Pentachlorophenol (PCP) Epidemiology and Cancer Risk

Pentachlorophenol (PCP)

RoseyFields evaluates the human epidemiological evidence for pentachlorophenol (PCP) as follows, drawing on occupational cohort studies, community and contamination-site studies, general-population biomonitoring data, and birth cohorts published in the peer-reviewed literature.

1. Study Populations and Data Sources

The human evidence base for PCP spans four broadly distinct types of study populations, each contributing a different kind of information and each subject to different limitations.

Occupational cohorts provide the highest-exposure, longest-duration data and the best statistical power for rare outcomes such as specific cancer subtypes. The largest and most extensively characterized is a cohort of 27,464 men employed for one year or more at 14 British Columbia sawmills between 1950 and 1995, in which dermal PCP and tetrachlorophenol (TCP) exposure was reconstructed from plant records and systematic interviews with senior employees, validated against urinary chlorophenol measurements, and cancer outcomes were ascertained through national mortality and cancer-incidence registries (Demers et al., 2006). A second key occupational data source is a cohort of 2,122 US PCP-production workers from four plants in the NIOSH Dioxin Registry, followed for mortality through 2005 (Ruder and Yiin, 2011), with an earlier analysis of a subset of this cohort reported separately (Ramlow et al., 1996). Smaller occupational studies include Swedish population-based case-control studies of lymphatic and soft-tissue cancers with detailed job-exposure-matrix-based exposure assessment (Hardell et al., 1994, 1995; Kogevinas et al., 1995), a Hawaii wood-treaters clinical/mortality study (Wood et al., 1983), and clinical/neurophysiological studies of small groups of chronically exposed timber and manufacturing workers (Triebig et al., 1987; Peper et al., 1999; Gorman et al., 2001).

Community and contamination-site studies examine populations exposed through environmental contamination rather than occupation, typically at lower per-person doses but often affecting entire communities including children and other sensitive subgroups. Examples include biomonitoring of residents near a former wood-treatment Superfund site in Pensacola, Florida (Karouna-Renier et al., 2007), and a series of Chinese ecological and case-control studies examining cancer risk among general community populations living in areas with historical, schistosomiasis-control-related PCP use and contamination of drinking-water sources along the Yangtze River (Cheng et al., 2015; Cui et al., 2017) and in case-control designs for specific cancers (Yang et al., 2021).

General-population biomonitoring studies characterize background exposure levels in the population at large and, where linked to health outcome data, allow evaluation of associations at the low end of the exposure distribution most relevant to non-occupational risk assessment. In the United States, this evidence comes principally from the National Health and Nutrition Examination Survey (NHANES) program, beginning with a subsample analysis of NHANES II urine specimens that found quantifiable PCP in 71.6% of the general population (geometric mean 6.3 ng/mL) (Kutz et al., 1992), with continued biomonitoring through more recent NHANES cycles. Comparable biomonitoring programs exist in other countries, including Germany’s Environmental Survey (GerES V) (Kolossa-Gehring et al., 2021).

Birth cohorts capture prenatal and early-life exposure and outcomes not well studied in adult occupational or community cohorts, including birth size, reproductive hormone status, and thyroid function in mothers and infants. Relevant cohorts include a Canadian Inuit (Nunavik) pregnancy cohort examining maternal and cord-blood thyroid hormones in relation to several persistent contaminants including PCP (Dallaire et al., 2009), a Swedish pregnancy cohort measuring PCP and related organohalogens in maternal and cord serum (Guvenius et al., 2003), and a large (n = 1,100) Chinese mother-newborn pair cohort examining maternal urinary PCP in relation to birth weight, length, and head circumference (Guo et al., 2016).

