Selecting experimental data for use in quantitative risk assessment: an expert-judgement approach

In many cancer risk assessments the experimental data used in statistical modeling are selected by applying generic guidelines. The guidelines exclude use of some types of experimental data and often appear arbitrary since rules rather than scientific judgments guide selection of data. This paper implements an alternative approach in which data are selected based on the judgments of practicing scientists. Eight such scientists were identified through an explicit selection procedure to help select data for use in a dose-response assessment of formaldehyde. Judgements about appropriate data sets were then elicited in personal interviews using a formal interview protocol. Appropriate data sets were fit to the multistage model and used as the basis for low-dose extrapolation. Low-dose risk estimates are shown to be sensitive to the selection of data, especially the treatment of benign tumors. The recommendations of the experts also differ in some respects from the choices made in previously published risk assessments. This suggests that scientific judgement may be an appropriate method to augment guidelines when a broad range of data is available. The paper argues that the expert judgment approach has some advantages that are worth considering.     

生物学机制模型在定量风险评估中的应用

对有毒有害化学物进行风险评估，是制定限量标准和保障人体健康的基础。近年来，生物学机制模型在风险评估领域中得到越来越广泛的应用。根据解剖、生物化学、生理、代谢动力学等知识及毒作用机制建立的模型，不仅能够定量地评定化学物毒作用的剂量-反应（效应）关系，解决低剂量暴露风险评估问题，也在很大程度上降低了将结果从动物外推到人的不确定性，提高了风险评估的可信度。本文主要综述报道生物学机制模型的特点及在定量风险评估中的应用。     

Projections of leukemia risk associated with occupational exposure to benzene

In 1982, White et al published an assessment of quantitative leukemia risk associated with lifetime occupational exposure to benzene. At about the same time, IARC (1982) published estimates of quantitative cancer risk associated with industrial chemicals. Benzene was one of the two chemicals selected by IARC for its risk estimation. This paper presents a summary of these assessments along with new study results demonstrating adverse effects on bone marrow and peripheral blood cells as a result of low-level benzene exposure. Mathematical extrapolations based on epidemiologic studies are consistent with a finding of significant risk of dying from leukemia under the current occupational permissible exposure limit of 10 ppm. Although a significant reduction of risk could be expected to be achieved by reducing exposure to 1 ppm, a significant risk may still remain. The uncertainty of the dose-response projections rests on the underlying estimates of relative risk of death from leukemia, the estimates of benzene exposure (dose), and the appropriateness of the mathematical model. Recent findings in experimental animals demonstrate chromosomal damage to bone marrow cells, significant depression of the bone marrow, and disturbances of immune system function as a result of less than 1 week of exposure to the current permissible benzene exposure limit of 10 ppm. This was the lowest dose tested. These experimental findings provide further evidence of a potentially significant risk of bone marrow proliferative cancer (leukemia) as a result of low-dose benzene exposure.     

基于剂量-反应模型的苯职业暴露健康风险定量评价方法研究

目的 初步建立一种评价职业暴露导致的健康风险的定量方法,为我国职业健康风险的定量研究提供理论依据.方法 根据健康风险评价的基本方法,利用剂量-反应模型计算个体暴露于剂量为d的某物质所导致的致癌概率.暴露物质的人体内剂量与致癌概率间的函数关系可用probit模型、logit模型、Weibull模型、单击模型、多击模型和多阶模型来表达.通过药代动力学分析或人类流行病学数据拟合,可以得到暴露剂量与人体内剂量间的函数关系,将其带入剂量-反应模型,即可得出不同环境浓度下暴露人群的致癌风险.结果 根据测得的环境苯浓度值,分别用2阶模型、单击模型和Weibull模型计算得到的剂量-反应曲线均为直线,2阶模型得出的致癌风险最高,Weibull模型得出的致癌风险最小,苯的平均浓度为7.25 mg/m3时,3个模型对应的风险值分别为7.50×10-4、1.83×10-4、2.31×10-6.结论 初步建立了一种定量评价职业健康风险的方法,可利用此方法对工作场所中的有害物质暴露情况进行分析,得出工人在此环境下的致癌风险.     

