Document 6RrDbRXQGk3ZbZ5qve5DZvN9o

r TO "OM USJECT B. M. G. Zwicker J. C. Schaap A. M. Fairlie FICLO POINT Oft AKRON DEPARTMENT * OUDO. NO. Cleveland FIELD POINT OR AKRON DEPARTMENT * mi na NO. Cleveland Comments on Peters and Monson and NIOSH Mortality Tables -- Proportional vs. Actual Mortality Analysis OAT* YOUR LITTER oat* This lxttkr August 22, 1974 SUMMARY: Proportional and actual mortality analyses are two methods commonly used to construct mortality tables. The proportional mortality table is useful in determining whether a population has an abnormal distribution of deaths amopg various diseases, but may be misleading as to whether the death rate of a specific disease is actually abnormal. The actual mortality table permits this latter determination. Peters and Monson use proportional mortality analysis to show that rates of brain, liver, lung, and total cancers are abnormally high in the Louisville and Calvert City plants mortality data. An actual mortality analysis by BFG indicates that only brain and liver are significantly in excess. Data from a NIOSH study of the Louisville plant indicates that brain, liver, and total cancer rates are again high. However, the method of mortality analy sis is not given. DISCUSSION: Dr. John M. Peters of Harvard University testified before OSHA on June 27, 1974. He and Dr. Richard Monson obtained data on 161 deceased workers from the Louisville and Calvert City BFG plants. They constructed a mortality table based on this data and the age distribution of the plants' employees (Table 1). Their major findings indicated (1) a 50% excess of total cancer deaths, (2) a 60% excess of lung cancer, and (3) a 320% excess of brain cancer. Liver cancer cases were ten times that expected. The above conclusions are based on a proportional mortality analysis. In addition, a study by NIOSH has indicated an excess of cancer deaths among 86 deceased employees of the Louisville plant.^ This data could also have been analyzed on a proportional mortality basis, but at this time we are not certain of this. The above conclusions are not consistent with those found in the TabershawCooper MCA study, nor with Table D of TCA's addendum to the BFG study. The discrepancy is apparently caused by the differences between the numbers of Tooiesez 1. "Louisville VC Study" to A. Vittone from M. N. Johnson ira-tm-a rev. 11/70 utho. in u.s.a. BFG26191 B. H. G. Zwicker August 22, 1974 Page 2 expected deaths calculated in the proportional' mortality analysis used above, and the actual mortality analysis method employed by TcA.^- The numbers of expected deaths in the actual mortality analysis are cal culated from the observed death rates of the U.S. male population having an identical age distribution. An expectation of, say, 40 cancer deaths in a plant's employee population means that 40 people will, on the average, have died of cancer, regardless of the rates of other diseases. Numbers of expected deaths in a proportional mortality analysis are found by computing the proportion of deaths in the age adjusted U.S. male popula tion, and multiplying these proportions by the total number of observed worker deaths. An expectation of, say, 40 cancer deaths in a proportional mortality table containing 200 total deaths would indicate a 20% proportion of cancer deaths in the age adjusted U.S. male population. It is our opinion that an actual mortality analysis reflects true relation ships between observed and expected death rates while the proportional mortality analysis can lead to serious distortions. The differences be tween these two analyses can be more easily seen in the following hypothe tical example. ^ Suppose 20Z of an age adjusted U.S. male population will, on the average, die of cancer, 60% of heart disease, and 20% of other causes. Also assume that the expected number of deaths calculated by the actual mortality analy sis is 200. Then the expectations of actual mortality are: Disease Cancer Heart Other Total Expected Number of Deaths 40 120 40 200 If 40, 120, and 40 plant people had actually died of cancer, heart disease, and other causes, respectively, the plant population would be considered "normal." However, suppose only 80 had died of heart disease, but still 40 of cancer and 40 of other causes. The actual mortality table would be: Disease Cancer Heart Other Total Observed 40 80 40 160 Expected 40 120 40 200 BFG26192 1. The term "actual mortality analysis" refers to the method used by TCA in determining the Standardized Mortality Ratio (SMR). 