Absolute Versus Relative Changes in Cardiac Troponins in the Diagnosis of Myocardial Infarction: A Systematic Review and Meta-Analysis

Ischemic heart disease (IHD) is one of the leading causes of death globally. Rapid diagnosis of myocardial infarction (MI) will enable earlier initiation of the treatment and improve patient outcomes. Practice guidelines for non-ST-elevation acute coronary syndromes by the American College of Cardiology (ACC)/American Heart Association (AHA) had listed the diagnostic performance of absolute versus relative changes in evidence gaps. We aimed to address this evidence gap by examining the diagnostic accuracy of absolute versus relative changes in cardiac troponins at various time intervals in diagnosing MI. Grey literature, conference abstracts, animal studies, and reports published before 2009 and in languages other than English were excluded. We included reports that investigated absolute or relative changes in highly sensitive cardiac troponin T (hs-cTnT) or sensitive/highly sensitive cardiac troponin I (s/hs-cTnI) assays after specific time intervals (1, 2, or 3 h) in patients presenting with symptoms suggestive of the acute coronary syndrome. After screening, we arranged the reports in 12 separate groups based on the variables for which the data was reported. Quality assessment of the diagnostic accuracy studies-2 (QUADAS-2) was used to assess the risk of bias in the included studies. The weighted summary area under the curve (AUC) was calculated for each pool. We then performed two-sided (or two-tailed) tests to compare independent receiver operating characteristic (ROC) curves. MedCalc version 20.106 (MedCalc Software Ltd., Ostend, Belgium) was used for all statistical analysis. We included eight reports with 23,450 patients in the meta-analysis. Weighted summary estimates and their respective 95% confidence intervals (CI) under random-effects model for ROC-AUC are as follows: absolute hs-cTnI at 1 h - 0.94 (95% CI: 0.922 to 0.959, p < 0.001); absolute hs-cTnT at 1 h - 0.921 (95% CI: 0.902 to 0.941, p < 0.001); absolute s/hs-cTnI at 2 h - 0.953 (95% CI: 0.926 to 0.980, p < 0.001); absolute hs-cTnT at 2 h 0.951 (95% CI: 0.940 to 0.962, p < 0.001); relative hs-cTnT at 2 h - 0.818 (95% CI: 0.733 to 0.903, p < 0.001); relative s/hs-cTnI at 2 h - 0.762 (95% CI: 0.726 to 0.798, p < 0.001); absolute hs-cTnI at 3 h - 0.967 (95% CI: 0.95 to 0.984, p < 0.001); absolute hs-cTnT at 3 h - 0.959 (95% CI: 0.950 to 0.968, p < 0.001); and relative hs-cTnT at 3 h - 0.926 (95% CI: 0.907 to 0.945, p < 0.001). P-values of comparison of absolute and relative changes are as follows: hs-cTnT at 1 h: <0.0001; hs-cTnI at 1 h: <0.0001; hs-cTnT at 2 h: 0.0024; s/hs-cTnI at 2 h: <0.0001; hs-cTnT at 3 h: 0.0022; and hs-cTnI at 3 h: 0.0005. Our analysis found absolute changes to be superior to relative changes in both hs-cTnT and s/hs-cTnI at 1, 2, and 3 h in the diagnosis of MI. There was no statistically significant difference in comparing s/hs-cTnI vs. hs-cTnT using absolute or relative changes at any time interval. Our findings suggest that future research investigating a potential 0 h/30 min algorithm should use absolute Δ over relative Δ. A suboptimal number of reports in the groups limited our ability to establish the robustness of the results. We did not receive any funding for this review.