In this paper the trivariate survival regression model for FGM family of distributions is const­ructed with marginal left-truncated logistic distributions. Two methods (using survival and hazard functions in the first case, and distributional density and ` `conditional'' survival function in the se­cond case) are used when constructing likelihood function for model parameter estimation. Const­ructed survival model was run with the data of the ` `KRIS'' (The Kaunas Rotterdam Intervention Study), which lasted for 22 years from 1972. The results show, that using second case for likehhood function construction gives better approximation for the data, because some additional conditions are considered.