The demographic and clinical characteristics of 4,284 men who met criteria for inclusion are summarized in and . Overall, 90% of the study population were white, 7% were African American, and 3% were represented in other ethnic groups (eg, Hispanic, Asian, Native American, or mixed). There was a significant association of age at diagnosis and ethnicity. Seventy-seven percent of African American men were diagnosed at younger than age 70 years compared with 68% of whites and 72% of men of other ethnicities. There were also significant differences in the type of site at which men received their treatment. Although the majority of patients in each group were treated in community-based practices, African Americans and men of other ethnicities were more likely to be treated at a VA facility. A higher percentage of whites and others received treatment at an academic institution than did African Americans (8% and 13% v 2%, respectively). A significant association was noted in the type of insurance reported by the various groups. A higher percentage of whites had Medicare than did African Americans and others (46% v 33% and 34%, respectively), likely because of the older age observed in the white cohort. There were significant differences in the highest level of education attained between the three ethnic groups. Thirty-six percent of African Americans and 48% of other men completed some education beyond high school level compared with 63% of white men.
Patient Clinical Characteristics and Primary Treatment Type
There were significant differences in risk classification between the ethnic groups. A lower percentage of African American men presented with low risk disease compared with white men (38% v 46%; P < .01), whereas a higher percentage of African American men presented with high-risk disease compared with white men (32% v 20%; P < .01). There were no significant differences between African Americans and the other ethnic groups. Although there was not a significant difference in the number of comorbidities reported between the ethnic groups, there was a statistically significant difference in health perception. Fewer African American men than white men rated their health as excellent (12% v 21%) or very good (28% v 41%). However, the majority of men in all groups rated their health as excellent, very good, or good.
There was a significant correlation between ethnicity and type of treatment, as African American and other ethnic groups were less likely to have radical prostatectomy (RP) than radiation therapy (RT; P = .03 and .04, respectively) or androgen deprivation therapy (ADT; P < .01 for both groups) compared with whites. Similar percentages of patients underwent active surveillance and other treatments.
We examined the effect of age, prognostic risk, comorbidities, health perception, insurance, education, and type of treatment facility on treatment and mortality. The variables that were associated with treatment in a univariate test—and also that, subsequently, in a multivariate model were independently associated with treatment or changed the effect of ethnicity on treatment—were included. Thus, a multinomial logistic regression model was designed that controlled for D'Amico risk level, age, health perception, number of comorbidities, education level, and insurance status. Treatment facility was not included, as it did not have a significant effect. The results are listed in . White men were 48% less likely to receive ADT than RP compared with African American men (OR, 0.52; 95% CI, 0.31 to 0.89; P = .02). White men also were less likely to receive RT than RP and were more likely to choose other treatment modalities instead of RP compared with African American men; however, this trend did not quite reach the level of statistical significance (P = .08 and .09, respectively). Men of other ethnicities also were more likely to receive ADT than RP compared with white men (P = .07). There were no significant differences in primary treatment between African American men and men of other ethnic groups.
ORs of Primary Treatment Type Versus RP
Because ADT typically is administered in patients with advanced age or higher-risk disease, we examined the association of ethnicity with type of treatment within risk subgroups. After analysis was adjusted for demographic characteristics, whites were 71% less likely to receive ADT than RP compared with African Americans for low-risk disease (OR, 0.29; 95% CI, 0.11 to 0.73; P = .009; ). Whites also were less likely to receive RT than RP for low-risk disease compared with African Americans, but this did not reach statistical significance (OR, 0.67; 95% CI, 0.41 to 1.09; P = .11). For high-risk disease, a similar association with ADT was seen, as whites were less likely to receive ADT than RP compared with African Americans (OR, 0.52; 95% CI, 0.23 to 1.15; P = .10). There were no significant differences between the other ethnic groups and African Americans or whites.
ORs of Primary Treatment Type Versus RP by Risk Level
To determine if the differences in treatment had an effect on prostate cancer–specific mortality, we performed a proportional hazards regression analysis on the ethnic groups for overall or prostate cancer–specific mortality (). Crude hazard ratio values demonstrated a nonsignificant tendency for worse overall and prostate cancer–specific mortality rates for African American men compared with white men (OR, 1.18; 95% CI, 0.82 to 1.70; P = .37; and OR, 1.81; 95% CI, 0.85 to 3.85; P = .12, respectively). However, when analysis was adjusted for risk classification, primary treatment, age, health perception, number of comorbidities, educational level, and type of insurance, the higher risk of overall and prostate cancer–specific mortality in African Americans disappeared (OR, 0.73; 95% CI, 0.47 to 1.14; P = .16; and OR, 0.37; 95% CI, 0.11 to 1.20; P = .10, respectively). No differences were detected between the other ethnic groups and whites or African Americans.
Crude and Adjusted Hazard Ratio of All-Cause and Prostate Cancer–Specific Mortality
To determine which factor or factors had the strongest effect on the risk of mortality in African Americans, a stepwise model was constructed in which each variable was added and the individual hazard ratios were compared (). Starting with the crude model testing only for ethnicity, there was a reduction in risk of mortality for ethnicity with the addition of each variable, except with the addition of the number of comorbidities, to the model. The final model, which included all variables, is what was used to determine the adjusted prostate cancer–specific mortality, as seen in . The variables with the strongest associations with mortality in the models were risk classification and type of treatment (P < .01 in all models). The type of insurance also had a significant effect until construction of the final model, for which it demonstrated a suggestion of correlation (P = .07).
Risk of Prostate Cancer–Specific Mortality by Race With Stepwise Adjustment