Does Diesel Exhaust Cause Human Lung Cancer?

Recent reviews of epidemiological evidence on the relation between exposure to diesel exhaust (DE) and lung cancer risk have reached conflicting conclusions, ranging from belief that there is sufficient evidence to conclude that DE is a human lung carcinogen (California EPA, 1994) to conclusions that there is inadequate evidence to support a causal association between DE and human lung cancer (Muscat and Wynder, 1995). Individual studies also conflict, with both increases and decreases in relative risks of lung cancer mortality being cited with 95% statistical confidence. On balance, reports of elevated risk outnumber reports of reduced risk. This paper reexamines the evidence linking DE exposures to lung cancer risk. After briefly reviewing animal data and biological mechanisms, it surveys the relevant epidemiological literature and examines possible explanations for the discrepancies. These explanations emphasize the distinction between statistical associations, which have been found in many studies, and causal associations, which appear not to have been established. Methodological threats to valid causal inference are identified and new approaches for controlling them are proposed using recent techniques from artificial intelligence (AI) and computational statistics. These threats have not been adequately controlled for in previous epidemiological studies. They provide plausible noncausal explanations for the reported increases in relative risks, making it impossible to infer causality between DE exposure and lung cancer risk from these studies. A key contribution is to show how recent techniques developed in the AI-and-statistics literature can help clarify the causal interpretation of complex multivariate data sets used in epidemiological risk assessments. Applied to the key study of Garshick et al. (1988), these methods show that DE concentration has no positive causal association with occupational lung cancer mortality risk.     

Statistical Issues in the Estimation of Assigned Shares for Carcinogenesis Liability

Congress is currently considering adopting a mathematical formula to assign shares in cancer causation to specific doses of radiation, for use in establishing liability and compensation awards. The proposed formula, if it were sound, would allow difficult problems in tort law and public policy to be resolved by reference to tabulated "probabilities of causation." This article examines the statistical and conceptual bases for the proposed methodology. We find that the proposed formula is incorrect as an expression for "probability and causation," that it implies hidden, debatable policy judgments in its treatment of factor interactions and uncertainties, and that it can not in general be quantified with sufficient precision to be useful. Three generic sources of statistical uncertainty are identified--sampling variability, population heterogeneity, and error propagation--that prevent accurate quantification of "assigned shares." These uncertainties arise whenever aggregate epidemiological or risk data are used to draw causal inferences about individual cases.     

Probability of Causation and the Attributable Proportion Risk

A uranium miner who smokes develops lung cancer: what is the probability that radiation, rather than tobacco, caused it? This paper briefly explains the principles and limits of probability models for which this question makes sense, and then shows how principles of risk accounting can be applied to obtain a solution to the general problem of attributing risk in the presence of joint, possibly interacting, causes. A procedure for calculating each factor's “share” in a jointly caused risk is proposed, and shown to be a generalization of the “probability of causation” concept. Problems of implementation and interpretation for the proposed attribution procedure are discussed, and illustrative error bounds are derived for a simple decision rule, in which probability of causation or attributable risk share calculations are made using aggregate data as a proxy for unknown individual data.     

From Ideal to Real Risk: Philosophy of Causation Meets Risk Analysis

A question has been raised in recent years as to whether the risk field, including analysis, assessment, and management, ought to be considered a discipline on its own. As suggested by Terje Aven, unification of the risk field would require a common understanding of basic concepts, such as risk and probability; hence, more discussion is needed of what he calls “foundational issues.” In this article, we show that causation is a foundational issue of risk, and that a proper understanding of it is crucial. We propose that some old ideas about the nature of causation must be abandoned in order to overcome certain persisting challenges facing risk experts over the last decade. In particular, we discuss the challenge of including causally relevant knowledge from the local context when studying risk. Although it is uncontroversial that the receptor plays an important role for risk evaluations, we show how the implementation of receptor‐based frameworks is hindered by methodological shortcomings that can be traced back to Humean orthodoxies about causation. We argue that the first step toward the development of frameworks better suited to make realistic risk predictions is to reconceptualize causation, by examining a philosophical alternative to the Humean understanding. In this article, we show how our preferred account, causal dispositionalism, offers a different perspective in how risk is evaluated and understood.     

