NEW ORLEANS—The current system of adverse event (AE) reporting on cooperative group oncology trials “has modest sensitivity and a demonstrable false positive rate,” a review of a phase 3 National Cancer Institute (NCI)-funded trial for pediatric acute myeloid leukemia (AML) at the 55th American Society of Hematology Annual Meeting and Exposition has found.
The report from The Children’s Oncology Group (COG) found that for clinical trial AAML0531, the sensitivity of AE reporting was relatively low for all but one AE and was less than 50% for nine of the 12 targeted toxicities, said Tamara P. Miller, MD, of the Division of Oncology, The Children’s Hospital of Philadelphia, Philadelphia, PA.
Positive predictive value was high for most toxicities; however, “it was not 100% for many, indicating that false-positive results occur for nearly all of the evaluated AEs,” she added.
The overall objective of the study was to improve the efficiency and effectiveness of toxicity reporting on clinical trials and, in the process, learn how to streamline AE reporting on clinical trials.
As in all cooperative oncology groups, COG trial clinical research associates report AEs via case report forms using the National Cancer Institute Common Terminology Criteria system. However, data are lacking “on the sensitivity, specificity, and positive and negative predictive values of AE reporting on pediatric or adult cooperative group oncology trials,” Dr. Miller explained.
Their study sought to determine these operating characteristics for AE reporting for the most recently completed clinical trial for de novo AML, a phase 3 study evaluating standard chemotherapy with or without gemtuzumab with standard chemotherapy, cytarabine, daunorubicin, and etoposide, which enrolled 1,022 patients between August 14, 2006, and June 15, 2010 at 171 treating cancers. For this trial alone, 11,791 AEs were reported, she said. Two abstractors performed chart abstraction for 164 patients (613 courses of chemotherapy for 12 targeted toxicities covering all of the major organ systems: cardiovascular (hypertension, hypotension), respiratory (hypoxia, adult respiratory distress syndrome [ARDS]), focal eosinophilic necrosis/gastrointestinal (anorexia, typhlitis), hematologic (disseminated intravascular coagulation [DIC], infectious disease [microbiologically proven Viridans group sepsis (VGS)], microbiologically proven invasive fungal infection), neurologic (pain, seizure), and renal (acute renal failure).
Presence or absence of each of the 12 toxicities was determined for each chemotherapy course, and sensitivity, specificity, positive predictive value, and negative predictive value determined by comparing COG data to the gold standard of chart abstraction data. Age, gender, race, and ethnicity were determined from COG data.
Compared with all patients enrolled on the trial, the 164 patients were younger (mean age, 8.3 vs. 9.4 years; P = 0.032), and a larger percentage was African American (16.5% vs. 10.4%; P = 0.054). No differences were observed in gender (female, 53.1% vs. 49.8%; P = 0.441) or ethnicity (Hispanic, 16.5% vs. 18.9%; P = 0.753).
In the COG AE report, rates of toxicities ranged from 0.3% of courses (seizure) to 12.5% of courses (anorexia). In the chart abstraction data, rates of toxicities ranged from 0.3% of courses (seizure) to 40.0% of courses (pain).
Sensitivity of COG AE report data was less than 50% for eight toxicities, including hypertension (27.3%), hypoxia (19.8%), ARDS (36.4%), anorexia (30.9%), typhlitis (28.6%), DIC (11.8%), VGS (41.8%), and pain (17.1%). The positive predictive value ranged from 40% (ARDS) to 97.7% (VGS).
“Cooperative group oncology trials have led to dramatic improvements in outcomes for children with cancer, but the current method of reporting adverse events is inefficient and potentially ineffective,” Dr. Miller said. “Work is ongoing to further refine these estimates and to develop improved AE reporting methodologies.”
She said AE reporting is underfunded in cooperative oncology groups: only $2,000 reimbursement is provided for all costs per patient. Excluding AEs, 360 data points per patient were captured on the AAML0531 study, or $5.55 per data point.
The next steps are to complete chart abstraction on the remaining 210 patients; examine AE-level and center-level variation, to determine if there is a pattern to non-reporting or if missing data occurs at random; and to determine if ICD-9 code and resource utilization data from external data sources increases sensitivity and positive predictive value of AE report.