Cadmium Epidemiology and Kidney and Lung Cancer Risk

RoseyFields evaluates the human epidemiological evidence for Cadmium as follows. This page summarizes the study populations and data sources underlying the human evidence base, organizes the full range of investigated health outcomes by strength of evidence, discusses study quality considerations relevant to weighing that evidence, and explains how RoseyFields integrates this evidence with toxicology to support causation opinions in litigation.

2. Overview of Study Populations and Data Sources

The human evidence base for cadmium is unusually broad even by the standards of well-studied metals, spanning four structurally distinct categories of study population.

Occupational cohorts

The strongest historical evidence for cadmium’s most severe effects (lung cancer, severe renal dysfunction) comes from occupational cohorts with sustained, high inhalation exposure: a U.S. cadmium recovery/production facility cohort followed from 1940 through multiple decades of re-analysis (Thun et al., 1985; Stayner et al., 1992a; Lamm et al., 1992, 1994; Sorahan and Lancashire, 1997); nickel-cadmium battery manufacturing cohorts in the United Kingdom and Sweden (Sorahan, 1987; Järup et al., 1998a); and cadmium smelter/alloy worker cohorts (Ades and Kazantzis, 1988; Kazantzis et al., 1992). These cohorts offer the exposure range and latency needed to detect cancer signals but are also the cohorts most affected by co exposure to other occupational carcinogens (particularly arsenic), a limitation discussed in detail in Section 5.

Environmental/community exposure cohorts

A second category comprises populations living in cadmium-contaminated environmental settings, most prominently residents of cadmium-polluted rice-growing regions of Japan (historically associated with Itai-Itai disease), China, and Thailand, studied for renal, skeletal, and cancer outcomes using urinary or blood cadmium as the exposure metric (Arisawa et al., 2001, 2007b; Jin et al., 2002, 2004a; Nordberg et al., 2002; Wang et al., 2003). These populations experience chronic, low-to-moderate dietary cadmium exposure — typically an order of magnitude or more below occupational inhalation exposure — and are directly relevant to community/contamination-site litigation matters.

General-population biomonitoring cohorts

A third category uses nationally representative general-population samples with cadmium biomarkers, most importantly the U.S. National Health and Nutrition Examination Survey (NHANES) and equivalent surveys in Sweden, Belgium, and Japan, examining associations between blood or urinary cadmium and outcomes including cardiovascular disease (Menke et al., 2009; Navas-Acien et al., 2005; Tellez-Plaza et al., 2008, 2010), bone density (Staessen et al., 1999; Alfvén et al., 2000, 2004; Schutte et al., 2008; Gallagher et al., 2008), renal biomarkers (Buchet et al., 1990; Chen et al., 2006a), and hormonally mediated conditions (Åkesson et al., 2008; Jackson et al., 2008). These studies extend the observed exposure-response relationship down to the general-population background range and, because they are drawn from representative samples, are the studies most directly relevant to characterizing risk in non-occupationally-exposed community and residential populations.

Clinical/case-control populations

A fourth, smaller category comprises clinical populations — most commonly infertility clinic patients — examined for reproductive and endocrine endpoints in relation to blood or seminal cadmium (Akinloye et al., 2006; Telišman et al., 2000; Jurasović et al., 2004). These studies are informative for hazard identification but, because they are drawn from a clinically selected (already infertile) population, are less directly generalizable to population-level risk than the cohort designs described above.

Using these four structurally different population types together strengthens causal inference for cadmium in a specific way: the renal and skeletal effects — cadmium’s best-established noncancer hazards — show a consistent dose-response relationship across occupational, environmental, and general-population designs spanning Europe, Asia, and North America, using a shared, validated biomarker (urinary cadmium) that permits direct quantitative comparison across studies of otherwise very different design. This cross population, cross-continent, single-biomarker consistency is considerably stronger evidence than any single study type could provide alone, and is the primary reason RoseyFields treats the renal and skeletal findings as the best-established human health effects of cadmium.

3. Health Outcomes by Body System and Strength of Evidence

The table below summarizes the full range of health outcomes investigated in the human cadmium literature, organized by RoseyFields’ assessment of evidentiary strength. Section 4 discusses the strong-evidence outcomes in greater depth.

4. Strong-Evidence Outcomes: Key Studies, Exposure Metrics, and Effect Sizes

Renal tubular dysfunction

Low-molecular-weight proteinuria — elevated urinary excretion of β2-microglobulin, retinol-binding protein, or human complex-forming protein (pHC/α1-microglobulin) — is the best-characterized and most consistently reproduced noncancer effect of cadmium exposure in humans, and is discussed in detail in RoseyFields’ companion cadmium risk assessment methodology page. Briefly, the exposure metric used across nearly all of the underlying studies is urinary cadmium (µg/g creatinine), and dose-response relationships have been independently derived and are broadly consistent across a Swedish battery-factory community cohort (Järup et al., 2000), a Belgian general-population cohort (Buchet et al., 1990), a Chinese occupational cohort (Chen et al., 2006a), and Japanese general-population cohorts (Uno et al., 2005; Suwazono et al., 2006), with adverse effect levels clustering in the range of approximately 1–10 µg cadmium/g creatinine depending on the specific biomarker, cut-off value, and population studied.

