This meta-analysis is based on 106 évaluations of response from 77 clinical studies about 5-fluorouracil (5FU) treatment with or without leucovorin (LV) in metastatic colorectal carcinoma. Overall, in naive patients, LV is associated with a médian response rate of 31% as compared with a 12% flgure with 5FU alone.Using a forward stepwise multilinear régression analysis, it is shown that as much as 44% of the variance in the reported response rates in naive patients can be accounted for by treatment-related variables (P < 0.001). The significant parameters are LV adjunction (partial R = 0.636), cumulative total 5FU dose (R = 0.344), and 5FU weekiy schedule (R = 0.246). In pretreated patients, the latter parameter is the only significant one (R = 0.443). Unexpectediy, LV administration behaves like an all-or-nothing governor, without any obvions dose-effect relationship.P rotracted 5FU infusion over weeks allows a mean cumulative drug delivery, 3 times higher than bolus regimens (21.3 vs 7.02 g m"^, P < 0.001) and may represent the best clinical approach to influence the 5FUrelated variables. Accordingly, it is suggested that SFU protracted infusion, titrated to the highest tolerable doses and potentiated with low doses of leucovorin, could represent the most efficacious way for using SFU in colorectal disseminated cancer. This hypothesis and its eventual impact on survival should be tested in randomized trials.