TABLE 3

Summary of Cox Regression Analysis on OS

	Univariate analysis			Multivariable analysis adjusting for clinical variables^*
Imaging parameter	Hazard ratio^†	95% CI	P	Hazard ratio^†	95% CI	P
Sum of SUVmax	1.01	1.006, 1.02	0.002	1.01	1.001, 1.02	0.053
Maximum SUVmax	1.11	1.03, 1.18	0.010	1.08	0.99, 1.16	0.085
Average SUVmax	1.13	0.99, 1.30	0.095	1.04	0.89, 1.22	0.60

↵* To adjust for clinical variables, all 6 clinical variables (age, PSA, alkaline phosphatase, pain medication, prior chemotherapy, and Gleason score at diagnosis) were included in multivariable Cox model, and linear prediction scores of failure were calculated for all patients. Linear prediction scores were then used to adjust for clinical variables in evaluation of contribution of sum of SUVmax, maximum SUVmax, and average SUVmax in predicting OS.
↵† For every unit increase in imaging parameter.
All P values are based on Cox regression models, as described in “Materials and Methods” section.