Researchers Question Credibility of Subgroup Analyses in Cancer Trials – Renal and Urology News

Subgroup analyses in cancer trials may not be credible, and their results should be interpreted with caution, researchers wrote in JAMA Network Open.

The researchers analyzed 379 phase 3 randomized trials with subgroup analyses and found that most claims for differential treatment effects were “low or very low credibility” when evaluated using the Instrument for Assessing the Credibility of Effect Modification Analyses (ICEMAN).

The 379 trials included a total of 331,653 patients. There were a median of 8 subgroup factors (range, 1-29) analyzed per trial and a median of 1 outcome evaluated in a subgroup analysis per trial (range, 1-8). Trial investigators evaluated 4148 subgroup effects in total and a median of 9 subgroup effects per trial (range, 1-58). Only 1 trial accounted for multiplicity of testing.

The trial investigators claimed a total of 101 differential treatment effects in 55 trials. ICEMAN ratings were “low credibility” in 69 cases, “very low credibility” in 25 cases, and “moderately credible” in 7 cases. None of the differential treatment effect claims were “highly credible.”

There were 5 cases in which interaction testing “suggested that chance may not explain the apparent effect modification (interaction P value range, ≤.01 to >.005) or was an unlikely explanation (interaction P ≤.005),” the researchers wrote.

They also noted that limited or indirect prior evidence supported the claim of a differential treatment effect in 40 cases, and strong prior evidence supported the claim in 9 cases.

Interaction tests were missing in 29 of the 55 trials claiming differential treatment effects. In an unadjusted analysis, industry-funded trials were more likely to report differential treatment effects in the absence of interaction testing (odds ratio [OR], 6.00; 95% CI, 1.32-42.88; P =.03).

Trials in which the primary endpoint was not met were also more likely to report differential treatment effects in the absence of interaction testing (OR, 4.47; 95% CI, 1.42-15.55, P =.01).

Of the 55 trials claiming differential treatment effects, 27 had forest plots that suggested differential treatment effects. The main effect was missing from the forest plot in 11 of the trials.

Based on these results, the researchers concluded that “oncology subgroup analyses should be interpreted with caution, and improvements to the quality of subgroup analyses are needed.”

Disclosures: Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of disclosures.

This article originally appeared on Cancer Therapy Advisor