Evaluation of mixed response in tumor size and survival in patients with rare cancers treated with dual checkpoint inhibitor therapy (DART SWOG S1609)

Background Mixed response, where different lesions within the same patient show discordant responses to treatment, remains poorly understood. To better understand the complex effects of mixed response on patient survival, we devised three different definitions of mixed response. This retrospective analysis provides the first evaluation of the association between mixed response and survival outcomes in patients with rare cancers treated with dual checkpoint blockade using ipilimumab plus nivolumab, based on data from 52 baskets in the DART SWOG S1609 trial. Methods We included 438 patients with Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1-measurable disease and at least two target lesions, after exclusions for ineligibility, early death, or missing data. Overall survival (OS) and progression-free survival (PFS) were compared using log-rank tests and Cox regression, stratified by basket and using a day 65 landmark. A mixed response was evaluated using three definitions: Method 1—RECIST discordance across lesions; Method 2—presence of ≥1 lesion with >5 mm increase or ≥1 with >5 mm decrease; and Method 3—same as Method 2 but with a 1 mm cut-off. Results Mixed response was significantly associated with worse OS and PFS using both Method 1 (OS: HR 1.80; PFS: HR 1.58) and Method 2 (OS: HR 1.55; PFS: HR 1.57) compared with “all SD/stable per lesion”. Among patients classified as “SD by RECIST”, those who exhibited a mixed per lesion response according to Method 1 had significantly worse OS (median 9.9 months (8.8–12.4)) than those with a non-mixed per lesion response (median 22.7 months (19.2–32.0)). Further stratification showed that any lesion increasing >5 mm was linked to worse outcomes (OS: HR >2.37, p<0.05). Conclusions Mixed response was significantly associated with worse survival outcomes in patients treated with dual immune checkpoint inhibitors, even among those with RECIST-defined stable disease. Our findings suggest that the “worst”-responding lesion drives prognosis, underscoring the limitations of RECIST in capturing clinically relevant heterogeneity. This study highlights the need to incorporate lesion-level assessment into immunotherapy decision-making and provides a foundation for guiding earlier transition to the next line therapies or other therapeutic options. Future studies integrating molecular biomarkers are warranted to refine response evaluation criteria and optimize immune checkpoint inhibitor-based strategies.