Heritability of age-at-onset of bone spavin in Icelandic horses estimated by survival analysis

Age-at-onset of bone spavin (BS) in horses was defined as a double-censored survival variable. A small material of binary responses to a radiographic examination of the distal tarsus in 439 Icelandic horses sired by 17 stallions was analysed. In order to test different models larger simulated data were generated according to an exponential proportional hazard model and subject to censoring. The resulting survivor functions were similar. The bias in survivor functions caused by double censoring in the material was reduced by use of the self-consistency (S-C) algorithm. Using a binary threshold model and the Weibull regression model, an analysis of age-at-onset of spavin resulted in severe underestimation of sire variance components in the simulated data. Survival analysis with Prentice and Gloecklerʹs (1978) model led to less biased estimates. Application of this method on the real data resulted in an effective heritability estimate of h2=0.33, which can be compared with an estimate of h2=0.1 based on an analysis of radiographic signs (RS) using the binary threshold model. These results indicate that the age-at-onset of RS, which reflect the predisposition of BS, is a trait with medium-high heritability.