Using multiple, diverse populations of this kind strengthens causal inference in several specific ways. First, occupational, community, and general-population studies span several orders of magnitude of exposure intensity, so consistent direction of effect across this range provides a form of internal dose-response replication that a single study cannot offer. Second, each population type carries a different, largely non-overlapping set of potential confounders and co-exposures: occupational cohorts are confounded chiefly by co-exposure to other wood-treatment chemicals and combustion by-products (dioxins/furans), community cohorts by other local environmental contaminants, and general-population biomonitoring studies by diet and lifestyle factors correlated with residual PCP body burden; an association that persists across all three is less likely to be fully explained by any single one of these confounding structures. Third, geographically and demographically distinct populations (Canadian, US, Swedish, and Chinese cohorts) differ in genetic background, diet, co-exposure profile, and healthcare/registry systems, so replication of an association across these populations reduces the likelihood that a single population-specific artifact (for example, a particular registry’s case-ascertainment practice) is driving the result. Finally, birth cohorts and biomonitoring studies extend the observed exposure-outcome relationship down into the range relevant to general-population risk assessment, complementing the higher-dose, longer-duration occupational data that anchors much of the strongest cancer evidence.

2. Overview of Health Outcomes Investigated in Humans, by Body System

The following overview organizes the human evidence by body system and classifies each outcome according to the strength and consistency of the epidemiological evidence, consistent with RoseyFields’ general approach to weight-of-evidence characterization.

2.1 Hematologic and Lymphoid Cancers — Strong, Consistent, Quantitatively Modeled Evidence

Non-Hodgkin’s lymphoma (NHL) and multiple myeloma have the strongest and most consistent epidemiological support of any PCP-associated health outcome. In the British Columbia sawmill cohort, duration of dermal PCP exposure showed a statistically significant, monotonic dose-response relationship with NHL mortality and incidence, with NHL mortality risk approximately twice as high in the highest exposure category as in the lowest (Demers et al., 2006). This finding is corroborated by an independent occupational cohort of US PCP-production workers, which found overall cancer mortality significantly elevated (SMR 1.17; 95% CI 1.05-1.31) with supporting evidence of elevated NHL risk (Ruder and Yiin, 2011; Ramlow et al., 1996), and by a series of Swedish population-based case-control studies using detailed job-exposure-matrix assessment, which found moderately to substantially elevated NHL risk associated with chlorophenol exposure (Hardell et al., 1994; Kogevinas et al., 1995) and a pooled analysis showing a significantly increased risk of soft-tissue sarcoma (Hardell et al., 1995). A systematic review of this literature concluded that, while PCP exposure has not been associated with total cancer incidence or mortality, the evidence for hematopoietic cancers specifically is considerable, consistent across multiple study locations and designs (Cooper and Jones, 2008).

2.2 Renal Cell (Kidney) Cancer — Strong to Moderate Evidence

The British Columbia sawmill cohort also found a statistically significant, dose-related association between PCP exposure and kidney cancer, one of the more robust site-specific findings in that study (Demers et al., 2006). This finding is of particular interest to RoseyFields because it shows some concordance with the mouse carcinogenicity bioassay, in which adrenal (not renal) tumors were the second major tumor type identified, and because kidney cancer risk was not similarly elevated across all of the smaller occupational studies, limiting the strength of this conclusion relative to the hematologic cancer findings above.

2.3 Liver Cancer — Suggestive, Directionally Consistent but Imprecise Evidence

The same British Columbia cohort reported an increased risk of liver cancer, with at least a doubling of risk observed across most exposure categories (Demers et al., 2006), and community-level studies in PCP-contaminated areas of China have similarly reported elevated liver and other digestive-system cancer risk associated with residential PCP exposure (Cui et al., 2017; Cheng et al., 2015). These findings show a notable degree of concordance with the animal bioassay data, in which hepatocellular tumors were the principal finding; however, the human liver cancer evidence carries more residual uncertainty than the hematologic cancer findings because of smaller case counts and more limited ability to rule out viral hepatitis, alcohol use, and other established liver cancer risk factors as confounders in these particular study populations.

2.4 Thyroid Cancer — Emerging, Suggestive Evidence

A Chinese case-control study found that PCP exposure in the highest quartile was associated with more than a threefold increase in thyroid cancer risk relative to the lowest quartile, with a statistically significant dose-response relationship also observed for two related chlorophenols (Yang et al., 2021). This is a newer and more limited body of evidence than the hematologic or liver cancer findings, but it is mechanistically consistent with the well-documented thyroid-hormone-disrupting activity of PCP discussed in Section 2.7 below.

