A new method based on the recently developed relative risk approach is introduced, and applied to data from Italian centenarian study (965 subjects aged from 13 to 109 years old) for investigating influences on longevity by Tyrosine Hydroxylase (TH) gene variability. The strategic parameterization enables the model to disentangle the various ways by which HUMTHO1.STR alleles (alleles 6, 7, 8, 9, 10*, 10, as defined according to the number of repeats) may contribute in reducing or increasing the hazard of death with different patterns of influences. Among all the alleles, we have found that allele 10* (10 imperfect repeats) shows a remarkable dominant and beneficial effect that reduces the log hazard of death in an additive manner. The results confirm that HUMTHO1.STR polymorphism is involved in the modulation of human longevity.