1,4-Dioxane Epidemiology and Cancer Toxicity

RoseyFields evaluates the human epidemiological evidence for 1,4-Dioxane as follows.

1. Study Populations and Data Sources

The human evidence base for 1,4-dioxane is structurally very different from — and substantially smaller than — the epidemiological databases RoseyFields summarizes for chemicals such as PFOA. Rather than several hundred published studies spanning multiple population types, the entire body of human data on 1,4-dioxane consists of a small number of case reports of severe acute occupational poisoning, a handful of controlled short-term exposure studies in healthy adult volunteers, and exactly two occupational cohort mortality studies. There are no published community or contamination-site health studies, no general-population biomonitoring program comparable to NHANES, and no birth cohort studies of 1,4-dioxane, despite the chemical’s frequent detection as a drinking water contaminant nationwide.

Case reports document acute, high-concentration occupational poisoning episodes from the 1930s through the 1950s, in which workers exposed to unspecified but clearly very high vapor concentrations of 1,4-dioxane developed fatal hemorrhagic nephritis, centrilobular liver necrosis, and central nervous system effects, with death typically occurring within one to two weeks of the onset of illness. These reports are informative as to the target organs affected by extreme acute exposure but cannot support quantitative dose-response analysis, since exposure concentrations were not measured precisely and the reports describe only a handful of individuals.

Controlled human volunteer studies, conducted between the 1930s and 2006, exposed small panels of healthy adult men and women to 1,4-dioxane vapor for periods ranging from one minute to six hours, at concentrations from roughly 20 ppm up to 5,500 ppm, and recorded acute sensory and irritant responses (eye, nose, and throat irritation; odor perception; and, in the most recent and best-controlled of these studies, self-reported symptom ratings, pulmonary function, nasal swelling, and inflammatory blood markers). These studies establish an approximate threshold for acute sensory irritation in humans and, in the most recent study, a no-observed-effect concentration for short-term exposure, but by design they cannot speak to chronic disease outcomes.

Occupational cohorts provide the only chronic, multi-year human mortality data available for 1,4-dioxane, and there are just two: a cross-sectional and mortality study of workers at a German 1,4-dioxane production facility, and a mortality follow-up study of workers at a Texas chemical manufacturing and processing facility. Combined, these two cohorts cover fewer than 250 workers. Both studies estimated historical exposure using a combination of area monitoring data (available only from the late 1960s/early 1970s onward) and the contemporaneous Threshold Limit Value, since no biomonitoring or personal exposure measurements existed for the earlier years of each cohort’s employment history.

General-population biomonitoring studies of the kind that anchor much of the epidemiological literature for other environmental chemicals do not exist for 1,4-dioxane. This is a direct consequence of the chemical’s toxicokinetics: 1,4-dioxane is rapidly absorbed, metabolized, and eliminated in humans, with a plasma half-life on the order of an hour and complete urinary elimination of the parent compound and its primary metabolite within roughly 24 hours of exposure ending. Because 1,4-dioxane does not persist in the body the way PFOA or other bioaccumulative chemicals do, a single blood or urine sample — the sampling approach used by NHANES-type surveys — would reflect only very recent exposure and would be of limited value for reconstructing longer-term exposure history, which likely explains why no general-population biomonitoring survey has incorporated 1,4-dioxane as a routine analyte. Likewise, no birth cohort or pediatric cohort has examined 1,4-dioxane exposure and child health outcomes, and no community-based health study of a population exposed to contaminated drinking water has been published, notwithstanding 1,4-dioxane’s status as one of the more frequently detected emerging drinking water contaminants nationally.

The general epidemiological principle that findings replicated across diverse, independently studied populations provide stronger support for causal inference than a single study or population applies with particular force here — precisely because that kind of replication is unavailable for 1,4-dioxane. With only two small occupational cohorts and no corroborating community, general-population, or birth-cohort data, the human evidence base cannot by itself establish, refute, or quantify a dose-response relationship for any chronic health outcome, a limitation that both cohort studies’ authors and EPA’s own toxicological review acknowledge directly. This stands in sharp contrast to chemicals with a mature, multi-population epidemiological literature, and it is the single most important framing point for any epidemiology-focused evaluation of 1,4-dioxane.

