The selection of the Comparison group began in 1957 at Majuro when the group was initiated with 86 individuals matched approximately for sex and age with the exposed group of 86 individuals. Members of the Comparison group were examinedperiodically thereafter at Rongelap or elsewhere along with members ofthe exposed Rongelap population. During 1958-59. after the return to Rongelap island. the numberof persons actively enrolled in the Comparison group was increased to about 150. During the following years up to 1974, another 31 persons were added. In 1974-76, to make up for more persons lost to followup or deceased, another 32 persons were added. No additions to the roster have been made since that time. Whenall enrollees are tallied. including those who have discontinued their participation in the annuai medical examinations, 227 persons have been examined at one time or anotheras part of the Comparison group. Although someof the group werelost to followup, there were 63 deaths recorded through 1987. Some deaths may have occurred in those lost to followup that were not brought to the attention of the Marshall Islands Medical Program. Furthermore, the death rate in subsequentiv added subgroups may not be the same as that for persons in 1957. There ts no way to determine if there is any bias introduced into mortality statistics as a consequence of these events which were beyond the control of the program. However, two points can be made. First. since it is cancer mortality which is specificallv in question, cancer deaths can be expressed in terms of total known deaths. thereby controlling to some extent for uncertainties in the determination of total deaths. Therefore, on the basis of information madeavailable to the Marshall Islands Medical Program. 8 of the 63 known deaths (13%) may have been due to malignant disease. In the United States cancer mortality accounts for 22% of total mortality (Silverberg and Lubera, 1987), and in the exposed Rongelap group it accounts for 19% of total mortality (5 of 26 deaths). Second, cancer deaths can be expressed in person/years of observation, thereby controlling somewhat for persons lost to followup. When this is done the cancer death rate for the 33-year observation period is 171/100,000 (8 possible cancer deaths in 4669 person/years) for the Comparison group overall and 187/100,000 (4 possible cancer deaths in 2136 person, years) for the 86 20048118 personsin the original 1957 Comparison group. The similarity of these numbers does not suggest the introduction of bias in death rates in subsequent additions in the Comparison population. For the Rongelap exposed population. which was Statistically similar in age and sex distribution to the Comparison group whenevaluated in 1982 (Adams et ai.. 1983), this number is 234, 100.000 (5 possible cancer deaths in 2139 person, years). The confirmed or presumptive cancer diagnoses in the Comparison group are given in Table 7, along with cancer deathsin the exposed Rongelap population. Table 8 contrasts the distribution of possible cancer deaths in the Comparison group according to vears of residence on Rongelap with that of the exposed population. One ofthe eight persons dving of possible cancer in the Comparison group was never known to be present on the island. Furthermore,six of the eight spent onlya short time on Rongelap. However.for those six that short time lay between 1958 and 1961, a period when residual radioactivity would have been higher than in subsequent years. One hundred fifty-one persons in the Comparison population were known to be on Rongelap at some time between 1958 and 1961. Of the six that ultimately died of possible cancer, four were among forty-two who were not on Ron- gelap after 1961, whereas two were among the one hundred-and-nine that were seen on Rongelap at a later date (Table 9). It is a statistical oddity that even the latter two individuals were found on Rongelap only onceafter 1961. There are several points that are relevant for those who would apply an epidemiologic analysis to these data: 1. Since the Marshall Islands Medical Program has not maintained a year-round medical presence on the different atolls where examinees may be found. causes of death were obtained in many instances from records and verbal aceounts of health aides and family members living on those atolls and from records and death certificates at the Ebeye and Majuro hospitals. Autopsies are rarely performed in the Marshall Islands. 2. Of the eight deaths thatclinically may have been cancer-related, confirmation by tissue diagnosis is available in only four. In the exposed Rongelap population only three of the five deaths attributed to cancer were confirmed.