Skeletal effects

Reduced bone mineral density and increased fracture risk show a similarly consistent cross-population pattern. A prospective Belgian population study found increased risk of forearm fracture in individuals with urinary cadmium levels above approximately 0.7 µg/g creatinine (Staessen et al., 1999). Swedish cohort studies found increased osteoporosis risk at urinary cadmium levels of 1.5 µg/g creatinine or greater in men over 60 (Alfvén et al., 2000) and increased distal forearm fracture risk with continued follow-up (Alfvén et al., 2004), while a separate Swedish analysis found no evidence that the bone effect was mediated through vitamin D metabolism, supporting a direct cadmium mechanism rather than a nutritional confound (Engström et al., 2009). Chinese cohort studies in cadmium-polluted areas found decreased forearm bone density (Wang et al., 2003) and combined low bone density with renal dysfunction in the same population (Nordberg et al., 2002), while a South African/Belgian population study and a Polish population study both found dose-related increases in bone resorption markers and decreased bone density at general-population background exposure levels (Schutte et al., 2008; Trzcinka-Ochocka et al., 2010). A U.S. NHANES-based analysis similarly found an association between urinary cadmium and osteoporosis (Gallagher et al., 2008). The consistency of this finding across four countries and multiple study designs, generally at exposure levels similar to those producing renal effects, is what elevates skeletal toxicity to the strong-evidence tier notwithstanding the historical complexity of the Itai-Itai disease cases, where nutritional co-factors (multiparity, low calcium/vitamin D/protein intake) were also present and complicate interpretation of that specific historical outbreak.

5. Study Quality Considerations

RoseyFields weighs the epidemiological evidence for cadmium in light of several recurring methodological limitations that vary substantially in severity across the outcome-specific literature.

Confounding by co-exposure and smoking

Confounding by arsenic co-exposure is the single most consequential and well-documented limitation in the cadmium lung cancer literature. The U.S. cadmium production cohort — the strongest single dataset for cadmium-lung cancer association — was, for part of its operating history, also an arsenic smelter, and successive re-analyses of the identical underlying cohort reached opposite conclusions depending on how arsenic exposure was modeled: Stayner et al. (1992a) concluded cadmium was independently associated with lung cancer risk, while Lamm et al. (1992, 1994), using a nested case-control design matched on hire date as an arsenic surrogate, concluded that arsenic exposure and smoking — not cadmium — were the operative risk factors, reporting that cases were eight times more likely to have smoked than controls. A later, more detailed reanalysis using actual job-history-based arsenic exposure reconstruction (Sorahan and Lancashire, 1997) found a significant cadmium-lung cancer trend only among workers who also had arsenic exposure, leaving the question of whether cadmium has any independent lung carcinogenic effect in humans genuinely unresolved even after four decades of reanalysis of the same underlying data. Smoking is a pervasive confounder throughout the cadmium literature more broadly, since smoking is itself a major cadmium exposure source and has independent, well-established associations with lung cancer, cardiovascular disease, and reduced fertility — several reproductive-endpoint studies discussed above did not adjust for smoking status at all (Akinloye et al., 2006; Telišman et al., 2000), materially limiting their interpretability.

Reverse causation

Reverse causation is a limited concern for the renal and skeletal endpoints, since there is no plausible biological pathway by which incipient renal or bone disease would increase cadmium absorption or urinary excretion. It is a more relevant concern for some of the endocrine/reproductive findings — for example, in infertility-clinic populations, where altered hormone levels or metabolic status conceivably could affect cadmium absorption, distribution, or excretion independent of any causal effect of cadmium on fertility.

Exposure misclassification

Exposure misclassification varies by biomarker and by population. Urinary cadmium is a well-validated, quantitatively reliable biomarker of cumulative body burden in the dose range below overt nephrotoxicity, and its use across the great majority of cadmium epidemiology studies is a genuine methodological strength relative to many other environmental chemicals lacking any validated exposure biomarker. However, at very high renal cadmium burdens associated with overt renal damage, the relationship between urinary cadmium and body burden becomes nonlinear (elimination half-time increases with increasing renal damage severity), meaning urinary cadmium can understate cumulative exposure specifically in the most heavily exposed, already-affected individuals — a limitation relevant to interpreting the upper end of dose-response curves in heavily contaminated environmental cohorts. Occupational cohort studies relying on historical air concentration measurements (rather than biomarkers) face larger exposure reconstruction uncertainty, given documented order-of-magnitude declines in workplace air cadmium concentrations over the study periods (e.g., from 1.53 mg/m³ in 1976 to 0.030 mg/m³ in 1990 at one facility).

