Chlorination By-Products in Drinking Water and Uterine Cancer Risk – A Study of 53 Thousand Women
Some news stories lend themselves all too readily to panic. "Drinking water causes cancer" is one of the clearest examples — because everyone drinks water, it is not something people choose, and there seems to be no easy way to avoid it. That is why a study like this needs to be reported without overstating or understating it.
A new study reported an association between long-term exposure to disinfection by-products formed when drinking water is chlorinated and uterine cancer risk . The most striking detail is that the association persisted at levels far below legal limits . In this article, we examine the finding, how robust the figures actually are, and what a woman should — and should not — do in response .
- It says: In the third of participants with the highest trihalomethane exposure, compared with the lowest-exposure group, uterine cancer risk was %18 higher, and the risk of estrogen-sensitive endometrioid tumors was %23 higher.
- It says: This association also persisted in analyses restricted to systems whose average levels remained below the legal upper limits .
- It does not say: "Tap water causes uterine cancer." This is an observational study; causation has not been demonstrated , and the authors explicitly acknowledge this.
- It does not say: "You should avoid chlorinated tap water." The study author states directly that the findings, given the vital role of water disinfection in preventing infectious diseases, do not justify recommending that women avoid chlorinated tap water altogether.
- This is an increase in relative risk. The absolute risk of uterine cancer remains low; an increase of %18–23 means a small difference at the individual level — but because exposure is so widespread, it may matter at the population level .
- No woman should change her water-drinking habits on the basis of this article or start anxiously searching for a filter or bottled water.
Uterine cancer is the most common gynecological cancer in the United States , and endometrial cancer (cancer of the uterine lining) accounts for more than %95 of uterine cancers. In recent years, both incidence and mortality have been increasing — whereas the overall trend in oncology has been the opposite.
The puzzle is this: Known risk factors — particularly obesity, reproductive history and hormone exposure — do not fully explain this increase. The frequency of the disease also varies markedly by geography. When the continuing increase and geographical variation are considered together, the first possibility that comes to mind is that environmental exposures may also play a role. Drinking water is a natural place to look: approximately %90 of the United States population obtains drinking water from public water systems — so exposure is almost universal.
Chlorine and similar disinfectants are essential for preventing waterborne infectious diseases. However, these substances react with naturally occurring organic matter in water (leaves, humic substances, algae), forming unwanted compounds. These are called disinfection by-products .
The two most closely monitored groups are trihalomethanes (THM) and haloacetic acids (HAA). It has been suggested that some of these compounds may interact with estrogen receptors or disrupt hormone signaling, while others may cause oxidative stress, cellular damage and DNA damage . The most common member of the THM group, chloroform, is classified by the International Agency for Research on Cancer (IARC) as Group 2B — possibly carcinogenic to humans . The association of this group of compounds with cancer has so far been studied most extensively in bladder cancer ; uterine cancer was not even on the agenda in this literature until recently.
- Prospective observational cohort: California Teachers Study — a cohort of teachers and administrators working in public schools in California.
- 53.100 women. Median age at enrollment in 1995–1996 was 50; all were cancer-free, had not undergone a hysterectomy, had lived at the same address for at least 10 years and had an address that could be linked to a public water system.
- An average of 19,4 years of follow-up (through December 31, 2020). During this period, 1038 cases of uterine cancer were diagnosed; 154 were of the non-endometrioid subtype.
- How was exposure estimated? Participants' home addresses were matched to monitoring data from the state's public water systems. 1990–2005 was used to calculate the total and individual THMs, as well as nitrate, arsenic and uranium concentrations, averaged over 15 years.
- A separate approach was used for HAA: Because historical HAA data were more limited, exposure was assessed using averages up to 2013 , and cancer follow-up in these analyses began in 2010. This is a key detail explaining why the HAA findings are more fragile.
- Mixture analysis: The effects of the compounds were evaluated not only individually but also together — because in real life people are not exposed to just one chemical.
