Outcome differences between PARAMEDIC2 and the German Resuscitation Registry: a secondary analysis of a randomized controlled trial compared with registry data

Background and importance There has been much discussion of the results of the PARAMEDIC2 trial, as resuscitation outcome rates are considerably lower in this trial than in country-level registries on out-of-hospital cardiac arrest (OHCA). Here, we developed a statistical framework to investigate this gap and to examine possible sources for observed discrepancies in outcome rates. Design Summary data from the PARAMEDIC2 trial were used as available in the publication of this study. We developed a modelling framework based on logistic regression to compare data from this randomized controlled trial and registry data from the German Resuscitation Registry (GRR), where we considered 26 019 patients treated with epinephrine for OHCA in the GRR. To account and adjust for differences in patient characteristics and baseline variables predictive for outcomes after OHCA between the GRR cohort and the PARAMEDIC2 study sample, we included all available variables determined at the arrival of EMS personnel in the modelling framework: age, sex, initial cardiac rhythm, cause of cardiac arrest, witness of cardiac arrest, CPR performed by a bystander, and the interval between emergency call and arrival of the ambulance at the scene (baseline model). In order to find possible explanations for the discrepancies in outcome between PARAMEDIC2 and GRR, in a second (baseline plus treatment) model, we additionally included all available variables related to the interventions of the EMS personnel (type of airway management, type of vascular access, and time to administration of epinephrine). Main results A patient cohort with baseline variables as in the PARAMEDIC2 trial would have survived to hospital discharge in 7.7% and survived with favourable neurological outcome in 5.0% in an EMS and health care system as in Germany, compared with 3.2 and 2.2%, respectively, in the Epinephrine group of the trial. Adding treatment-related variables to our logistic regression model, the rate of survival to discharge would decrease from 7.7 (for baseline variables only) to 5.6% and the rate of survival with favourable neurological outcome from 5.0 to 3.4%. Conclusion Our framework helps in the medical interpretation of the PARAMEDIC2 trial and the transferability of the trial's results for other EMS systems. Significantly higher rates of survival and favourable neurological outcome than reported in this trial could be possible in other EMS and health care systems.