Variance Estimators for Three "Probabilities of Causation"

This article introduces the definitions of three "probabilities of causation" suggested by Pearl (1999), which are used to evaluate the causal effect of an exposure on a disease in epidemiological studies. Pearl (1999) and Tian and Pearl (2000a, 2000b) provided identification formulas for three "probabilities of causation" from statistical data under some assumptions. In order to examine the estimation accuracy problem, this article derives variance estimators for three "probabilities of causation" correspondent to each case in Pearl (1999) and at the same time clarify their properties. In addition, we conduct simulation experiments and show that the proposed method can approximate sufficiently to the variance of "probabilities of causation." The results of this article provide a complete framework for using "probabilities of causation" effectively in order to analyze responsibility and susceptibility in epidemiological studies.     

The Cancer Risk Associated with Residential Exposure to Soil Containing Radioactive Coal Combustion Residuals

Coal combustion residuals (CCRs) are composed of various constituents, including radioactive materials. The objective of this study was to utilize methodology on radionuclide risk assessment from the Environmental Protection Agency (EPA) to estimate the potential cancer risks associated with residential exposure to CCR‐containing soil. We evaluated potential radionuclide exposure via soil ingestion, inhalation of soil particulates, and external exposure to ionizing radiation using published CCR radioactivity values for ²³²Th, ²²⁸Ra, ²³⁸U, and ²²⁶Ra from the Appalachia, Illinois, and Powder River coal basins. Mean and upper‐bound cancer risks were estimated individually for each radionuclide, exposure pathway, and coal basin. For each radionuclide at each coal basin, external exposure to ionizing radiation contributed the greatest to the overall risk estimate, followed by incidental ingestion of soil and inhalation of soil particulates. The mean cancer risks by route of exposure were 2.01 × 10⁻⁶ (ingestion), 6.80 × 10⁻⁹ (inhalation), and 3.66 × 10⁻⁵ (external), while the upper bound cancer risks were 3.70 × 10⁻⁶ (ingestion), 1.18 × 10⁻⁸ (inhalation), and 6.15 × 10⁻⁵ (external), using summed radionuclide‐specific data from all locations. The upper bound cancer risk from all routes of exposure was 6.52 × 10⁻⁵. These estimated cancer risks were within the EPA's acceptable cancer risk range of 1 × 10⁻⁶ to 1 × 10⁻⁴. If the CCR radioactivity values used in this analysis are generally representative of CCR waste streams, then our findings suggest that CCRs would not be expected to pose a significant radiological risk to residents living in areas where contact with CCR‐containing soils might occur.     

Could Removing Arsenic from Tobacco Smoke Significantly Reduce Smoker Risks of Lung Cancer?

If a specific biological mechanism could be determined by which a carcinogen increases lung cancer risk, how might this knowledge be used to improve risk assessment? To explore this issue, we assume (perhaps incorrectly) that arsenic in cigarette smoke increases lung cancer risk by hypermethylating the promoter region of gene p16INK4a, leading to a more rapid entry of altered (initiated) cells into a clonal expansion phase. The potential impact on lung cancer of removing arsenic is then quantified using a three‐stage version of a multistage clonal expansion (MSCE) model. This refines the usual two‐stage clonal expansion (TSCE) model of carcinogenesis by resolving its intermediate or “initiated” cell compartment into two subcompartments, representing experimentally observed “patch” and “field” cells. This refinement allows p16 methylation effects to be represented as speeding transitions of cells from the patch state to the clonally expanding field state. Given these assumptions, removing arsenic might greatly reduce the number of nonsmall cell lung cancer cells (NSCLCs) produced in smokers, by up to two‐thirds, depending on the fraction (between 0 and 1) of the smoking‐induced increase in the patch‐to‐field transition rate prevented if arsenic were removed. At present, this fraction is unknown (and could be as low as zero), but the possibility that it could be high (close to 1) cannot be ruled out without further data.     