Assessing the cancer risk from foods

Assessing the carcinogenicity of a compound and then determining what dosage is appropriate for human beings is a complex process. Carcinogens act either by altering deoxyribonucleic acid (DNA) or by promoting the growth of already altered cells. Carcinogenicity is evaluated with the use of structural analysis, in vitro mutagenesis assays, epidemiological findings, and dose-response studies in laboratory animals. In the animal studies, high doses are administered. Once a compound is found to be carcinogenic, the dose that will pose an acceptable cancer risk to human beings--one in a million--must be extrapolated from the high-dose data. This virtually safe dose (VSD) will be the allowable dosage for human contact. The extrapolation from high-dose animal studies to a VSD for human beings is based on the models for carcinogenic mechanisms. Debate exists as to how many interactions with DNA the carcinogen must have to initiate neoplastic growth and whether there exists a threshold for carcinogenic action below which there is no risk of cancer. The extrapolation model that is chosen greatly affects the VSD. Knowing how this extrapolation to a VSD is done will help dietitians better understand how allowable levels for carcinogens in foods are determined.     

Estimates of lifetime lung cancer risks resulting from Rn progeny exposure

Data on five mining populations exposed to Rn progeny have been used to estimate the lifetime risk of lung cancer resulting from occupational and environmental exposure under current standards. Slopes of dose-response relations for lung cancer show a tendency to decrease with increasing dose. Our best estimate of curvilinearity is given by raising dose to the power 0.92 +/- 0.07, but the improvement in fit beyond simple linearity is not significant. On the other hand, the addition of a cell-killing term significantly improves the fit of the linear model. In any event, linear extrapolation is unlikely to underestimate the excess risk at low doses by more than a factor of 1.5. However, these inferences about curvilinearity are highly subject to error from the choice of reference populations, dosimetry, and latency. Under the linear-cell-killing model, our best estimate of excess relative risk is 2.28 +/- 0.35 per 100 working level month (WLM) (a doubling dose of 44 WLM). Attributable risks in these five studies range from 3.4-17.8 per 10(6) person-yr WLM-1. Risks from Rn progeny appear to interact with age and smoking in a form intermediate between additive and multiplicative. The "relative risk" model is therefore preferable for projecting lifetime risks, but life-table projections are described for a wide variety of assumptions. Our best estimate of the effect of a 50-yr occupational exposure to 4 WLM yr-1 is 130 excess lung cancer deaths per 1000 persons (0.65 per 1000 person-WLM), with a range from 60-250 per 1000. Similar calculations for lifetime exposure to an additional 0.02 working level (WL) beyond normal background produces an estimate of 20 excess lung cancers per 1000 persons.     

老年营养风险指数与老年癌症及非癌症患者住院时间的关系研究

目的 应用老年营养风险指数(GNRI)评估患者的营养风险,探讨其与老年癌症及非癌症患者住院时间的关系。方法 选择老年住院患者37 267例,将其分为癌症组和非癌症组,应用GNRI评估患者入院时的营养风险;以患者死亡及出院为观察终点,住院时间(d)作为临床结局指标,采用边际结构模型探讨GNRI与老年癌症及非癌症患者住院时间的关联性。结果 超过一半(56.3%)的老年住院患者具有不同程度的营养风险;与非癌症患者相比,癌症患者GNRI水平(91.0±10.2)及无营养风险患者的比例(26.8%)更低,而具有低(24.6%)、中(29.0%)、高(19.7%)营养风险患者的比例均更高,差异均具有统计学意义(P＜0.05)。在控制其他混杂因素后,边际结构模型分析结果显示在癌症和非癌症患者中,住院时间均随营养风险程度的升高而延长,具有高营养风险的癌症患者住院时间最长,高达19.1(95% CI:17.5~20.8)d;在不同的营养风险分组中,癌症患者住院时间(14.5~19.1 d)均高于非癌症患者(10.1~15.2 d),差异均具有统计学意义(P＜0.05)。结论 GNRI 适用于老年癌症及非癌症患者的营养风险筛查评估,由GNRI评估的营养风险越高,患者住院时间越长。     