2. Overview of Human Health Outcomes Investigated, by Body System

Because the human evidence base for 1,4-dioxane is limited to acute case reports, short-term volunteer studies, and two small occupational mortality studies, the “strong, consistent, quantitatively modeled evidence” tier that RoseyFields is able to populate for chemicals with larger epidemiological literatures is, for 1,4-dioxane, essentially empty from a human-epidemiology standpoint. The tiers below reflect that reality candidly.

Outcomes with strong, consistent, quantitatively modeled human epidemiological evidence: none. No health outcome has been quantitatively modeled from human data for 1,4-dioxane; both the EPA’s chronic oral reference dose and its oral and inhalation cancer slope factors are derived entirely from animal bioassay data (discussed in Section 6), not from epidemiological studies. This is an important structural difference from RoseyFields’ PFOA page, where several outcomes (hepatic injury, immune suppression, cardiovascular effects, developmental effects, and kidney/testicular cancer) meet this tier based on human data.

Outcomes with suggestive, qualitative, but non-quantifiable evidence in humans, drawn from the acute case reports and volunteer studies, include: acute hepatotoxicity (centrilobular liver necrosis) and acute nephrotoxicity (hemorrhagic nephritis) following very high-concentration acute occupational exposure; central nervous system effects (headache, agitation, coma, and post-mortem evidence of cerebral perivascular changes) in the same fatal poisoning cases; and acute sensory/mucous membrane irritation (eye, nose, and throat irritation) at moderate-to-high short-term vapor concentrations in controlled volunteer studies, with a no-observed-effect concentration identified in the best-controlled of these studies. These findings are directionally consistent with the target organs (liver, kidney, and — for inhalation — nasal and respiratory epithelium) identified in chronic and subchronic animal studies, which is a meaningful point of cross-species coherence even though the human observations themselves come only from a handful of acute, high-dose case reports rather than from a population-based or dose-responsive epidemiological design.

Outcomes investigated with null findings in the two available occupational cohort mortality studies include: overall (all-cause) mortality; cancer mortality/incidence generally; liver disease and cirrhosis; kidney disease; hematological abnormalities (complete blood count, differential); serum liver enzyme and renal function panels in actively employed and retired workers; and chromosomal aberrations in peripheral lymphocytes. Both occupational cohort studies found observed mortality and cancer incidence comparable to expected rates based on general-population reference statistics, and neither study identified clinical, hematological, or biochemical abnormalities attributable to chronic occupational 1,4-dioxane exposure. As discussed in Section 5, however, these are underpowered null findings from small cohorts with short average exposure durations and latency periods, not findings that affirmatively rule out an association — a distinction both the original study authors and EPA’s review emphasize.

3. Discussion of the Available Human Evidence: Studies, Exposure Metrics, and Findings

Because no outcome meets RoseyFields’ “strong evidence” bar from human data alone for 1,4-dioxane, this section departs from the format used on RoseyFields’ PFOA and other pages with mature epidemiological literatures. Rather than presenting effect sizes for quantitatively modeled outcomes, it summarizes what the available human studies actually found, organized by study.

Acute occupational poisoning case reports. Two published case series describe fatal acute occupational poisoning from 1,4-dioxane. One describes four fatal cases of hemorrhagic nephritis and centrilobular hepatic necrosis in workers acutely exposed by inhalation to high (unspecified) vapor concentrations, with death occurring five to eight days after onset of illness; autopsy findings suggested the kidney injury, rather than the liver injury, was the proximate cause of death. The other describes a single fatal case in a worker exposed by combined inhalation and dermal routes over a one-week period, with measured workplace air concentrations of 208–650 ppm (mean approximately 470 ppm); this worker developed severe abdominal pain, renal failure, headache, hypertension, agitation, and coma, and autopsy revealed centrilobular hepatic necrosis, hemorrhagic renal cortical necrosis, and perivascular brain changes with focal demyelination, changes the case report’s author attributed to secondary anoxia and cerebral edema rather than direct neurotoxicity. Because both reports involve only a handful of individuals with imprecisely characterized exposure, they establish that extremely high acute 1,4-dioxane exposure is capable of causing severe, sometimes fatal, liver, kidney, and CNS injury in humans, but they cannot be used to characterize a dose-response relationship or to draw conclusions about chronic, lower-level exposure.