Diagnostic and biomarker heterogeneity

A wide range of cut-off values (233–1,129 µg/g creatinine for urinary β2-microglobulin alone) has been used across studies to define “abnormal” biomarker levels, and different studies have used different renal biomarkers (β2-microglobulin, retinol-binding protein, pHC, N-acetyl-β-glucosaminidase) with different sensitivities to tubular damage, complicating direct comparison of adverse effect levels across studies even where the underlying urinary cadmium exposure metric is identical. RoseyFields’ risk assessment methodology page discusses how ATSDR addressed this heterogeneity through a formal meta-analysis of dose-response functions rather than relying on any single study or cut-off value.

Statistical power and study design limitations for cancer endpoints

Outside of lung cancer, most cadmium-cancer epidemiology involves single studies of specific tumor types (prostate, bladder, breast, endometrial, pancreatic) with limited case numbers and no independent replication, meaning that any individual positive or null finding should be weighted cautiously pending confirmation in additional populations. Several of the community/environmental cohort studies of cadmium polluted areas of Japan found no excess in overall or site-specific cancer mortality despite clearly elevated exposure biomarkers, a pattern of non-replication that RoseyFields considers material context when any single positive cancer-site finding is offered in isolation.

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

EPA and IARC have reached materially different weight-of-evidence classifications for cadmium carcinogenicity based on substantially the same underlying dataset. IARC classifies cadmium as Group 1, carcinogenic to humans, based on its determination that there is sufficient evidence of carcinogenicity in both human and animal studies; the U.S. Department of Health and Human Services (National Toxicology Program) has likewise classified cadmium and certain cadmium compounds as known human carcinogens. EPA’s IRIS assessment, by contrast, classifies cadmium as Group B1, a probable (not known) human carcinogen, based on its own characterization of the human evidence as “limited” rather than sufficient — citing the relatively low standardized mortality ratios observed and, most importantly, the unresolved arsenic-confounding problem discussed in Section 5. This divergence is not a matter of one agency having access to different data; it reflects a genuine difference in how much weight two expert bodies placed on the same conflicting reanalyses of the U.S. cadmium worker cohort. RoseyFields discloses both classifications and their underlying rationale in any matter where cadmium’s carcinogenic classification is material, rather than presenting either agency’s conclusion as the uncontested scientific consensus.

7. Use of Epidemiological Evidence Alongside Toxicology to Support Causation Opinions

RoseyFields integrates the epidemiological evidence summarized above with the toxicological and mechanistic evidence (addressed in RoseyFields’ companion cadmium toxicology and risk assessment methodology pages) to support both general and specific causation analyses.

For general causation of renal tubular dysfunction and skeletal effects, RoseyFields relies on the cross population, cross-continent consistency described in Sections 2 and 4 — a body of evidence considerably stronger and more directly quantitative than is available for most environmental chemicals, because cadmium’s validated urinary biomarker allows direct comparison of dose-response relationships across otherwise very different study populations and designs. This is the strongest evidentiary foundation available to RoseyFields for a cadmium general causation opinion and is the basis for the ATSDR chronic oral and inhalation MRLs described in RoseyFields’ risk assessment methodology page.

For lung cancer general causation, RoseyFields’ position is necessarily more qualified, reflecting the genuine, unresolved scientific disagreement between EPA and IARC and the documented arsenic-confounding problem in the single strongest underlying cohort. RoseyFields presents this as an area of legitimate scientific dispute rather than settled causation, and specifically flags for any testifying expert that reliance on the Thun et al. (1985)/Stayner et al. (1992a) cadmium-lung cancer association should address, rather than ignore, the contrary Lamm et al. (1992, 1994) and Sorahan and Lancashire (1997) reanalyses of the same cohort.

For specific causation, RoseyFields uses the quantitative urinary-cadmium dose-response relationships from Section 4, combined with the biomarker-based dose reconstruction methodology described in RoseyFields’ risk assessment page, to evaluate whether a claimant’s estimated or measured cadmium body burden falls within the range associated with elevated risk in the epidemiological literature — an approach that is unusually well-supported for cadmium relative to chemicals lacking a validated cumulative-exposure biomarker. Where a claimed outcome falls in the suggestive-but-inconsistent or null tiers (Section 3), RoseyFields’ causation opinions are correspondingly more qualified, and RoseyFields explicitly identifies the specific sources of inconsistency (co-exposure confounding, smoking, small case numbers, lack of replication) that limit confidence in any given finding.

References