Purpose: Adjuvant! is a well-validated, widely-used analytic model that provides 10-year recurrence and mortality risk predictions for women with breast cancer considering adjuvant therapies, i.e. endocrine or chemical therapies. One study found that, among patients who were at low risk for mortality, those randomized to reviewing Adjuvant! printouts with their doctors chose adjuvant therapy less frequently than controls. Were these patients influenced by the educational content of the printouts, or did some other dimension persuade them to avoid adjuvant therapy? We conducted secondary analyses to explore whether the Adjuvant! printouts influenced patient choice through educational or other means.
Method: The original study randomized 58 oncologists from 14 oncology practices in Cleveland, and San Antonio to review either Adjuvant! printouts (intervention) or general information pamphlets (control) with their patients. After the visit, patients reported their estimate of their 10-year baseline prognosis (survival without adjuvant therapy). Study personnel reviewed charts and recorded whether or not patients took adjuvant therapy. As a measure of patient knowledge, we calculated whether patients were accurate within 5% of the baseline Adjuvant! estimate. For those patients who were at relatively low risk (estimated 10-year survival > 85%), we asked: 1. Were Adjuvant! printouts associated with increased frequency of patient accuracy? 2. Was patient accuracy associated with frequency of adjuvant therapy? 3. Was there any residual association between use of printouts and adjuvant therapy after controlling for patient accuracy? We answered these questions by constructing frequency tables and testing the corresponding null hypotheses using chi-square statistics with one degree of freedom, at a two-sided 5% significance level.
Result: Twenty-eight of 234 women in the intervention and 20 of 171 in the control group were low-risk. For these patients: 1. Use of Adjuvant! printouts was associated with significantly increased frequency of patient accuracy (57% vs 25%, p=0.04). 2. Accurate patients were significantly less likely to take adjuvant therapy (11% vs 38%, p=0.04). 3. In subset analyses, neither the high (p=0.11) nor low (p=0.57) accuracy patients exhibited significant associations between use of printouts and adjuvant therapy. Therefore there was no residual association between use of printouts and adjuvant
Conclusion: Adjuvant! printouts appear to influence treatment choices by improving patient knowledge of prognosis, rather than through other persuasive mechanisms.
Candidate for the Lee B. Lusted Student Prize Competition