Small sample size of genetic data is often a limiting factor for desirable accuracy of estimated genetic effects on age-specific risks and survival. Longitudinal non-genetic data containing information on survival or disease onsets of study participants for whom the genetic data were not collected may provide an additional "reserve" for increasing the accuracy of respective estimates. We present a novel method for joint analyses of "genetic" (covering individuals for whom both genetic information and mortality/morbidity data are available) and "non-genetic" (covering individuals for whom only mortality/morbidity data were collected) subsamples of longitudinal data. Our simulation studies show substantial increase in the accuracy of estimates in such joint analyses compared to analyses based on genetic subsample alone. Application of this method to analysis of the effect of common apolipoprotein E (APOE) polymorphism on survival using combined genetic and non-genetic subsamples of the Framingham Heart Study original cohort data showed that female, but not male, carriers of the APOE e4 allele have significantly worse survival than non-carriers, whereas empirical analyses did not produce any significant results for either sex.