In clinical trials and other biomedical research there is an interest in summarizing the treatment effect on endpoints across key stratification factors. For example, assessing the change in blood glucose level between treatment and placebo across the gender strata: males and females. There are several ways, such as inverse variance approach or harmonic means approach, to quantify the treatment effect across the strata by combining the estimates from different strata. Mehrotra and Railkar (Statistics in Medicine, 2000) proposed the minimum risk weighting method to estimate the treatment effect across the strata for two binomial proportions. We propose here a simple likelihood based method (combining the likelihood from different strata) to estimate the overall difference in proportions between two treatment groups across strata. Our simulation results for the coverage, size and the expected length of the confidence interval of the proposed method are comparable to the minimum risk approach. We also demonstrate our method using a real world data.