The long-awaited results of the pivotal phase III trial of autologous dendritic cell vaccination in patients with newly diagnosed glioblastoma have been published in JAMA Oncology in November 2022.1 This trial started recruitment in 2007 and was initially designed as a randomized (2:1) comparative trial of DCVax added to standard of care versus placebo, with progression-free survival as the primary endpoint. Patients in the placebo arm (n = 99) were allowed to cross over to receive vaccine treatment after tumor progression, and patients in the vaccination arm were allowed to continue treatment. Median progression-free survival was 6.2 months (95% CI, 5.7–7.4) and 7.6 months (95% CI, 5.6–10.9) for DCVax and placebo, respectively.
During the conduct of the trial, the limitations and difficulties of assessing progression-free survival became apparent, and, after completion of enrollment, the initial design was modified to restrict analyses to the overall survival outcomes of the 232 patients initially randomized to DCVax, which were to be compared with pooled historical control data inferred from published data. An increase in survival of almost 3 months with DCVax compared with that observed in historical controls has been reported (19.3 vs 16.5 months; HR, 0.80; 98% CI, 0.00–0.94).
To evaluate these results in the context of other trials, a few unique characteristics need to be highlighted. The outcome is reported from randomization (day 1 of vaccination), which occurred over 3 months after initial surgery, during which time patients were treated with standard temozolomide and radiation and underwent apheresis and vaccine production. This 3-month interval inevitably leads to a lead-time bias — patients with the worst prognosis progressing within 3 months of diagnosis were not included. Although not detailed in this publication, the authors reported in 2018 that, of 1599 screened patients, 1202 had glioblastoma and were eligible. Vaccine production or apheresis was unsuccessful (reasons being not enough tumor lysate or tissue not processed, etc) in 201 patients, (presumed) radiological or clinical disease progression occurred in 412 patients, and an additional 75 patients did not participate due consent withdrawal or investigator decision. Ultimately, only 331 patients (28%) out of 1202 eligible patients at initial surgery were randomized.
Patient characteristics are not reported in detail (some information can be found in Supplement 2). All patients had undergone prior tumor resection (fresh tumor tissue was required for vaccine production!), and, in the “historical control” trials, up to 13% of tumors were nonresectable. MGMT was methylated in 41% of patients. The outcome of the patients included in the placebo arm (n = 99) is not reported yet; an analysis of the 64 patients in the placebo arm who received DCVax at the time of presumed progression (or pseudoprogression) indicates a median survival of 13.2 months (95% CI, 9.7–16.8) from the date of clinical or radiological progression.
The authors created a synthetic control looking at trials that randomized patients only after the end of radiochemotherapy and pooled the patients included in the respective control arms (n = 1366). Individual patient data were not available for the controls; thus, the information had to be extracted and extrapolated from published data. The presumed benefit of DCVax seems to be restricted to patients with MGMT-methylated tumors (Supplement 2, eFigure 1). The authors do not provide an explanation for this observation nor for the counterintuitive benefit seen in the elderly (age ≥65 years) and in those with only partially resected tumors.
The results are intriguing; yet, it remains to be demonstrated that the effect observed is truly due to DCVax rather than a mere reflection of patient selection. There is a strong movement to allow for synthetic controls to replace randomized trials and limit the number of patients to be treated as controls, particularly in diseases like glioblastoma, where better treatments are urgently needed and progress has been slow. But, what magnitude of an effect do we want to see by accepting historical external controls in lieu of randomized trials for proof of efficacy? Being too stringent carries the risk of rejecting a potentially effective treatment strategy that could ultimately unfold its full potential if used in an optimized context or in combination, whereas criteria that are too relaxed will expose patients to ineffective treatments, toxicity, and costs and will ultimately hamper the development of more effective innovative therapies. Timely and appropriate trial design with meaningful and reproducible endpoints, efficient enrollment, and adequate funding remain the most efficient way to provide patients access to better treatments with proven efficacy, approved and reimbursed.
Roger Stupp reports having acted as a paid consultant to Northwest Biotherapeutics, sponsor and manufacturer of the DCVax trial discussed here.