Controlled acute volunteer exposure studies. A series of small controlled-exposure studies, spanning from 1930 to 2006, exposed healthy adult volunteers to defined 1,4-dioxane vapor concentrations for short durations and recorded symptom and physiological responses. Early studies conducted for U.S. and European government and industrial hygiene laboratories found eye, nose, and throat irritation at concentrations at or above roughly 280–300 ppm for exposures of several minutes, with vertigo reported at higher concentrations (5,500 ppm for one minute), while lower concentrations (up to 1,000–2,000 ppm for brief exposures, or 200 ppm for fifteen minutes) were generally tolerated without significant symptoms. A study conducted for a chemical manufacturer exposed four healthy adult men to 50 ppm for six hours and observed only mild eye irritation, along with diminishing odor perception over the exposure period; this study also characterized 1,4-dioxane pharmacokinetics in humans. The most methodologically rigorous of these studies, conducted in Sweden with institutional ethics review, exposed twelve healthy men and women to 20 ppm 1,4-dioxane vapor at rest for two hours and found no statistically significant effect on self-reported symptom ratings, blink frequency, pulmonary function, nasal swelling, or plasma inflammatory markers (C-reactive protein, interleukin-6), with only subjective ratings of solvent odor significantly elevated during exposure; a preliminary screening phase of the same study found no self-reported symptoms at concentrations up to 20 ppm, which the investigators used as a tentative no-observed-adverse-effect concentration. Collectively, these studies support an approximate acute sensory-irritation threshold in the low hundreds of ppm and a no-effect concentration around 20 ppm for short-term exposure, but — like the case reports — they address acute, not chronic, exposure and were not designed to detect disease outcomes.

Occupational cohort mortality studies. The German cohort study evaluated 74 workers at a 1,4-dioxane production facility — 24 actively employed at the time of the investigation, 23 former employees still with the company, and 27 retired or deceased workers — with employment durations ranging from three to 41 years and contemporaneous air concentrations measured at 0.06–0.69 ppm (with a retrospective simulation of pre-1969 conditions estimating historical concentrations as high as 7.2 ppm). Active and former workers underwent clinical examination, chest X-ray, hematology, and serum biochemistry; no pathological findings, malignant disease, liver enlargement, or jaundice were identified, and renal function tests and urinalysis were normal. Serum transaminase elevations were observed in 16 of 47 workers tested, but the investigators attributed this to chronic heavy alcohol consumption (more than 80 grams per day) reported by these workers, not to 1,4-dioxane exposure. Chromosomal aberration analysis in a subset of six actively employed workers and six comparison individuals found no differences in aberration frequency. Mortality analysis of the full 74-worker cohort, covering 1,840.5 person-years of observation, found 12 observed deaths against 14.5 expected based on age-specific German national mortality rates; two cancer cases were identified (one epithelial carcinoma, one leukemia), neither considered the primary cause of death, and cancer-specific standardized mortality ratios did not differ significantly from the reference population.

The Texas cohort study followed workers at a chemical manufacturing plant and a separate processing unit five miles away, employed between 1954 and 1975, divided into a 100-worker manufacturing cohort and a 65-worker processing cohort; exposure was estimated at below 25 ppm for essentially all workers in both cohorts using a combination of the contemporaneous 50 ppm Threshold Limit Value and available area/personal monitoring data (available only from 1968 in the manufacturing area and 1974 in the processing area). Manufacturing-area workers were co-exposed to several other chemicals, and processing-area workers were co-exposed to vinyl chloride. Twelve total deaths were identified across both cohorts (seven manufacturing, five processing), compared with 9.8 expected based on Texas age-race-sex-specific mortality rates; three deaths were attributed to cancer (stomach carcinoma and alveolar cell carcinoma in the manufacturing cohort, a mediastinal tumor in the processing cohort) against 1.7 expected, though the two manufacturing-cohort cancer cases both had a positive heavy-smoking history and the processing-cohort case occurred in a worker exposed for only one year. A sensitivity analysis restricted to manufacturing workers exposed more than two years found six observed deaths against 4.1 expected, again not statistically significant using a conditional Poisson approach. The study authors themselves flagged the analysis’s central limitation: mean exposure duration was under five years (under two years for 43% of the cohort), and the latency period between first exposure and the mortality analysis was under ten years for 59% of workers — too short, by conventional epidemiological standards, to expect a chronic disease or cancer signal to emerge even if a true elevated risk existed. The study authors recommended a longer-latency follow-up study; none has been published in the decades since.