Effects of dinoseb on the life cycle of Daphnia magna: modeling survival time and a proposal for an alternative to the no-observed-effect concentration

Risk assessment is in urgent need of more accurate toxic effect endpoints than those currently in use, especially for low concentrations. Often such endpoints are estimated by analysis of variance, linear interpolation, or smoothing. As these statistical methods are not always satisfactory, some authors have proposed to describe the entire dose-response curves by fully formalized parametric regression models whose parameters have toxicological meaning. These models allow a better evaluation of pollutant effects, including inter- and extrapolation to any other than the measured effect values. Following this line, a four-parameter logistic regression model (standard model) was fitted to survival data of Daphnia magna under pesticide (dinoseb) stress. The heterogeneity of the variance was taken into account with a both-sides logarithmic transformation. Besides the standard model, a hormesis and a threshold model were tested too. These two others models have been described in the literature and might better represent the dose-response function we are looking for. All three models showed a good fit to our data, and the statistics gave no hints as to which model is the most appropriate. As no evidence was seen for hormesis or for the existence of a threshold concentration, we used the simplest, namely, the standard model, for most of our calculations. Model calculations allow the quantification of the effects on individuals' longevity as well as on mean survival time of the population. We used them to define a no-effect value, the statistical-no-effect concentration (SNEC). The SNEC is based on the confidence bands of the modeled regression and represents the highest value for which an effect is statistically not different from the control. The SNEC is an alternative to classical endpoints, like the no-observed-effect concentration (NOEC) or the low-effect concentrations (e.g., EC10, EC5, EC1).     

Factors that modify risks of radiation-induced cancer

The collective influence of biologic, physical, and other factors that modify risks of radiation-induced cancer introduces uncertainties and assumptions that limit precision of estimates of human cancer risk that can be calculated for populations exposed to low-dose radiation. The important biologic characteristics include the tissue sites and cell types, baseline cancer incidence, latent periods, time-to-tumor recognition, and individual host (e.g., age and sex) and competing etiologic influences. Physical factors include radiation dose, dose rate, and radiation quality. Statistical factors include time-response projection models, risk coefficients, and dose-response relationships. Sources that modify risk also include other carcinogens and biologic factors (e.g., hormonal conditions, immune status, hereditary factors). Discussion includes examples of known influences that modify radiation-associated cancer risks and how they have been dealt with in the risk-estimation process, including extrapolation to low doses, use of relative risk models, and other uncertainties.     

Radiation dose and leukaemia risk: general relative risk techniques for dose-response models in a matched case-control study

Generalized relative risk functions were used to model radiation dose-response information from a large matched case-control study of leukaemia occurring after treatment for cervical cancer. Models suggested by radiobiological theory were investigated and compared to standard analyses of categorical dose-response to the linear model. Local radiation doses to each of fourteen bone marrow compartments for each patient were incorporated into the models, and the corresponding risks were summed. Conditional maximum likelihood methods were used to estimate risk parameters. Unique features of the analysis include modelling both induction and reduction of risk as a function of radiation dose absorbed by different parts of the body within individuals. Detailed statistical aspects of these analyses are presented and discussed.     