"High exposure" in this study does not mean a contaminated water supply, as one might imagine. In the cohort, the median total THM concentration was only 4,56 µg/L . The highest third started at 15,58 µg/L , and the maximum value was 108,66 µg/L.
For comparison: the upper limit for total THM in the United States is 80 µg/L, , and in Türkiye and the European Union it is 100 µg/L. In other words, the threshold for the "high-exposure" group is less than even one fifth of the legal limit. This makes the finding interesting from two perspectives: on the one hand, it raises the question of whether current limits are sufficiently protective; on the other, it also raises the question of how plausible it is to expect a biological effect at such low levels. The honest answer is that both remain open questions.
HR 1,18 (95% CI 1,02–1,38) · trend p=0,06
HR 1,23 (95% CI 1,05–1,46) · trend p=0,01
Legal limit: United States 80 · Türkiye/EU 100 µg/L
| Exposure | Outcome | Hazard ratio (95% CI) | Trend p | Interpretation |
|---|---|---|---|---|
| Total THM | All uterine cancers | 1,18 (1,02–1,38) | 0,06 | The comparison between exposure groups is nominally significant, but the dose–response trend is borderline |
| Total THM | Endometrioid | 1,23 (1,05–1,46) | 0,01 | The study's most consistent finding |
| Chloroform | All uterine cancers | 1,18 (1,02–1,38) | 0,07 | Among individual THMs, the clearest signal |
| Chloroform | Endometrioid | 1,27 (1,07–1,50) | 0,01 | Stronger in the hormone-sensitive subtype |
| HAA5 | Non-endometrioid | 1,86 (1,06–3,25) | 0,01 | Very wide confidence interval; only 154 cases, limited exposure data — the most fragile finding |
| Mixture (IQR increase) | All uterine cancers | 1,21 (0,96–1,53) | — | not significant |
| Mixture (IQR increase) | Endometrioid | 1,34 (1,04–1,74) | — | Main contributors: chloroform and dibromochloromethane |
| Nitrate · arsenic · uranium | All outcomes | no association identified | — | An important negative finding — the study can distinguish between them |
In the cohort, 1038 cases of uterine cancer occurred among 53.100 women over approximately 19 years — roughly 20 cases per 1000 women. Applying an 18% relative increase to that baseline gives approximately 23–24 cases per 1000 women . In other words, over roughly two decades, the difference is on the order of 3–4 additional cases per 1000 women .
This is a rough illustration — the study does not report absolute rates specific to each exposure group, and the actual difference may be smaller or larger. But it is sufficient to put the scale of the difference in perspective. This is exactly how the study author frames it: an increase on this scale may not be meaningful for an individual woman; however, because exposure to disinfection by-products encompasses almost everyone , it may be important at the population level . The distinction between relative and absolute risk may be the most fundamental tool for reading cancer news accurately.
1 · The most robust: The association with endometrioid tumors. For both total THM (1,23) and chloroform (1,27), the confidence interval does not include 1, and the dose–response trend is also significant (p=0,01). The mixture analysis is also significant for the same subtype (1,34). Multiple analyses pointing in the same direction are more convincing than a single analysis.
2 · The borderline finding: The result for all uterine cancers. The comparison between exposure groups appears nominally significant (1,02–1,38, which does not include 1), but the dose–response trend test gives p=0,06 — above the conventional threshold. Overlooking this distinction is a common mistake: "There is a difference between two groups" and "risk increases steadily as exposure increases" are not the same thing. The latter is stronger evidence for causation, and it has not been fully demonstrated here.
3 · The most fragile: The association between HAA5 and non-endometrioid tumors. The figure is striking (1,86 — an increase of 86%), but the confidence interval ranges from 1,06 to 3,25. That means "the actual effect could be 6% or 225%." Moreover, this analysis is based on only 154 non-endometrioid cases and more limited exposure data covering a shorter period . It would be wrong to make this finding the headline.
