physical activity, diet pattern, CRP, hypertension, heart disease, stroke, and diabetes. We applied multiple imputations to the missing covariates in Cox models using the R package “mice.”⁴² In brief, predictive mean matching was used for continuous variables (socioeconomic level, alcohol consumption, physical activities, diet score, BMI, and CRP), whereas logistic regression was used for categorical variables (race, education level, current employment status, and smoking status). The process was repeated five times, and five imputed datasets were obtained. These five imputed datasets were entered into the Cox model individually, and the results were then pooled to one estimate. —, no data; BMI, body mass index; CI, confidence interval; CRP, C-reactive protein; HR, hazard ratio; Ref, reference.

^a p For interaction corresponded to the chi-square test statistic for the likelihood ratio test comparing models with and without interaction between water hardness and potential risk factors.

Additional Analyses

The associations between domestic water hardness and all-cause cancer were similar to the primary results when we excluded participants with missing values on covariates (Table S3), performed competing risk regression model (Table S4), restricted analyses to individuals living at the same residential address for $\ge 10$ y (67.5% of participants) (Table S5), and further adjusted for natural environment percentage (Table S6).

We further classified domestic water hardness into two groups following the WHO classification of ${\text{CaCO}}_3$ concentration: soft water ($<200\mathrm{mg}/\mathrm{L}$) and hard water ($\ge 200\mathrm{mg}/\mathrm{L}$). In multi-adjusted Cox regression models, in comparison with participants exposed to domestic water with a ${\text{CaCO}}_3$ concentration $<200\mathrm{mg}/\mathrm{L}$, those exposed to domestic water with a concentration of $\ge 200\mathrm{mg}/\mathrm{L}$ had a higher risk of incident all-cause cancer ($\text{HR}=$1.07, 95% CI: 1.05, 1.09), as shown in Figure S2. For cause-specific cancer, participants exposed to domestic water with a ${\text{CaCO}}_3$ concentration $\ge 200\mathrm{mg}/\mathrm{L}$ had a higher risk of malignant melanoma ($\text{HR}=$1.12; 95% CI: 1.04, 1.21), female breast cancer ($\text{HR}=$1.13; 95% CI: 1.08, 1.19), prostate cancer ($\text{HR}=$1.18; 95% CI: 1.14, 1.23), bladder cancer ($\text{HR}=$1.11; 95% CI: 1.03, 1.20), non-Hodgkin lymphoma ($\text{HR}=$ 1.17; 95% CI: 1.07, 1.28), multiple myeloma ($\text{HR}=$1.14; 95% CI: 1.01, 1.28), and hematological malignancies ($\text{HR}=$1.09; 95% CI: 1.03, 1.15) but with lower risk of lung cancer ($\text{HR}=$0.89; 95% CI: 0.84, 0.94) in comparison with those exposed to domestic water with ${\text{CaCO}}_3$ concentration $<200\mathrm{mg}/\mathrm{L}$.

Discussion

In this population-based cohort study of 447,996 UK Biobank participants, a nonlinear association between domestic water hardness and all-cause cancer was observed. Participants exposed to hard water ($>120–180\mathrm{mg}/\mathrm{L}$) had a lower risk of esophagus cancer, stomach cancer, colorectal cancer, lung cancer, female breast cancer, prostate cancer, multiple myeloma, and hematological malignancy in comparison with those exposed to soft water ($0–60\mathrm{mg}/\mathrm{L}$). Participants exposed to very hard water ($>180\mathrm{mg}/\mathrm{L}$) had a higher risk of malignant melanoma, bladder cancer, non-Hodgkin lymphoma, female breast cancer, prostate cancer, and hematological malignancies. Moreover, we observed a positive linear relationship between water hardness and bladder cancer.

Domestic water with high mineral content (hard water) is associated with health outcomes, such as chronic degenerative diseases,⁴⁴ cardiovascular diseases,⁴⁵ mortality due to acute myocardial infarction,⁴⁶ rectal cancer,¹⁵ colon cancer,¹⁴ esophageal cancer,²⁰ breast cancer,²² and pancreatic cancer.⁴⁷ However, the investigations do not prove causality, and hence the evidence is being debated. Further, the relationship between domestic water hardness and cancer remains fragmented and demands further investigation. Our study attempted to fill this gap in literature. We found that there was a U-shaped relationship between water hardness and all-cause cancer, suggesting a potentially beneficial association between moderate water hardness and the incidence of all-cause cancer.

