Diagnostic accuracy of NUDT15 gene variants for thiopurine-induced leukopenia: a systematic review and meta-analysis

We herein conducted a systematic review and meta-analysis of published studies to estimate diagnostic accuracy of NUDT15 gene polymorphisms for detection of thiopurine-induced leukopenia. Eligible studies were identified through a comprehensive search on PubMed, Web of Knowledge, Cochrane and OpenGrey datasets up to April 2018. The methodological quality of eligible studies was assessed using the QUADAS-2 criteria. The diagnostic odds ratio (DOR) was used as a single measure of diagnostic performance. Sixteen studies including a total of 3538 thiopurine-treated patients fulfilled inclusion criteria for the systematic review. Among these, 16 studies were available for the meta-analysis of rs116855232, 6 studies for rs186364861 and 5 studies for rs554405994 of NUDT15. A higher DOR was found for rs116855232 (8.44, 95% CI: 5.46-13.03), as compared to rs554405994 (4.336, 95% CI 2.924-6.429) or rs186364861 (2.742, 95% CI 1.453-5.175). Results of meta-regression analysis showed that incidence of leukopenia (relative DOR: 0.96; 95%CI: 0.93-1.00, p = 0.037) and leukopenia onset (late vs early leukopenia, relative DOR: 0.41, 95% CI 0.20-0.85, p = 0.0189) significantly influenced diagnostic accuracy of rs116855232. Subgroup analysis for rs186364861 and rs554405994 revealed a significant DOR for early-onset leukopenia (rs186364861: 4.04, 95% CI 1.78-9.20; rs554405994: 2.94, 95% CI 1.74-4.95), but not for late-onset leukopenia (rs186364861: 1.52, 95% CI 0.52-4.43; rs554405994: 2.02, 95% CI 0.93-4.40). The present meta-analysis points to rs116855232, rs554405994 and rs186364861 of NUDT15 as clinically relevant predictors of thiopurine-induced leukopenia. Nevertheless, prospective studies of genotype-guided dosing of thiopurines are warranted to prove clinical benefit and cost-effectiveness of pretreatment NUDT15 gene testing across different populations.