P11.28.A HELPING THE DEAD TO TEACH THE LIVING: ADVICE FROM FOUR DECADES OF LEPTOMENINGEAL DISEASE RANDOMIZED CONTROLLED TRIALS

Abstract BACKGROUND Despite 50 years of intensive clinical and laboratory research, the prognosis for patients with leptomeningeal disease (LMD) remains dismal. Recently, potentially paradigm-changing insights into the cellular and molecular anatomy of the leptomeningeal space, and the genomic underpinnings of tumors metastasizing to the cerebrospinal fluid, coupled with novel therapeutic strategies have fueled a renaissance of interest in this disease. Examining the study design, patient-level data, and challenges from all prior randomized controlled trials (RCTs) may help maximize the efficiency and impact of future trials. MATERIAL AND METHODS Using a PRISMA-compliant search with multiple databases, 8 qualifying phase III RCTs were identified. Information regarding study design, patient demographics, eligibility, follow-up, treatment, and outcomes were recorded. Patient-level data was acquired from study sponsors or published Kaplan Meier curves. Summary statistics and sample size calculations were computed using Cox Proportional Hazards and random effects models and an inverse variance technique. RESULTS The 8 randomized phase III studies (1987 -2019) enrolled 429 patients and analyzed 412 (51.5 [25.9] patients/study) from 169 institutions (2.5 patients/institution) over an average of 5.1 years/trial. Two patients were screened for each patient enrolled. Only 1 study achieved its accrual goal; only 2 provided a sample size calculation. Treatment arms included, MTX (4), DepoCyt (4), cytarabine (3), Thiotepa (1), systemic therapy plus DepoCyt or MTX (2) or systemic therapy alone (2). Two studies (52 patients) included only lymphoma LMD patients and 2 (108 patients) only breast cancer LMD patients. Overall, 87% of patients had solid tumors and 13% had lymphoma. Primary endpoints included OS (1 study, 34 patients), neurologic PFS (2 studies, 124 patients), LMD-related PFS (1 study, 73 patients), cytologic response (2 studies, 87 patients), and a composite response measure (2 studies, 94 patients). Median OS was 101 (86,124) days for all patients, 85.2 (63,156) for lymphoma, and 103 (87.6, 142) for solid tumor patients. Only 1 study (outcome measure cytologic response) provided class I evidence. All other studies and endpoints provided class III or IV evidence. Sample size modeling suggests that only NM-related PFS is a practical endpoint for future studies in light of accrual challenges. CONCLUSION The study design of RCTs for LM has been inadequate, outcome measures suboptimal, patient selection and treatment comparisons extremely narrow, there are no pediatric or leukemia randomized trials, and substantial accrual challenges have dramatically limited study size. All of these shortcomings are potentially remediable, and should be addressed in the next generation of RCTs.