Scholars Award: Three Methodological Rules in Risk Assessment

在风险评估的第一阶段或危险识别阶段，经常存在与各种因素（如特定的手机、药品或杀虫剂）相关的假定风险的争议。不同科学家同时发表在权威期刊上的结论，往往在某些危害的存在上相互矛盾。虽然政治或议程驱动的资金来源有助于解释一些冲突，但许多冲突似乎源于对用于解释流行病学统计和风险数据的因果推论的方法上的分歧。至少10年来，这些冲突被称为“流行病学战争”，其中一个焦点是对用于评估风险数据的3种方法规则的分歧。这就是流行病学证据规则（EER），根据该规则，对危害的因果推断需要（人类）流行病学数据，而不仅仅是动物或实验室数据；统计显著性规则（SSR），根据该规则，只有在有统计学显著证据表明存在危害（p或= 0.05）时，才应该拒绝无效或无效应假设；以及相对风险规则（RRR），根据该规则，危险识别（声称某些物质造成了给定的伤害）需要证明相对风险至少为2。知识价值。该项目评估接受、拒绝或修改这3条方法学规则在推进流行病学风险评估的各种认知目标方面的有效性，有7个主要目标。这些是(1)提供EER、SSR和RRR优势的方法学概述；(2)使用(1)来评估每条规则在方法上的弱点；(3)评估每条规则是否适用的不同事实情况；(4)评估接受/拒绝每条规则有助于实现哪些认知和实践目标；(5)通过接受/拒绝各项规则，确定符合何种流行病学划界标准；(6)调查每条规则的哪些修订可能符合不同的流行病学划界标准、实际情况和认知目标；(7)利用审稿人的批评来修正项目发现。在阐明方法论规则（如Doppelt， Laudan, Schmaus），因果解释（如Cartwright， Glymour, Suppes），统计哲学和实验推断（如Mayo），特别是在医学（如Clouser， Kyburg, Schaffner）方面，科学哲学家已经做了极好的基础研究。然而，这个项目是必要的，因为没有科学哲学家调查过EER、SSR和RRR，无论是与流行病学风险评估还是其划分标准、事实背景和认知目标相关。PI资格包括数学和科学哲学学位；3名博士后（经济学、生物学、水文地质学）；30年关于风险评估方法的资助/出版物（但不包括项目的3个方法规则）；并参与撰写了许多美国国家科学院的风险评估研究，包括1996年的经典著作《理解风险》。更广泛的影响。在美国环保署科学顾问委员会的许多年里，PI评估了许多风险评估中使用的流行病学方法/规则，特别是农药。关键的项目方法（书目研究、概念分析、评估方法假设、评估结果）在PI早期的资助/出版物中使用。在相互冲突的风险评估（针对32种有机磷农药，2006年在美国重新注册）中检查EER、SSR和RRR——由于科学顾问委员会的工作，PI对这些评估很熟悉——PI将评估13种假设，这些假设是关于划分标准、事实背景和认知目标如何帮助解释方法规则上的冲突。

At the first or hazard-identification stage of risk assessment, there are frequent controversies over supposed risks associated with various agents - such as particular cell phones, pharmaceuticals, or pesticides. Often conclusions of different scientists, published concurrently in refereed journals, contradict each other regarding the existence of some hazard. While politics or agenda-driven funding sources help explain some conflicts, many appear to arise from methodological disagreements over the causal inferences used to interpret epidemiological statistics and risk data. For at least 10 years, these conflicts have been known as "the epidemiology wars" - one focal point of which is disagreement over 3 methodological rules used to assess risk data. These are the epidemiological-evidence rule (EER), according to which causal inferences about harm require (human) epidemiological data, not merely animal or laboratory data; the statistical-significance rule (SSR), according to which the null or no-effect hypothesis ought to be rejected only if there is statistically significant evidence of harm (p or = 0.05); and the relative-risk rule (RRR), according to which hazard identification (alleging that some agent has caused a given harm) requires evidence of a relative risk of at least 2.Intellectual Merit. The project - evaluating how effective accepting, rejecting, or amending these 3 methodological rules is, in advancing various cognitive goals of epidemiological risk assessment - has 7 main objectives. These are (1) to provide a methodological overview of strengths of EER, SSR, and RRR; (2) to use (1) to assess each rule's methodological weaknesses; (3) to evaluate different factual circumstances on which using/not using each rule is contingent; (4) to assess which cognitive and practical goals are served by accepting/rejecting each rule; (5) to determine which epidemiological demarcation criteria are served by accepting/ rejecting each rule; (6) to investigate which amendments to each rule might fit different epidemiological demarcation criteria, factual circumstances, and cognitive goals; and (7) to use reviewers' criticisms to revise the project findings. In explicating methodological rules (e.g., Doppelt, Laudan, Schmaus), causal explanation (e.g., Cartwright, Glymour, Suppes), philosophy of statistics and experimental inference (e.g., Mayo), especially in medicine (e.g., Clouser, Kyburg, Schaffner) - philosophers of science have done superb basic research. However, this project is needed because no philosophers of science have investigated EER, SSR, and RRR, relative either to epidemiological risk assessment or to its demarcation criteria, factual context, and cognitive goals. PI qualifications include degrees in mathematics and in philosophy of science; 3 post-docs (economics, biology, hydrogeology); 3 decades of grants/publications on risk-assessment methods (but not on the project's 3 methodological rules); and coauthoring many US National Academy of Sciences risk-assessment studies, including the 1996 classic, Understanding Risk.Broader Impact. In many years on US EPA's Science Advisory Board, the PI has evaluated epidemiological methods/rules used in scores of risk assessments, especially for pesticides. Key project methods (bibliographical research, conceptual analysis, assessing methodological assumptions, evaluating consequences) were used in the PI's earlier grants/publications. Examining EER, SSR, and RRR in conflicting risk assessments (of 32 organophosphate pesticides, done for 2006 US re-registration) - with which the PI is familiar, because of Science Advisory Board work - the PI will assess 13 hypotheses about how demarcation criteria, factual contexts, and cognitive goals help explain conflicts over methodological rules.