Fig. 1

Client-server architecture of DataSHIELD. The diagram shows four study sites/servers (DC) each having data stored in the Original.DB. The analyst (client) sends commands from the analysis computer (AC) to each study site to request the specific data (Assigned.Data) to be analyzed. This could be all the variables or specific variables stored in Analysis.DB. R commands are also sent from the analysis computer to every study telling it to create survival objects and fit the Cox proportional hazard model. Each site responds to instructions sent by creating the survival object and fitting the model. This fitting is carried out in the R environment of each study. The coefficient matrices, standard errors, and odds ratios from each site are then pooled and meta-analyzed using fixed optimization methods, and only non-disclosive statistics are returned to the analyst