2.5 Soft-Tissue Sarcoma and Other Cancer Sites — Suggestive but Inconsistent Evidence

Soft-tissue sarcoma shows a less consistent pattern than the hematologic cancers: a pooled analysis of Swedish case-control studies found a significantly increased risk (Hardell et al., 1995), but the British Columbia sawmill cohort — considerably larger and, by most assessments, more rigorously designed with respect to exposure assessment — did not find a significant association with soft-tissue sarcoma specifically (Demers et al., 2006), and statistical power to evaluate this comparatively rare cancer was limited in most other studies. Multiple myeloma shows a similar pattern of positive but not universally replicated findings (Demers et al., 2006). Notably, and consistent across nearly every well-designed study, total cancer incidence and mortality have not been found to be elevated in PCP-exposed populations overall (Demers et al., 2006; Ruder and Yiin, 2011; Cooper and Jones, 2008) — the epidemiological signal for PCP carcinogenicity is specific to a handful of cancer sites rather than reflecting a generalized increase in cancer risk.

2.6 Immune System — Suggestive Evidence

Human blood PCP levels have been associated with impaired immune function in two independent clinical studies: lymphocyte mitogen response was impaired in a dose-dependent manner in patients with elevated blood PCP levels, most markedly above 20 µg/L (Daniel et al., 1995), and altered immune response measures were separately reported in PCP-exposed individuals by McConnachie and Zahalsky (1991). These findings are qualitatively consistent with the animal immunotoxicity evidence (impaired antibody responses at PCP doses well below those causing overt systemic toxicity) but derive from small clinical samples rather than large, population-based studies, so the overall body of human immune evidence is best characterized as suggestive rather than strong.

2.7 Endocrine (Thyroid) System — Suggestive Evidence

In a Canadian Inuit (Nunavik) pregnancy cohort, maternal plasma PCP concentration was negatively associated with cord plasma free thyroxine (fT4) concentration in neonates (β = -0.59, p = 0.02), a finding the study authors flagged as warranting further investigation given the biological plausibility established in animal and in vitro data (Dallaire et al., 2009). No association was found between PCP and thyroid hormone levels in the same infants at 7 months of age, indicating this effect, if real, may be limited to the perinatal period. This finding is consistent with, and adds human evidence to, the animal literature’s demonstration that PCP consistently suppresses circulating thyroid hormones.

2.8 Reproductive and Developmental Outcomes — Suggestive Evidence

A large Chinese birth cohort (1,100 mother-newborn pairs) found that higher maternal urinary PCP concentration during pregnancy was associated with lower birth weight (37 g lower per standard-deviation increase in log-transformed PCP concentration; 95% CI -64, -10) and smaller head circumference (Guo et al., 2016). A separate Chinese birth cohort study found that maternal urinary PCP concentration during pregnancy was significantly associated with altered maternal and infant sex-steroid hormone and gonadotropin concentrations, with some effects differing by infant sex, consistent with PCP’s characterization as an endocrine-disrupting compound. These findings are directionally consistent with, though smaller in magnitude than, some of the developmental effects seen in higher-dose animal studies (Section 4.3 of RoseyFields’ companion toxicology summary), and represent some of the only available human data at exposure levels within the general-population range.

2.9 Nervous System — Inconsistent, Mixed Evidence

The human neurological evidence is genuinely mixed. A study of women with long-term residential and occupational exposure to wood-preserving chemicals containing PCP and lindane found associations with neurobehavioral performance deficits (Peper et al., 1999), and case-series and poisoning-assessment literature has long described acute and subacute neurological symptoms following high-level PCP exposure (Gorman et al., 2001). By contrast, a longitudinal study of chronically exposed workers found no measurable change in peripheral nerve conduction velocity over a multi-year follow-up period (Triebig et al., 1987) — a clean null finding for a specific, objectively measured neurological endpoint. RoseyFields characterizes the human neurotoxicity evidence as inconsistent: the subjective/neurobehavioral literature shows some signal, but the more objective physiological measure (nerve conduction) does not, and most studies in this area involve small numbers of technical-grade-exposed workers, limiting the ability to separate PCP’s own contribution from that of co-occurring dioxin/furan contaminants.

