Methylene Chloride Risk Assessment and Cancer Risk

RoseyFields conducts Methylene Chloride risk assessments using the following methodology.

Chemical Identity, Regulatory Background, and Exposure Sources

Methylene chloride (dichloromethane, DCM; CASRN 75-09-2) is a low-molecular-weight chlorinated aliphatic solvent that is volatile, non-flammable under most use conditions, and highly effective at dissolving oils, resins, and polymers. These properties have made it a workhorse industrial solvent for more than half a century, and they are also the reason it is readily inhaled and dermally absorbed by workers and consumers. RoseyFields’ methodology begins from the chemical’s toxicokinetic profile: methylene chloride is rapidly absorbed by inhalation and ingestion, distributed systemically, and metabolized by two competing hepatic pathways—an oxidative, cytochrome P450 (CYP2E1)-mediated pathway that generates carbon monoxide (and consequent carboxyhemoglobinemia), and a glutathione S-transferase theta-1 (GST-T1)-mediated conjugative pathway that produces reactive, DNA-damaging intermediates (Andersen et al., 1991). Both pathways are central to the dose-response modeling described below.

Historically, methylene chloride has been used as a paint and coating stripper, a metal cleaner and vapor degreaser, an aerosol propellant, a solvent in pharmaceutical and polyurethane foam manufacturing, and an extraction solvent in food processing (e.g., decaffeination, hops and spice extraction). Occupational exposure has been most extensively documented among furniture strippers, degreasers, painters, and workers in cellulose triacetate fiber and photographic film-base manufacturing—the industrial cohorts that anchor much of the human epidemiologic evidence discussed in Section 6, below. Consumer and bystander exposure has historically occurred through use of paint strippers and aerosol products in poorly ventilated spaces, a pathway associated with numerous acute fatalities.

The regulatory landscape for methylene chloride has tightened substantially. OSHA’s Methylene Chloride standard (29 C.F.R. § 1910.1052), promulgated in 1997, set a permissible exposure limit (PEL) of 25 ppm as an 8-hour time-weighted average, a short-term excursion limit (STEL) of 125 ppm over 15 minutes, and an action level of 12.5 ppm—a substantial reduction from the prior 500-ppm ceiling. More recently, EPA finalized a risk management rule under the amended Toxic Substances Control Act (TSCA) in April 2024 (effective July 8, 2024) that prohibits all consumer uses of methylene chloride and most industrial and commercial uses, including paint and coating removal, with a phased compliance timeline; narrow exemptions remain for certain closed-system industrial applications (e.g., battery-separator and other chemical manufacturing, laboratory use, solvent welding) subject to a Workplace Chemical Protection Program. Methylene chloride is also listed as a hazardous air pollutant under the Clean Air Act and a hazardous substance under CERCLA, and it is regulated as a hazardous constituent under RCRA at contaminated sites. Because RoseyFields’ assessments frequently arise in a litigation or regulatory-compliance context—historical occupational exposure claims, vapor intrusion investigations, or Superfund/brownfield cost-recovery matters—the toxicity values described in this methodology (the RfD, RfC, oral slope factor, and inhalation unit risk) are applied against exposure reconstructions specific to each matter’s facts, as discussed in Section 7.

Point-of-Departure Selection

RoseyFields’ selection of points of departure (PODs) for methylene chloride tracks the weight-of-evidence approach used by EPA’s National Center for Environmental Assessment, but the underlying toxicological support rests on a specific set of well-conducted, peer-reviewed bioassays, which RoseyFields’ assessors examine directly rather than relying on the regulatory summary alone.

For the noncancer oral endpoint, the critical study is the two-year drinking-water bioassay in F344 rats conducted by Serota and colleagues (Serota et al., 1986a), in which 85 rats/sex/dose (135 controls) were administered methylene chloride in drinking water at mean doses ranging from 6 to approximately 260 mg/kg-day. The critical effect was hepatic—a dose-related, statistically significant increase in foci/areas of cellular alteration in the liver, first apparent after 78 weeks of treatment and progressing through week 104. The study authors identified 5 mg/kg-day as a NOAEL and 50 mg/kg-day as a LOAEL for these lesions. This study was selected as the principal study because it is a well-conducted, peer-reviewed chronic bioassay with four dose groups plus concurrent controls, adequate statistical power, and a biologically coherent, monotonic dose-response relationship; it is corroborated by a subchronic drinking-water study in rats and mice showing similar hepatocellular vacuolation and necrosis at higher doses (Kirschman et al., 1986).

