OBJECTIVES:

We aimed to systematically examine the degree of concordance between clinician-reported adverse events (AEs) and patient-reported outcomes (PROs) in supporting the same treatment arm in cancer randomized controlled trials (RCTs). We also examined how authors summarized this information into treatment recommendations.

METHODS:

We conducted a literature search in PubMed/Medline and Cochrane Library to identify all RCTs published between January 2004 and April 2017 in the highest incidence cancer types: breast, colorectal, lung and prostate cancer. In each RCT, we determined the number of PRO scales reporting a statistically significant difference between arms and the number of symptomatic AEs (any grade) reported at least 5% more frequently in one treatment arm versus the other. We then classified PROs and symptomatic AEs as better, equal or worse in the experimental arm as compared to the control arm. Authors’ recommendations were extracted from text and framework method was used for qualitative data analysis.

RESULTS:

We identified 207 RCTs. In 74.4% of studies AEs were assessed using the NCI-CTCAE, while the EORTC QLQ-C30 was the most frequently used PRO questionnaire (47.3%). Concordance between PROs and clinician-reported symptomatic AEs was found in only 35.8% of RCTs. In 50.2% of studies, clinician-reported AEs were more frequent in the experimental arm, while PROs supported that arm. The opposite situation was found in 14.0% of trials. Sixty-eight percent of RCTs did not publish a secondary paper focusing on PROs. Out of these studies, when a concordance was found authors made treatment recommendations based only on clinical outcomes in 38.6% of cases. In the discordant RCTs, this percentage increased to 47.6%.

CONCLUSIONS:

We found a substantial discordance between clinician-reported AEs and PROs in supporting the same treatment arm in RCTs. We also found that authors’ recommendations are more likely to be based only on clinical outcomes in discordant RCTs.