2.10 Cardiovascular System — Null or Inadequate Evidence

The largest occupational mortality cohort found no significant elevation in ischemic heart disease mortality among PCP-exposed workers, including in subgroup analyses of workers with combined PCP and trichlorophenol/TCDD exposure (Ruder and Yiin, 2011). No dedicated PCP-specific cardiovascular epidemiological study with quantitative exposure-response modeling was identified beyond this mortality analysis; the available human evidence for cardiovascular effects is best characterized as null with respect to mortality, though this reflects the absence of dedicated clinical or biomarker-based cardiovascular studies rather than a well-powered negative finding across multiple study designs.

2.11 Renal Function (Non-Cancer) and Hepatic Enzymes — Inadequate Data, with Limited Suggestive Signal

Beyond the kidney cancer finding described in Section 2.2, there is little dedicated human epidemiological literature on non-malignant renal function in relation to PCP exposure specifically. For the liver, elevated serum liver enzymes have been documented in some PCP-exposed worker populations in the older clinical/occupational-health literature (Cheng et al., 1993), consistent with animal hepatotoxicity findings, but this evidence largely predates modern epidemiological methods and does not include formal exposure-response modeling comparable to the cancer studies described above.

2.12 Dermatologic Effects — Established but Confined to High-Level Occupational Exposure

Chloracne, a distinctive acneiform skin condition associated with dioxin-class compounds, has been documented in PCP-production workers (O’Malley et al., 1990). Because chloracne is understood to result primarily from the dioxin/furan contaminants of technical-grade PCP rather than PCP itself, and because it has been observed only in high-level occupational settings, RoseyFields treats this as an established but narrow occupational finding rather than a general population health concern.

3. Study Quality Considerations

Confounding by co-exposure is the most significant and most frequently discussed limitation across the PCP epidemiological literature, because technical-grade PCP inherently contains chlorinated dibenzo-p-dioxin and dibenzofuran contaminants, and because occupational PCP use very often co-occurred with exposure to other chlorophenols (particularly tetrachlorophenol) and, in some cohorts, to 2,3,7,8-TCDD from co-located trichlorophenol production (Ruder and Yiin, 2011). This is precisely why the British Columbia sawmill cohort’s exposure assessment — which specifically distinguished dermal PCP from tetrachlorophenol exposure using validated, job-specific estimates — carries particular evidentiary weight (Demers et al., 2006), and why systematic reviews specifically comparing PCP-specific findings against dioxin-cohort findings have concluded that contaminant confounding is an unlikely full explanation for the observed hematologic cancer associations (Cooper and Jones, 2008; Collins et al., 2007).

Reverse causation is a more limited concern for PCP than for some other chemicals because most of the strongest cancer findings derive from cohort studies with exposure assessed from employment records predating disease onset by years to decades, rather than from case-control designs relying on post-diagnosis exposure recall; however, some of the smaller Swedish case-control studies and biomonitoring-based studies are more susceptible to this concern, and the absence of an association in a Swedish case-control study of NHL conducted after PCP had already been banned (Hardell and Eriksson, 1999) illustrates how changes in background exposure over time complicate later case-control designs specifically.

Exposure misclassification varies considerably across study designs. Occupational cohorts using detailed job-exposure matrices validated against urinary biomarkers (Demers et al., 2006) minimize this concern relative to studies relying on job title alone or on retrospective self-report. Biomonitoring and birth-cohort studies using a single spot urine or serum PCP measurement are subject to exposure misclassification from PCP’s relatively short human half-life (on the order of 17-20 days; see RoseyFields’ companion toxicology summary) relative to the months-to-years exposure windows relevant to chronic disease outcomes, meaning a single measurement may not reliably represent long-term average exposure, particularly for outcomes like cancer with long latency periods.

Statistical power and rare-outcome precision limit several of the site-specific cancer findings. Soft-tissue sarcoma and multiple myeloma are both comparatively rare, so even the largest available cohort (Demers et al., 2006) has had limited precision for some of these specific tumor types, and several smaller studies have had almost no statistical power at all — for example, a nested case-control study within an international occupationally exposed cohort had only three PCP-exposed NHL cases, all from a single facility, precluding any generalizable conclusion from that particular substudy (Kogevinas et al., 1995).

