Article Text

OP0153 Validation Study of the Prediction Model for Methotrexate Hepatotoxicity Composed of 13 SNPS Selected by DMET Microarray Profiling in Japanese RA Patients
  1. S. Kumagai1,
  2. G. Tsuji1,
  3. Y. Ichise1,
  4. R. Umeda1,
  5. Y. Uemura1,
  6. Y. Hagiwawa2,
  7. H. Uga2,
  8. H. Kurata2,
  9. K. Misaki3
  1. 1Center for Rheumatic DIseases, Shinko Hospital
  2. 2Central Research Laboratories, Sysmex Corporation, Kobe
  3. 3Rheumatology, Kurashiki Central Hospital, Kurashiki, Japan


Background Efficacy and toxicity of methotrexate (MTX) differs among individual patients, suggesting the influence of genetic variations in enzymes associated with MTX metabolism and folate metabolic pathway. We had investigated 1,971 polymorphisms of 246 enzymes/transporters for 33 RA patients with hepatotoxicity and 38 patients without (the learning cohort) by using the DMET microarray (Affimetrix Inc.) and direct-sequencing (DS) method. We had developed a predictive model of hepatotoxicity by 12 SNPs in 12 genes with sensitivity and specificity of 97% and 89%, respectively, and reported the results in the 2013EULAR meeting.

Objectives To validate our predictive model for adverse events of MTX by another cohort (the validation cohort), and to reduce the number of SNPs in the model by multiple logistic regression methods

Methods More than 300 RA patients treated with MTX were newly recruited from Kurashiki Central Hospital and Shinko Hospital. Forty four patients with hepatotoxicity (AST/ALT >40 IU/L with MTX ≤8 mg/week) were enrolled from the cohort and also 56 patients without hepatotoxicity (AST/ALT ≤40 IU/L with MTX ≥8 mg/week). Genotyping: DNA samples were extracted from whole blood samples, and genotyped 13 polymorphisms by RT-PCR (Applied Biosystems, Inc.) and DS methods. Selected genes consisted of 3 folate metabolism-related genes and 10 genes related to pharmacokinetics. Statistical analysis: We firstly confirmed the validity of genes selection by comparing frequencies of genotypes in the validation cohort by ROC analysis. Multiple logistic regression analysis was used to reduce the number of SNPs and to construct the new model made of fewer SNPs.

Results We added one SNP chosen by cross validation method to the 12 SNPs chosen by stepwise method, and finally established the predictive model for MTX hepatotoxicity which consisted of 13 SNPs from the learning cohort. First, when the validation cohort was tested for the 13 SNPs predictive model by ROC analysis, the model could discriminate patients with hepatotoxiciy in the cohort with AUC, sensitivity, and specificity of 0.84984, 80%, and 86%, respectively (Figure 1). Next, SNPs in enzymes (MTRR and ADORA2A) which were reportedly associated with MTX efficacy/toxicity and selected as members of the 13 genes in the learning cohort were not shown by themselves to a significant association with hepatotoxicity in the validation cohort. Finally, by multiple logistic regression analysis, we obtained a predictive model by a combination of only 4 SNPs, which could discriminate hepatotoxicity with sensitivity and specificity of 68% and 77%, respectively.

Conclusions The predictive model for MTX hepatotoxicity established by the leaning cohort composed by 13 SNPs in genes related to drug metabolism/transport was shown to have a sensitivity of 80% and a specificity of 86% by the validation cohort. By multiple logistic regression analysis, the number of SNPs in the model could be reduced to 4 SNPs with passable sensitivity and specificity.

Disclosure of Interest None declared

DOI 10.1136/annrheumdis-2014-eular.4204

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.