Document g2qBV46o3VJq4VZMV7V7wxjV

MEMORANDUM TO: Dr. Richard Clapp cc: Dr. Peter Montague CONFIDENTIAL FROM: Paul Merrell RE: -Gaffey v. Montague COMMUNICATION PRIVILEGED__ AND__ CONFIDENTIAL DATE: November 16, 1993*I I have asked Peter Montague to use his scientific skills to i review Dr. Gaffey's study and see if he might be able to identify some leads for you to further develop, in order to hold his expenses down and reduce the demands on your time. Enclosed please find his first cut. It would be extremely helpful if you might look this over and let me know when you might be available for a conference call among yourself, Peter, and me, hopefully tomorrow. If not tomorrow, Friday would be the next day I am able to do it because of my need to travel to Eugene on Thursday. I caution that I or one of Peter's other attorneys need to be part of any communications between him and you in regard to the issues of this lawsuit, both to ensure that the communications remain privileged and to get the trial lawyers up to speed. I'd appreciate it if you could call me in the morning to set a time for the telephone conference, which we will initiate. Best regards from me and from Carol. REENSVEAIRRCOHNFMOUENNDTAATLION P.O. Box 5036 Annupolis, MD 21403-7036 Telephone (410) 263-1584 Econet: erf Internet: erf@ igc.apc.org Fax (410) 263-8944 November 16, 1993 Mr. Paul Merreil 7493 East Five Rivers Road Tidewater, OR 9 7 3 9 0 confidential Dear Paul, In preparing my defense in Gaffey v. Environmental Research Foundation et al., I have been reviewing the basis for some of the statements in my newsletter #171, which Mr. Gaffey found defamatory. One of the offending statements occurs at the bottom of column 1, pg. 1: "In fact, excess cancers have occurred, but it appears that the data have been manipulated to hide the facts." [Emphasis in the original. 1 The statement has two parts: (a) excess cancers have occurred; and (b) it appears that the data have been manipulated to hide the facts. At the moment, I am focused entirely on lung cancer in the Nitro cohort, as reported by Zack/Gaffey. I will refer to two studies, Zack/Gaffey1 and Zack/Suskind.12 Regarding the first part of the statement, I am attaching two short analyses I have done, based on Zack/Gaffey's data. 1 merely calculated the probability that the apparently-increased lung cancers in the Nitro cohort were due to chance. Gaffey tested for significance at the 0.05 level; if he had selected 0.20 as his level of significance, as one might justifiably do in an exploratory analysis,3 he would have found significantly-elevated increases in lung cancer in the period 1970-1977. 1 Judith A. Zack and William R. Gaffey, "A Mortality Study of Workers Employed at the Monsanto Company Rant in Nitro, West Virginia,'' in Richard E. Tucker, Alvin L. Young, and Allan P. Gray, editors, Human and Environmental Health Risks of Chlorinated Dioxins and Related Compounds (New York: Plenum Press, 1983), pgs. 575-591. 2 Judith A. Zack and Raymond R. Suskind. "The Mortality Experience of Workers Exposed to Tetrachlorodibenzodioxin in a Trichlorophenol Process Accident," Journal of Occupational Medicine Vol. 22 No. 1 (January, 1980), pgs. 11-14. 3 John T. Roscoe, Fundamental Research Statistics for the Behavioral Sciences (Second Edition] (Fort Worth: Holt, Rinehart and Winston, 1975), pg. 182, says most research is conducted at the .01 and -05 levels of significance but "...in exploratory research, the .10 and .20 levels may be more appropriate." CONFIDENTIAL Since Gaffey's selection of 0.05 is an arbitrary choice, I think my two analyses strengthen my case for saying that cancers have increased. The second part of the offending phrase, about the data being manipulated, is somewhat more complicated. In Table 9 of the Zack/Gaffey study, we see the heart of the matter: he conducted PM R analyses for the "exposed" workers and the "non-exposed" workers. In the "exposed" workers, Zack/Gaffey found a PM R of 168 (3.57 expected and 6 observed). Among the "non-exposed" workers, Zack/Gaffey reported a PMR of 125 (6.42 expected, 8 observed). However, we