Goulet J R, MacKenzie T, Levinton C, Hayslett J P, Ciampi A, Esdaile J M
Department of Medicine, Montreal General Hospital, PQ, Canada.
J Rheumatol. 1993 Jan;20(1):59-65.
An inception cohort of 87 patients with lupus nephritis was evaluated using a classification tree regression technique. Four relevant outcomes were studied: (1) renal insufficiency (serum creatinine > 5.0 mg/dl); (2) renal failure; (3) death due to renal involvement; and (4) any death due to systemic lupus erythematosus. All 4 outcomes could be predicted by one or more renal severity measures (serum creatinine, 24-h urinary protein excretion, nephrotic syndrome, or duration of prior renal disease), and among those with nonsevere renal disease, with a single disease activity measure (the National Institutes of Health or le Riche index). In general 3 prognostic groups (high, intermediate and low risk) could be identified for each outcome. Our results demonstrate the value of regression tree techniques in studies of prognosis and are compatible with a hypothesis of the interaction of disease activity with organ damage in lupus nephritis.
采用分类树回归技术对一个由87例狼疮性肾炎患者组成的起始队列进行了评估。研究了四个相关结局：（1）肾功能不全（血清肌酐>5.0mg/dl）；（2）肾衰竭；（3）因肾脏受累导致的死亡；（4）因系统性红斑狼疮导致的任何死亡。所有这4个结局都可以通过一种或多种肾脏严重程度指标（血清肌酐、24小时尿蛋白排泄量、肾病综合征或既往肾脏疾病持续时间）来预测，而在非严重肾脏疾病患者中，则可通过单一疾病活动指标（美国国立卫生研究院或勒里什指数）来预测。一般来说，每个结局都可以确定3个预后组（高、中、低风险）。我们的结果证明了回归树技术在预后研究中的价值，并且与狼疮性肾炎中疾病活动与器官损伤相互作用的假说相符。
