Title: An alternative test for trend in exposure-response analysis.
Author: Dosemeci M, Benichou J
Journal: J Expo Anal Environ Epidemiol 8(1):9-15
Year: 1998
Month: January
Abstract: The Mantel-extension chi-square test for overall trend is commonly used in evaluating dose-response relationships in epidemiologic studies. As illustrated by others, it is not an optimal test for a monotonic dose-response. In addition, the result of this test depends heavily upon the scores assigned to the exposure categories. The selection of the score may have a substantial impact on the test statistic and consequently, on the interpretation of the study results. A monotonic dose-response relationship exists when risk increases (or decreases) with increasing exposure. Although the Mantel-extension test is one of the most cited test for trend, it is not a sensitive test to reflect the incremental risk change of a dose-response relationship which may potentially generate misleading results. We propose an alternative test for the evaluation of monotonic dose-response, which reconfigurates the Mantel-Haenszel chi square test for dichotomous exposure series into a polychotomous exposure series. The proposed test generates chi values that are sensitive to incremental increase (or decrease) in risk at each exposure category and does not require exposure scores. The test is a simple summation of Mantel-Haenszel chi statistics obtained in for 2 x 2 tables over adjacent exposure categories (i.e., the sum of Mantel-Haenszel chi values between categories 1 and 2, 2 and 3, 3 and 4 and 1-1 and 1). Several examples are presented to illustrate that the proposed test generates more realistic chi values than those obtained by the Mantel-extension test for trend.