OBJECTIVES:  Evaluate the impact on preference estimates of changing the substitutability of treatment alternatives in DCE questions. METHODS:  Adult members of an online panel, completed a web-enabled DCE survey (N=811). Respondents were asked to imagine they had a gene variant associated with risks of health problems. Two constructed interventions to reduce the risk of health problems were described in addition to watchful waiting: preventative surgery and medication. Using a three-alternative labeled experimental design, respondents were allowed to either choose between surgery and watchful waiting (n=193) (Group 1), medication and watchful waiting (n=217) (Group 2), and surgery or medication in addition of watchful waiting (n=401) (Group 3). A random-parameters logit (RPL) model was used to estimate separate preferences for respondents in each of the three groups. Compensating variation (CV) for the average surgery and medication were calculated for each group to measure the expected benefit of adding each intervention to the list of alternatives to prevent future health problems. RESULTS:  Surgery and medication CV values were similar when estimated with and without a second type of intervention in the choice questions. We found that CV varied with the likelihood and severity of future health problems. For example, among respondents who considered only the possibility of choosing medications in the choice questions, CV ranged between $32 and $103, based on the risk level and the severity of the health problem. In general, having only one type of intervention in the choice questions was associated with greater CV for the least valuable intervention. Only with high risk of mild health problems was surgery more desirable than medications. CONCLUSIONS:  The study results provide some assurance that DCE models can be robust to incomplete specifications of relevant substitutes in choice questions, even in cases where clinically-relevant alternatives are imperfect substitutes.