Methodology for Determining the Appropriateness of a Linear Dose‐Response Function

Microbial food safety risk assessment models can often at times be simplified by eliminating the need to integrate a complex dose‐response relationship across a distribution of exposure doses. This is possible if exposure pathways lead to pathogens at exposure that consistently have a small probability of causing illness. In this situation, the probability of illness will follow an approximately linear function of dose. Consequently, the predicted probability of illness per serving across all exposures is linear with respect to the expected value of dose. The majority of dose‐response functions are approximately linear when the dose is low. Nevertheless, what constitutes “low” is dependent on the parameters of the dose‐response function for a particular pathogen. In this study, a method is proposed to determine an upper bound of the exposure distribution for which the use of a linear dose‐response function is acceptable. If this upper bound is substantially larger than the expected value of exposure doses, then a linear approximation for probability of illness is reasonable. If conditions are appropriate for using the linear dose‐response approximation, for example, the expected value for exposure doses is two to three logs₁₀ smaller than the upper bound of the linear portion of the dose‐response function, then predicting the risk‐reducing effectiveness of a proposed policy is trivial. Simple examples illustrate how this approximation can be used to inform policy decisions and improve an analyst's understanding of risk.     

Transmissible Spongiform Encephalopathy Risk Assessment: The UK experience

Transmissible spongiform encephalopathy (TSE) risk assessments undertaken in the United Kingdom have mainly had the objective of determining the risks posed to humans from exposure to the causal agents associated with bovine spongiform encephalopathy (BSE) and variant Creutzfeld-Jakob disease (vCJD). In this article, I examine 19 of these risk assessments published to date and consider how their results might be influenced by underlying model assumptions and methodology. Three separate aspects common to all the assessments are infective load estimation, exposure pathway identification, and risk estimation. These are each discussed in detail.     

Bounding Poorly Characterized Risks: A Lung Cancer Example

For diseases with more than one risk factor, the sum of probabilistic estimates of the number of cases caused by each individual factor may exceed the total number of cases observed, especially when uncertainties about exposure and dose response for some risk factors are high. In this study, we outline a method of bounding the fraction of lung cancer fatalities not due to specific well-studied causes. Such information serves as a "reality check" for estimates of the impacts of the minor risk factors, and, as such, complements the traditional risk analysis. With lung cancer as our example, we allocate portions of the observed lung cancer mortality to known causes (such as smoking, residential radon, and asbestos fibers) and describe the uncertainty surrounding those estimates. The interactions among the risk factors are also quantified, to the extent possible. We then infer an upper bound on the residual mortality due to "other" causes, using a consistency constraint on the total number of deaths, the maximum uncertainty principle, and the mathematics originally developed of imprecise probabilities.     

Multi-Stage Model Estimates of Lung Cancer Risk from Exposure to Diesel Exhaust, Based on a U.S. Railroad Worker Cohort

A California Environmental Protection Agency (Cal/EPA) report concluded that a reasonable and likely explanation for the increased lung cancer rates in numerous epidemiological studies is a causal association between diesel exhaust exposure and lung cancer. A version of the present analysis, based on a retrospective study of a U.S. railroad worker cohort, provided the Cal/EPA report with some of its estimates of lung cancer risk associated with diesel exhaust. The individual data for that cohort study furnish information on age, employment, and mortality for 56,000 workers over 22 years. Related studies provide information on exposure concentrations. Other analyses of the original cohort data reported finding no relation between measures of diesel exhaust and lung cancer mortality, while a Health Effects Institute report found the data unsuitable for quantitative risk assessment. None of those three works used multistage models, which this article uses in finding a likely quantitative, positive relations between lung cancer and diesel exhaust. A seven-stage model that has the last or next-to-last stage sensitive to diesel exhaust provides best estimates of increase in annual mortality rate due to each unit of concentration, for bracketing assumptions on exposure. Using relative increases of risk and multiplying by the background lung cancer mortality rates for California, the 95% upper confidence limit of the 70-year unit risks for lung cancer is estimated to be in the range 2.1 x 10(-4) (microg/m3)(-1) to 5.5 x 10(-4) (microg/m3)(-1). These risks constitute the low end of those in the Cal/EPA report and are below those reported by previous investigators whose estimates were positive using human data.     

