Growth velocity assessment in paediatric AIDS: smoothing, penalized quantile regression and the definition of growth failure

The analysis of growth records in paediatric anti-HIV clinical trials plays an important role in trial evaluation. Growth failure may be a manifestation of progressive disease or treatment toxicity, and is commonly specified as a major trial outcome event indicating poor treatment performance. Despite new therapeutic advances against HIV proliferation in infected patients, accurate monitoring and interpretation of somatic growth in paediatric AIDS remains clinically important, in light of uncertainties regarding relationship between viral load reductions and achievement of favourable somatic growth profiles. Our aim in this paper is to construct a criterion for growth failure that discriminates patients whose risk of death subsequent to growth failure is elevated to a clinically significant degree. To construct the criterion, individual growth curves and velocities are modelled using loess smoothing, penalized likelihood quantile regressions are fit to model age-specific growth velocity distributions for gender-stratified cohorts, and proportional hazards model selection is conducted to identify features of velocity series that are informative on the survival distribution. The resulting growth-failure criterion is expected to be useful for disease staging in resource-limited medical environments Where T-cell counts and viral load measures are unavailable. Copyright (C) 2003 John Wiley Sons, Ltd.