Principal component analysis of systemic lupus erythematosus (SLE): A proposal for handling data with many missing values
✍ Scribed by Hiromu Yawo; Yoshiya Shinagawa; Yasuko Shinagawa; Tokugoro Tsunematsu
- Publisher
- Elsevier Science
- Year
- 1981
- Tongue
- English
- Weight
- 798 KB
- Volume
- 14
- Category
- Article
- ISSN
- 0010-4809
No coin nor oath required. For personal study only.
✦ Synopsis
Data from 383 Japanese patients with systemic lupus erythematosus (SLE), collected by the schedule method, were analyzed by the principal component analysis. The new method, "combination of variates," reduced the number of missing data. Quantitative indices u&o1 in evaluating the clinical status of the patients were extracted. The 6rst principal component correlated with the prognosis. Eighty-four patients with SLE were classified into four subgroups accord@ to the second and third principal components. The subgroups were compared with 23 patients with progressive systemic sclerosis or dermatomyositis.
Systemic lupus erythematosus (SLE) presents various clinical signs and symptoms, and this pleomorphic status makes difbculty for evaluation and effective therapy. The greater part of patient-related information is qualitative. Such data are termed dichotomous or binary, that is, a certain clinical finding is present or absent. The objective classifkation criteria of SLE have been proposed by the American Rheumatism Association (ARA) (2). The ARA criteria have facilitated the screening of patients, converting the clinical picture into one of quantitative indices. Analysis of the life table implies that there is a certain pattern in the clinical findings which determines the prognosis of a patient (2,3). Are there any quantitative indices that might estimate the prognosis? Categorization of patients has been attempted with a numerical handling of clinical findings, e.g., factor analysis by h4ustakallio et al. (4), and cluster analysis by Miyahara (5).