Endometrioid tumors account for the great majority of uterine cancers; they are estrogen-sensitive and generally have a more favorable course. They classically develop in the setting of unopposed estrogen exposure . Non-endometrioid tumors (serous, clear-cell, carcinosarcoma) are less common, develop largely independently of estrogen and are more aggressive.
This is where the finding's biological consistency lies: the THM signal is pronounced in the estrogen-sensitive subtype and absent in the estrogen-independent subtype. This is consistent with the hypothesis that "these compounds may affect hormone signaling." However, the same reasoning struggles to explain the HAA finding: why would HAA show the opposite pattern and be associated with the estrogen-independent subtype? This internal inconsistency means we need to remember that at least one finding could be due to chance. When a study examines numerous compounds and numerous subtypes, a few would be expected to produce "significant" findings by chance.
- The association persisted in sensitivity analyses restricted to levels below the legal limits. This indicates that the finding was not driven by a few heavily contaminated water systems.
- The results did not vary significantly according to menopausal status, smoking, hormone therapy use or most body mass index categories. In other words, the finding was not driven by a single subgroup.
- The negative controls worked: for nitrate, arsenic and uranium, examined using the same method, no association was found. If the method were systematically biased, everything would be expected to appear risky.
- A biologically plausible difference between subtypes (a stronger signal in estrogen-sensitive tumors).
- And most importantly: this is the second study. See below.
- The finding of a single observational study is always met with skepticism — and should be. Chance, hidden confounders or a characteristic specific to that cohort may have produced the result.
- A result in the same direction in a different state, a different period and a different group of women makes those possibilities less likely. In the editorial's words, the association has now become "substantially harder to dismiss" .
- However, there is a large gap between "hard to dismiss" and "proven." This is still only the second epidemiological study, and the authors explicitly identify this as a limitation.
- The problem of a shared weakness: both studies estimated exposure not through individual measurements but through the average values in the water system serving the home address . The error inherent in this approach could operate in the same direction in both studies. Truly independent confirmation would come from a different method of measuring exposure.
This may be the most instructive section of the article. Three experts looking at the same data reach three different conclusions — and each is consistent within their own framework. This is a good example of what scientific uncertainty looks like.
She describes this as a very important study , because millions of people in the United States and billions worldwide are exposed to these compounds in drinking water every day.
As a result, she argues that we should try to minimize disinfection by-products .
She says the findings provide "some evidence" that current standards may not be sufficiently protective .
However, she emphasizes that this study alone cannot establish a new regulatory threshold : that requires replication, mechanistic support and assessment of geographical variation.
She states that regulatory decisions should be based on the entire body of evidence from numerous studies, different exposure settings and populations .
She says the findings should not change the advice doctors give their patients : the absolute risk of endometrial cancer remains low , and the evidence for established risk factors such as obesity, reproductive characteristics and hormone exposure is much stronger.
Richardson's question is "What should water utilities do?" — and, given that reducing by-products carries almost no cost, even a tentative signal is enough to justify action. Jones's question is "Should regulation change?" — which requires a much higher threshold of evidence. McCullough's question is "What should I tell a patient in the consulting room?" — and there the answer is clear: this finding should not take precedence over risk factors for which the evidence is much stronger. Whether a finding is "important" depends on whom you ask and what you ask them. The same data carry different weight for different decision-makers.
- Reducing organic matter in source water. The by-product results from chlorine reacting with organic matter; less raw material means less product.
- Improving treatment before disinfection — removing the organic load before chlorination.
- Optimizing disinfectant selection and dose; avoiding unnecessary excess without compromising microbiological safety.
- Reducing residence time in the water distribution network — by-products continue to accumulate the longer water remains in the pipes.
- And the common feature of all these measures: none is something an individual can do at home. Water quality is a collective issue; individual anxiety and individual purchases are not the solution to a population-level problem.
