Document text
Machine-read from the scan, so expect some errors. Lines that couldn’t be read are left out.
Page 1
Health Sciences Research and Applications S ranch Division of Biomedical and En v iron nen tal Research Atmic Energy Commission Ina way, it's a pity that the meeting on the X zn cu so et a1 study was not for a day and one - helf. much more confident this a or ning about what I think, and what I would reconmend. But i t ' s perhaps just as well that we didn ' t meet again -i fall five of us had cae t of in,but different,op in im,that rjould ' not have helped. After o night ' s s + eep,I felt so Ret men dati ons
Page 2
Sanders means well but knows nothing &out radiation Mancuso has not been working a t it, 2. Fly recommendation is contained in it emL If you cannot 01: will not follow this line but instead propose to aMt in ue this contractor,thenI have some re com end at ions. These detailed, and 1 would not want t o impose them upon a new investigator,since he should be s me one who has himself the expertness to devise his mm plans and freer to make suggestions since I know that my voice is but one of five: But the Pi:tsburgh group do need sone inst--.llctions; Ifeel a. Forget about the A and B controls, (1) They are not ideal in any case. (2) Waiting for them creates inordinz- delays - the need t o assem5le the whole roster before c5oo sing them has prevented a Zirst paper on, say, Hanfcrrrd. the process would have to be repeated enc Ek s sly,with a ccm pan ying delays. (3) Presumably one will want t o contirr=rre to monitor - cancer mortal ir j. Although I sure th a ~z cort ali ty for am not convinced that the bias is impor- for cancer, care u he the r this be mortality ratios,or I;= tables or mortality rates - but let it be a standard zez hod so that results can b e compared with the results of others, of analysis like " Percent of Years Lost" are wanted also, they should be in addition to,notin place ... They should then cme to a real is ric
Page 3
assessment of whether they can o r cannot hope to demonstrate risks of that order of magnitude (or even twice as large) in the material available. more, at best,than give reassurance that risks are not inordinately large and the study should be structured to that limits " for risks, and this would be valuable if it did not turnout that the confidence intervals were so wide as t o be risks are not a relevant basis for designing the study. Ifnot, then the study can do no It may be possible to obtain atleast "upper confidence In this connection, Gokan's projections for cancer I hope this is useful for your pr poses.