Retransplante de fígado: validação externa de um modelo matemático preditivo de sobrevida
Data
2013
Autores
Mattos, Rogerio Obregon de 
Orientadores
Linhares, Marcelo Moura
Tipo
Tese de doutorado
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Introdução: A necessidade de transplante de figado nos diversos paises tem suplantado e muito a oferta de enxertos. Em geral, a sobrevida dos doentes RETX e dos enxertos e menor em relacao ao TX primario. Portanto, e de grande importancia determinar, no periodo pre-operatorio, quais doentes teriam uma melhor sobrevida no RETX, evitando, desta maneira, o RETX oinutilo e aumentando os beneficios de um recurso escasso. Em um estudo previo, Linhares et al. (2006) desenvolveram um modelo matematico preditivo de mortalidade precoce em pacientes retransplantados, modelo que foi validado internamente por Mattos et al. (2012). No entanto, para que um modelo matematico tenha aceitacao e efiCiência clinica e necessaria a sua validacao externa em uma populacao diferente da que o modelo foi construido. Objetivo: Validar externamente o modelo matematico preditivo de sobrevida, criado por Linhares et al. (2006). Metodo: Estudaram-se, retrospectivamente, 217 doentes submetidos a retransplante hepatico no periodo compreendido entre janeiro/2000 e dezembro/2008, em quatro centros Europeus. Os dados dos doentes foram obtidos da base de dados computadorizada da ELTR e dos Servicos participantes. Foram colhidas as seguintes variaveis do receptor: idade, creatinina, urgencia do retransplante e tempo entre o transplante e o retransplante. Foram utilizadas as curvas de sobrevida de Kaplan-Meier e a curva ROC com objetivo de validar o modelo matematico. Intervalos de confianca de 95% foram estimados. Pela tecnica de log-rank, compararam-se as curvas de sobrevida, quando dicotomizadas em dois decis (alto e baixo risco) e quando essas foram dicotomizadas em tercis, de acordo com os pontos de cortes do modelo original: alto risco (> 32), medio risco (24 - 32) e baixo risco (< 24). Resultados: Os resultados deste trabalho mostram igual tendencia de sobrevida em relacao ao trabalho de Linhares et al. (2006) no que se refere as variacoes biologicas, quando comparamos as curvas de sobrevida para cada uma das quatro variaveis estudadas entre ambas as amostras. Quando comparadas, as areas abaixo da curva ROC (aROC) das amostras de derivacao (0,733) e validacao (0,593), observou-se uma diferenca estatisticamente significante (p= 0,005), mostrando uma baixa relacao de sensibilidade e especificidade entre as duas curvas. Observou-se semelhanca nas curvas de sobrevida de Kaplan-Meier das duas amostras, tanto dicotomizadas pelo cut-off (ponto de corte = 30), como pela distribuicao em tercis (<24, 24-32 e > 32). Conclusao: Com este trabalho foi possivel validar externamente, atraves das curvas de sobrevida de Kaplan-Meier, o modelo matematico preditivo de sobrevida no retransplante de figado proposto por Linhares et al. (2006), porem a validacao atraves da curva ROC, a aROC, foi considerada de fraca habilidade discriminatoria.
Introduction: The need of liver transplantation in several countries has been far higher than the availability of grafts. In general, the survival of RETX individuals and grafts is much lower compared to the primary TX. Thus, it is very important to determine preoperatively which individuals might have longer survival from the RETX, thus avoiding the “useless” RETX and enhancing the benefits of a scarce resource. In a previous study, Linhares et al. (2006) developed a predictive mathematical model of early mortality in re-transplantation individuals, which was internally validated by Mattos et al. (2012). Objective: To externally validate the predictive mathematical model of survival designed by Linhares et al. (2006). Method: This retrospective study was conducted on 217 individuals submitted to liver retransplantation from January 2000 to December 2008 in four European centers. Data on the individuals were obtained from the computer database of ELTR and participating centers. The following variables were obtained on the receptor: age, creatinine, urgency of retransplantation and time between transplantation and retransplantation. The Kaplan-Meier survival curve and ROC curve were used to validate the mathematical model. Confidence intervals of 95% were estimated. The survival curves were compared by the log-rank technique, dichotomized in two deciles (high and low risk) and in tertiles, according to the cutoff points of the original model: high risk (> 32), medium risk (24 - 32) and low risk (< 24). Results: The present results evidenced similar tendency of survival compared to the study of Linhares et al. (2006) concerning the biological variations, when survival curves were compared for each of the four variables analyzed between both samples. When compared, the areas below the ROC curve (aROC) of derivation (0.733) and validation samples (0.593) presented significant difference (p= 0.005), revealing low relationship of sensitivity and specificity between the two curves. Similarity was observed in Kaplan-Meier survival curves of the two samples, both dichotomized by the cutoff (cutoff point = 30) and distribution in tertiles (<24, 24-32 and > 32). Conclusion: This study allowed external validation by the Kaplan-Meier survival curves of the predictive mathematical model of survival in liver retransplantation proposed by Linhares et al. (2006). However, validation through the ROC curve, the aROC, evidenced weak discrimination ability.
Introduction: The need of liver transplantation in several countries has been far higher than the availability of grafts. In general, the survival of RETX individuals and grafts is much lower compared to the primary TX. Thus, it is very important to determine preoperatively which individuals might have longer survival from the RETX, thus avoiding the “useless” RETX and enhancing the benefits of a scarce resource. In a previous study, Linhares et al. (2006) developed a predictive mathematical model of early mortality in re-transplantation individuals, which was internally validated by Mattos et al. (2012). Objective: To externally validate the predictive mathematical model of survival designed by Linhares et al. (2006). Method: This retrospective study was conducted on 217 individuals submitted to liver retransplantation from January 2000 to December 2008 in four European centers. Data on the individuals were obtained from the computer database of ELTR and participating centers. The following variables were obtained on the receptor: age, creatinine, urgency of retransplantation and time between transplantation and retransplantation. The Kaplan-Meier survival curve and ROC curve were used to validate the mathematical model. Confidence intervals of 95% were estimated. The survival curves were compared by the log-rank technique, dichotomized in two deciles (high and low risk) and in tertiles, according to the cutoff points of the original model: high risk (> 32), medium risk (24 - 32) and low risk (< 24). Results: The present results evidenced similar tendency of survival compared to the study of Linhares et al. (2006) concerning the biological variations, when survival curves were compared for each of the four variables analyzed between both samples. When compared, the areas below the ROC curve (aROC) of derivation (0.733) and validation samples (0.593) presented significant difference (p= 0.005), revealing low relationship of sensitivity and specificity between the two curves. Similarity was observed in Kaplan-Meier survival curves of the two samples, both dichotomized by the cutoff (cutoff point = 30) and distribution in tertiles (<24, 24-32 and > 32). Conclusion: This study allowed external validation by the Kaplan-Meier survival curves of the predictive mathematical model of survival in liver retransplantation proposed by Linhares et al. (2006). However, validation through the ROC curve, the aROC, evidenced weak discrimination ability.
Descrição
Citação
MATTOS, Rogério Obregón. Retransplante de fígado: validação externa de um modelo matemático preditivo de sobrevida. 2013. 88f. Tese (Doutorado em Ciência CirúrgicaInterdisciplinar) – Escola Paulista de Medicina, Universidade Federal de São Paulo. São Paulo, 2013.