4. Application to Quantitative Risk Assessment

Because neither occupational cohort study provided data adequate to characterize a dose-response relationship for any noncancer or cancer endpoint, EPA’s Integrated Risk Information System (IRIS) toxicological review — the primary quantitative risk assessment for 1,4-dioxane — derived both its chronic oral reference dose and its oral and inhalation cancer slope factors entirely from chronic animal bioassay data, not from epidemiology. The chronic oral reference dose (0.03 mg/kg-day) is based on a no-observed-adverse-effect level for liver and kidney toxicity in a two-year drinking-water study in male rats. The oral and inhalation cancer slope factors are similarly based on tumor incidence data (liver, nasal, peritoneal, and mammary tumors) from multiple two-year rat, mouse, and guinea pig bioassays. EPA’s own review states plainly that the two available human occupational studies do not provide sufficient information to quantify either noncancer or cancer risk, a conclusion the Agency’s independent external peer reviewers unanimously affirmed.

5. Study Quality Considerations

Statistical power and cohort size. Both occupational cohort studies are very small by epidemiological standards — 74 and 165 workers, respectively — with correspondingly small numbers of observed deaths (12 and 12) and cancer cases (2 and 3). Cohorts of this size have limited statistical power to detect even a substantially elevated relative risk for any but the most common causes of death, and essentially no power to detect a moderate elevation in a rare cancer. A negative (null) finding from a cohort this size is properly interpreted as inconclusive rather than as affirmative evidence of safety, a point both original study authors and EPA’s toxicological review make explicitly.

Latency and exposure duration. As detailed in Section 3, the Texas cohort’s average exposure duration was under five years, and a majority of workers had fewer than ten years of latency between first exposure and the mortality analysis. Because many chronic diseases, and cancer in particular, can have latency periods of twenty years or more from first relevant exposure to clinical diagnosis or death, a mortality analysis conducted this soon after exposure began is structurally unable to capture a substantial fraction of the disease burden that might eventually emerge in this cohort, regardless of the true underlying risk.

Co-exposure to other chemicals. Both cohorts involved substantial co-exposure to other chemicals that complicates attribution of any observed (or unobserved) effect specifically to 1,4-dioxane. In the Texas cohort, processing-area workers were co-exposed to vinyl chloride, a known human carcinogen, and manufacturing-area workers were co-exposed to several other unspecified chemicals; a null finding in this setting could reflect either a true absence of effect or a masking/confounding effect from co-exposures that were not separately modeled. In the German cohort, the elevated serum transaminase levels observed in roughly a third of tested workers were attributed by the study authors to heavy alcohol consumption rather than 1,4-dioxane exposure — a plausible alternative explanation, but one that also illustrates how a chronic lifestyle exposure common in an occupational population can complicate interpretation of clinical biomarker findings in either direction.

Exposure misclassification. Neither cohort had access to biomonitoring or personal exposure monitoring data covering the full employment history of its workers; both relied on a combination of the contemporaneous regulatory Threshold Limit Value and area monitoring data that, in each case, was available only starting years after the cohorts’ employment began (1968–1969 in the German cohort’s simulation approach, 1968/1974 in the Texas cohort). Historical exposure reconstruction of this kind is subject to substantial uncertainty, particularly for the retired and deceased workers in the German cohort and the earliest-hired workers in the Texas cohort, and any resulting misclassification would generally be expected to bias observed associations toward the null rather than away from it.

Healthy-worker effect. As with essentially all occupational cohort mortality studies, both cohorts are subject to the healthy-worker effect: individuals who remain employed, particularly in physically demanding industrial roles, tend to be healthier on average than the general population used as the comparison group, which can mask a true excess in mortality or morbidity by depressing the observed-to-expected ratio even in the presence of a real occupational hazard.

Absence of general-population and community data. As discussed in Section 1, there is no published epidemiological study of a general population or drinking-water-contaminated community exposed to 1,4-dioxane, which means the two occupational cohorts — despite their limitations — represent the entire available epidemiological record. There is no opportunity, as there is for chemicals with richer literatures, to assess whether findings (or null findings) replicate across populations with different exposure routes, co-exposure profiles, or demographic characteristics, which meaningfully limits the confidence that can be placed in either a positive or a null conclusion drawn from this literature.

Reverse causation. Reverse causation — where the health outcome itself alters the exposure measurement rather than the reverse — is not a significant concern in this literature, primarily because the available studies did not attempt to correlate a continuous biomarker of exposure with a continuous health measure in the way that, for example, some persistent-chemical epidemiology does; the mortality-ratio design used in both occupational cohorts is not vulnerable to this particular bias, though it carries the other limitations described above.