For the noncancer inhalation endpoint, the critical study is the two-year inhalation bioassay in Sprague-Dawley rats by Nitschke and colleagues (Nitschke et al., 1988a), which exposed 90 rats/sex/group to 0, 50, 200, or 500 ppm methylene chloride for 6 hours/day, 5 days/week. The critical effect—hepatic lipid vacuolation and multinucleated hepatocytes—was statistically significant in females at 500 ppm, with a similar, non-significant trend in males. This study was preferred over a comparable rat inhalation bioassay by Burek and colleagues (Burek et al., 1984), which examined a higher, less informative exposure range (500–3,000 ppm) and reached a broadly consistent conclusion regarding hepatic vacuolation as the sensitive, precursor lesion.

For carcinogenicity, the critical oral dataset is the concurrent two-year drinking-water bioassay in B6C3F1 mice (Serota et al., 1986b), which found statistically significant, dose-related increases in hepatocellular adenomas and carcinomas in male mice at 125, 185, and 250 mg/kg-day. For the inhalation cancer endpoint, the critical dataset is the National Toxicology Program’s two-year inhalation bioassay in B6C3F1 mice (NTP, 1986; published in peer-reviewed form as Mennear et al., 1988), which identified dose-related increases in both hepatocellular tumors and bronchoalveolar (lung) tumors at 2,000 and 4,000 ppm—the only bioassay in the database demonstrating carcinogenicity at two tissue sites and by two exposure routes, and therefore the most robust basis for quantitative cancer risk estimation.

RoseyFields’ assessors do not rely on administered (external) dose as the dose metric for any of these endpoints. Because methylene chloride’s toxicity is driven by internally generated, tissue-specific metabolites rather than the parent compound itself, both the noncancer and cancer PODs are derived from internal dose metrics using benchmark dose (BMD) modeling, described in Section 3, and PODs are expressed as the first-percentile human-equivalent dose (HED) or concentration (HEC)—a health-protective choice that accounts for population variability in metabolic activation, rather than relying on a central-tendency or single “average” internal dose estimate.

Pharmacokinetic Modeling and Dose Conversion

Because methylene chloride’s carcinogenic and hepatotoxic effects are mediated by tissue-specific metabolites rather than the parent compound, RoseyFields’ methodology places physiologically based pharmacokinetic (PBPK) modeling at the center of the dose-response analysis, rather than treating it as a peripheral adjustment.

The rat PBPK model used to translate the Serota et al. (1986a) and Nitschke et al. (1988a) bioassay exposures into internal dose metrics is the model originally published by Andersen and colleagues (Andersen et al., 1991), which simulates methylene chloride’s disposition and its partition between the CYP-mediated (carbon monoxide-generating) and GST-mediated (DNA-reactive) metabolic pathways, expressed as milligrams of methylene chloride metabolized via the relevant pathway per liter of target tissue per day. Liver lesion incidence data from Serota et al. (1986a) and Nitschke et al. (1988a) were fit to dichotomous benchmark-dose models (using internal dose, not administered dose, as the independent variable) to derive a rat-internal BMDL10 (the 95% lower bound on the internal dose associated with a 10% extra risk of the critical lesion).

