Development and validation of a model that predict

Authors: Nicolas Penel JeanPierre Delord MarieEdith Bonneterre Thomas Bachelot Isabelle RayCoquard JeanYves Blay Laurent Benjamin Pascal Cécile Borel Thomas Filleron Antoine Adenis Jacques Bonneterre

Publish Date: 2009/02/10

Volume: 28, Issue: 1, Pages: 76-82

PDF Link

Abstract

Objective Selecting patients for phase 1 studies remains challenging Given the lack of clear and reliable guidance for the estimation of life expectancy we retrospectively assessed predictive factors of early death within 90 days following inclusion among these patients Methods Two hundred fiftyseven consecutive cancer patients enrolled in phase I studies investigating cytotoxics at Oscar Lambret Cancer Center and Institut Claudius Regaud were included in the development database Univariate and multivariate analyses logistic regression model were undertaken to determine the prognostic factors A probability tree described the rate of early death in the different prognostic subgroups This prognostic model was then evaluated on a second independent cohort of 128 patients treated at Léon Bérard Cancer Center Results The median overall survival was 84 months in the dataset population and the rate of early death was 15 In multivariate analysis the two prognostic factors for early death were albumin 38 g/l OR = 521 and lymphocytes 700/mm3 OR = 388 According to these two parameters three prognostic subgroups were defined with early death rates of respectively 8/121 6 19/119 16 and 13/17 76 In the validation dataset the rates of early death according to three prognostic groups were 13/68 19 20/57 35 and 3/3 100 respectively Conclusion We do not recommend the enrolment of patients with albumin level below 38g/l and lymphocytes count below 700/mm3 in phase 1 trial investigating cytotoxics Our model is helpful to discriminate “patients with reasonable life expectancy” as defined in most phase 1 protocols

Keywords: