"INTRODUCTION In January 1972@ Tabershaw-Cooper Associates@ Inc.@ under the sponsorship of the American Petroleum Institute (API)@ began a retrospective cohort mortality study of workers in a sample of United States petroleum refineries. The purpose of the study was to evaluate the general assumption that the health of the American petroleum refinery worker is better than that of his counterpart in the general population and@ if this assumption was invalid@ to characterize the variance from the general population and the factors associated with it. The general approach was to identify a group of 20@000 workers who had been employed in refineries in the past and to determine whether they were alive. The principal findings of the study were: ? The standardized mortality ratio (SMR) for all causes of death was 69@ significantly less than expected. ? For virtually every specific cause examined@ the number of observed deaths was approximately the same or lower (frequently substantially lower) than the expected number of deaths from the mortality experience of the comparable U.S. population. For many of these specific causes of death@ the SMRs were significantly lower than 100. ? The one cause of death with an SMR somewhat@ although not significantly@ above 100 was neoplasms of the lymphatic and hematopoietic system other than leukemia. A subsequent analysis showed that the excess of observed-to-expected deaths was confined entirely to employees hired between 1943 and 1951; it appeared that a disproportionate number of these individuals were hired in 1944 and 1945@ compared with the experience of the res t of the study cohort. Subsequent to that analysis@ the vital status of the study cohort was updated by SRI International through December 31@ 1976. The principal findings of the study update were: ? The number of deaths from all causes remained significantly lower than the expected number. ? For almost every cause of death@ the SMRs were near 100 or substantially lower. Many of these SMRs were significantly lower than 100. ? The ten deaths from benign neoplasms (a cause not examined in the Tabershsw-Cooper Associates analysis) represented a 37% excess over the number expected from the mortality experience of the U.S. white male population. This SMR had decreased over time and was not statistically significant. The computer program used did not permit calculation of the SMR for benign brain neoplasms@ but the SMR for brain cancer was 88. ? A more specific analysis was made of neoplasms of the lymphatic and hematopoietic system other than lymphoma. The SMR for lymphosarcoma and reticu10sarcoma was 107@ and that for Hodgkin's disease was 84. However@ for cancer of other lymphs tic tissue@ the SMR was 164 and had increased over time and become statistically significant. In fact@ the number of deaths from cancer of other lymphs tic tissue during 1975 and 1976 was over three times the expected number@ which suggested that either a chance cluster occurred or the SMR might continue to rise markedly. An analysis of the trend by date of hire showed that for cancer of other lymphatic tissue@ there was a marked clustering of risk by hire date with follow-up through December 31@ 1971@ with an SMR of 210 for those hired between 1940 and 1944. However@ with an additional 5 years of follow-up@ the SMR for those hired In that 5-year interval decreased@ and the SMRs for those individuals hired earlier and later increased so that the clustering by date of hire nearly disappeared. There are several limitations to the original analysis performed by Tabershaw-Cooper Associates and the earlier update carried out by SRI International. In each case@ vital status information was first obtained from the plants. Later@ vital status information was requested from the Social Security Administration@ but only for those whose vital status was then unknown. Theoretically@ this procedure could be criticized. Some deceased study cohort members may have been erroneously reported by the refineries as alive@ and the correct vital status would not have been discovered. In addition@ several important and useful data elements either were missing (such as termination date and race) or had not been updated in a long time (such as employment status and vital status). The goals of this second update@ in addition to updating the vital status through December 31@ 1980@ are to obtain information on race and termination date to improve the quality of the analysis@ to have the Social Security Administration provide independent verification of vital status@ to calculate through special analyses SMRs for benign neoplasms of the brain@ unspecified (benign or malignant) neoplasms of the brain@ and multiple nyeloma@ and to prepare and submit a paper for publication to a peer-reviewed journal. In the original analysis@ the maximum follow-up interval for any individual in the cohort was less than 10 years. With the additional follow-up@ the maximum follow-up has doubled. The additional data elements allow control of the analysis for race@ thus eliminating a confounding variable@ and the ability to calculate accurately the duration of employment. The ascertainment of vital status for all study cohort members through the Social Security Administration effectively answers the theoretical argument that SMRs are low because plants erroneously reported deceased individuals as alive. Finally@ the more specific analyses for subsets of benign neoplasms and cancer of other lymphatic tissue allow a more complete analysis of the two causes of"" death found to be in excess at the end of the first update."