Validation of the “smart” minimum FFR Algorithm in an unselected all comer population of patients with intermediate coronary stenoses

Barry Hennigan, Nils Johnson, John McClure, David Corcoran, Stuart Watkins, Colin Berry, Keith G. Oldroyd

Using data from a commercial pressure wire system (St. Jude Medical) we previously developed an automated "smart" algorithm to determine a reproducible value for minimum FFR (smFFR) and confirmed that it correlated very closely with measurements made off-line by experienced coronary physiology core laboratories. In this study we used the same "smart" minimum algorithm to analyze data derived from a different, commercial pressure wire system (Philips Volcano) and compared the values obtained to both operator-defined steady state FFR and the online automated minimum FFR reported by the pressure wire analyser. For this analysis, we used the data collected during the VERIFY 2 study (Hennigan et al. in Circ Cardiovasc Interv, doi: 10.1161/CIRCINTERVENTIONS.116.004016 ) in which we measured FFR in 257 intermediate coronary stenoses (mean DS 48%) in 197 patients. Maximal hyperaemia was induced using intravenous adenosine (140 mcg/kg/min). We recorded both the online minimum FFR generated by the analyser and the operator-reported steady state FFR. Subsequently, the raw pressure tracings were coded, anonymised and 256/257 were subjected to further off-line analysis using the smart minimum FFR (smFFR) algorithm. The operator-defined steady state FFR correlated well with smFFR: r = 0.988 (p < 0.001), average bias 0.008 (SD 0.014), 95% limits of agreement -0.020 to 0.036. The online automated minimum FFR also correlated well with the smFFR: r = 0.998 (p < 0.001), average bias 0.004 (SD 0.006), 95% limits of agreement -0.016 to 0.008. Finally, the online automated minimum FFR correlated well the operator-reported steady state FFR: r = 0.988 (p < 0.001), average bias 0.012 (SD 0.014), 95% limits of agreement -0.039 to 0.015. In 95% of lesions studied (244/256), the operator reported steady-state FFR, smFFR, and online automated minimum FFR agreed with each other to within 0.04, which is within the previously reported test/retest limits of agreement of FFR reported by an experienced core lab. Disagreements >0.05 among methods were rare but in these cases the two automated algorithms almost always agreed with each other rather than with the operator-reported value. Within the VERIFY 2 dataset, experienced operators reported a similar FFR value to both an online automated minimum (Philips Volcano) and off-line "smart" minimum computer algorithm. Thus, treatment decisions and clinical studies using either method will produce nearly identical results.