Converting the rodent internal dose to a human-equivalent value requires two further steps. First, because the relevant dose metric is a metabolic rate—and clearance of these metabolites is presumed slower per unit tissue volume in humans than in rats—the rat internal BMDL10 is divided by an allometric scaling factor based on body weight to the three-quarter power (operationalized as [BWhuman/BWrat]^0.25 ≈ 4.09), yielding a human-equivalent internal BMDL10. Second, that scaled internal dose is converted into an external human-equivalent dose or concentration using a probabilistic human PBPK model adapted from David and colleagues (David et al., 2006), which was calibrated using hierarchical Bayesian methods and Monte Carlo sampling to characterize population variability in human metabolic parameters, including polymorphic GST-T1 activity. This model generates a full distribution of human-equivalent doses rather than a single point estimate; RoseyFields’ assessors use the first percentile of that distribution—rather than the mean—as the POD for the noncancer RfD and RfC, ensuring that individuals with higher-than-average metabolic activation capacity are not systematically underprotected.

The analogous mouse-to-human PBPK translation for the cancer endpoints uses a related but distinct mouse PBPK model (Marino et al., 2006), which similarly partitions methylene chloride metabolism between the lung- and liver-specific GST pathways relevant to the tumor sites observed in the Serota et al. (1986b) and NTP (1986)/Mennear et al. (1988) bioassays. As with the noncancer analysis, mouse-internal BMDL10 values are scaled to humans by an allometric body-weight factor (approximately 7-fold for the mouse-to-human comparison) and then propagated through the David et al. (2006) human PBPK model to generate distributions of cancer potency estimates specific to the GST-T1+/+ (homozygous active) genotype—the subpopulation expected to be most sensitive to methylene chloride’s mutagenic metabolites. RoseyFields’ assessors flag this genotype-specific framing as directly relevant to inter-individual susceptibility arguments in litigation, since GST-T1 activity (and null-genotype prevalence) varies meaningfully across ethnic populations (Sherratt et al., 2002).

Uncertainty Factor Derivation

RoseyFields applies a composite uncertainty factor (UF) of 30 to both the oral RfD and the inhalation RfC, reflecting the same underlying areas of uncertainty in each case, consistent with EPA’s standard five-domain UF framework.

An interspecies uncertainty factor (UFA) of 3 (10^0.5) is applied to both values. This is a reduced factor—rather than the full default of 10—because the PBPK modeling described in Section 3 already resolves the toxicokinetic component of interspecies uncertainty (i.e., differences in absorption, distribution, metabolism, and excretion between rats and humans are explicitly modeled rather than assumed). The residual UFA of 3 covers only the toxicodynamic possibility that human liver tissue is intrinsically more sensitive than rat liver tissue to an equivalent internal dose of reactive metabolite, a question the PBPK model cannot resolve.

An intraspecies (human variability) uncertainty factor (UFH) of 3 (10^0.5) is likewise applied to both values, again a reduced default. The probabilistic human PBPK model (David et al., 2006) already characterizes toxicokinetic variability across the human population via Monte Carlo sampling of metabolic parameters (including GST-T1 polymorphism), which is why the first percentile of the HED/HEC distribution—rather than the median—is used as the POD. The residual UFH of 3 accounts for toxicodynamic variability among sensitive human subgroups that the PBPK model does not capture.

No LOAEL-to-NOAEL uncertainty factor (UFL) is applied to either value. Because benchmark dose modeling was used to derive the POD directly from the full dose-response curve (rather than from a single NOAEL/LOAEL pair), the LOAEL-to-NOAEL extrapolation concern is addressed through the choice of benchmark response level itself, which RoseyFields’ assessors treat as representing a minimum biologically significant change in liver histology.

No subchronic-to-chronic uncertainty factor (UFS) is applied to either value, because both principal studies (Serota et al., 1986a; Nitschke et al., 1988a) are full chronic-duration (2-year) bioassays; no temporal extrapolation is required.

A database uncertainty factor (UFD) of 3 is applied to the oral RfD and a separate UFD of 3 to the inhalation RfC, each reflecting specific, identified gaps rather than a generic database “penalty.” For the oral RfD, the UFD reflects the absence of a two-generation oral reproductive study (only a limited two-generation inhalation study is available, and it did not maintain continuous exposure through gestation and lactation, per Nitschke et al., 1988b) and the absence of any oral neurodevelopmental toxicity study—an evidence gap of particular concern because methylene chloride crosses the placental barrier (Anders and Sunram, 1982; Withey and Karpinski, 1985) and CYP2E1 activity, which generates the neuroactive metabolite carbon monoxide, is relatively high in the developing human fetal brain (Brzezinski et al., 1999; Johnsrud et al., 2003; Hines, 2007). For the inhalation RfC, the UFD of 3 reflects the same reproductive/developmental data limitations plus a single acute (3-hour) inhalation study demonstrating pulmonary immunosuppression in mice at a concentration below the RfC’s POD (Aranyi et al., 1986); because no chronic or repeated-exposure immunotoxicity study is available to confirm or refute the relevance of that acute finding, RoseyFields treats it as an unresolved database gap rather than dismissing it.