Participants exposed to hard water were associated with a lower risk of esophageal cancer, stomach cancer, and colorectal cancer compared with those exposed to soft water. A matched case–control study conducted in Taiwan on the relationship between hard water and gastrointestinal cancer reported that hard water was associated with a lower risk of esophageal cancer²⁰ in comparison with soft water. Another matched case–control study in Taiwan showed a significant negative relationship between drinking water hardness and stomach cancer mortality.¹³ Two other matched case–control studies from Taiwan showed a significant inverse relationship between drinking water hardness and colon cancer and rectal cancer mortality.^14,15 The results of our present study are consistent with those reported in the above-mentioned studies. The potential benefit effects of water hardness and cancers of esophagus, stomach, and colorectal tract may be explained by the mechanisms underlying the biological effects of calcium ions in domestic water. Several mechanisms have been proposed to explain the roles of calcium in reducing colorectal cancer risk, including binding of long-chain fatty acids and bile acid in the small intestine, thereby protecting colonic epithelial cells from mutagens, as well as effects on cell proliferation and differentiation, apoptosis, angiogenesis, and cell-cycle regulation.^48–54 Our findings highlight important associations between domestic water hardness and human health, warranting further investigations. Further, animal studies are needed to examine the association between water hardness and gastrointestinal cancer.

Multiple epidemiological studies have examined the association between drinking water and bladder cancer risk. A matched case–control study conducted in Taiwan reported a positive correlation between the concentration of trihalomethanes in drinking water and risk of death from bladder cancer.⁵⁵ Evlampidou et al. found that trihalomethane levels in drinking water in the European Union contribute to the burden of bladder cancer.⁵⁶ Two case–control studies, one conducted by Beane et al.⁵⁷ and one a US study,⁵⁸ demonstrated a positive association between bladder cancer and disinfection by-products in drinking water. Moreover, a case–control study showed that the risk of bladder cancer increased with the intake level of beverages made with tap water.⁵⁹ In addition, a population-based case–control study supported an association between low-to-moderate levels of arsenic in drinking water and bladder cancer risk in New England.⁶⁰ Our observations also suggested a positive relationship between water hardness and bladder cancer. Further experimental studies are needed to explore our findings.

Lung cancer is the leading cause of cancer deaths globally. Cheng et al. in a cross-sectional study found that women who had a high intake of magnesium through drinking water had a lower risk of lung cancer risk, but similar significant associations were not seen with calcium levels in drinking water.²¹ To our knowledge, limited prospective studies have been conducted to assess the hardness of domestic water on lung cancer risk. Our findings demonstrated that higher concentrations of calcium and magnesium in domestic water were associated with lower risk of developing lung cancer, which is slightly inconsistent with the above analysis that calcium levels in drinking water was not associated with lung cancer.²¹ One potential explanation may be that in our study, calcium concentrations in domestic water exceeding $80\mathrm{mg}/\mathrm{L}$ (29.5% of the population) were linked to lower risk of lung cancer. But for the aforementioned analysis, the calcium levels in drinking water ranged from 0 to $81\mathrm{mg}/\mathrm{L}$, thus limiting the exploration of potential association between higher calcium concentrations ($>81\mathrm{mg}/\mathrm{L}$) and the incidence of lung cancer.

Our findings suggested that participants exposed to hard water had a lower risk of breast cancer, whereas those exposed to very hard water had a higher risk of breast cancer (in comparison with soft water). Previous studies have investigated the associations between calcium and magnesium levels and the breast cancer risk. Yang et al.²² showed an inverse association between calcium and magnesium levels in drinking water and breast cancer mortality. Wu et al.⁶¹ recruited 1,050 cases and 1,229 controls in a case–control study and showed that dietary magnesium negatively affected breast cancer risk directly and indirectly through regulating CRP levels. In our study, female breast cancer was significantly associated in a U-shaped association with calcium but not with magnesium. These results are contradictory to those of previous studies,^22,61 which might be due to the varying sample size, prospective study design, confounding factors, and the methods used to measure the water hardness. Experimental data in animals suggested that an adequate intake of calcium has antiproliferative and pro-differentiation actions on mammary gland cells of rats,^62–64 whose potential mechanism involves a combination of calcium and vitamin D working together to regulate breast cell growth and differentiation,⁶⁵ potentially lowering breast cancer risk at moderate levels of intake.

Furthermore, our results suggested that participants exposed to hard water had a lower risk of prostate cancer, whereas those exposed to very hard water had a higher risk of prostate cancer (in comparison with soft water exposure). This is the first prospective study to explore the associations between domestic water hardness and prostate cancer. One matched case–control study in Taiwan including 1,364 participants reported that a high intake of magnesium levels through drinking water might have a protective effect against mortality due to prostate cancer, whereas calcium levels in drinking water did not appear to provide a similar protective effect.⁶⁶ In contrast, our study showed a nonlinear relationship between the calcium levels in domestic water and prostate cancer. One possible explanation may be that calcium is important for human health, with moderate calcium levels in domestic water helping to prevent prostate cancer development. Further large-scale prospective and experimental studies are needed to verify these findings.

