The Pitfalls of Early Publication of Data in Acute Myeloid Leukemia: A Report from the Eastern Cooperative Oncology Group (ECOG).

Abstract

Abstract Results from clinical trials may change with time. Data presented in initial abstracts or early reports are often different from what appears in subsequent publications, sometimes due to a more rigorous analysis (as in initial response rates) or to maturity of data (as in reports of survival). Survival curves usually provide projected data based on Kaplan-Meier analyses. Such data may change with time when unexpected events occur (as, for example, relapse after T-cell depleted allogeneic transplants) or more patients advance to the initially projected time point. For AML, several groups have reported that there are very few relapses beyond 3 years. At issue is the optimal time for reporting these survival data. Table 1 reports data from 6 consecutive clinical trials addressing induction and post-remission questions in AML patients performed by ECOG. These data compare the 3-year overall survival (OS) and disease-free survival (DFS) for the identical cohorts of patients as presented at 1 year and at 3 years after completion of the study accrual. PC 486 was a phase II study of induction therapy followed by allogeneic or autologous transplant without prior consolidation. E 3489 was a US intergroup study in which patients received post-induction therapy using an allogeneic transplant (allo), an autologous transplant (auto) or chemotherapy (chemo). E1490 was a study in older adults of induction and consolidation with GM-CSF or placebo. E2491 was an intergroup study comparing ATRA induction versus standard chemotherapy in patients with APL. After consolidation, patients were randomized to maintenance with ATRA versus observation (obs). E3993 was a study in older adults comparing priming with GMCSF or placebo. E4995 was a phase II study of intensified induction and consolidation followed by an autologous transplant. For the same cohorts of patients the 3-year OS tends to decrease as the time from completion of study accrual increases. When looking at the DFS, while a decrease may also occur, in several studies there was an improved DFS with time (for example, in E2491, chemo to ATRA, or in E3993 with GM-CSF priming). There were no changes in either OS or DFS beyond 3 years (data not shown). Table 1: Three-year OS and DFS Presented at Different Time Points OS (%) DFS (%) Study One year Three years One year Three years PC 486 auto or allo HSCT 42 → 38 45 → 39 E 3489 allo HSCT 50 → 50 48 → 47 auto HSCT 47 → 46 39 → 39 chemo 56 → 51 36 → 36 E 1490 Induction / consolidation GM-CSF 23 → 22 23 → 24 placebo 15 → 12 16 → 15 E2491 APL induction+ maintenance ATRA + ATRA 87 → 84 72 → 73 ATRA + obs 83 → 80 61 → 60 chemo + ATRA 81 → 75 43 → 58 chemo + obs 54 → 45 16 → 18 E 3993 Induction/consolidation GM-CSF priming 13 → 11 5 → 10 placebo 14 → 14 9 → 9 E 4995 Induction / consolidation / auto HSCT 54 → 48 63 → 53 In conclusion: For large cooperative group studies, treatment results in AML presented one year from conclusion of study accrual are unlikely to dramatically alter the overall results. Nonetheless, some differences may occur, both in OS and DFS, warranting caution in interpreting the data and comparing with other published reports. Such differences are likely to be even greater if data are presented upon completion of study accrual, prior to one year - although, in a cooperative group setting, this should be balanced against the need to release data for future planning. Publication of interim data, before conclusion of the study accrual, may be misleading and are to be strongly discouraged. Three years appear to be the appropriate time for the publication of definitive survival data.