How to deal with the data from sports centres in Piacenza?

In their Heart manuscript, Aschieri et al report an impressive impact of the availability of automated external defibrillators (AEDs) on survival after sudden cardiac arrest (SCA) in amateur sports centres.1 For this observational study, the authors analysed data from a registry on all out-of-hospital SCA cases from the Piacenza area (Italy). Using the predetermined variable ‘location’, they selected 26 cases. As AEDs were gradually introduced over an 18-year period, a comparison could be made according to the on-site presence of an AED. Neurologically intact survival rates in sports centres with and without on-site AED were 93% (14/15) and 9% (1/11), respectively.

At first glance, this impressive difference should lead to patently clear conclusions. A closer look at the methodology and the data, however, urges caution as at least four (often interrelated) issues potentially leading to bias should be mentioned. First, the AEDs were not allocated at random since the sports centres decided whether or not to start a local AED project, implying that all results are prone to bias. Second, the time of collapse was not registered. Although the authors state that all cases were witnessed and that the emergency medical services (EMS) dispatch centre was never alerted for pre-arrest symptoms, one cannot exclude that in sports centres with an on-site AED the collapse to call interval was shorter, for example, because a local campaign aiming at an optimal AED use also emphasised the importance of an early call to the dispatch centre. Third, the gradual introduction of AEDs (ie, only 36 devices in 2007 vs 207 in 2016) is reflected by an increasing proportion of SCA cases in AED-equipped sites (ie, only 1 out of 8 cases between 1999 and 2007 vs 14 out of 18 between 2008 and 2016). Consequently, any intervention improving the quality of prehospital and in-hospital care in the Piacenza area, for example, more accurate dispatch of EMS units, introduction of telephone-assisted basic life support (BLS) and more aggressive coronary interventions in resuscitated patients, favoured AED-equipped sites. Fourth, it is not clear why there were only eight cases in the period 1999–2007 and 18 between 2008 and 2016. Therefore, selection bias cannot be excluded.

Undoubtedly, many uncertainties are inevitable when using prospectively registered data to study the effects of a particular intervention such as AED use by lay people. Only through a well-designed randomised controlled trial (RCT), most of the aforementioned issues can be avoided. In this respect, Aschieri et al clearly mention that the reported observational data do not prove causality between the on-site availability of an AED and high survival. Fortunately, the authors were able to provide some details on individual cases. Indeed, the on-site presence of an AED does not necessarily imply that this device saved the patient’s life. This is obvious when the on-site AED is not connected or when no shock is delivered before EMS arrival. Furthermore, the magnitude of the beneficial effect of the on-site AED in an individual patient is related to the interval between the delivery of the first shock by a lay volunteer and EMS arrival, that is, a 10%–12% decrease in survival rate for every minute of delay.2 Applying these principles to the reported data showed that there were only eight SCA cases where an on-site AED was used by a lay bystander, and that the interval between AED connection by the lay responder and EMS arrival in these eight cases ranged from 2 to 10 min.

Our closer look at the raw data on the one hand suggests that the observed huge difference in survival rates in sports centres with and without AED (ie, 93% vs 9%) does not reflect the real surplus value of an on-site AED. On the other hand, the authors have done much to show that some patients benefited from the on-site AED.

What about other studies on public access defibrillation?

Similar methodological problems and flawed analyses potentially leading to an overestimation of the benefit of AED-related projects are to be looked after in other observational studies. Two examples may illustrate this issue.

In a project implementing a text message (TM) alert system for trained volunteers, Pijls et al compared two subgroups: 291 cases where at least one volunteer arrived on the scene before EMS arrival in order to perform BLS or to defibrillate (group 1) versus 131 cases where no volunteer could be sent to the patient (group 2).3 The survival rates were 27.1% and 16.0% in groups 1 and 2, respectively. From these figures, the authors concluded that the TM alert system is effective in increasing survival. As in the Piacenza study, the authors minimised the potential impact of the non-randomisation of the included cases, the search for indicators for any kind of bias was limited and a critical analysis on the role of the TM alert system among the 79 survivors in group 1 was lacking (although some raw data are available to the authors). From the overall data on group 1, we could only find that the TM responders were the first to start BLS in only 24.7% and the first to connect an AED in only 26.8%. Furthermore, the first recorded rhythm was shockable in 59.9% (implying that an AED brought to the patient by a TM responder was of no use in 40.1%). From these figures, it seems reasonable to assume that in a substantial number of the 79 survivors, the TM responders contributed not or only minimally to the favourable outcome. Unfortunately, the authors were unable to register the arrival times of the TM responders. Therefore, one cannot take into account the difference in survival rate for every minute of BLS and defibrillation delay. All these elements lead to the conclusion that the reported surplus value of the TM alert system may be too great and that more refined analyses are possible.

Using a large, prospective, population-based registry of out-of-hospital non-traumatic SCA cases, Kiyohara et al compared 351 cases with AED application by the public (group 1) with all other 9627 cases (group 2).4 Their main finding was a significantly different 1-month survival with a favourable neurological outcome: 19.4% in group 1 versus 3.0% in group 2. For group 1, many factors associated with a better outcome are mentioned: younger age, more cases of presumed cardiac origin, more witnessed cases and more cases with bystander CPR (implying that public access defibrillation by itself is a marker for better outcome after SCA). Regrettably, the authors did not try to weight the surplus value of earlier defibrillation via public-access AEDs against the impact of the outbalanced prognostic factors. Once again, the beneficial effect of the public-access AEDs could have been estimated more accurately by analysing some elementary data in the survivors. Was the AED connected to the patient? Did the AED deliver a shock? How long was the AED connection to EMS arrival interval?

Merits of the Piacenza study and suggestions for future research on publicaccess defibrillation

From many studies, it is beyond any doubt that AEDs can be used properly by lay people and that this approach has the potential to save numerous lives.2 5Aschieri et al provide additional arguments that public access defibrillation is feasible and useful in amateur sports centres.1

At present, the most intriguing questions related to public access AED projects concern cost-effectiveness and effective ways to implement these projects in particular settings.5–7 As for many reasons RCTs are often impossible, science in this field has to proceed principally via observational studies. In this editorial, warnings are expressed against the acceptance of causality between good outcome and the presence of an on-site AED in observational studies. Furthermore, arguments are given to add analyses on very simple AED-related data in all surviving patients. Was the device connected? Was a shock delivered? Was the spontaneous circulation restored before EMS arrival? How much time elapsed between AED connection and EMS arrival?7 Regarding the latter question, algorithms (based on parameters as witnessed collapse, bystander BLS, collapse to AED connection interval and AED connection to EMS arrival interval) are needed for a proper quantification of the surplus value of public access defibrillation in a particular patient. In this respect, it should be stressed that all research should be based on data gathered in accordance with the internationally accepted Utstein definitions.8