Biodynamic prediction of neoadjuvant chemotherapy response: Results from a prospective multicenter study of predictive accuracy among muscle-invasive bladder cancer patients.

BACKGROUND: Biodynamic signatures (temporal patterns of microscopic motion within a 3-dimensional tumor explant) offer phenomic biomarkers that are highly predictive for therapeutic response. OBJECTIVE: By utilizing motility contrast tomography, which provides a simple, fast assessment of motion patterns in living tissue, we evaluated the predictive accuracy of a biodynamic drug response classifier in muscle-invasive bladder cancer (MIBC) patients undergoing neoadjuvant chemotherapy (NAC). DESIGN, SETTING, AND PARTICIPANTS: One hundred five consecutive bladder cancer patients suspected of having MIBC were screened in a multi-institutional prospective observational study (NCT03739177) from July 2018 to June 2020, of whom, 30 completed NAC and radical cystectomy. INTERVENTION(S): Biodynamic signatures from treatment-naïve fresh bladder tumor specimens obtained after transurethral resection were measured in living tumor fragments challenged by standard-of-care cytotoxins. Patients received gemcitabine and cisplatin or dose-dense methotrexate, vinblastine, doxorubicin, and cisplatin per institutional guidelines and were followed through radical cystectomy. OUTCOMES MEASUREMENTS AND STATISTICAL ANALYSIS: A 4-level classifier was developed to predict pathologic complete response (pCR) vs. incomplete response utilizing a one-left-out cross-validation protocol to minimize over-fitting. Area under the curve evaluated predictive utility. RESULTS: Thirty percent (9 of 30) achieved pCR. Utilizing the 4-level classifier, biodynamically "favored" (scoring ≥ 3) and "strongly favored" (scoring 4) regimens accurately predicted pCR at rates of 66.7% (4 of 6 patients) and 100% (4 of 4 patients), respectively. Biodynamically "favored" scores predicted pCR with 88% sensitivity and 95% negative predictive value, P < 0.0001. Only 5.0% (1 of 20 patients) achieved pCR from regimens scoring 1 or 2, indicating poor to no response from NAC. Area under the receiver operating curve was 96% (95% Confidence Interval: 79%-99%, P < 0.0001). Future direction involves validating this model prospectively. PRINCIPAL CONCLUSIONS: Biodynamic scoring accurately predicts response in MIBC patients receiving NAC and holds promise to substantially improve the scope of appropriate management intervention.