Therapeutic interventions for childhood cancer: An umbrella review of randomized evidence

Article type
Authors
Lavasidis G1, Markozannes G1, Panagiotou O2, Trikalinos NA3, Petridou ET4, Voorhies K5, Ntzani EE1
1Evidence-based Medicine Unit, Department of Hygiene and Epidemiology, University of Ioannina School of Medicine
2Department of Health Services, Policy & Practice, Brown University School of Public Health
3Division of Oncology, Department of Medicine, Washington University in St Louis
4Department of Hygiene, Epidemiology and Medical Statistics, School of Medicine, National and Kapodistrian University of Athens
5Department of Biostatistics, Brown University School of Public Health
Abstract
Background: Treatment advancements in pediatric cancer have substantially improved prognosis of the second most common cause of child mortality in developed countries. However, the strength of the supporting evidence has not yet been thoroughly evaluated.

Objectives: To critically appraise the evidence in the field, we performed an umbrella review of meta-analyses (MAs) of randomized controlled trials (RCTs) examining the efficacy and safety of therapeutic interventions for pediatric malignancies.

Methods: We searched PubMed from inception to July 2019 for relevant MAs. For each study, we re-estimated the summary effect size using random and fixed effect models, as well as 95% confidence and prediction intervals. We further estimated the between-study heterogeneity using the I2 metric, assessed evidence of small-study effects and excess significance bias and evaluated the replicability, consistency and quality (AMSTAR2 instrument) of the evidence.

Results: Eleven MAs assessing 54 comparisons and 240 individual study estimates were eligible. The median number of RCTs/MA was 3 (range 2-16), with a median of 841 participants/MA (range 275-8873). Acute lymphoblastic leukemia (ALL), notably the commonest leukemia type, was the most frequently investigated cancer (29 MAs), followed by sarcomas (12 MAs). The most commonly assessed outcomes were event-free survival and overall survival (22 and 8 MAs, respectively). The summary random effects were statistically significant (P75%). There were limited indications for small-study effects (n=1) and excess significance bias (n=1). There was evidence of replicability/consistency of the treatment effect in 7 MAs. Based on previously applied evidence grading criteria (P-value, I2, largest study, prediction intervals, small-study effects, excess-significance), the most concrete evidence pertained to the use of methotrexate and vincristine plus prednisone pulses for ALL, both improving event-free survival. The evidence pertaining to other cancer types was relatively weak. Finally, the quality of the included studies was generally low.

Conclusions: We found few small MAs of randomized evidence of varying consistency assessing therapeutic interventions for pediatric cancer focusing mainly on ALL and usually assessing event-free survival as an outcome. Randomized evidence stemming from adult populations seems to serve as a valuable indirect evidence backup feeding informed clinical decisions. More and better-quality RCTs as well as MAs of individual patient data are needed to increase certainty and precision in the care of pediatric cancer patients.

Patient or healthcare consumer involvement: No.