4.2. Poor reporting of harms in RCTs, particularly for non-pharmacological interventions
Evidence across various medical domains suggests the reporting of harms in clinical trials has been especially inadequate and receives less attention than efficacy outcomes (110-113). As one group of authors noted, “Reporting harms may cause more trouble and discredit than the fame and glory associated with successful reporting of benefits” (114). Breau also observed that, in general, “[t]rialists may not evaluate adverse outcomes because they believe the safety of the intervention has already been established. However, this assumption is often invalid since the adverse effects of an intervention may differ depending on the indication or population subjected to treatment” (57).
To help remedy this, the CONSORT group issued a statement extension in 2004 focusing on harms (56). This extension consisted of a 22-item checklist that researchers should consider in the process of designing, carrying out, and publishing their studies. The checklist includes “clarify[ing] how harms-related information was collected”, “list[ing] addressed adverse events with definitions for each”, and “describ[ing] for each arm the participant withdrawals that are due to harms and their experiences with the allocated treatment” (56).
Harms reporting for nonpharmacologic RCTs is generally inferior to that for pharmacologic RCTs. A study focusing on the reporting of harm in RCTs of mental health interventions found that no report of nonpharmacologic treatment trials adequately reported harms (115). Another group of researchers compared the reporting of harm in pharmacologic (n=119) and nonpharmacologic (n=74) RCTs of treatments for rheumatic disease (74). Pharmacologic treatment reports included information related to harms more often than nonpharmacologic treatment reports. This information consisted of collection methods, blinded assessment, reporting of adverse events, causal relationship between the treatment and adverse events, withdrawals due to the events, and severity of the events. A greater proportion of the space in the results section was allocated to harms in pharmacologic than nonpharmacologic treatment reports. These differences remained with adjustment for sample size, medical area, funding, and multicenter trials. Fewer than half of the nonpharmacologic treatment trials assessed reported any harm-related data at all.
The authors commented (74): “Presupposed lower toxicity profiles of nonpharmacologic treatment, such as exercise therapy, complementary and alternative medicine, and behavioral interventions, could explain a lower interest in the evaluation of adverse events. However, most therapy entails the risk for adverse events, including serious events.” The aforementioned CONSORT statements provide a framework against which to evaluate the quality of RCTs. Reporting of harms from trials of nonpharmacologic trials should be as systematic as the reporting recommended for pharmacologic trials.