Epidemiology : Incorporation into Pesticide Risk Assessment and Management: Just when you thought it was SAFE to go back into the Agency… Daniel A. Goldstein, M.D. Director, Medical Sciences and Outreach, Monsanto Current Chair, CLA-EHHWG (Epidemiology and Human Health Working Group )
We are used to a animal-data-driven, quantitative- risk-assessment approach- but INTUITION tells the PUBLIC that the street lamp shines HERE: Epidemiology Exposure-Based Risk ◦ HUMAN data Assessment ◦ REAL health effects ◦ Toxicology studies Intuition tells us, ◦ Exposure assessment the light is over ◦ Risk assessment here!! BUT….. ◦ Risk management
In reality, Epidemiology faces serious limits to knowledge: Short term outcomes- relatively easy (How many people who ate peanut butter got salmonella?) Long term or delayed outcomes are difficult (cancer, parkinsons, developmental, etc.) ◦ Metrics of past exposure are limited. ◦ Recall is subject to significant bias. ◦ Diagnosis can be challenging and unreliable. ◦ Alternative causes and contributors may be difficult to identify and exclude. ◦ Ecological studies (lacking individual exposure data) are problematic.
The problem is aggravated by “ Hypothesis Generating ” studies . (We don’t know what we are looking for…. So we will look for everything.) Hypothesis T esting Hypothesis Generating Single or few a-priori Multiple hypotheses: hypotheses to be tested. Computing required….. Thought required….. Agents X Outcomes X Subgroups* = Design considers known or 10’s – 1000’s of hypotheses! suspect confounders and alternative hypotheses. 1 in 20 hypotheses will test positive even if no real association exists Within limits of model, can define likelihood of a “false Design considers only general positive” result. (Typically < 5%) confounders (age, smoking). Requires “confirmatory” studies- but usually NOT done! * Age, sex, genotype, etc.
The Ag Health Study (AHS slides adapted from Alavanja M. , AHS update, APPCO April 6 2009) Prospective cohort study of 89,658 pesticide applicators & spouses. 82% of target population enrolled 1993-1997. Little loss to follow-up (<2%). Cancer incidence and mortality updated annually. Comprehensive exposure assessment information on 82 pesticides collected at three points in time. Questionnaire exposure assessment evaluated with field measurements of pesticides. Buccal cells collected on >35,000 study subjects.
And they have been busy….. Herbicides: Alachlor – trends for leukemia and myeloma Atrazine – no associations Butylate – prostate Cyanazine – no associations Dicamba – trend with colon and lung cancer EPTC – leukemia and colon Glyphosate – multiple myeloma Imazethapyr – bladder and colon Metolachlor – deficit for prostate, excess for lung Pendimethalin – lung, trend with rectal cancer Trifluralin – colon
Insecticides: Aldicarb – colon Aldrin – breast, deficit for colon Carbaryl – melanoma and a deficit for prostate Carbofuran – prostate, trend for lung cancer Chlordane/heptachlor– rectum, breast, leukemia Chlorpyrifos – brain, rectum, lymphohematopoietic, trend for lung Coumaphos - prostate Diazinon – trends for leukemia and lung Dichlorvos – no associations Dieldrin – lung, breast Fonofos – leukemia, prostate Lindane – NHL Malathion – breast, deficit for melanoma Methyl bromide – prostate Permethrin - prostate Phorate – prostate among family history Toxaphene – melanoma Fungicides: Captan – no associations Chlorothalonil – no associations
Pressure on the Agency The AHS team has indicated that their data alone are sufficient to establish causation. Existence of studies like AHS puts EPA under some pressure to look at and incorporate epidemiologic data in risk-assessment. NGO’s and the public have been / will come in behind the use of epidemiological data. There are LOTS MORE STUDIES OUT THERE- past, present and future! (Can YOU say National Children’s Study?)
