Defining clinically relevant risk threshold for predicting sudden cardiac death

Abstract There is no consensus on a risk threshold for sudden cardiac death (SCD) that could be used in practical design and evaluation of prediction models and decisions regarding implantable cardioverter-defibrillator (ICD) therapy. Baseline assumptions for a simulation framework were derived from previous randomized controlled trials ( n = 18) to identify minimal average SCD risk threshold that would translate to mortality benefit by ICD therapy also considering the effect of competing non-sudden mortality. ICD efficacy to prevent SCDs and other data for simulations were estimated using inverse-variance weighted meta-analysis of included trials. Number needed to treat (NNT) was evaluated over a 5- and 10-year horizons (≤ 20 defined as clinically relevant). Correlation analysis confirmed average annual SCD incidence in trial populations as the key factor associating with ICD therapy effectiveness to reduce mortality (Pearson’s r = 0.653, p < 0.01). In a simulation assuming 5% annual non-sudden mortality (pooled estimate of included RCTs) and a 56% (48–62%) efficacy for ICDs to reduce SCDs or similar events, 3% average annual SCD risk (≈ 12% over 5 years) emerged as the lowest practical threshold even after controlling for excess (overlapping) mortality among those saved successfully from SCD by ICD therapy. In populations with very low incidence of non-sudden mortality (1%), the lowest practical threshold over a longer 10-year time horizon was 1.3% (assuming no overlapping mortality). Even under substantial competing risk, populations with an average 3% (or higher) annual SCD incidence most likely would benefit from ICD therapy. Lower threshold may be considered only in populations with very low non-sudden mortality.