Statistical analysis
Individual mean and standard deviation were used to produce the pooled
estimate. In absence of reporting of standard deviation, we recalculate
it from confidence interval limits if available, or from the
1st and 3rd quartiles or from the
range using approximation formula [5]. Inverse variance weighting
was used for pooling individual estimation of mean,
95th end 97.5th percentile of the
incubation period. A random-effects model was used. Heterogeneity
between studies was assessed using I2 statistics, with
an I2 of more than 50% indicating substantial
heterogeneity [6]. We explored sources of heterogeneity with a
meta-regression for continuous explicative covariables and by subgroup
analysis for qualitative explicative covariables. Age, gender and study
design were considered as candidate explicative covariables. All data
analyses were done using R (version 3.6.3) with R package meta and the
metamean function. Results of this real-time meta-analysis are shared on
the following site: InCoVid-Lyon [7].