Table 2

Correlation between domains and disease activity, as assessed by the PhGA scale, and regression coefficients and domain weights obtained with the least-median-of-squares robust regression model with an MM estimator

			Multivariate modelling
	Bivariate analysis		Least median of square with MM estimator
Domains	Correlation with PhGA	p Value	Regression coefficient	Standard error	Weight	p Value
Constitutional	0.106	0.005	0.704	0.163	3	<0.001
Lymphadenopathy	0.134	<0.001	0.817	0.089	4	<0.001
Glandular	0.067	0.078	0.407	0.091	2	<0.001
Articular	0.063	0.095	0.489	0.068	2	<0.001
Cutaneous	0.156	<0.001	0.559	0.097	3	<0.001
Pulmonary	0.170	<0.001	1.066	0.115	5	<0.001
Renal	0.125	<0.001	1.090	0.170	5	<0.001
Muscular	0.156	<0.001	1.193	0.191	6	<0.001
PNS	0.197	<0.001	0.944	0.117	5	<0.001
CNS	0.159	<0.001	0.936	0.157	5	<0.001
Hematological	0.041	0.277	0.361	0.126	2	0.004
Biological	0.073	0.053	0.206	0.080	1	0.010

For bivariate analysis, Pearson's correlation coefficient (r) were obtained between the physician global assessment (PhGA) scale and each domain; for each domain, scores ranged from 0 ‘no activity’ to 3 ‘high activity’. All domains were entered in multivariate regression with the least-median-of-squares model with an MM estimator. R²=0.30 for the model. All domains were significantly associated with disease activity (as defined by the 0–10 PhGA numerical scale) in the multivariate model. The weights were derived from the regression coefficient of the multivariate model.
CNS, central nervous system; MM, modified maximum likelihood; PNS, peripheral nervous system.