The aim of this study was to determine the most appropriate set of weights with which to calculate the number of sound-equivalent teeth (T-Health index) against perceived oral health, which was used as a proxy of oral health status. This study used data from 5,057 dentate subjects, > or = 30 yr of age, who ere participating in the Finnish Health 2000 Survey. Subjects provided information on socio-demographic characteristics, behaviours and perceived oral health, and had a clinical examination. The T-Health index was calculated by assigning different weights to missing, decayed, filled, and sound teeth. Thirty-six alternative sets of weights were evaluated. The most appropriate set of weights was judged by the strength of the adjusted association between the T-Health index and levels of perceived oral health in ordinal logistic regression models and by the invariance of this association according to the extent of restorative treatment (non-significant statistical interaction). Among the 36 sets of weights used to calculate the T-Health index, assigning twice the weight of a decayed tooth to a filled tooth whilst keeping the weight for a filled tooth < or = 0.20 provided the strongest association with levels of perceived oral health and did not vary according to the extent of restorative treatment.