Statistical analysis
We coded the 22 STROBE items as article section and article sub-section
items attributing a dichotomous YES/NO score according to a conservative
(all sub-items defined by the STROBE explanatory article were satisfied)
and liberal (at least 50% of sub-items were satisfied) criteria. Items
without sub-items, i.e. the STROBE items on discussion and funding, were
coded the same for the conservative and liberal score. Afterwards, the
qualitative nature of STROBE was taken into account describing the 22
dichotomous scores by means of percentages. We merged results regarding
multiple STROBE items belonging to the same article section by means of
medians. Finally, we performed a multiple correspondence analysis, which
is an analogue of principal component analysis applied to contingency
tables from categorical data. We applied this technique to investigate
the multivariate association between STROBE items and their correlation
with study design. The correspondence analyses was performed using the
FactoMineR package of the R software. Results are reported using
graphical representations as bar charts of frequencies, scatter plot of
percentages over publication years and correspondence analysis plot of
the first two dimensions.