Sociodemographic and clinical predictors of participation in two randomized trials: Findings from the collaborative ocular melanoma study COMS ReportNo. 7

Collecting sociodemographic and clinical data concerning patients who choos e not to enroll in a randomized clinical trial can be useful in assessing t he feasibility of attaining sample size goals during the course of a trial. It can also aid in addressing the extent of generalizability of trial find ings after a trial has ended. The Collaborative Ocular Melanoma Study (COMS ) consists of two multicenter randomized clinical trials to evaluate the ef fectiveness of radiotherapy in comparison to standard enucleation (removal of the affected eye) in prolonging survival of patients with choroidal mela noma. One trial is for patients with large tumors and the other trial is fo r patients with medium-sized tumors. The same baseline sociodemographic and clinical data were collected for both enrolled patients and eligible patie nts who did not enroll in the randomized trials during the first 3 years of patient recruitment. Partial information on nonrandomized patients was col lected thereafter. Recruitment ended in the large tumor and medium tumor tr ials on December 31, 1994, and July 31, 1998, respectively. From November 1 986 through July 1998, 6906 patients with choroidal melanoma were evaluated for the randomized trials, of whom 4191 (61%) were eligible for enrollment . Logistic regression methods were used to identify factors predictive of t rial participation. Sociodemographic factors that appeared to be associated (p<0.15) with participation in the univariable models in the medium tumor trial were age 60 years or older, less than college education, nonmanageria l occupation, current smoking, and residing in the same state as a COMS cli nical center. In the large tumor trial, males, individuals who were not col lege-educated, and individuals residing in the same state as a COMS clinica l center were more likely to enroll. In both trials, clinical determinants of participation were larger tumor dimensions and initial visual acuity wor se than 20/20 in the study eye. In multivariable regression models, the var iables that were significantly predictive of enrollment (P<0.05) in at leas t one of the trials were older age, residence in the same state, larger tum or basal diameter, and worse initial visual acuity in the study eye. Knowle dge of possible sociodemographic and clinical predictors of differentials i n patient participation for nonenrolled patients may help to refine patient education and recruitment strategies for future trials. Patient enrollment in clinical trials may be increased by heightened physician awareness of s uch predictors, strategies for addressing these differences, and enhanced c ommunication between physicians and patients.