Validation of Duke Criteria for Endocarditis Assessment after 30 years: A retrospective cohort study

Background: The Duke criteria have shaped the way infectious endocarditis (IE) is diagnosed in the last 30 years. This study aims to evaluate the present-day accuracy, reliability and limits of the Duke criteria in diagnosing IE by using our latest cohort of 163 patients.

Methods: A retrospective cohort study was conducted on 163 consecutive patients who presented at the University Hospital in Ulm (Germany) with clinical suspicion of IE between 2009 and 2019. With patients' medical records we differentiated between a definitive endocarditis (DIE), a possible endocarditis (PIE) and a rejected endocarditis (RIE). The original Duke criteria were applied to each case, and compared with the final discharge diagnosis. We then tried to identify more valid IE criteria.

Results: The validity of the Duke Criteria, particularly its major criteria, improves with the length of the clinical stay. This effect is the most pronounced within the cardiac imaging criterion (RIE 33.3%, PIE 31.6% and DIE 41.9%, p = 0.622 at admission and RIE 53,3%, PIE 68,4%, DIE 92.2%, p < 0.001 at discharge). Therefore, at admission, overall sensitivity and specificity of the Duke Criteria were respectively 29.5% and 91,2% in the DIE group. At discharge, sensitivity in the DIE group rose to 77.5% and specificity decreased to 79.4%. Within all new screened metrics that we observed when comparing DIE and RIE groups, the only ones showing an additional impact on the diagnosis (p < 0.2) were microhematuria (p = 0.124), leucocyturia, (p = 0.075), age (p = 0.042) and the lack of a rheumatoid disease (p = 0.011). In multivariate regression, only  microhematuria qualified as a new potential parameter, when measured at admission as a sixth minor criterion. Sensitivity could be increased by up to 4 pointsmwith the same specificity.

Conclusion: Even with the technological breakthroughs of the last 30 years in imaging and lab methods our study findings suggest that the Duke criteria continue to hold value in the accurate assessment of infective endocarditis in a contemporary patient population. The criteria demonstrated good sensitivity and specificity in diagnosing IE in our cohort. Nevertheless, their sensitivity only rises after time, while a fast diagnosis or rejection is essential for the patient’s outcome. Therefore, diagnostic criteria that allow a faster assessment like hematuria and metagenomic next generation sequencing (mNGS) could be at the forefront of improving the diagnosis and therapy of IE both in native and prosthetic valves.