Risk assessment and oncodynamics of ethylene oxide as related to occupational exposure

Two rat inhalation bioassays have been integrated into the risk assessment on the carcinogenicity of ethylene oxide (EO). The carcinogenic findings as well as relevant metabolism and pharmacokinetic data are reviewed. Brain tumors were selected as the endpoint for the assessment of risk because of the indication that adverse effects on the nervous system, related to EO exposure, were consistent across species. Two methods, time-exposure concentration product and area under the plasma concentration-time curve (AUC) are used as a basis for calculating effective dose. Scaling of the dose to man from both rat and dog is explored based on pharmacokinetic studies. Two different mathematical risk extrapolation models, the probit and the multi-stage, are used to estimate the cancer risk for daily exposures to EO of 1.8 microgram/liter over a working lifetime. The use of AUC as a basis for dose from a daily exposure of 1.8 microgram/liter over a working lifetime gives the higher risk rates (90-142/10,000 workers). The implication of the simulated dose using plasma concentrations versus the time-concentration product approach is discussed in relation to threshold effects.     

A meta-analysis of the relation between cumulative exposure to asbestos and relative risk of lung cancer.

OBJECTIVES: To obtain summary measures of the relation between cumulative exposure to asbestos and relative risk of lung cancer from published studies of exposed cohorts, and to explore the sources of heterogeneity in the dose-response coefficient with data available in these publications. METHODS: 15 cohorts in which the dose-response relation between cumulative exposure to asbestos and relative risk of lung cancer has been reported were identified. Linear dose-response models were applied, with intercepts either specific to the cohort or constrained by a random effects model; and with slopes specific to the cohort, constrained to be identical between cohorts (fixed effect), or constrained by a random effects model. Maximum likelihood techniques were used for the fitting procedures and to investigate sources of heterogeneity in the cohort specific dose-response relations. RESULTS: Estimates of the study specific dose-response coefficient (kappa 1.i) ranged from zero to 42 x 10(-3) ml/fibre-year (ml/f-y). Under the fixed effect model, a maximum likelihood estimate of the summary measure of the coefficient (k1) equal to 0.42 x 10(-3) (95% confidence interval (95% CI) 0.22 to 0.69 x 10(-3)) ml/f-y was obtained. Under the random effects model, implemented because there was substantial heterogeneity in the estimates of kappa 1.i and the zero dose intercepts (Ai), a maximum likelihood estimate of k1 equal to 2.6 x 10(-3) (95% CI 0.65 to 7.4 x 10(-3)) ml/f-y, and a maximum likelihood estimate of A equal to 1.36 (95% CI 1.05 to 1.76) were found. Industry category, dose measurements, tobacco habits, and standardisation procedures were identified as sources of heterogeneity. CONCLUSIONS: The appropriate summary measure of the relation between cumulative exposure to asbestos and relative risk of lung cancer depends on the context in which the measure will be applied and the prior beliefs of those applying the measure. In most situations, the summary measure of effect obtained under the random effects model is recommended. Under this model, potency, k1, is fourfold lower than that calculated by the United States Occupational Safety and Health Administration.     

Hormesis: why it is important to toxicology and toxicologists

This article provides a comprehensive review of hormesis, a dose-response concept that is characterized by a low-dose stimulation and a high-dose inhibition. The article traces the historical foundations of hormesis, its quantitative features and mechanistic foundations, and its risk assessment implications. The article indicates that the hormetic dose response is the most fundamental dose response, significantly outcompeting other leading dose-response models in large-scale, head-to-head evaluations. The hormetic dose response is highly generalizable, being independent of biological model, endpoint measured, chemical class, and interindividual variability. Hormesis also provides a framework for the study and assessment of chemical mixtures, incorporating the concept of additivity and synergism. Because the hormetic biphasic dose response represents a general pattern of biological responsiveness, it is expected that it will become progressively more significant within toxicological evaluation and risk assessment practices as well as have numerous biomedical applications.     