The resulting composite UF for both the RfD and RfC is UFA (3) × UFH (3) × UFD (3) ≈ 30 (the individual half-log factors of 3 combine multiplicatively and are rounded to a single significant figure by convention).

Derivation of the Oral RfD and Inhalation RfC

Applying the POD and composite UF described above, RoseyFields’ methodology derives:

Oral RfD = 6 × 10⁻³ mg/kg-day, calculated as the first-percentile human-equivalent dose of 0.19 mg/kg-day (derived from the Serota et al., 1986a, rat liver lesion data via the Andersen et al., 1991, and David et al., 2006, PBPK models) divided by a composite UF of 30.

Inhalation RfC = 0.6 mg/m³, calculated as the first-percentile human-equivalent concentration of 17.2 mg/m³ (derived from the Nitschke et al., 1988a, rat liver lesion data via the same PBPK modeling framework) divided by a composite UF of 30.

Both values are corroborated, though not directly relied upon, by independent human epidemiologic data. Candidate RfCs calculated from two occupational neurobehavioral studies—1.2 mg/m³ based on Cherry et al. (1981) and 1.8 mg/m³ based on Lash et al. (1991)—are of the same order of magnitude as the 0.6 mg/m³ value derived from the animal liver-toxicity dataset, which RoseyFields’ assessors view as an important cross-validation of the rodent-based POD, notwithstanding the substantial limitations of those occupational studies (small cohorts, imprecise historical exposure reconstruction, and absence of standardized neurobehavioral testing batteries).

Carcinogenicity Classification and Quantitative Cancer Risk Values

RoseyFields’ assessors classify methylene chloride as likely to be carcinogenic in humans, operating by a mutagenic mode of action. That classification rests on convergent evidence across three lines of data. First, animal bioassays demonstrate carcinogenicity at two tissue sites in mice—liver and lung tumors following inhalation exposure (NTP, 1986; Mennear et al., 1988) and liver tumors following oral exposure (Serota et al., 1986b)—together with supporting evidence of benign and, less consistently, malignant mammary tumors in rats (Nitschke et al., 1988a; Burek et al., 1984; Maltoni et al., 1988) and rare astrocytomas/gliomas in rats (Nitschke et al., 1988a). Second, genotoxicity and mechanistic studies establish that methylene chloride’s carcinogenicity is mediated by glutathione S-transferase theta-1 (GST-T1)-catalyzed metabolic activation to DNA-reactive intermediates: mutagenicity in bacterial and mammalian assays is consistently GST-dependent (Thier et al., 1993; DeMarini et al., 1997; Graves and Green, 1996; Graves et al., 1996), and in vivo genotoxic effects (micronuclei, chromosomal aberrations, sister chromatid exchange, DNA strand breaks) localize specifically to the liver and lung—the two tissue sites where tumors occur (Allen et al., 1990; Graves et al., 1995; Sasaki et al., 1998). Third, occupational cohort and case-control studies in humans report associations between methylene chloride exposure and several cancer types, most consistently brain cancer (Hearne and Pifer, 1999; Heineman et al., 1994; Tomenson, in press), liver and biliary tract cancer (Lanes et al., 1990; Lanes et al., 1993), and non-Hodgkin lymphoma (Seidler et al., 2007; Miligi et al., 2006; Wang et al., 2009; Barry et al., 2011), with weaker supporting evidence for multiple myeloma (Gold et al., 2011). RoseyFields’ assessors weigh these epidemiologic findings carefully: the underlying cohorts are small, statistical power for rare cancers is limited, and exposure reconstruction in several studies (e.g., Cocco et al., 1999) relies on relatively crude job-history or death-certificate data—yet the consistency of the brain-cancer and liver/biliary-cancer signals across multiple independent cohorts, combined with the mechanistic plausibility supplied by the GST-T1 pathway data, supports a “likely” rather than merely “suggestive” weight-of-evidence characterization.