OSHA''s Permissible Exposure Limits: Regulatory Compliance Versus Health Risk

Workplace exposures to airborne chemicals are regulated in the U.S. by the Occupational Safety and Health Administration (OSHA) via the promulgation of permissible exposure limits (PELs). These limits, usually defined as eight-hour time-weighted average values, are enforced as concentrations never to be exceeded. In the case of chronic or delayed toxicants, the PEL is determined from epidemiological evidence and/or quantitative risk assessments based on long-term mean exposures or, equivalently, cumulative lifetime exposures. A statistical model was used to investigate the relation between the compliance strategy, the PEL as a limit never to be exceeded, and the health risk as measured by the probability that an individual's long-term mean exposure concentration is above the PEL. The model incorporates within-worker and between-worker variability in exposure, and assumes the relevant distributions to be log-normal. When data are inadequate to estimate the parameters of the full model, as it is in compliance inspections, it is argued that the probability of a random measurement being above the PEL must be regarded as a lower bound on the probability that a randomly selected worker's long-term mean exposure concentration will exceed the PEL. It is concluded that OSHA's compliance strategy is a reasonable, as well as a practical, means of limiting health risk for chronic or delayed toxicants.     

Modeling the Modification of the Risk of Radon-Induced Lung Cancer by Environmental Tobacco Smoke

The presence of environmental tobacco smoke (ETS) in homes has been implicated in the causation of lung cancer. While of interest in its own right, ETS also influences the risk imposed by radon and its decay products. The interaction between radon progeny and ETS alters the exposure, intake, uptake, biokinetics, dosimetry, and radiobiology of those progeny. The present paper details model predictions of the various influences of ETS on these factors in the U.S. population and provides estimates of the resulting change in the risk from average levels of radon progeny. It is predicted that the presence of ETS produces a very small (perhaps unmeasurable) increase in the risk of radiation-induced tracheobronchial cancer in homes with initially very high particle concentrations for both active and never-smokers, but significantly lowers the risk in homes with initially lower particle concentrations for both groups when generation 4 of the lung is considered the target site. For generation 16, the presence of ETS generally increases the radon-induced risk of lung cancer, although the increase should be unmeasurable at high initial particle concentrations. The net effect of ETS on human health is suggested to be a complicated function of the initial housing conditions, the concentration of particles introduced by smoking, the target generation considered, and the smoking status of exposed populations. This situation precludes any simple statements concerning the role of ETS in governing the incidence of lung cancer in a population.     

Premature Deaths,Statistical Lives,and Years of Life Lost: Identification,Quantification, and Valuation of Mortality Risks

Mortality effects of exposure to air pollution and other environmental hazards are often described by the estimated number of “premature” or “attributable” deaths and the economic value of a reduction in exposure as the product of an estimate of “statistical lives saved” and a “value per statistical life.” These terms can be misleading because the number of deaths advanced by exposure cannot be determined from mortality data alone, whether from epidemiology or randomized trials (it is not statistically identified). The fraction of deaths “attributed” to exposure is conventionally derived as the hazard fraction (R – 1)/R, where R is the relative risk of mortality between high and low exposure levels. The fraction of deaths advanced by exposure (the “etiologic” fraction) can be substantially larger or smaller: it can be as large as one and as small as 1/e (≈0.37) times the hazard fraction (if the association is causal and zero otherwise). Recent literature reveals misunderstanding about these concepts. Total life years lost in a population due to exposure can be estimated but cannot be disaggregated by age or cause of death. Economic valuation of a change in exposure-related mortality risk to a population is not affected by inability to know the fraction of deaths that are etiologic. When individuals facing larger or smaller changes in mortality risk cannot be identified, the mean change in population hazard is sufficient for valuation; otherwise, the economic value can depend on the distribution of risk reductions.     

A regression-based approach for estimating primary and secondary particulate matter intake fractions.

