Data synthesis and analysis
The main results of the SR are described narratively and tabulated as summary of findings. Data were pooled and meta-analysed using Review Manager (Review Manager V.5.3, Cochrane Collaboration, Oxford, UK) using the random effects model approach. For binary outcomes pooled relative risk ratios and rate ratios (RRs) were calculated. For continuous outcomes mean differences (MDs) with 95% confidence intervals (CI), were used. If mean or standard deviations (SD), or changes of mean and SDs from baseline were not reported, standard errors (SE), CI, or the correlation coefficient were used. Where multiple arms were compared to a common placebo arm, SE were adjusted to avoid the unit of analysis error (22).
The magnitude of heterogeneity between the included studies was calculated using the Higgings’ I2 statistic interpreted according to the Cochrane Handbook guidelines (23). To account for clinical heterogeneity, subgroup analysis was predefined if possible by different doses of dupilumab, age and ROB. The median estimate reported in the control arms was used as baseline risk to estimate absolute effects. For the economic evidence, results are summarized narratively and tabulated, including the cost, incremental effectiveness, ICERs and the degree of uncertainty.