RoseyFields applies two quantitative cancer potency values, both derived using PBPK-based internal-dose modeling and linear low-dose extrapolation from a benchmark dose (consistent with a mutagenic mode of action, for which a non-threshold, linear default is scientifically appropriate):

Oral slope factor (OSF) = 2 × 10⁻³ per mg/kg-day, derived from male mouse liver tumor incidence data (Serota et al., 1986b) using the Marino et al. (2006) mouse PBPK model to estimate internal liver GST-metabolism dose, benchmark-dose modeling to derive a mouse-internal BMDL10 (39.6 mg metabolized/L liver tissue/day), allometric scaling to a human-equivalent internal BMDL10, and the David et al. (2006) probabilistic human PBPK model to translate that internal potency into an external-dose slope factor for the GST-T1+/+ (most-sensitive) genotype. This oral slope factor should not be extrapolated to doses exceeding its underlying point of departure (a BMDL10 of 60 mg/kg-day expressed as external dose), above which the fitted dose-response model, rather than linear extrapolation, better characterizes the data.

Inhalation unit risk (IUR) = 1 × 10⁻⁸ per µg/m³, derived from combined male mouse liver and lung tumor incidence data (NTP, 1986; Mennear et al., 1988) using the same PBPK-modeling and benchmark-dose framework, with liver- and lung-specific internal dose metrics combined into a single unit risk estimate. As with the OSF, this IUR should not be extrapolated above its underlying POD (a BMDL10 of 7,700 mg/m³ expressed as external concentration).

Because methylene chloride acts through a mutagenic mode of action, RoseyFields’ cancer risk calculations incorporate EPA’s age-dependent adjustment factors (ADAFs)—10-fold for exposure ages younger than 2 years, 3-fold for ages 2 to younger than 16 years, and 1-fold (no adjustment) for ages 16 and older—reflecting the generally accepted principle that early-life exposure to mutagenic carcinogens carries disproportionately greater carcinogenic risk per unit dose. Age-specific risks are calculated separately for each exposure age window and summed across the exposure period of interest, rather than applying a single potency value across an entire lifetime exposure scenario.

Application in Practice

RoseyFields’ risk assessors deploy these toxicity values through standard EPA risk-assessment architecture, adapted to the fact pattern of each engagement.

For noncancer hazard evaluation, assessors calculate a Hazard Quotient (HQ) for each relevant exposure pathway as the ratio of the reconstructed chronic daily intake (or air concentration) to the RfD (or RfC): HQ = CDI / RfD for oral/dermal pathways, or HQ = EC / RfC for inhalation. Where a case involves exposure to methylene chloride alongside other chemicals affecting the same target organ (commonly the liver, given methylene chloride’s hepatotoxic profile), assessors sum the individual HQs into a Hazard Index (HI) for that target organ, consistent with EPA’s guidance on cumulative noncancer risk assessment; an HI exceeding unity signals a potential for adverse effects that warrants further pathway-specific analysis rather than an automatic finding of harm.

For cancer risk evaluation, assessors calculate excess lifetime cancer risk (ELCR) as ELCR = CDI × OSF (oral) or ELCR = EC × IUR (inhalation), with age-specific ELCR values calculated separately for each ADAF-defined age window when childhood exposure is at issue and summed to a total lifetime estimate. RoseyFields’ assessors are careful to flag, in every report, that the OSF and IUR represent upper-bound (not central-tendency or “most likely”) risk estimates, and that neither should be applied to exposure levels exceeding the points of departure identified in Section 6.

