Figure 1 . Meta-analysis of placebo-controlled randomized trials
of fluvoxamine (2x100 or 3x100 mg/day over 10 to 15 days) in adult,
non-vaccinated symptomatic mild COVID-19 outpatients evaluating the
effects on disease progression. Implemented are frequentist and Bayesian
random-effects pooling methods used also in the meta-analysis by Lee et
al. 3 [restricted maximum likelihood estimator of
across study variance in the frequentist analysis, and weakly
informative neutral prior for the effect – 0 for ln(RR) and 0.355 for
its standard deviation – and half-cauchy with scale 0.10 for the
heterogeneity parameter]. The differences vs. the published
meta-analyses 3, 4 are in that: (i) it includes data
from the Korean trial (Seo et al. 8) and (ii) uses
Hartung-Knapp-Sidik-Jonkman correction to calculated frequentist
confidence intervals, as recommended 9. A.Meta-analysis of study-defined primary outcomes (explained in the text).
Data for Stop COVID 1 5, TOGETHER 7and the Korean trial (Seo et al. 8) are taken from the
respective publications. Data for Stop COVID 2 are not publicly
available and were taken from the meta-analysis by Lee et al.3. B . Meta-analysis of hospitalizations. Data
for TOGETHER trial 7 and the Korean trial8 are taken from the respective publications. Data for
Stop COVID 1 and 2 trials are taken from the meta-analysis by Lee et al.3 – the principal investigator of the Stop COVID
trials is one of the co-authors, hence data should be considered
accurrate.
Bayesian analysis was performed using package bayesmeta10 in R (as in the published meta-analysis3), frequentist analysis was performed using packagemeta (11) in R.