Application of benzo(a)pyrene and coal tar tumor dose-response data to a modified benchmark dose method of guideline development

Assessment of cancer risk from exposure to polycyclic aromatic hydrocarbons (PAHs) has been traditionally conducted by applying the conservative linearized multistage (LMS) model to animal tumor data for benzo(a)pyrene (BaP), considered the most potent carcinogen in PAH mixtures. Because it has been argued that LMS use of 95% lower confidence limits on dose is unnecessarily conservative, that assumptions of low-dose linearity to zero in the dose response imply clear mechanistic understanding, and that "acceptable" cancer risk rests on a policy decision, an alternative cancer risk assessment approach has been developed. Based in part on the emerging benchmark dose (BMD) method, the modified BMD method we used involves applying a suite of conventional mathematical models to tumor dose-response data. This permits derivation of the average dose corresponding to 5% extra tumor incidence (BMD0.05) to which a number of modifying factors are applied to achieve a guideline dose, that is, a daily dose considered safe for human lifetime exposure. Application of the modified BMD method to recent forestomach tumor data from BaP ingestion studies in mice suggests a guideline dose of 0.08 microg/kg/day. Based on this and an understanding of dietary BaP, and considering that BaP is a common contaminant in soil and therefore poses human health risk via soil ingestion, we propose a BaP soil guideline value of 5 ppm (milligrams per kilogram). Mouse tumor data from ingestion of coal tar mixtures containing PAHs and BaP show that lung and not forestomach tumors are most prevalent and that BaP content cannot explain the lung tumors. This calls into question the common use of toxicity equivalence factors based on BaP for assessing risk from complex PAH mixtures. Emerging data point to another PAH compound--H-benzo(c)fluorene--as the possible lung tumorigen.     

Cigarette smoking and bronchial carcinoma: dose and time relationships among regular smokers and lifelong non-smokers.

In a 20-year prospective study on British doctors, smoking habits were ascertained by questionnaire and lung cancer incidence was monitored. Among cigarette smokers who started smoking at ages 16-25 and who smoked 40 or less per day, the annual lung cancer incidence in the age range 40-79 was:0.273X10(-12). (cigarettes/day+6)2. (age--22.5)4.5. The form of the dependence on dose in this relationship is subject not only to random error but also to serious systematic biases, which are discussed. However, there was certainly some statistically significant (P less than 0.01) upward curvature of the dose-response relationship in the range 0-40 cigarettes/day, which is what might be expected if more that one of the "stages" (in the multistage genesis of bronchial carcinoma) was strongly affected by smoking. If a higher than linear dose-response relationship exists between dose per bronchial cell and age-specific risk per bronchial cell, this may help explain why bronchial carcinomas chiefly arise in the upper bronchi, for dilution effects might then protect the larger areas lower in the bronchial tree.     

A radiobiological basis for setting neutron radiation safety standards

Present neutron standards, adopted more than 20 yr ago from a weak radiobiological data base, have been in doubt for a number of years and are currently under challenge. Moreover, recent dosimetric re-evaluations indicate that Hiroshima neutron doses may have been much lower than previously thought, suggesting that direct data for neutron-induced cancer in humans may in fact not be available. These recent developments make it urgent to determine the extent to which neutron cancer risk in man can be estimated from data that are available. Two approaches are proposed here that are anchored in particularly robust epidemiological and experimental data and appear most likely to provide reliable estimates of neutron cancer risk in man. The first approach uses gamma-ray dose-response relationships for human carcinogenesis, available from Nagasaki (Hiroshima data are also considered), together with highly characterized neutron and gamma-ray data for human cytogenetics. When tested against relevant experimental data, this approach either adequately predicts or somewhat overestimates neutron tumorigenesis (and mutagenesis) in animals. The second approach also uses the Nagasaki gamma-ray cancer data, but together with neutron RBEs from animal tumorigenesis studies. Both approaches give similar results and provide a basis for setting neutron radiation safety standards. They appear to be an improvement over previous approaches, including those that rely on highly uncertain "maximum" neutron RBEs and unnecessary extrapolations of gamma-ray data to very low doses. Results suggest that, at the presently accepted neutron dose limit of 0.5 rad/yr, the cancer mortality risk to radiation workers is not very different from accidental mortality risks to workers in various nonradiation occupations. The neutron dose estimated to produce 2.5% lifetime risk (maximum radiation-induced risk presently accepted by the ICRP and NCRP for occupational exposure) is in the range 10-156 rad; the most probable dose is 30 rad, which is not very different from 25 rad, the presently accepted lifetime dose limit for fast neutrons.     

