Table 2.

Methylation cluster risk membership retains prognostic significance by multivariate analysis

		Univariable		Multivariable
	n (%)	HR (95% CI)	P*	HR (95% CI)	P*
Model 1: Clinical features and cluster risk retained by multivariable analysis
Cluster risk group
High vs low	52 (48)	1.95 (1.24-3.08)	.003	2.02 (1.25-3.27)	.004
Clinical features
Karyotype			.13
Abnormal (not complex) vs normal	33 (31)	1.35 (0.82-2.24)	.24	1.43 (0.85-2.39)	.174
Complex vs normal	9 (8)	2.20 (0.93-5.20)	.071	2.79 (1.15-6.75)	.023
Bone marrow blasts, %			.014
5-10 vs <5	23 (21)	2.14 (1.22-3.77)	.008	2.18 (1.23-3.86)	.007
11-30 vs <5	23 (21)	1.77 (1.01-3.09)	.046	1.40 (0.79-2.49)	.252
Model 2: Clinical features, somatic mutations, and cluster risk retained by multivariable analysis
Cluster risk group
High vs low	52 (48)	1.95 (1.24-3.08)	.003	1.60 (0.95-2.71)	.076
Clinical features
Karyotype			.13
Abnormal (not complex) vs normal	33 (31)	1.35 (0.82-2.24)	.24	1.30 (0.75-2.28)	.353
Complex vs normal	9 (8)	2.20 (0.93-5.20)	.071	2.91 (1.16-7.28)	.022
Bone marrow blasts, %			.014
5-10 vs <5	23 (21)	2.14 (1.22-3.77)	.008	2.08 (1.16-3.73)	.014
11-30 vs <5	23 (21)	1.77 (1.01-3.09)	.046	1.55 (0.85-2.83)	.15
Somatic mutations
RUNX1 mutated vs not mutated	19 (18)	2.12 (1.22-3.68)	.008	1.97 (1.07-3.64)	.031
EZH2 mutated vs not mutated	9 (8)	2.91 (1.37-6.18)	.005	2.19 (0.91-5.24)	.079
TP53 mutated vs not mutated	9 (8)	3.16 (1.41-7.08)	.005	2.40 (0.96-6.01)	.062

^*P values for individual categories within variables were calculated using the Wald test. P values for full variables correspond to a log-rank test. Multivariable models were constructed by optimizing the Aikake Information Criterion, which is why some multivariable P values are <.05.