Regulation: In Türkiye, the quality of water for drinking and domestic use is regulated by the Regulation on Water Intended for Human Consumption . According to the limit values in its annex, the upper limit for total trihalomethanes is 100 µg/L — the same as in the European Union. The same regulation sets limits of 10 µg/L for arsenic, 50 mg/L for nitrate, 10 µg/L for bromate . The federal limit in the United States for total THM is 80 µg/L, and for the sum of five haloacetic acids it is 60 µg/L.
How this relates to the study: the "highest-exposure" group in this study started at 15,58 µg/L . So the question is not about a breach of the limits — whether levels far below those limits could also have an effect . This challenges the regulatory approach itself, and there is no answer yet.
And let us be honest — we cannot answer this question for Türkiye. Trihalomethane levels vary markedly between water systems; the determining factors are the organic matter content of the source water, the chlorine dose used and the time the water spends in the network. The absence of a publicly accessible, systematic and up-to-date nationwide map of THM levels makes it difficult to assess how far this study's findings apply to conditions in Türkiye. The most meaningful step an interested citizen can take is to request the analysis results published by their local water authority or ask for them to be made public — a much more effective action than buying an individual filter.
"Tap water causes uterine cancer! Chlorinated water increases cancer risk in women — even below the legal limits. One by-product raises the risk by up to 86% . Experts warn: filter your water."
This is an observational cohort study; causation has not been demonstrated. The increase is relative and small (18–23%); the absolute risk remains low. For all uterine cancers, the dose–response trend is borderline (p=0,06). The figure most likely to make a headline, 86%, is the study's most fragile finding — it is based on 154 cases, with a confidence interval between 1,06 and 3,25. This is only the second epidemiological study. And most importantly: the study author explicitly states that recommending women avoid chlorinated tap water is not justified.
- It is an observational cohort study; causation cannot be established, and residual confounding cannot be excluded.
- Exposure was not measured individually. The average values from the network serving the women's home addresses were used; individual water intake, filter use, preference for bottled water and consumption at work are unknown. This inevitably means exposure misclassification.
- Temporal uncertainty: as the authors note, the 15-year average exposure window overlaps with the first 10 years of follow-up.
- The HAA analyses are markedly weaker: more limited historical data, follow-up starting in 2010, and only 154 non-endometrioid cases. The confidence intervals are consequently very wide.
- The dose–response trend test for all uterine cancers gives p=0,06 — above the conventional significance threshold.
- Numerous compounds and numerous tumor subtypes were examined; with multiple comparisons, some findings would be expected to occur by chance. This possibility is raised by the biological inconsistency between the THM–endometrioid and HAA–non-endometrioid findings.
- The cohort predominantly consists of white, educated women working in education in California; generalizability is limited, and the exposure range is low.
- Chemicals do not occur in isolation. Drinking-water contaminants occur alongside other environmental and social exposures; separating the effect of a single compound is difficult.
- This is only the second epidemiological study on the subject; replication is needed.
- This article contains no recommendations and does not recommend changing water-consumption habits.
1 · Do not stop drinking tap water. The protection provided by chlorinated water is concrete, measured and substantial; the risk discussed here is preliminary and small. Reversing these priorities would clearly be harmful. This is exactly what the study author herself says.
2 · If you still want to do something, there is a reasonable, low-risk step: a water filter documented/certified to reduce total trihalomethanes or volatile organic compounds . But know its limits — a filter does not address exposure while bathing. These compounds can be inhaled with vapor during a shower or absorbed through the skin. A filter therefore does not eliminate the problem entirely.
3 · Focus your efforts where the evidence is stronger. For endometrial cancer, the strongest modifiable risk factor remains overweight and obesity — and the evidence for this is many times stronger than the evidence concerning water. Weight management, regular physical activity and monitoring thickening of the uterine lining make a measurable difference.
