Statistical analysis
Reporting frequencies were estimated for the following variables: gender, age group, seriousness, mortality, HCQ dose, HCQ treatment length, HCQ indication, concomitantly used medications, as well as type of ADR grouped by MedDRA system and organ classification (SOC). SPSS v24 and Excel programme were used for this purpose.
When possible, daily doses of HCQ were estimated based on posology. Incomplete posologies that did not allow daily dose to be estimated were considered unknown. Daily doses were categorised into three levels: low (400 mg /day), medium (400-600 mg/day), and high (800 mg/day). Also, total administered dose was estimated and categorised as follows: low (≤ 50.000 mg), medium (51,000-200,000 mg), and high (> 200,000 mg)
For each period, the disproportionality analysis was carried out based on the FEDRA database “Signal generation module”. The Reporting Odds Ratio (ROR) was obtained to estimate disproportionality of the different ADRs reported with HCQ [11]. We used a case/non case approach to determine the strength of the association between HCQ and whichever ADR. The lower limit of the 95% confidence interval was -ROR. We considered a potential HCQ-ADR association when - ROR was higher than 1 and there were at least 3 cases reported. Standardised MedDRA Queries (SMQ) from MedDRA were used to identify ADRs. SMQs are validated pre-determined sets of MedDRA terms grouped together that are associated with a specific medical condition [12] . We identified the HCQ-ADR associations with at least 3 reported cases and with the lower confidence interval (IC 95%) larger than 1 in at least one of the periods of interest.
To explore potential interactions, Omega (Ω) statistic was obtained. This analysis compares the observed relative reporting rate of a given ADR, granted that the reports include the co-existence of two drugs, with its expected value estimated based on the relative reporting rates of the ADR for each drug separately. The observed number of cases is the number of reports in FEDRA for the two drugs together that include the ADR of interest. When Ω is positive, then the drug-drug-ADR association is reported more often than expected [13]. The lower limit of the 95% credibility interval is referred to as Ω -025. We considered a potential drug-drug interaction to exist when Ω-025>0.
For the disproportionality analysis, exposure to a given drug was defined as the recording of the drug in a report, whether or not it was suspected of causing the reaction.