Osborne J S, Shy C M, Kaplan B H
Michigan State University, College of Human Medicine, Kalamazoo 49001.
Am J Epidemiol. 1990 Jul;132(1 Suppl):S87-95. doi: 10.1093/oxfordjournals.aje.a115794.
This study investigated a reported cluster of cancer deaths in the small rural community of Bynum, North Carolina. Residents felt the proportion of deaths involving cancer had been increasing since the mid-1960s. To address this concern, cancer mortality was investigated from 1947 to 1985 to determine 1) if the proportion of cancer deaths had increased since the mid-1960s, 2) if it differed from what should be expected based on comparison with the reference population of the state of North Carolina, and 3) if observed mortality was within the bounds of statistical probability. Results indicated that the proportion of cancer deaths remained relatively constant from 1947 to 1964 (ranging from 9% to 14%) but increased steadily after 1965 to a high of 58% (1980-1985). Standardized proportionate mortality ratios adjusted for age, sex, race, and calendar time indicated that from 1975 to 1985 cancer deaths were 2.4 to 2.6 times greater than expected. Cancer mortality exceeded the upper limit of 95 percent Poisson confidence intervals from 1975 to 1985, which suggested that the excess proportion of cancer deaths was not likely to represent random case clustering. These findings serve to illustrate the feasibility of and a methodology for assessing reports of clusters in small populations.
本研究调查了北卡罗来纳州拜纳姆这个小乡村社区中报告的一组癌症死亡病例。居民们感觉自20世纪60年代中期以来，涉及癌症的死亡比例一直在上升。为解决这一担忧，对1947年至1985年期间的癌症死亡率进行了调查，以确定：1）自20世纪60年代中期以来癌症死亡比例是否上升；2）与北卡罗来纳州的参考人群相比，该比例是否存在差异；3）观察到的死亡率是否在统计概率范围内。结果表明，1947年至1964年期间癌症死亡比例相对稳定（在9%至14%之间），但1965年后稳步上升，在1980 - 1985年达到58%的高位。经年龄、性别、种族和日历时间调整后的标准化比例死亡比表明，1975年至1985年期间癌症死亡人数比预期高出2.4至2.6倍。1975年至1985年期间癌症死亡率超过了泊松95%置信区间的上限，这表明癌症死亡人数的超额比例不太可能代表随机的病例聚集。这些发现有助于说明评估小群体中聚集性报告的可行性和方法。
