2018 年 20 巻 2 号 p. 57-65
Objective:The objectives were to assess regional differences in the safety outcomes of telaprevir-based triple therapy(T/PR) in Japan and evaluate a suitable generalized linear mixed model for estimating regional differences.
Design and Methods:This study targeted individuals infected with genotype 1 chronic hepatitis C virus registered in a nationwide Japanese interferon database from December 2009 to August 2015. The rate of dropout from treatmentattributable to adverse events was calculated in every prefecture where ≥ 20 cases were reported. We constructed the following four models and evaluated the best-fit model based on Akaike information criterion (AIC) and Bayesian information criterion (BIC):1)prefecture as a fixed-effect,2)prefecture and identified confounding factors as fixed-effects,
3)prefecture as a random-effect,and 4)prefecture as a random-effect and identified confounding factors as fixed-effects.
Results:A total of 25,989 individuals from 38 prefectures were registered during the study period;among them,1,591 from
18 prefectures were included as the study population. The dropout rate ranged from 7.0 to 23.1%among 17 prefectures.
The model considering prefecture as a random-effect and confounding factors as fixed-effects showed the best-fit for the databased on both the AIC (1,108.06)and BIC (1,113.41).
Conclusion:It is difficult to determine if regional differences exist in the safety outcomes of T/PR in Japan because of the limited number of cases. However, the model using prefecture as a random-effect and other confounding factors as fixed-effects would be suitable for estimating parameters that reflect the influence of the prefecture. Further studies using the model would help inform chronic hepatitis C treatment.