6. Independent Expert Panel and Agency Weight-of-Evidence Conclusions

Every major regulatory and scientific body that has evaluated 1,4-dioxane’s carcinogenic potential has reached its conclusion primarily or entirely on the basis of animal bioassay data, explicitly characterizing the human epidemiological evidence as inadequate or limited — a marked contrast to chemicals where independent panels have found the human data itself sufficient to support a “probable” or “known” causation determination. EPA’s IRIS program classifies 1,4-dioxane as “likely to be carcinogenic to humans,” a determination the Agency states is based on “(1) inadequate evidence of carcinogenicity in humans, and (2) sufficient evidence in animals” — specifically, liver tumors across three rat strains, two mouse strains, and guinea pigs, together with nasal, peritoneal (mesothelioma), mammary, kidney, and other tumors observed in chronic rat bioassays. EPA’s external Science Advisory Board peer review panel considered this classification directly; five of six reviewers affirmed it, while the dissenting reviewer argued for the weaker “possible human carcinogen” descriptor, citing (among other factors) the complete absence of increased cancer incidence in the available human epidemiology. The International Agency for Research on Cancer similarly classifies 1,4-dioxane as Group 2B, “possibly carcinogenic to humans,” explicitly on the basis of “inadequate evidence in humans and sufficient evidence in experimental animals” — a classification one tier below the “probably carcinogenic” (Group 2A) or “carcinogenic” (Group 1) tiers IARC reserves for chemicals with stronger human evidence. The National Toxicology Program lists 1,4-dioxane as “reasonably anticipated to be a human carcinogen,” NIOSH characterizes it as a “potential occupational carcinogen,” the American Conference of Governmental Industrial Hygienists classifies it as a “confirmed animal carcinogen with unknown relevance to humans” (Group A3), and California’s Proposition 65 program lists it as a carcinogen — each of these determinations, like EPA’s and IARC’s, rests on the animal bioassay database rather than on an affirmative reading of the human occupational mortality studies. No independent expert panel has issued a “probable link” or comparable determination between 1,4-dioxane and any specific human disease based on epidemiological data, in contrast to the C8 Science Panel’s determinations for PFOA.

7. Use of the Epidemiological Evidence to Support Causation Opinions

For 1,4-dioxane, epidemiology plays a substantially more limited role in supporting causation opinions than it does for chemicals with mature, multi-population human literatures, and RoseyFields’ approach to both general and specific causation for this chemical necessarily differs accordingly. General causation opinions for 1,4-dioxane’s hazard potential rest overwhelmingly on the animal toxicological and mechanistic evidence — the consistent induction of liver, nasal, and other tumors across multiple rodent species, strains, and exposure routes, together with dose-related non-neoplastic liver and kidney findings in subchronic and chronic studies — rather than on epidemiological demonstration of an association in exposed human populations. The available human data play a narrower, corroborative role: the acute occupational poisoning case reports demonstrate that the same organ systems identified as targets in animal studies (liver, kidney, and, for inhalation exposure, the nasal/respiratory epithelium) are also affected in humans following sufficiently high exposure, which supports the biological plausibility and human relevance of the animal findings even though it cannot independently establish a dose-response relationship or a chronic disease risk at lower exposure levels. The two occupational mortality studies, properly characterized as underpowered and inconclusive rather than as evidence of safety, do not meaningfully strengthen or weaken the animal-based general causation case in either direction; they simply establish that no epidemiological signal has been detected in a limited investigation, not that no risk exists.

This evidentiary structure has direct implications for specific causation analysis in individual cases. Because there is no epidemiological dose-response literature to anchor a quantitative individual risk estimate — no relative risk, odds ratio, or hazard ratio derived from human data comparable to what is available for chemicals like PFOA — specific causation opinions for 1,4-dioxane rely correspondingly more heavily on animal-derived toxicity values (the oral reference dose and cancer slope factor described in Section 4), individual exposure reconstruction, and case-specific medical and differential-diagnosis analysis, together with the general biological plausibility support the human case reports and volunteer studies provide for the relevant target organs. Experts opining on causation for 1,4-dioxane-related disease should expect, and be prepared to explain, why the human epidemiological record is thin relative to other environmental chemicals, and should be careful to characterize the two available occupational cohort studies accurately — as underpowered and inconclusive, not as negative or exculpatory findings — consistent with how EPA’s own toxicological review, its external peer reviewers, and the original study authors themselves have characterized this same evidence.

Bibliography