Table 2.

Results derived from multivariable Cox regression models^a

	Model A: No medications included		Model B: Corticosteroid included		Model C: AZA/MTX included
	HR (95% CI)	P value	HR (95% CI)	P value	HR (95% CI)	P value
Sex (male referent)	0.14 (0.04 to 0.58)	0.006	0.10 (0.01 to 1.27)	0.076	0.09 (0.02 to 0.52)	0.007
Age at diagnosis (years)	1.07 (1.04 to 1.11)	0.000	1.06 (1.01 to 1.12)	0.018	1.11 (1.06 to 1.16)	0.000
Race (white referent)
Black	3.94 (1.48 to 10.46)	0.006	1.5 (0.3 to 7.4)	0.61	9.5 (2.42 to 37.09)	0.001
Other or unknown	2.03 (0.66 to 6.29)	0.22	0.86 (0.19 to 4.0)	0.85	5.6 (1.3 to 24.3)	0.021
Diagnosis (PM referent)	0.78 (0.24 to 2.54)	0.681	1.14 (0.26 to 5.1)	0.86	0.30 (0.1 to 1.5)	0.141
Sex-diagnosis interaction	6.41 (1.22 to 33.7)	0.028	12.5 (0.65 to 241.5)	0.10	15.7 (1.9 to 128.0)	0.010
IV corticosteroid (vs oral)	-	-	5.6 (1.6 to 20.3)b	0.008	-	-
MTX (vs AZA)	-	-	-	-	0.96 (0.3 to 2.8)	0.938

^aAZA = azathioprine; CK = creatine kinase; IIM = idiopathic inflammatory myopathy; ILD = interstitial lung disease; IV = intravenous; MTX = methotrexate; PM = polymyositis. ^bSteroid association was restricted to the white subset of the study population when stratified models were used. Inclusion of initial CK level, comorbid ILD or cancer (ever or within 3 years of IIM diagnosis) in additional models did not substantively affect results.