Kosyakovsky L, Austin P, Ross H, et al. - Summarized by David L. Brown, MD

Using administrative data (ICD-10 codes), this retrospective study attempted to determine the impact of early coronary angiography on patients who presented with acute heart failure (HF) and at least one factor suggesting an ischemic etiology including prior myocardial infarction (MI), troponin elevation, or angina on presentation. The study was conducted from 2010 to 2013 in 70 emergency departments in Ontario, Canada. Patients were included if they met the Framingham criteria for HF and had an elevation in brain natriuretic peptide (BNP) or N-terminal pro-BNP levels.

Patients with a primary diagnosis of acute coronary syndrome were excluded, as were patients with an admission for acute MI, coronary angiography, or revascularization within the 3 months prior to admission. Additional exclusions included palliative care patients and those with chronic kidney disease, anaphylactic contrast allergy, stroke within 3 months, active bleeding or severe thrombocytopenia, metastatic malignancy, or the highest degree of frailty as determined by the Hospital Frailty Risk Score. Early coronary angiography was defined as angiography performed within 14 days of presentation. To prevent immortal time bias, patients who died in the first 14 days following admission were also excluded.

The co-primary outcomes were all-cause and cardiovascular mortality at 2 years. In an attempt to adjust for differences in clinical characteristics between the nonrandomized groups, inverse-probability-of-treatment-weighting (IPTW) was used. IPTW assigns a weight to each patient which is the inverse of the probability of receiving the treatment received (angiography or not).

Cox proportional hazard regression models were used to estimate hazard ratios for each outcome. Kaplan-Meier survival curves the adjusted log-rank test were used to compare all-cause mortality between groups after IPTW. The final cohort included 2,994 patients of whom 1,567 (52.3%) underwent angiography and 1,427 (47.7%) did not. There was no loss to follow up. Of the patients who underwent coronary angiography, 59% had obstructive CAD and 18% underwent revascularization within 90 days. Both all-cause (26% vs. 34%; HR 0.74, 95% CI: 0.61-0.90) and cardiovascular mortality (15% vs. 20%; HR 0.72, 95% CI: 0.56-0.93) were significantly lower in the early angiography group.

The authors claim that this data should inform decision-making about coronary angiography in acute HF. Specifically, they state that early angiography “should be strongly considered for those with HF and suspicion of underlying ischemic heart disease.” This recommendation represents a dangerous overreach regarding the ability of retrospective administrative data to inform decision-making. Despite the robust statistical efforts to adjust for and reduce confounding from selection bias, no methodology replaces randomization because only randomization balances both measured and unmeasured confounders between groups. The results of the current study are entirely compatible with confounding by the frequent observation that lower-risk patients are more likely to undergo invasive procedures than higher-risk patients (selection bias). Furthermore, at 2 years only 13% more patients in the early angiography group underwent revascularization than in the control group. Thus, the authors are left attributing the bulk of a ~25% relative reduction in mortality to a diagnostic test. Unfortunately, no amount of statistical adjustment of retrospective administrative data can recapitulate the real-time bedside calculus involved in the assessment of the risk-benefit ratio of a specific treatment or test for an individual patient. Thus, this study is only hypothesis-generating and should not inform evidence-based clinical practice.