Warning: fopen(/home/virtual/epih/journal/upload/ip_log/ip_log_2024-10.txt): failed to open stream: Permission denied in /home/virtual/lib/view_data.php on line 95 Warning: fwrite() expects parameter 1 to be resource, boolean given in /home/virtual/lib/view_data.php on line 96
Evidence of a dose-response effect is fundamental in making a causal interpretation of an association in epidemiological research. The analysis method known as tests for trend includes score test methods used in traditional analysis, as well as logistic models, which have been widely used since they were proposed in the 1960s, and the likelihood ratio for testing a hypothesis using a log-linear model. Although the score test method using Pearson statistics involves relatively simple calculations, it becomes complicated when adjusting for multiple variables and becomes practically impossible for the qualitative analysis of continuous variables. However, more powerful and diverse dose-response effect analyses are possible if linear models are used. In fact, it is possible to adjust for the effects of multiple variables simultaneously, test for the qualitative associations of a specific variable, and investigate and adjust for the effect of variables that show a proxy effect or interactions with that specific variable in testing the linear trend. Moreover, using such models has the advantage of eliminating randomly selected exposure levels, which can give rise to statistical bias and inconsistency. In this review, we present the theoretical background of conventional score test methods and their application in assessing dose-response relationships, the theoretical aspects and methods of interpreting the results of trend analysis using logistic models, and practical procedures with examples using log-linear models for trend analysis. Programs for SAS and the GLIM statistical package are also included to assist in the application of these methods.