Sicherheitsfaktoren in der Risikobewertung von Chemikalien

Risk assessment of chemicals is based on data from experimental exposure of animals. Departing from the dose/concentration at which no toxic effects have been observed in animals (no observed adverse effect level, NOAEL), the dose/exposure in humans that will not result in toxic effects is derived by extrapolation done in two steps.Step one is the extrapolation from animal to man and, traditionally, a safety factor of 10 is used to account for differences in toxicokinetics and in toxicodynamics.Likewise, a safety factor of 10 is used in step two, which takes differences into account between the “median” human and the whole population including the sensitive subpopulation.There is increasing awareness that substance-specific factors should be used if data exist.Some regulators apply additional factors to account for uncertainty in the data and model uncertainties. The concept of safety factors is used to derive “safe” levels of exposure or “safe” doses.Application of the margin of safety (MOS) approach is different in which the distance is evaluated between a relevant level of toxicity (NOAEL) in animals and the level of exposure in humans.This approach is used when,e.g., food contamination with levels higher than safe levels have been found.The MOS approach is also used in chemical risk assessment. As the public often does not understand the assumptions behind safe levels, lay people associate levels higher than the safe levels with acute health risks.     

State-of-the-Science Workshop Report: Issues and Approaches in Low-Dose–Response Extrapolation for Environmental Health Risk Assessment

Low-dose extrapolation model selection for evaluating the health effects of environmental pollutants is a key component of the risk assessment process. At a workshop held in Baltimore, Maryland, on 23–24 April 2007, sponsored by U.S. Environmental Protection Agency and Johns Hopkins Risk Sciences and Public Policy Institute, a multidisciplinary group of experts reviewed the state of the science regarding low-dose extrapolation modeling and its application in environmental health risk assessments. Participants identified discussion topics based on a literature review, which included examples for which human responses to ambient exposures have been extensively characterized for cancer and/or noncancer outcomes. Topics included the need for formalized approaches and criteria to assess the evidence for mode of action (MOA), the use of human versus animal data, the use of MOA information in biologically based models, and the implications of interindividual variability, background disease processes, and background exposures in threshold versus nonthreshold model choice. Participants recommended approaches that differ from current practice for extrapolating high-dose animal data to low-dose human exposures, including categorical approaches for integrating information on MOA, statistical approaches such as model averaging, and inference-based models that explicitly consider uncertainty and interindividual variability.     

A practical guide to dose-response analyses and risk assessment in occupational epidemiology

Dose-response modeling in occupational epidemiology is usually motivated by questions of causal inference (eg, is there a monotonic increase of risk with increasing exposure?) or risk assessment (eg, how much excess risk exists at any given level of exposure?). We focus on several approaches to dose-response in occupational cohort studies. Categorical analyses are useful for detecting the shape of dose-response. However, they depend on the number and location of cutpoints and result in step functions rather than smooth curves. Restricted cubic splines and penalized splines are useful parametric techniques that provide smooth curves. Although splines can complement categorical analyses, they do not provide interpretable parameters. The shapes of these curves will depend on the degree of "smoothing" chosen by the analyst. We recommend combining categorical analyses and some type of smoother, with the goal of developing a reasonably simple parametric model. A simple parametric model should serve as the goal of dose-response analyses because (1) most "true" exposure response curves in nature may be reasonably simple, (2) a simple parametric model is easily communicated and used by others, and (3) a simple parametric model is the best tool for risk assessors and regulators seeking to estimate individual excess risks per unit of exposure. We discuss these issues and others, including whether the best model is always the one that fits the best, reasons to prefer a linear model for risk in the low-exposure region when conducting risk assessment, and common methods of calculating excess lifetime risk at a given exposure from epidemiologic results (eg, from rate ratios). Points are illustrated using data from a study of dioxin and cancer.