2004 52?! B. M. G. Zvicker August 22, 1974 Page 3 The rate of heart disease is 33% less than normal. Cancer and other diseases occur at normal rates. The overall death rate is 20% less than normal. The proportional mortality table does not lead to the same conclusion. Multiplying the respective proportions (.2, .6, .2) by the total nuntoer of observed deaths (160), the proportional mortality table is: Disease Cancer Heart Other Total Observed 40 80 40 160 Expected 32 96 32 160 Cancer and other diseases are now indicated as 25% in excess. Heart disease is 17% less than normal. This apparent increase in the rate of cancer is actually due to the lower incidence of heart disease. The Peters and Monson table is a proportional mortality table. Much of the cancer excess indicated in their Table 1 is apparently due to the lower incidence of heart disease found at Louisville and perhaps also at Calvert City. The MCA study and the BFG addendum Indicate that vinyl chloride workers in general and also Louisville employees have significantly lower death rates than comparable age adjusted U.S. male populations. Taking this lower mortality into account, the Peters and Monson table can be corrected to express actual rather than proportional mortality death rates.i This correction is only approximate since it is based just on the Louisville workers included in Table D of the addendum, and not on the combined Louis ville and Calvert City plant populations. The corrected data are displayed in Table 2. Major conclusions are: (1) Total deaths due to all cancers are not significantly higher than expected (41 actual vs. 35.1 expected). (2) Liver and biliary cancers are significantly in excess (8 actual vs. 1.0 expected). (3) Cancer of the brain is significantly in excess (5 actual vs. 1.5 expected). ^ ^ (N Q l< PQ (4) Cardiovascular and renal diseases are significantly lower than expec ted (74 actual vs. 98.1 expected). 03 1, Corrected by multiplying each expected number of deaths in the proportional mortality table by 1.257 expected number of deaths among Louisville employees with El 1.1-3.0 m actual number of deaths among Louisville employees with El 1.1-3.0 108.17 86 03 O CJ B. M. G. Zwicker August 22, 1974 Page 4 The NIOSH data may have also been analyzed on 'a proportional mortality basis. If so, the same correction factor (1.257) could be applied: Cause of Death All Malignancies Liver Malignancy Brain Malignancy Leukemia and Related Malignancy All Others PROPORTIONAL MORTALITY Observed Expected Obs/Exp 26 16.3 1.6 * 6 .4 15.0 * 3 .57 5.3 * 3 1.6 1.9 14 13.69 1.0 ACTUAL MORTALITY Observed Expected Obs/Exp 26 20.5 1.3 6 .5 12.0 * 3 .7 4.3 * 3 2.0 1.5 14 17.2 0.8 * Significant with 95% confidence The actual mortality analysis indicates that liver and brain malignancies are significantly in excess. The proportional mortality analysis also indi cates (incorrectly) that the "all malignancy" category is abnormally high. 3537004 A. M. Fairlie T AMF/JCS:cls cc: A. Vittone R. D. Scott J. L. Nelson W. J. Wilcox / W. C. Becker M. N. Johnson J Ce Schddf R. W. Strassburg E. B. Katzenmeyer, Jr. BFG26194 E. W. Harrington 0. F. Beckmeyer R. M. Kreager - A. L. Schultz R. L. Bowles TABLE 1 OBSERVED AND EXPECTED DEATHS IN VINYL CHLORIDE WORKERS a ICD Number ^ -- 140-205 150-159 155 162,163 193 200-205 -- 330-334 400-468 800-998 960-969 _M MM-- Cause of Death All All cancer Digestive Liver & Biliary Tract Lung Brain Lymphatic & Hematopoetic Other c CNS Vascular Circulatory External Suicide All Other d Observed Expected Obs/Exp 161 41 13 8 13 .5 5 161.0 1.0 27.9 1.5 8.3 1.6 0.8 10 7.9 1.6 1.2 4.2 3.4 1.5 5 8 66 22 10 24 7.1 9.5 . 68.6 24.5 5.3 30.5 0.7 0.8 1.0 0.9 1.9 0.8 a. Expected numbers based on age-time specific proportional mortality rates for U.S. white males ' b. International Classification of Diseases, 7th Revision c. Nasopharynx - 1, prostate - 1, kidney - 1, thyroid - 1, undetermined - 1 d. Includes 2 cirrhosis (4.2 expected) S004C5S TABLE 2 Actual Mortality Analysis Approximate Observed and Expected Deaths In Louisville and Calvert City VC1 Workers, Adjusting for Overall Death Rate of Non-Exposed Age Specific U.S. Male Population I CD Cause of Death All Causes 140-205 All Cancer 150-159 Digestive 155 Liver and Biliary 162,163 Lung 193 Brain 200-205 Lymphatic and Hematopoetic Other Cancers 330-334 Cardiovascular 400-468 and Renal 330-334 CNS Vascular 400-468 Circulatory 800-998 External 960-969 Suicide All Other (Inc. 2 Cirrhosis, 5.3 Expected) Obs erved 161 41 13 8 13 5 5. 5 74 8 66 22 10 24 Expected 202.4 35.1 10.4 1.0 9.9 1.5 4.3 8.9 98.1 11.9 86.2 30.8 6.7 38.3 Obs/Exp .80 * 1.17 1.25 8.00 * 1.31 3.33 * 1.16 .56 .75 * .67 .77 * .71 1.49 .63 * 9oo4S?: * Significant with 95Z confidence BFG26196