Homozygote CRIM1 variant is associated with thiopurine-induced neutropenia in leukemic patients with both wildtype NUDT15 and TPMT

Table 4 Comparison of star-allele-based diplotyping vs. the gene-wise variant burden (GVB) method for predicting thiopurine toxicity in pediatric ALL subjects

Phase	Method	Molecular phenotype	Last-cycle 6-MP DIP			Sensitivity	Specificity	PPV	NPV	Accuracy
Phase	Method	Molecular phenotype	≤ 25%	> 25%	Total	Sensitivity	Specificity	PPV	NPV	Accuracy
Discovery	CPIC NUDT15 and TPMT metabolizer	PM + IM	10	46	56	0.526	0.796	0.179	0.952	0.775
	CPIC NUDT15 and TPMT metabolizer	NM	9	179	188
	GVB^NUDT15,TPMT	≤ 0.3	10	42	52	0.526	0.813	0.192	0.953	0.791
	GVB^NUDT15,TPMT	> 0.3	9	183	192
	GVB^{NUDT15,TPMT,CRIM1}	≤ 0.3	11	32	43	0.579	0.858	0.256	0.960	0.836
	GVB^{NUDT15,TPMT,CRIM1}	> 0.3	8	193	201
		Total	19	225	244
Replication	CPIC NUDT15 and TPMT metabolizer	PM + IM	13	11	24	0.520	0.784	0.542	0.769	0.697
	CPIC NUDT15 and TPMT metabolizer	NM	12	40	52
	GVB^NUDT15,TPMT	≤ 0.3	13	10	23	0.520	0.804	0.565	0.774	0.711
	GVB^NUDT15,TPMT	> 0.3	12	41	53
	GVB^{NUDT15,TPMT,CRIM1}	≤ 0.45	16	10	26	0.640	0.804	0.615	0.820	0.750
	GVB^{NUDT15,TPMT,CRIM1}	> 0.45	9	41	50
		Total	25	51	76
Combined	CPIC NUDT15 and TPMT metabolizer	PM + IM	23	57	80	0.523	0.794	0.288	0.913	0.756
	CPIC NUDT15 and TPMT metabolizer	NM	21	219	240
	GVB^NUDT15,TPMT	≤ 0.3	23	52	75	0.523	0.811	0.307	0.914	0.772
	GVB^NUDT15,TPMT	> 0.3	21	224	245
	GVB^{NUDT15,TPMT,CRIM1}	≤ 0.45	28	60	88	0.636	0.783	0.318	0.931	0.763
	GVB^{NUDT15,TPMT,CRIM1}	> 0.45	16	216	232
		Total	44	276	320

Prediction accuracies for the last-cycle 6-MP DIP of star-allele-based Clinical Pharmacogenetics Implementation Consortium (CPIC) practice guidelines on NUDT15 and TPMT were compared with the quantitative GVB^NUDT15,TPMT and GVB^{NUDT15,TPMT,CRIM1} methods in the discovery, replication, and combined cohorts. GVB cutoffs were determined by maximizing Youden’s index
IM intermediate metabolizer, PM poor metabolizer

ISSN: 1479-5876