Background: That mortality is not increased in rhGH-treated patients when adjusting for birth characteristics was recently published (1). When applying a developed mortality model of the general population, the observed and expected deaths in rhGH-treated IGHD, ISS and SGA patients (n=3847) where 21 and 21.99, respectively. The model includes gender, age, calendar year, gestational age (GA), birth lengthSDS (BLSDS), birth weightSDS (BWSDS) and congenital malformations. When the model only adjusted for gender, age and calendar year (as for traditional SMR) the number of expected deaths was 14.68.
Objective and hypotheses: How much of the expected deaths in rhGH-treated patients can be explained by the different birth characteristics.
Method: Multiple stepwise regressions of the Poisson mortality model was performed, evaluating the results of different combinations of GA, BLSDS and BWSDS in addition to gender, age and calendar year with interactions between variables included as in the published model (1). Malformations were not included in these models causing the exact numbers to differ slightly (<2%) from previous calculations (1).
Results: As much as 35% of the expected number of deaths (7.9/22.4) can be explained by different combinations of GA, BLSDS and BWSDS as visualized in the figure below. Depending on the order the variables are included, the number of deaths explained by GA became 30,40,43,50%; explained by BLSDS became 20,40,43%; and explained by BWSDS became 17,30% (see figure).
Conclusion: As much as 35% of expected deaths among GH-treated children can be explained by various combinations of birth characteristics; GA, BLSDS and BWSDS; valuable information when validating the Swedish mortality model in countries missing information. References: 1) Albertsson-Wikland, K et al. JCEM 2016:101. doi: 10.1210/jc.2015-3951
10 - 12 Sep 2016
European Society for Paediatric Endocrinology