Widespread use of automated culturing systems over the last several decades necessitated the reevaluation of optimal blood culture sample collection protocols originally created in the late 1970s and 1980s. In a 2004 single-center retrospective study of 37,568 blood cultures obtained from adult patients tested on an automated culturing system, researchers evaluated the relationship between several testing parameters and pathogen recovery.¹ A blood culture was defined as blood obtained from one venipuncture divided equally and inoculated into a “set” of aerobic and anaerobic culture media bottles. Only positive culture
results thought to represent clinically relevant bloodstream infection were included. Culture positivity was defined based on the total number consecutive cultures drawn from the same patient in a 24-hour period. Given the association between infective endocarditis (IE) and degree of bacteremia, results were stratified by those with and without a clinical diagnosis of IE (based on medical chart review by treating physicians).

Of patients without clinical IE who had ≥3 cultures drawn, diagnostic yield increased with the number of consecutive cultures: 65% (106/163) of bloodstream infections were identified with the first culture, whereas 80.4% (131/163), 95.7% (156/163), and 100% (163/163) were identified with the second, third, and fourth cultures, respectively. Patients with clinical IE tended to have earlier blood culture positivity: 88.8% (16/18) had bloodstream infections identified with the first culture and 94.4% (17/18) with two cultures. Study caveats include single-center design and potential influence of local microbiologic patterns or clinical practice standards.

In 2007, a retrospective study replicated this analysis using similar sample collection, laboratory techniques, and study design in two geographically independent hospitals.² Only patients with ≥3 blood cultures drawn within a 24-hour period were included, and results were stratified by number of causative microorganisms identified (unimicrobial versus polymicrobial) but not clinical diagnosis of IE. This study corroborated previous findings, showing that diagnostic yield increases with the number of blood culture samples obtained. Of unimicrobial bloodstream infections for which ≥3 blood cultures were obtained, 73.1% (460/629) were identified with first culture, 89.7% (564/629) with two cultures, 98.2% (618/629) with three cultures, and 99.8% (628/629) with four cultures.

Neither study addressed timing of antimicrobial therapy in relation to culture acquisition, potentially confounding results, and detail about the timing of consecutive cultures within a 24-hour period was limited (e.g., frequently occurred within 30 minutes of each other in the first study; timing not reported in the second study). Nonetheless, based on the finding that higher numbers of consecutive blood cultures increase diagnostic yield, the 2015 American Heart Association (AHA)/Infectious Diseases Society of America (IDSA) endocarditis
guidelines recommend that ≥3 sets of cultures be obtained from different venipuncture sites for patients with clinical suspicion of IE with the first and last samples drawn at least 1 hour apart (class I, level of evidence A).³

This question was addressed in a 2016 single-center retrospective analysis⁴ of 790 cases of suspected native-valve IE in which patients underwent both TTE and TEE within a 7-day period. Patients with high-risk clinical features such as prior valve repair or replacement, complex congenital disease, history of heart transplant, or indwelling devices were excluded. The authors compared two analysis approaches to TTE results: (1) a standard analysis involving presence or absence of vegetation versus (2) a set of strict negative rule-out criteria (moderate or better quality of ultrasound; normal valve anatomy; no valvular stenosis or sclerosis; at most trivial valvular regurgitation; at most mild, simple pericardial effusion; absence of implanted hardware or a central venous catheter; no vegetation). TEE was used as the gold-standard comparison.