Table 2.

Multinomial model without imputation of missing variables, comparing men who remain in the active surveillance program, men who left the program on a clinician's recommendation, and men who self-elected to leave the program. Men with missing variables are excluded from the regression model^a,b

	RRR (95 % CI)c
	Left on physician's recommendation	Self-elected to leave
Age at diagnosis
40 to <60 years	1 (reference)	1 (reference)
60 to <70 years	0.97 (0.60–1.57)	0.55 (0.30–1.02)
≥70 years	1.08 (0.63–1.85)	0.16 (0.06–0.43)
PSA level at diagnosis
<4 ng/mL	1 (reference)	1 (reference)
4 to <10 ng/mL	1.71 (1.17–2.48)	1.95 (1.09–3.50)
≥10 ng/mL	0.80 (0.38–1.66)	0.74 (0.21–2.65)
Race/ethnicity
White	1 (reference)	1 (reference)
Black	1.24 (0.64–2.42)	0.67 (0.19–2.37)
Other	1.28 (0.50–3.27)	0.43 (0.05–3.42)
Family history of prostate cancer
No	1 (reference)	1 (reference)
Yes	1.10 (0.75–1.62)	1.24 (0.69–2.21)
Family history of any cancer
No	1 (reference)	1 (reference)
Yes	1.17 (0.83–1.65)	1.26 (0.74–2.15)
Education
Less than college	1 (reference)	1 (reference)
Bachelor's degree	1.06 (0.65–1.73)	0.55 (0.26–1.14)
Graduate degree	0.82 (0.52–1.29)	0.64 (0.34–1.21)
Marital status
Not married	1 (reference)	1 (reference)
Married or partnered	0.86 (0.53–1.38)	0.40 (0.15–1.03)

CI confidence interval, RRR relative risk ratio

^aExcludes patients who withdrew from the program without being treated (n = 31), are no longer in the Johns Hopkins program but are being followed elsewhere (n = 69), were lost to follow-up (n = 15), or are deceased (n = 46)

^bIn the multinomial model, the reference category is men who remain in the active surveillance program. The models simultaneously adjust for all covariates listed in the table. Only completely ascertained cases are included in the regression model

^cBold text indicates significance at the p < 0.05 level