4 · And this is the truly life-saving point: after menopause, any vaginal bleeding is abnormal and must be evaluated. Before menopause, a marked change in menstrual patterns, bleeding between periods or unusual discharge also warrants consulting a doctor. Uterine cancer is largely treatable when detected at an early stage — and this single sentence saves far more lives than the entire debate about water.
5 · Finally, a matter of perspective: It is a good thing that studies like these are conducted and published. Uterine cancer is increasing, and we do not fully understand why. Seriously investigating environmental exposures is the only way to solve that puzzle. But treating an early research signal as an established danger benefits neither those women nor science.
Endometrial cancer incidence and mortality are increasing, and this rise cannot be fully explained by established risk factors. Geographical variation has brought environmental exposures into focus. Drinking-water disinfection by-products are an area in which exposure is almost universal , so even small relative risks can create a large attributable burden . This study is the second prospective cohort analysis on the subject; like the 2022 Iowa findings, it produced results in the same direction .
California Teachers Study; n=53.100, mean follow-up 19,4 years, 1038 uterine cancers (154 non-endometrioid). Highest tertile of total THM: all uterine cancers HR 1,18 (95% CI 1,02–1,38; trend p=0,06); endometrioid HR 1,23 (1,05–1,46; p=0,01). Chloroform: 1,18 (1,02–1,38), and for endometrioid tumors 1,27 (1,07–1,50; p=0,01). HAA5 → non-endometrioid 1,86 (1,06–3,25). Mixture analysis (IQR increase): endometrioid 1,34 (1,04–1,74); all uterine cancers 1,21 (0,96–1,53, not significant); main contributors were chloroform and dibromochloromethane. No association for nitrate, arsenic or uranium. The associations persisted in sensitivity analyses restricted to systems with averages below federal limits; they did not differ significantly by menopausal status, smoking, hormone therapy or most BMI categories. Median total THM in the cohort was 4,56 µg/L; the highest tertile started at 15,58 µg/L.
Observational design — causation cannot be demonstrated. Exposure is address-based; individual consumption, filter use and workplace exposure are unknown. The exposure window overlaps with the first 10 years of follow-up. HAA analyses rely on limited historical data and 154 non-endometrioid cases; confidence intervals are very wide. For all uterine cancers, trend p=0,06. Because numerous compounds and subtypes were examined, there is a multiple-comparison problem ; the association of THM with endometrioid tumors and HAA with non-endometrioid tumors is biologically inconsistent and suggests that at least one may be due to chance. The cohort is predominantly white and educated; the exposure range is low — generalizability is limited. Both existing cohorts share the same method of estimating exposure , which limits the evidentiary value of "replication." This is only the second epidemiological study.
Clinical counseling should not change. Risk assessment and prevention for endometrial cancer should continue to focus on factors with much stronger evidence — obesity, unopposed estrogen exposure, reproductive history and Lynch syndrome . Patients should not be advised to avoid tap water. For an anxious patient, a certified THM/VOC filter may be offered as a low-risk, reasonable option; however, it should be explained that it does not address exposure while bathing. The highest-value clinical message remains unchanged: prompt evaluation of postmenopausal bleeding, abnormal uterine bleeding and abnormal discharge. At the public health level , the meaningful target is treatment and distribution-network management that reduces by-product formation without compromising microbiological safety .
Will the finding be replicated in different populations and using different methods of measuring exposure ? Can a biological mechanism be demonstrated in humans — will the estrogen-signaling hypothesis be confirmed? Are current limits really sufficiently protective, or should the threshold be lowered? Which exposure window is decisive (early adulthood? the menopausal transition?)? How much does exposure through bathing/showering contribute , and how will it be measured? Is the HAA–non-endometrioid finding real or due to chance? And the most practical question for us: What are trihalomethane levels in Türkiye's water systems, and why are these data not systematically available to the public?
