O'Connor G T, Olmstead E M, Zug K, Baughman R D, Beck J R, Dunn J L, Seal P, Lewandowski J F
Department of Medicine, Dartmouth-Hitchcock Medical Center, Hanover, NH 03736.
Arch Dermatol. 1989 Sep;125(9):1209-17.
To determine whether liver function tests and clinical and demographic information would predict methotrexate-associated hepatotoxicity, we identified 78 patients who had undergone 147 liver biopsies associated with methotrexate therapy for psoriasis. The joint sensitivity of aspartate aminotransferase, alkaline phosphatase, and total bilirubin values in detecting abnormal results from a biopsy specimen obtained after treatment was .86; the predictive value of negative test results was .93. A logistic regression model significantly predicted the presence of abnormal (grade III or higher) liver biopsy specimen results. The concordance index was .92 (perfect, 1.0). Regression coefficients may be used along with information from a specific patient to calculate the predicted probability of an abnormal result from a liver biopsy specimen after treatment. We conclude that this multivariate risk estimation model significantly predicts the likelihood of positive findings from liver biopsy specimens in this patient population. The clinical use of this model awaits further validation.