Underpinning both calculations is site- or case-specific exposure and dose reconstruction: assessors combine available monitoring data (industrial hygiene sampling, indoor air or vapor intrusion measurements, biomonitoring where available) with documented work histories, product-use patterns, and standard or site-specific exposure factors (intake rate, body weight, exposure frequency and duration, averaging time) following EPA’s Risk Assessment Guidance for Superfund (RAGS) framework, to generate the chronic daily intake or exposure concentration that is then compared against the RfD/RfC/OSF/IUR values derived above. In litigation and regulatory-compliance matters—toxic tort and product liability claims, workers’ compensation disputes, and Superfund or state-cleanup cost-recovery actions—this full evidentiary chain (peer-reviewed critical study → PBPK-based internal dose modeling → benchmark dose → uncertainty factors → RfD/RfC or OSF/IUR → case-specific exposure reconstruction → HQ/HI or ELCR) is documented transparently so that each step can be independently verified and defended under Daubert or analogous evidentiary standards. RoseyFields’ assessors routinely perform sensitivity analyses around key modeling assumptions (e.g., alternative dose metrics, alternative PODs, GST-T1 genotype assumptions) to characterize how robust a given risk conclusion is to reasonable alternative scientific judgments—an analysis that is often dispositive in adversarial proceedings.

Key Uncertainties and Sensitive Populations

Several uncertainties merit particular attention when RoseyFields’ assessors apply these values to a specific case. Genetic polymorphism in GST-T1 activity is the most consequential: because the cancer potency values are explicitly derived for the GST-T1+/+ (homozygous active) genotype—the population subgroup most capable of metabolically activating methylene chloride to its DNA-reactive intermediate (Sherratt et al., 2002)—individuals who are GST-T1 null or heterozygous may in principle be considerably less susceptible to the mutagenic cancer pathway, a consideration that can be case-relevant where genotype data are available but that also means the applied potency values are conservative (protective) defaults for a mixed-genotype population.

Early-life and in-utero exposure represents a second significant area of uncertainty. Methylene chloride crosses the placental barrier (Anders and Sunram, 1982; Withey and Karpinski, 1985), and the developing human fetal brain expresses relatively high CYP2E1 activity relative to the liver (Brzezinski et al., 1999; Johnsrud et al., 2003; Hines, 2007), raising a plausible—but not conclusively demonstrated in humans—concern for neurodevelopmental effects mediated by the carbon monoxide metabolite, particularly since single-dose animal developmental studies at high exposure concentrations have reported decreased offspring birth weight and altered behavioral habituation (Bornschein et al., 1980; Hardin and Manson, 1980). No dedicated developmental neurotoxicity study exists in the oral or inhalation database for this chemical, which is precisely why a database uncertainty factor is applied rather than omitted. The ADAF framework partially, but not completely, compensates for this gap by assuming greater per-unit cancer risk in early life; it does not address potential noncancer neurodevelopmental risk in the same way.

Individuals with compromised cardiopulmonary or hematologic status are a further sensitive subgroup, given that methylene chloride’s CYP-mediated metabolite, carbon monoxide, adds to endogenous carboxyhemoglobin burden; persons with pre-existing anemia, cardiovascular disease, or who smoke may experience clinically significant carboxyhemoglobin elevations at exposures that would be inconsequential in healthy adults, even though the PBPK modeling underlying the RfD and RfC predicts that carbon monoxide levels at those regulatory exposure levels are insufficient to cause neurodevelopmental toxicity in typical populations.

Finally, RoseyFields’ assessors note several structural uncertainties in the toxicological database itself: the epidemiologic cohorts underlying the human cancer evidence are small and have limited statistical power to detect effects for rare cancers such as brain or biliary tract cancer; the human PBPK model, while calibrated using rigorous Bayesian population methods, is applied deterministically (using mean parameter values) in the final risk calculations, which does not fully propagate parameter uncertainty into the reported point estimates; and the immunotoxicity database consists of a single acute-exposure study (Aranyi et al., 1986), leaving chronic immunomodulatory risk essentially uncharacterized. RoseyFields’ assessors disclose each of these limitations explicitly in case-specific work product, consistent with the firm’s practice of distinguishing well-supported quantitative conclusions from areas requiring qualitative, case-by-case professional judgment.

Bibliography