know from Table 11 of Zack/Gaffey and Table 10 of Zack/Suskind that four workers listed as "exposed" in Zack/Suskind's Table 10 are listed as "Not Exposed" in Zack/Gaffey's Table 11. In Zack/Suskind's Table 10 these workers are (counting from the top of the table), Nos. 1, 2, 5, and 7. In Zack/Gaffey's .table 11, these same workers appear as (counting from the top), Nos. 5, 6, 9, and 22 I will refer to these 4 workers as "misclassified." If we re-classify the misclassified workers, Zack/Gaffey's Table 8 changes accordingly. The number of deaths among the "exposed" becomes 62 (not 58) and the number of deaths among the non-exposed becomes 100 (not 104). These changes cause additional changes in the "Percent" column of Table 8; the percent of deaths among the "exposed" becomes 38.0 (not 35.6), and the percent of deaths among the "non-exposed" becomes 61.4 (not 63.8). I note that 3 of these misclassified workers died of lung cancer. If we re classify these 4 workers in Zack/Gaffey's Table 9, moving them from the "non-exposed" column to the "exposed" column, then the lung cancer figures in Table 9 would change. The new numbers would be 9 (not 6) lung cancers observed among the "exposed" workers, causing the PMR to change from 168 to 252. Among the non-exposed, we would now have 5 (not 8) lung cancers observed among the "non-exposed" workers, causing the PM R to change from 125 to 78. I am unable to go further with this analysis. Having re-classified these workers and calculated new PMRs, I would like to calculate the probability that the new PM Rs were due to chance. Unfortunately, I can't understand Gaffey's method for calculating the 9 5 % confidence interval for any given PMR; he describes his method at the top of pg. 579 but I don't see where one gets the basic data he calls "confidence limits for the % of observed deaths" and I don't understand the denominator, which Gaffey 2 gives as exp* CONFIDENTIAL To me, exp means "e raised to the power of," but power of what? Or does Mr. Gaffey intend exp to mean "expected deaths?" i need to talk to Dick Clapp about this. Could you please set up a conference call with Dick and you and me? I want to discuss how to calculate confidence intervals for the revised P M R s or, perhaps, how to calculate the standard error of a PMR. M y goal is to calculate various confidence intervals and/or standard errors that will help me understand quantitatively what the revised PM Rs tell us about the likelihood of real lung cancer increases having occurred among the Nitro cohort. (The direction of the change induced by Mr. G affey's misclassification is already clear.) By this means, I hope to learn something Quantitative about the effect of Mr. Gaffey's misclassification of 4 exposed workers. I also need Dick to check my methods and arithmetic in the two attached analyses of S M R s among the entire Nitro cohort. Thanks very much for your help with this. Sincerely, Peter Montague 3 CONFIDENTIAL sm FO R LU N G CANCER. 1955-1977. IN THE ENTIRE NITRO W ORKER I a r In their study of Nitro workers, Zack/Gaffey conducted an analysis of cause-specific standardized mortality ratios (SM R s) for 15 selected cancer sites among the entire Nitro worker cohort during the period 1955-1977.4 This paper focuses on their analysis of lung cancer. With 9.91 lung cancers expected and 14 observed, they reported that the S M R for lung cancer w as 141 [Zack/Gaffey Table 41. Zack/Gaffey highlighted this finding by saying, "Although not statistically significant, the S M R for lung cancer appears to be elevated." W hat is the likelihood that this apparently-elevated lung cancer mortality w as due to chance? Zack/Gaffey suggest that the proper w ay to calculate the standard error is: 100 * sqrtlno. of observed deaths) standard error = ---------------------------------------------------- no. of expected deaths By this formula, the standard error for lung cancer is: (100 3.741/9.91 = 37.8 The SM R , which is 141, differs from 100 by 41, which is 41/37.8 = 1.09 standard errors. In the table of areas under the standard normal distribution, we find