The mechanisms underlying the association between domestic water hardness and the risk of developing cancer are multifactorial and incompletely understood. Hard water has no known adverse health effect, and the potential health effects of water hardness primarily stem from the presence of dissolved salts, particularly calcium and magnesium.⁶⁷ Our findings showed a U-shaped relationship between the concentrations of ${\text{CaCO}}_3$ in domestic water and the incidence of all-cause cancer. Specifically, high levels of ${\text{CaCO}}_3$ (until it reaches an optimal level) has a protective effect against cancer. After the optimal concentration is achieved, the protective effect against cancer gradually decreases with further increases in ${\text{CaCO}}_3$ concentration. We observed a U-shaped relationship between calcium concentrations in domestic water and all-cause cancer, and no significant associations were observed with magnesium concentrations. Hence, the potential effects of water hardness and all-cause cancer risk might be explained by the mechanisms underlying the biological effects of calcium ions. Calcium plays a crucial role in regulating the cellular processes of proliferation and apoptosis in both malignant and immune cells.²⁴ Cellular machinery important in tumor progression is often driven or influenced by changes in calcium ions. Hard water could provide an important supplementary intake to total calcium,⁶⁷ and moderate calcium levels in domestic water might help to prevent cancer development. In addition, apart from calcium, domestic water contains various other dissolved metals, such as aluminum, barium, strontium, iron, zinc, and manganese.¹⁶ Given an increase in the salts and the resultant hardness of water, the ion concentrations for these metals might also increase. The genotoxic properties of aluminum are well documented, including its ability to induce DNA alterations and epigenetic changes,⁶⁸ which may counteract the protective effect of calcium against cancer. Animal studies are required to understand the underlying mechanisms. In addition, measurement inaccuracies may be another potential explanation for the nonlinear relationship between water hardness and all-cause cancer. Further large-scale prospective and experimental studies will be needed to verify our findings.

The current study presents several strengths, including a large-scale sample size and a long follow-up period, as well as the use of standardized protocols for data collection and multiple resources for disease diagnosis. However, the current study also has several potential limitations. First, one potential limitation of our study is the possibility of changes in exposure levels over time due to factors such as changes in water source and treatment methods after the baseline examination, which could have influenced our risk estimates. Yet, using data from a single year is insufficient to estimate long-term exposure to water hardness and to evaluate its potential impact on cancer risk. In addition, data on domestic water hardness were measured before the baseline. If participants move to a different address, the exposure levels of domestic water hardness could be changed at the time of the baseline examination, potentially influencing our risk estimates. Moreover, postcodes were linked to corresponding water-supply companies in UK Biobank. Individuals may obtain their drinking water from various sources (e.g., home, work, or bottled water), which may not match the water hardness levels associated with their residential postcode. Such discrepancies could lead to a misclassification of an individual’s actual exposure to water hardness. Furthermore, using postcode-based water hardness data does not account for individual variations in water consumption habits and treatment methods (e.g., water softeners in the home). Therefore, our data may not accurately reflect individual exposures, potentially leading to biased or inaccurate conclusions. Second, participants’ addresses at baseline were used to estimate the water hardness levels, and potential residential mobility may cause the exposure misclassification. However, we restricted our analyses to individuals living at the same residential address for $\ge 10$ y and found that the results were similar to our primary results. Third, there might be participants who rely on private wells, which may cause the exposure misclassification. Fourth, as with many observational studies, there is a possibility of bias caused by unmeasured confounding factors, despite conducting multiple sensitivity analyses. Fifth, data on other water contaminants (e.g., nitrate, metals, disinfection by-products) are not available in the UK Biobank, limiting our ability to further explore whether these contaminants modify the association between hard water and the risk of cancer. Sixth, additional information regarding accurate measurements for water hardness taken at specific sampling points was unavailable to us. Furthermore, regional variations in water hardness, influenced by geological factors and local water treatment practices, can affect the applicability of our study’s findings to the relationship between water hardness and all-cause cancers across different areas. Seventh, the identification of cancer cases relied on medical records and cause of death during follow-up, which may have underestimated cancer incidence among individuals without recorded symptoms. In addition, the participants who volunteered in the UK Biobank cohort might be more health-conscious than those who did not participate, potentially leading to an underestimation of cancer incidence.⁶⁹ Because of the observational nature of our study, the findings do not imply causality. Finally, the study primarily included participants of White British descent, which limits the generalizability of the findings to other races.

Despite advances in understanding the pathophysiology of cancer, clinical treatment of most cancers continues to be suboptimal. Therefore, identifying the preventable risk factors for cancer is of high priority. Our findings support a potentially beneficial association between hard water and the incidence of all-cause cancer. From a public health perspective, because water is essential for life, even small potential health benefits associated with drinking water may have important public health implications. Further clinical trials will be necessary to assess whether the observed associations are causal.

In summary, our findings demonstrate that the association between domestic water hardness and all-cause cancer is nonlinear. In addition, domestic water hardness was associated with 11 out of 22 cause-specific cancers, including cancers of the esophagus, stomach, colorectal tract, lung, breast, prostate, and bladder, as well as non-Hodgkin lymphoma, multiple myeloma, malignant melanoma, and hematological malignancies. Furthermore, we observed a nonlinear relationship between calcium concentrations in domestic water and all-cause cancer. Our findings support a potentially beneficial association between moderate domestic water hardness ($>120–180\mathrm{mg}/\mathrm{L}$) and the incidence of cancer.

Conclusions and opinions are those of the individual authors and do not necessarily reflect the policies or views of EHP Publishing or the National Institute of Environmental Health Sciences.