Multiple imputation procedures allow the rescue of missing data: An application to determine serum tumor necrosis factor (TNF) concentration values during the treatment of rheumatoid arthritis patients with anti-TNF therapy

• IRENE SCHIATTINO ^[1] ; RODRIGO VILLEGAS ^[1] ; ANDREA CRUZAT ^[1] ; JIMENA CUENCA ^[1] ; LORENA SALAZAR ^[1] ; OCTAVIO ARAVENA ^[1] ; BÁRBARA PESCE ^[1] ; DIEGO CATALÁN ^[1] ; CAROLINA LLANOS ^[1] ; MIGUEL CUCHACOVICH ^[1] ; JUAN C AGUILLÓN ^[1]
  1. [1] Universidad de Chile

     Universidad de Chile

     Santiago, Chile

     
• Localización: Biological Research, ISSN-e 0717-6287, ISSN 0716-9760, Vol. 38, Nº. 1, 2005, págs. 7-12
• Idioma: inglés
• Enlaces
  • Texto completo
• Resumen

  • Longitudinal studies aimed at evaluating patients clinical response to specific therapeutic treatments are frequently summarized in incomplete datasets due to missing data. Multivariate statistical procedures use only complete cases, deleting any case with missing data. MI and MIANALYZE procedures of the SAS software perform multiple imputations based on the Markov Chain Monte Carlo method to replace each missing value with a plausible value and to evaluate the efficiency of such missing data treatment. The objective of this work was to compare the evaluation of differences in the increase of serum TNF concentrations depending on the ­308 TNF promoter genotype of rheumatoid arthritis (RA) patients receiving anti-TNF therapy with and without multiple imputations of missing data based on mixed models for repeated measures. Our results indicate that the relative efficiency of our multiple imputation model is greater than 98% and that the related inference was significant (p-value < 0.001). We established that under both approaches serum TNF levels in RA patients bearing the G/A ­308 TNF promoter genotype displayed a significantly (p-value < 0.0001) increased ability to produce TNF over time than the G/G patient group, as they received successively doses of anti-TNF therapy.