Case-control studies are often used to explore the relationship between the rare disease and potential risk factors. When the confounding factors are difficult to be quantified, matching designs are used to control the confounding effects. But this results in highly stratum-specific intercepts. Breslow & Day (1980) adopt the conditional approach to eliminate the intercepts. In this paper, we introduce a new goodness-of-fit test which is proposed by Chen & Wang (2010). This test statistic is constructed by the difference between the estimates of the second moment of covariate which estimated by the conditional m.l.e. and nonparametric m.l.e., respectively. We assess the performance of the proposed method through simulation studies, and analyze two real datasets for illustration.