Table 1.

Cox proportional hazards regression model, modeling time to dropout from treatment, as a function of naltrexone treatment assignment.

	Coefficient	Standard Error	Chi-Square	p-value
Low dose	−1.87	0.70	7.15	0.0075
High Dose	−2.43	0.81	8.95	0.0028
Urine	−1.01	0.56	3.21	0.073
Low dose * urine	2.22	0.89	6.17	0.013
High dose * urine	1.37	1.07	1.63	0.20

Placebo (N = 17) is the reference group, against which the low dose-192 mg (N = 19), and high dose-384 mg (N = 21) conditions are contrasted. Urine toxicology, measured at twice-weekly clinic visits, was scored as a dichotomous covariate in this analysis as positive if one or more urines was positive for opioids during the trial, and negative otherwise (negative is the reference group). Values in the table are the regression coefficients for each term in the model, corresponding significance levels, and the point estimate of the hazard ration and its 95% confidence limits. The significant low-dose naltrexone-by-urine interaction term, indicates that the effect of low-dose naltrexone differs between patients with vs. without a positive urine; when urine is positive dropout is similar (and high) on low-dose naltrexone and placebo, while when urines are all negative, dropout rate is low on low-dose naltrexone (compared to high dropout on placebo) and approaches the low dropout rates for the high-dose naltrexone condition.