EPA Has begun to address the issue: Existing Risk Assessment guidance (water program) details limitations and outlines assessment principles for epi data. New Epidemiology Branch created in Health Effects Division under Mary Manibusan, with expertise “imported” from water program. White paper and internal review process underway, and FIFRA SAP being developed. HED Director, Tina Levine, has an understanding and appreciation of the limitations of bio-statistical methods. New Agency web-page indicates a Weight of Evidence approach….
Agricultural Health Study – EPA’s Role and Plans Current as of March 5, 2009 NOTE- selected excerpts only, emphasis (in red) added… Developing a science-based process In anticipation of results from current AHS studies, EPA is developing an approach for incorporating epidemiology information into human health risk assessments for pesticide chemicals. As a basis for this approach, EPA is using existing Agency and international guidance including: U.S. EPA 2005 Guidelines for Carcinogen Risk Assessment Reference Dose guidance IPCF Framework for Analysing the Relevance of a Cancer Mode af Action for Humans and Case-Studies (PDF) EPA relies on multiple lines of evidence to evaluate the safety of a pesticide, including animal toxicology studies as well as other sources of information such as how effects are caused (mechanisms of action), use patterns, what happens to the pesticide in the environment and how long it remains (environmental fate and persistence), food residue levels, and human exposure potential. Scientists evaluate all available data and weigh the evidence rather than relying on any one study. Incorporating epidemiology data into this process will expand the range of the evaluation. EPA’s approach to incorporating the AHS findings into existing hazard and exposure information will focus on: how the pesticide may cause potential harm (i.e., the pesticide’s mode of action ) how the body handles the compound once exposed what the compound does when it is in the body An evaluation of exposure pathways, route (e.g., dermal, inhalation, oral), and duration will be an important component of the proposed approach. Another aspect of the approach will be to compare results of the AHS to those obtained from toxicology studies done with rodents and other experimental animals . There are a variety of sophisticated models and tools available, which EPA will use as available and appropriate to evaluate risk. Obtaining peer review and involving the public EPA will present its scientific approach to integrating exposure, toxicology, and epidemiology in a white paper, which will be available along with draft case studies for review by the FIFRA Scientific Advisory Panel in 2009. EPA will also solicit comments from the public on the proposed approach. http://www.epa.gov/pesticides/health/ag-health.html
And- Ag Health does admit at least some limitations…. General observations (adapted from 2009 AHS update cited above, emphasis added) No pesticide associated with all/ many cancers Cancer associations do not cluster by chemical class Some associations occur for cancers not elevated among farmers (colon, rectum, lung, bladder) Cancer excesses are small. This could indicate they are: - due to chance or confounding - real, but impact weak - real, but impact reduced because of misclassification Exposure assessments need to be improved Need to replicate analyses
Epidemiology is Adapting to Criticisms Focus on single agents Alternative exposure assessment techniques Focus on high risk or susceptible populations Agents: Paraquat and Maneb. Exposure assessment: California geographical use reporting system. Genetic sub-classification by Dopamine Transporter variants. Study has serious limitations, but findings are not easily dismissed by “traditional” objections
Developing a Response- Process Coordinate efforts with EU Epi Team Continue conversations with HED ◦ Engagement in white paper review/ comment opportunity as appropriate. ◦ Engagement with pending FIFRA SAP Consider engagement with broader epidemiology community to: ◦ More fully understand the applications and limitations of epidemiology. ◦ Encourage movement toward hypothesis selection and refinement followed by focused hypothesis testing.
Developing a Response- Focus Quality of Design and Use of Weight of Proper Interpretation Evidence Approach ◦ Hypothesis testing vs ◦ Bradford-Hill Criteria generation ◦ Classical animal testing ◦ Exposure assessment ◦ Mechanism of Action ◦ Model selection ◦ Exposure pathways ◦ Bias, confounding ◦ Dose-response based ◦ Alternative causes risk assessment
Discussion
Recommend
More recommend