One of the common challenges for life cycle impact assessment and risk assessment is the need to estimate the population exposures associated with emissions. The concept of intake fraction (a unitless term representing the fraction of material or its precursor released from a source that is eventually inhaled or ingested) can be used when limited site data are available or the number of sources to model is large. Although studies have estimated intake fractions for some pollutant-source combinations, there is a need to quickly and accurately estimate intake fractions for sources and settings not previously evaluated. It would be expected that limited source or site information could be used to yield intake fraction estimates with reasonable accuracy. To test this theory, we developed regression models to predict intake fractions previously estimated for primary fine particles (PM2.5) and secondary sulfate and nitrate particles from power plants and mobile sources in the United States. Our regression models were able to predict pollutant-specific intake fractions with R2 between 0.53 and 0.86 and equations that reflected expected relationships (e.g., intake fraction increased with population density, stack height influenced the intake fraction of primary but not secondary particles). Further analysis would be needed to generalize beyond this case study and construct models applicable across source categories and settings, but our analysis demonstrates that inclusion of a limited number of parameters can significantly reduce the uncertainty in population-average exposure estimates.     

Quantitative Links Between Arsenic Exposure and Influenza A (H1N1) Infection‐Associated Lung Function Exacerbations Risk

The objective of this study was to link arsenic exposure and influenza A (H1N1) infection‐induced respiratory effects to assess the impact of arsenic‐contaminated drinking water on exacerbation risk of A (H1N1)‐associated lung function. The homogeneous Poisson process was used to approximate the related processes between arsenic exposure and influenza‐associated lung function exacerbation risk. We found that (i) estimated arsenic‐induced forced expiratory volume in 1 second (FEV₁) reducing rates ranged from 0.116 to 0.179 mL/μg for age 15–85 years, (ii) estimated arsenic‐induced A (H1N1) viral load increasing rate was 0.5 mL/μg, (iii) estimated A (H1N1) virus‐induced FEV₁ reducing rate was 0.10 mL/logTCID50, and (iv) the relationship between arsenic exposure and A (H1N1)‐associated respiratory symptoms scores (RSS) can be described by a Hill model. Here we showed that maximum RSS at day 2 postinfection for Taiwan, West Bengal (India), and the United States were estimated to be in the severe range of 0.83, 0.89, and 0.81, respectively, indicating that chronic arsenic exposure and A (H1N1) infection together are most likely to pose potential exacerbations risk of lung function, although a 50% probability of lung function exacerbations risk induced by arsenic and influenza infection was within the mild and moderate ranges of RSS at day 1 and 2 postinfection. We concluded that avoidance of drinking arsenic‐containing water could significantly reduce influenza respiratory illness and that need will become increasingly urgent as the novel H1N1 pandemic influenza virus infects people worldwide.     

Environmental Tobacco Smoke Revisited: The Reliability of the Data Used for Risk Assessment