the probability that a lung cancer S M R of 141 is due to chance is 0.276. In other words, the chance that the S M R of 141 is due to an actual increase in lung cancer among Nitro workers (and not due to chance) is 72.4 percent. Thus Zack/G affey's analysis of the entire Nitro cohort reveals that the chances are better than 2-out-of-3 that an actual increase of lung cancer occurred among the Nitro workers during the period 1955-1977 (compared to the normal occurrence of lung cancer among the American public). -Peter Montague, 11-15-93 4 Judith A. Zack and Wiliam R. Gaffey, "A Mortality Study of Workers Employed at the Monsanto Company Plant in Nitro, West Virginia," in Richard E. Tucker, Alvin L. Young, and Allan P. Gray, editors. Human and Environmental Heeith Risks of Cfiforinated Dioxins and Related Compounds (New York: Plenum Press, 1983}, pgs. 575-591. 4 * conrdchttal SM FOR LUNG CANCER. 1970-1977. W THE ENTIRE NITRO WORKER W mW St Zack/Gaffey5 calculated cause-specific standardized mortality ratios (SM R s) for 15 selected cancer sites among the entire Nitro cohort during the period 1970-1977. Mere we examine lung cancer. With 5.52 lung cancers expected and 10 observed, Zack/Gaffey reported that the S M R is 181 Itheir Table 5]. What is the likelihood that this apparent increase w as due to chance, and not due to a real increase in lung cancer among the Nitro workers? Zack/Gaffey calculate the standard error of the S M R as: 100 * sqrtlno. of observed deaths) standard error ------------------------------------------------------ no. of expected deaths By this formula, the standard error for lung cancer is: (100 # 3.16)/5.52 = 57.3 The S M R , which is 181, differs from 100 by 81, which is 81/57.3 = 1.42 standard errors. In the table of areas under the standard normal distribution, we find the probability that a lung cancer S M R of 181 is due to chance is 0.168. In other words, the chance that the S M R of 141 is due to an actual increase in lung cancer among Nitro workers (and aot due to chance) is 83.2 percent. Thus Zack/G affey's analysis of the entire Nitro cohort reveals that the chances are better than 4-out-of-5 that an actual increase of lung cancer occurred among the Nitro workers during the period 1970-1977 (compared to the normal occurrence of lung cancer among the American public). -Peter Montague, 11-15-93 6 Judith A. Zack and William R. Gaffey, "A Mortality Study of Workers Employed at the Monsanto Company Plant in Nitro, West Virginia,* in Richard E. Tucker, Alvin L. Young, and Allan P. Gray, editors, Human and Environmental Health Risks of Chlorinated Dioxins and Related Compounds (New York: Plenum Press, 1983), pgs. 575-591. 5 Telephone: (503) PAUL MERRELL Lawyer 7493 East Five Rivers Road Tidewater, Oregon 97390 528-7151 Telefax: (503) 528-7105 ea j< ,To (k RtJlClf. P. _____From : W r-C 'v 'P .L / sh eet No. ( o / ? - H lZ ^ ^ rz transmitting from (503) 528-7105. Date : M . / lC / V Pages (including this sheet) Message: __ C C ___ < P estcA_a ^ If you experienced any difficulties in receiving this transmission, please call (503) 528-7151. N o t e s We now have a separate fax line and can receive telefax messages automatically. If the message is urgent, please follow-up with a telephone message because the telefax receiver is in a remote location. ^,fc:CT*********** -COMM. JOURNAL- ******************** DATE NDU-17-1993 ***** TIME 00: 06 **** p. 1 MODE = MEMORY TRANSMISSION START=N0V-17 00=05 END=N0U-17 00=08 FILE NO.= 053 NO. COM SPEED/NTUK STATION NAME/ TELEPHONE NO. PAGES PRG.ND. PROGRAM NAME 001 OK a 14102638944 002/002 -ALDER HILL FORM ***************************************C FAX-310 02.183* - 503 528 7105- ************ :jkf:)oM<*cfojcioK -COMM. JDLJRNAL- ^^^^^tcKKKtoKKK^oK^>K DATE N O V-16 -1 9 9 3 tetctctoK T IM E 23= 5 6 P .l MODE = MEMORY TRANSMISSION START=N0U-16 23=52 END=N0U-16 23=56 FILE ND.= 052 NO. COM SPEED/NTUK STATION NAME/ TELEPHONE ND. PAGES PRG.ND. PROGRAM NAME 001 OK a 16174B20617 007/007 -ALDER HILL FARM *>lcfc*5tcfofc**3fc***5iGfofcik5iGicK5icf3k^ FAX--310 U 2 ,185* -- 503 528 7105-- =K^cK^cfc^cicfcHc}cfok