Carmelli D, Zhang H, Swan G E
Health Sciences Program, SRI International, Menlo Park, CA 94025, USA.
Epidemiology. 1997 Jul;8(4):378-83. doi: 10.1097/00001648-199707000-00005.
We used tree-structured survival analysis (TSSA), a computer-intensive method of classification, to determine prospectively the relation of the body mass index and the waist-to-calf circumference ratio to coronary heart disease and cancer mortality in 3,155 middle-aged men initially free of these diseases. Applied to coronary heart disease mortality, TSSA identified seven subgroups that differed in profile of risk factors and associated survival. Among the seven subgroups, a small subgroup of older, obese, normotensive men (N = 71) experienced an exceptionally high risk of coronary heart disease deaths over the 33 years of follow-up (34%), similar to the risk of 36% experienced by a larger subgroup (N = 387) of men of similar ages who were less obese and had higher blood pressure levels. We also observed a higher overall risk of coronary heart disease deaths during follow-up (10.3% vs 5.3%) in younger centrally obese men who had low blood pressure levels compared with their counterparts of similar age who were less obese. When applied to mortality from cancer of all sites, TSSA identified five subgroups that differed in survival distributions and profile of risk factors. A subgroup of younger, centrally obese, and ever-smoker men experienced a higher risk of cancer deaths than their counterparts who were less obese (14% vs 8%). Results from these analyses demonstrate the usefulness of a tree-structured analysis for classification of subjects into high- and low-risk survival subgroups.
我们采用树状结构生存分析（TSSA）这一计算机密集型分类方法，前瞻性地确定了体重指数和腰臀围比与3155名起初无冠心病和癌症的中年男性冠心病及癌症死亡率之间的关系。将TSSA应用于冠心病死亡率分析时，识别出了七个亚组，这些亚组在危险因素特征及相关生存率方面存在差异。在这七个亚组中，一个由年龄较大、肥胖且血压正常的男性组成的小亚组（N = 71）在33年的随访期内冠心病死亡风险极高（34%），这与年龄相仿、肥胖程度较低但血压较高的一个较大亚组（N = 387）所经历的36%的风险相似。我们还观察到，与年龄相仿但肥胖程度较低的男性相比，血压较低的年轻中心型肥胖男性在随访期间冠心病死亡的总体风险更高（10.3%对5.3%）。当将TSSA应用于所有部位癌症的死亡率分析时，识别出了五个亚组，这些亚组在生存分布和危险因素特征方面存在差异。一个由年轻、中心型肥胖且曾吸烟者组成的亚组比肥胖程度较低的对应亚组经历了更高的癌症死亡风险（14%对8%）。这些分析结果证明了树状结构分析在将研究对象分类为高风险和低风险生存亚组方面的有用性。