- Drinking water contaminant exposures and risk of uterine cancer. JAMA Network Open. 2026;9(7):e2624391. doi:10.1001/jamanetworkopen.2026.24391 (California Teachers Study; n=53.100; prospective observational cohort.)
- Accompanying editorial. JAMA Network Open. 2026. (Joint assessment of the two cohorts.)
- Drinking water disinfection byproducts, ingested nitrate, and risk of endometrial cancer in postmenopausal women. Environ Health Perspect. 2022. doi:10.1289/EHP10207 (Iowa Women's Health Study; the first signal.)
- IARC Monographs. Chloroform — Group 2B (possibly carcinogenic to humans) assessment. International Agency for Research on Cancer.
- US EPA. Stage 1 and Stage 2 Disinfectants and Disinfection Byproducts Rules. (Upper limit for total trihalomethanes 80 µg/L; sum of five haloacetic acids 60 µg/L.)
- Regulation on Water Intended for Human Consumption, Annex-1 — chemical parameters and limit values. Official Gazette, February 17, 2005 (with amendments). (Total trihalomethanes 100 µg/L; arsenic 10 µg/L; nitrate 50 mg/L; bromate 10 µg/L.)
- Exposure to drinking water trihalomethanes and risk of cancer: a systematic review of the epidemiologic evidence and dose–response meta-analysis. 2025. (Bladder cancer is the most firmly established cancer association for disinfection by-products.)
- Brooks M. Drinking water contaminants tied to uterine cancer risk. Medscape Medical News, July 24, 2026. (Assessments by the researcher and independent experts.)
- Related reading (drozdogan.com): Uterine cancer: symptoms, stages and treatment · Obesity and uterine cancer · Thickening of the uterine lining — endometrial hyperplasia · What is a carcinogen? · The distinction between relative and absolute risk · Focusing on the real causes of cancer · Microplastics and health · The 8 most preventable types of cancer · Immunotherapy for uterine cancer.
Editor's note: This content is intended for general information only; it is not medical advice, a diagnosis or a treatment recommendation. The reported data were checked against the primary source and presented without direct quotation. This is an observational cohort study and does not demonstrate causation; disinfection by-products in drinking water have not been proven to cause uterine cancer. The study author describes the findings as "a convincing but still preliminary association" and does not consider it justified to recommend that women avoid chlorinated tap water — water disinfection plays a vital role in preventing infectious diseases. The reported increases are increases in relative risk; the absolute risk of uterine cancer remains low, and the difference at an individual level is small. Exposure was estimated not by individual measurements but by the average concentrations in the public water system serving the home address; individual water consumption, filter use, preference for bottled water and workplace exposure are unknown. The dose–response trend test for all uterine cancers gives p=0,06, above the conventional significance threshold; the most statistically consistent findings concern the endometrioid subtype . The finding for haloacetic acids (HR 1,86) is the study's most fragile result; it is based on only 154 non-endometrioid cases and more limited historical exposure data, with a very wide confidence interval (1,06–3,25), and this headline-grabbing figure has the lowest evidentiary value. Because numerous compounds and tumor subtypes were examined, the possibility of chance findings arising from multiple comparisons should not be overlooked; the association of trihalomethanes with endometrioid tumors and haloacetic acids with non-endometrioid tumors is biologically inconsistent. The cohort predominantly consists of white, educated women working in education in California, and its generalizability is limited; measured exposure levels are also low (median total trihalomethanes 4,56 µg/L). This is only the second epidemiological study on the subject, and replication is needed; the two existing cohorts share the same address-based method of estimating exposure. The limit values reported in this article depend on regulation and may change over time; consult the relevant official sources for current values. This article contains no recommendations on water consumption, filtration or treatment and provides no reason for any woman to change her water-drinking habits. If symptoms such as postmenopausal vaginal bleeding, abnormal uterine bleeding or unusual discharge occur, consult a doctor without delay. Consult the doctor who knows you for every decision concerning your health.
Was this article helpful?
Your feedback helps us improve our content.
