Test for trend: evaluating dose-response effects in association studies