Multiple comparisons and selective reporting are a general concern whenever a study examines many cancer sites or health outcomes simultaneously, as several of the larger cohort and ecological studies do; RoseyFields gives the greatest weight to associations that were pre-specified as hypotheses of interest (as in Ruder and Yiin, 2011, which stated a priori hypotheses based on prior human and animal findings) and that have been independently replicated in at least one other population, rather than to isolated positive findings for outcomes not otherwise implicated by the toxicological or prior epidemiological literature.

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

The International Agency for Research on Cancer (IARC) classified PCP as a Group 2B agent, “possibly carcinogenic to humans,” based on its 1991 review, and subsequently reevaluated PCP in a 2016 Monograph, at which point IARC upgraded the classification to Group 1, “carcinogenic to humans,” citing sufficient evidence in experimental animals and, notably, sufficient mechanistic evidence combined with strong supporting human evidence for non-Hodgkin’s lymphoma specifically. The US National Toxicology Program’s Report on Carcinogens has similarly concluded that PCP is “known to be a human carcinogen,” a conclusion the Report on Carcinogens profile grounded specifically in the strength and consistency of the non-Hodgkin’s lymphoma findings across the British Columbia sawmill cohort, the Michigan/US production-worker cohort, and the Swedish case-control literature, while characterizing the evidence for other cancer sites (multiple myeloma, soft-tissue sarcoma, kidney cancer) as weaker and more limited by statistical power. EPA’s own weight-of-evidence determination under its 2005 Guidelines for Carcinogen Risk Assessment similarly classifies PCP as “likely to be carcinogenic to humans,” drawing on both the animal bioassay evidence and this same body of human epidemiological findings.

RoseyFields notes that these expert bodies reached materially convergent conclusions using different evaluative frameworks and, in IARC’s case, at two different points in time separated by 25 years, during which additional human data (principally the extended follow-up of the British Columbia cohort) accumulated — a pattern of strengthening classification over time as the epidemiological record matured, rather than a static or contested classification.

5. Application of Epidemiological Evidence to Litigation Causation Opinions

RoseyFields uses the epidemiological evidence summarized above, alongside the animal toxicological and mechanistic evidence described in its companion toxicology summary, to support both general and specific causation analyses in PCP-related litigation.

For general causation — whether PCP exposure is capable of causing a given disease in humans at all — RoseyFields relies primarily on the hematologic cancer findings (Section 2.1), which satisfy several of the classical Bradford Hill considerations frequently applied in causation analysis: strength of association (approximately two-fold elevated risk in the highest exposure category), consistency (replication across the Canadian, US, and Swedish cohorts using different study designs), a monotonic exposure-response gradient (Demers et al., 2006), biological plausibility supported by the animal carcinogenicity bioassay and oxidative-metabolite mechanism described in RoseyFields’ toxicology summary, and temporality (cohort designs with exposure assessed well in advance of disease onset). For outcomes with more limited or inconsistent human evidence (soft-tissue sarcoma, cardiovascular disease, most neurological endpoints), RoseyFields’ general causation opinions rely more heavily on the animal toxicological evidence and mechanistic plausibility, with the epidemiological evidence characterized candidly as supportive but not independently sufficient.

For specific causation — whether PCP exposure caused a given disease in a particular individual — the epidemiological literature’s primary contribution is establishing the relevant background and exposed-population risk estimates that inform probabilistic causation analyses (e.g., relative risk and attributable fraction calculations), while individual case evaluation necessarily also incorporates that person’s reconstructed exposure history, latency period, competing risk factors, and differential diagnosis considerations that population-level epidemiology cannot resolve on its own. RoseyFields is careful to note, in any specific-causation analysis, which health outcomes have quantitatively modeled human dose-response data available (Section 2.1-2.2) versus which rely more heavily on qualitative human signal plus animal/mechanistic data (most other outcomes in Section 2), since courts and opposing experts frequently focus scrutiny on this distinction. Because occupational exposure histories in litigation frequently involve co-exposure to other chlorophenols, wood-treatment chemicals, and dioxin/furan contaminants, RoseyFields’ case-specific work also draws on the contaminant-confounding analyses described in Section 3 to help isolate, where the underlying exposure and job history permit, the portion of risk attributable to PCP specifically as opposed to co-occurring exposures.

References