Background: Adaptive Conjoint Analysis (ACA) is increasingly applied in studies of patient treatment preferences, with encouraging results [e.g. Fraenkel et al, Ann Rheum Dis 2004]. This indirect technique elicits preferences for treatment attributes by asking respondents to tradeoff between these. Its similarity to real-life decision-making is said to make it more valid for preference assessment than direct techniques (e.g. Treatment Tradeoff Method, TTM). ACA presents pairwise comparisons and tailors these to the individual's responses. Attribute-level utilities are derived by OLS regression analysis.

Purpose: To investigate whether ACA is a viable option as decision support tool within oncology, by examining the (a) feasibility and test-retest reliability of ACA, (b) consistency of ACA findings, and (c) concordance of findings between ACA and TTM.

Methods: Randomly selected disease-free rectal cancer patients are interviewed who participated in a multicenter trial assessing the benefit of adding pre-operative radiotherapy (PRT) to surgery [Kapiteijn et al., NEJM 2001]. Treatment preferences are assessed based on tradeoffs between side effects (fecal incontinence, sexual problems) and local control (and survival, ACA). Half of the respondents are asked to perform the ACA-task twice. Using TTM, minimally-desired benefit (local control) from PRT+surgery is assessed, given pre-defined risks of side effects. Based on ACA-utilities and TTM, and given the costs and benefits of PRT, respondents' true treatment preference are computed.

Results: As of May 25th, 40 respondents have been recruited. All respondents except one completed the ACA-task. Fifteen (39%) performed the ACA-task twice, at 17±4 days interval. The 17/39 respondents who were presented with one or two dominant choices preferred the best option in 22/27 (81%) cases. Respondents' attribute-level utilities were consistent for 92% (local control), 74% (survival), 64% (incontinence), and 51% (sexual problems). Incontinence was most important in determining preference (30%±13), followed by local control (28%±9), survival (23%±10), and sexual problems (19%±9). Over time, mean importance scores were stable (paired t-tests ns) yet individual scores varied greatly (Pearson's r 0.05-0.39). Comparing ACA- and TTM-based preferences, only 17/36 respondents were congruent.

Conclusion: Findings suggest stable results at group-level yet unstable results at individual-level. Treatment preferences clearly diverge depending on assessment method. It is as yet questionable whether ACA is suitable for assessing individual treatment preferences within oncology.