Programmed cell death protein 1 (PD-1) blockade is only effective in a minority of patients, prompting the search for combinatorial therapies that increase responses. Identifying effective combinations requires lengthy testing and so far has shown few successes. To accelerate progress Voorwerk and colleagues (Nat Med. 25(6):920-8, 2019) used an adaptive trial design to compare 4 short-course therapies (radiotherapy, cyclophosphamide, cisplatin and doxorubicin) for their ability to improve the tumor immune microenvironment and enhance responses to subsequent PD-1 blockade in women with metastatic triple negative breast cancer, a disease with low response rate to PD-1 blockade. They reported the first phase of the trial that enrolled 12 to 17 patients per arm to “pick the winner” induction treatment. Higher objective response rates (ORR) compared to no induction were observed only in the arm containing doxorubicin, which proceeded to phase II. These results raise a number of questions about testing local versus systemic induction treatments and whether sequencing with PD-1 blockade is appropriate in light of evidence supporting concomitant treatment, at least for radiotherapy. Small imbalances in baseline characteristics can also influence results obtained with limited numbers of patients per arm. We hope that these considerations will help future adaptive, signal-finding combination immunotherapy studies.