Several epidemiological studies have found a weak, but consistent association between lung cancer in nonsmokers and exposure to environmental tobacco smoke (ETS). In addition, a purported link between such exposure and coronary heart disease (CHD) has been of major concern. Although it is biologically plausible that ETS has a contributory role in the induction of lung cancer in nonsmoking individuals, dose-response extrapolation-supported by the more solid database for active smokers-gives an additional risk for lung cancer risk that is more than one order of magnitude lower than that indicated by major positive epidemiological studies. The discrepancy between available epidemiological data and dosimetric estimates seems, to a major part, to reflect certain systematic biases in the former that are difficult to control by statistical analysis when dealing with risks of such low magnitudes. These include, most importantly, misclassification of smoking status, followed by inappropriate selection of controls, as well as certain confounding factors mainly related to lifestyle, and possibly also hereditary disposition. A significant part of an association between lung cancer and exposure to ETS would disappear, if, on the average, 1 patient out of 20 nonsmoking cases had failed to tell the interviewer that he had, in fact, recently stopped smoking. In the large International Agency for Research on Cancer (IARC) multicenter study even lower misclassification rates would abolish the weak, statistically nonsignificant associations that were found. In the former study an apparent significant protective effect from exposure to ETS in childhood with respect to lung cancer later in life was reported, a most surprising finding. The fact that the mutation spectrum of the p53 tumor suppressor gene in lung tumors of ETS-exposed nonsmokers generally differs from that found in tumors of active smokers lends additional support to the notion that the majority of tumors found in ETS-exposed nonsmokers have nothing to do with tobacco smoke. The one-sided preoccupation with ETS as a causative factor of lung cancer in nonsmokers may seriously hinder the elucidation of the multifactorial etiology of these tumors. Due to the high prevalence of cardiovascular disease in the population, even a modest causal association with ETS would, if valid, constitute a serious public health problem. By pooling data from 20 published studies on ETS and heart disease, some of which reported higher risks than is known to be caused by active smoking, a statistically significant association with spousal smoking is obtained. However, in most of these studies, many of the most common confounding risk factors were ignored and there appears to be insufficient evidence to support an association between exposure to ETS and CHD. Further, it seems highly improbable that exposure to a concentration of tobacco smoke at a level that is generally much less than 1% of that inhaled by a smoker could result in an excess risk for CHD that-as has been claimed-is some 30% to 50% of that found in active smokers. There are certainly valid reasons to limit exposure to ETS as well as to other air pollutants in places such as offices and homes in order to improve indoor air quality. This goal can be achieved, however, without the introduction of an extremist legislation based on a negligible risk of lung cancer as well as an unsupported and highly hypothetical risk for CHD.     

Quantitative Assessment of the Risk of Lung Cancer Associated with Occupational Exposure to Refractory Ceramic Fibers

We present the results of a quantitative assessment of the lung cancer risk associated with occupational exposure to refractory ceramic fibers (RCF). The primary sources of data for our risk assessment were two long-term oncogenicity studies in male Fischer rats conducted to assess the potential pathogenic effects associated with prolonged inhalation of RCF. An interesting feature of the data was the availability of the temporal profile of fiber burden in the lungs of experimental animals. Because of this information, we were able to conduct both exposure–response and dose–response analyses. Our risk assessment was conducted within the framework of a biologically based model for carcinogenesis, the two-stage clonal expansion model, which allows for the explicit incorporation of the concepts of initiation and promotion in the analyses. We found that a model positing that RCF was an initiator had the highest likelihood. We proposed an approach based on biological considerations for the extrapolation of risk to humans. This approach requires estimation of human lung burdens for specific exposure scenarios, which we did by using an extension of a model due to Yu. Our approach acknowledges that the risk associated with exposure to RCF depends on exposure to other lung carcinogens. We present estimates of risk in two populations: (1) a population of nonsmokers and (2) an occupational cohort of steelworkers not exposed to coke oven emissions, a mixed population that includes both smokers and nonsmokers.     

Improving Causal Inferences in Risk Analysis

Recent headlines and scientific articles projecting significant human health benefits from changes in exposures too often depend on unvalidated subjective expert judgments and modeling assumptions, especially about the causal interpretation of statistical associations. Some of these assessments are demonstrably biased toward false positives and inflated effects estimates. More objective, data‐driven methods of causal analysis are available to risk analysts. These can help to reduce bias and increase the credibility and realism of health effects risk assessments and causal claims. For example, quasi‐experimental designs and analysis allow alternative (noncausal) explanations for associations to be tested, and refuted if appropriate. Panel data studies examine empirical relations between changes in hypothesized causes and effects. Intervention and change‐point analyses identify effects (e.g., significant changes in health effects time series) and estimate their sizes. Granger causality tests, conditional independence tests, and counterfactual causality models test whether a hypothesized cause helps to predict its presumed effects, and quantify exposure‐specific contributions to response rates in differently exposed groups, even in the presence of confounders. Causal graph models let causal mechanistic hypotheses be tested and refined using biomarker data. These methods can potentially revolutionize the study of exposure‐induced health effects, helping to overcome pervasive false‐positive biases and move the health risk assessment scientific community toward more accurate assessments of the impacts of exposures and interventions on public health.