The heterogeneity of symptom reporting across study sites: a secondary analysis of a randomised placebo-controlled multicentre antimalarial trial.

INTRODUCTION: Symptoms reported following the administration of investigational drugs play an important role in decisions for registration and treatment guidelines. However, symptoms are subjective, and interview methods to quantify them are difficult to standardise. We explored differences in symptom reporting across study sites of a multicentre antimalarial trial, with the aim of informing trial design and the interpretation of safety and tolerability data. METHODS: Data were derived from the IMPROV trial, a randomised, placebo-controlled double blinded trial of high dose primaquine to prevent Plasmodium vivax recurrence conducted in eight study sites in Afghanistan, Ethiopia, Indonesia and Vietnam. At each follow up visit a 13-point symptom questionnaire was completed. The number and percentage of patients with clinically relevant symptoms following the administration of primaquine or placebo, were reported by study site including vomiting, diarrhoea, anorexia, nausea, abdominal pain and dizziness. Multivariable logistic regression was used to estimate the confounder-adjusted site-specific proportion of each symptom. RESULTS: A total of 2,336 patients were included. The greatest variation between sites in the proportion of patients reporting symptoms was for anorexia between day 0 and day 13: 97.3% (361/371) of patients in Arba Minch, Ethiopia, reported the symptom compared with 4.7% (5/106) of patients in Krong Pa, Vietnam. Differences attenuated slightly after adjusting for treatment arm, age, sex, day 0 parasite density and fever; with the adjusted proportion for anorexia ranging from 4.8% to 97.0%. Differences between sites were greater for symptoms graded as mild or moderate compared to those rated as severe. Differences in symptom reporting were greater between study sites than between treatment arms within the same study site. CONCLUSION: Despite standardised training, there was large variation in symptom reporting across trial sites. The reporting of severe symptoms was less skewed compared to mild and moderate symptoms, which are likely to be more subjective. Trialists should clearly distinguish between safety and tolerability outcomes. Differences between trial arms were much less variable across sites, suggesting that the relative difference in reported symptoms between intervention and control group is more relevant than absolute numbers and should be reported when possible. TRIAL REGISTRATION: Clinicaltrials.gov: NCT01814683; March 20th, 2013.