The results of chemotherapy are not fully satisfactory in many cases, particularly solid cancers. Therefore, it seems useful if the effective anticancer drugs can be selected for each patient using the screening methods. In such a background, we developed a new anticancer-drug sensitivity testing method that overcame several defects in the existing method, which is the collagen gel droplet embedded culture drug sensitivity test (CD-DST) and satisfies the following requirements: a high success rate, ability to assay biopsy specimens, and quantification of the anticancer effects without contamination with fibroblasts. Under the biological exposure condition, the sensitivities of the cancers were similar to their clinical response rate and there was a statistically significant correlation between clinically reported response rates and the response rates obtained by CD-DST (P < 0.01). Furthermore, the true positive rate was 79.8%, and the true negative rate was 88.8%. Sensitivity and specificity were 88.2% and 80.6%, respectively, resulting in an overall predictive accuracy of 84.1% (154/183). The CD-DST not only shows high predictive accuracy for humans, but because of the high correlation between the results of in vitro and nude mice assays, if it also is practical as a pre-clinical screening, it can easily provide these predictions.