Incorporating intermediate binary responses into interim analyses of clinical trials: a comparison of four methods