Document QkxaxZZn5zxvOr4yd74bGbwM7

OPR 13 '97 19S41 TO Shah-Olfl FROM APPLIED EPIDEMIOLOGY T-452 P.02 APPLIED EPIDEMIOLOGY INC. TO: FROM: DATE: RE; Drs. Hasmukh Shah and Jonathan Ramlow Kenneth A. Mundt, PhJ April 14,1997 University of Louisville Case-Control Study on Vinyl Chloride Exposure and Brain Cancer Following are several thoughts on what we might seek to accomplish with Dr. Tamburro and his team in Louisville on April 18, 1997. 1. To date, we have not seen a detailed protocol which clearly addresses all the items specified in CMA's Guidelines for Good Epidemiological Practices (GEP). At our last meeting. Dr. Tamburro suggested that he will prepare a more complete study protocol after the upcoming meeting, so that he may incorporate appropriate ideas from our discussions. Suggested Action: Establish a time linefor preparation, review and approval ofthe finalprotocol. Peer review and sign-offshould take place as soon as possible. 2. Additional elements of the GEP's should be addressed, including quality control (which is apparently quite high with regard to the database), standard operating procedures, various review mechanisms, etc. Suggested Action: Review the GEP's in detail to Identify specific areas to be addressed 3. Several of the external reviewers' comments are still germane, and have not been appropriately addressed. Some of these points should be addressed directly in the revised protocol. For those determined to be inappropriate or for which a decision is made not to incorporate suggested changes, an explanation should be made in a separate memorandum. This memorandum should accompany the protocol when re submitted to the reviewers. P.0. BOX 2424 AMHERST, MA 01004 (413) 256'3556 FAX (413) 256-3503 CMA 121202 APR 13 *97 19:42 TO Shah-CMA FROM APPLIED EPIDEMIOLOGY T-452 P.03 (Drs. Shah and Ramkw, page 2) SuggestedAction: Request that Dr. Tamburro prepare individual responses to the points raised by the external reviewers, Drs. Checkowqy and Marsh, and incorporate any that are determined to be appropriate directly in the protocol A time linefor this action should be established A. One ofthe main concerns ofthe reviewers (as well as Dr. Ramlow and myself) is that the proposed statistical methodology (Oreenberg/Tamburro method) may not be adequate to reveal an association between brain cancer and VCM/PVC, should one exist. Part of this concern stems from the small anticipated sample size, leading to low statistical power, and pan from the lack of documentation ofthe performance of these techniques outside of extremely strong and highly specific associations (such as VCM and angiosarcoma of the liver). Suggested Action: Discuss extending the proposed analytical methodology to include classical casecontrol analyses (logistic regression) inparallel with the proposed approach. Statisticalpower will remain a problem. Perhaps estimates ofpower should be determinedfor each analytical approach to enhance the interpretation ofthe study results. ,5 A less tangible concern about the study is whether it will be seen as credible by the scientific community. On the other hand, the study has been developed as a casecontrol study nested in a larger cohort; however, primary assessment of the cohort giving rise to the cases apparently has not occurred. For example, does the set of cases proposed for the case-control study represent an excess of such cancer cases (deaths)? Are there other patterns of specific causes of death which might help us understand the context in which the brain cases were identified (e.g., a marked deficit of cardiovascular disease deaths)? In many ways, the surveillance database constructed by Dr. Tamburro and his team is impressive, and much of its value will not be realized in the current case-control effort. Suggested Action: Evaluate the role this case-control study plays in the context of the broader research objectives beingpursued by this team, with the aim ofenhancing the totalpackage. Ifthe case-control study is perceived by the scientific community as apreliminary strategyfor assessing the potential risk ofinterest, what next steps might be considered? Linking it with other planned analyses/studies might help capitalize on this unseen value. Perhaps conducting a proper cohort analysisfirst or in parallel would be informative. CMA 121203