LAUSR

Patients with diabetes mellitus (DM) may have elevated levels of high-sensitive cardiac troponin (hs-Tn) despite acute myocardial ischemia being present. However, it is unclear whether this constrains diagnostic strategies in patients with suspected acute myocardial infarction (MI). We aimed to assess the diagnostic performance of the European Society of Cardiology (ESC) 0/1 hour (h) and 0/3h-algorithms comparing patients with and without DM and to derive optimized cut-offs. We prospectively enrolled patients with symptoms suggestive of MI in two large clinical cohorts and measured hs-TnI at admission (baseline) and 1 (cohort A) and 3h (cohort A+B) thereafter. Patients with ST-elevation MI were excluded. Patients were stratified based on a diagnosis of DM at baseline. Final diagnoses were adjudicated independently by two cardiologists using all clinically available information, including hs-TnT, but blinded to hs-TnI values. Our primary outcomes of interest were safety of rule-out (defined by sensitivity and negative predictive value [NPV]), accuracy of rule-in (defined by specificity and positive predictive value [PPV]) and the overall performance (% of patients adjudicated to either rule-out or -in). For optimized cut-offs, a NPV >99.0% and a PPV >75.0% were targeted. DM was prevalent in 563 (15.29%) of 3683 included patients. MI was more prevalent among patients with DM (137 [24.3%] vs. 498 [16.0%], p<0.001). Using the ESC 0/1h-algorithm (Figure), rule-out was safe in diabetics (p for sensitivity = 1.00) with higher NPV in non-diabetics (p<0.001), while the proportion of patients ruled-out was smaller in diabetics (22.3% vs. 41.8%). Accuracy of rule-in was significantly lower in diabetics (specificity p=0.0035, PPV p=0.48), with a higher rule-in rate of patients with DM (29.5% vs. 21.8%). Using the ESC 0/3h-algorithm, safety of rule-out was lower in both groups compared to the ESC 0/1h-algorithm, with again higher NPV for non-DM (sensitivity p=0.18, NPV p<0.001) and a higher proportion of non-DM ruled-out (65.9% vs. 75.2%). Accuracy of rule-in was significantly lower for patients with DM (specificity p=0.0094, PPV p=0.87). Cut-off adjustment to yield pre-defined accuracy measures resulted in: 4ng/L at baseline or 6ng/L with a delta of 2ng/l for rule–out and 90ng/L or a delta of 10ng/L for rule-in with the ESC 0/1h algorithm; for the ESC 0/3h-algorithm cut-offs were 5ng/L with a delta of 20% for rule-out and 50ng/L with a delta of 20% for rule-in. Application of the ESC 0/1h- and 0/3h-algorithms in diabetic patients provided reduced safety and accuracy for rule-out and rule-in of MI, respectively. Use of alternative cut-offs resulted in improved diagnostic safety and accuracy. Abbott Diagnostics, German Center of Cardiovascular Research, German Heart Foundation, Else-Kröhner-Stiftung angegeben