The nonlinear relationship between high density lipoprotein cholesterol (HDL-C) and triglycerides (TG) was examined in two different populations: (1) survivors of myocardial infarction in the /I-Blocker Heart Attack Trial; and (2) a reference population from a health maintenance organization. An empirical evaluation of three mathematical models was done for males and females in each of the two populations. On the basis of higher R2 values and lower standard errors and coefficients of variation of the estimates for the three models, models ( 1) [HDL = Y0 + ((K)(Y,))/(K + TG)] and (2) [HDL = C,, + (C,,,)exp(( -B) (TG))] were preferred. Parameter Y, (and C,) is the minimum value for HDL when TG becomes infinitely large, Y, (andC,,,) is the range of the HDL between TG = 0 and co, and K (and B) is a rate-of-change constant. Comparison of curves between males and females in each population revealed differences in the parameters, indicating that separate models are needed. Narrow confidence bands at the low TG levels for each sex and population group suggest better estimations would be obtained at these values than at elevated TG concentrations.