Dubin N, Pasternack B S, Strax P
Cancer Detect Prev. 1984;7(2):87-102.
Breast cancer risk factors were examined in a screened population expected to exhibit selection bias. Data were obtained for 1,383 breast cancer cases and 2,543 randomly selected controls screened at the Guttman Breast Diagnostic Institute between 1968 and 1979. Relative risk (RR) estimates were found to be elevated for the usual factors: greater age, greater weight, early menarche, late first live birth or nulliparity, late menopause, and family history of breast cancer. Adjustment for screening variables (number of screenings, breast symptomatology, and year of first screening) removed substantial bias in RR estimates. The unadjusted analysis tended to reduce the magnitude of RR for family history and weight among screenees in general and for age among women screened only once. Although the unadjusted analysis showed no effects for either previous breast biopsy or religion, adjustment for screening variables indicated that among rescreened women, previous biopsy, Jewish and Catholic religion were associated with increased risk; among women screened once only, these characteristics were associated with decreased risk of breast cancer.
在一个预计存在选择偏倚的筛查人群中对乳腺癌风险因素进行了研究。获取了1968年至1979年间在古特曼乳腺诊断研究所接受筛查的1383例乳腺癌病例和2543例随机选择的对照的数据。发现常见因素的相对风险（RR）估计值有所升高：年龄较大、体重较重、初潮较早、首次生育较晚或未生育、绝经较晚以及有乳腺癌家族史。对筛查变量（筛查次数、乳腺症状和首次筛查年份）进行调整后，RR估计值中的偏差大幅减少。未经调整的分析往往会降低一般筛查人群中家族史和体重的RR值，以及仅接受过一次筛查的女性中年龄的RR值。尽管未经调整的分析显示既往乳腺活检或宗教信仰对RR均无影响，但对筛查变量进行调整后表明，在再次接受筛查的女性中，既往活检、犹太教和天主教信仰与风险增加相关；在仅接受过一次筛查的女性中，这些特征与乳腺癌风险降低相关。
