Fifty-nine RCTs (24,083 participants) were included in the review; trial sample size ranged from 30 to 3047 participants. Double-blind procedures were reported by 21 trials and low risk of bias for allocation concealment was reported by nine trials. Concealment methods were not reported by 41 trials; high risk of bias was reported for nine trials. Assessment for the use of intention-to-treat data was not reported.
Additional Gram-negative coverage versus mainly Gram-positive coverage: Compared to mainly Gram-positive coverage, statistically significant lower risks of postoperative pneumonia (RR 0.66, 95% CI 0.51 to 0.86, 24 trials) and all-cause mortality (RR 0.69, 95% CI 0.50 to 0.93, 18 trials) were found for antibiotics with additional Gram-negative coverage. The effect was statistically strongest for pneumonia when Gram-negative coverage was provided by a β-lactam (second or third-generation cephalosporin in most studies). No statistically significant heterogeneity or small-study effects were observed.
Prophylaxis duration: Statistically significant differences were found for trials that compared the same or different antibiotics in the long versus short-duration study arms for the following outcomes: deep sternal wound infection (RR 1.49, 95% CI 1.08 to 2.06, 15 trials); any sternal wound infection (RR 1.37, 95% CI 1.14 to 1.64, 20 trials); Gram-positive surgical site infection (RR 1.30, 95% CI 1.01 to 1.66, 12 trials); and surgical intervention for surgical site infection (RR 1.48, 95% CI 1.04 to 2.09, seven trials). Subgroup analyses by short duration arms indicated that the significant differences were due to short prophylaxis durations 24 or less hours post-operation but not short durations over 24 hours post-operation.
Glycopeptide versus β-lactam: Compared to β-lactams of the same duration, the risk for any surgical site infection was lower with glycopeptides but this was of borderline statistical significance (RR 0.74, 95% CI 0.56 to 0.99, six trials). Risk of surgical intervention for surgical site infection was statistically significantly lower with glycopeptides than with β-lactams of the same duration (RR 0.20, 95% CI 0.07 to 0.52, two trials).
Glycopeptides statistically significantly increased risks for deep sternal wound infection compared to β-lactams of a longer duration (RR 1.90, 95% CI 1.14 to 3.19, 3 trials), any sternal wound infection compared to β-lactams overall (RR 1.33, 95% CI 1.08 to 1.64, 9 trials) and to β-lactams of a longer duration (RR 1.77, 95% CI 1.36 to 2.32, three trials).
Cephalosporin-based versus penicillin-based prophylaxis regimen: Weak statistical significance was found for a lower risk of bacteraemia with cephalosporin-based prophylaxis regimens (RR 0.48, 95% CI 0.24 to 0.99, three trials, Ι²=52%).
Antibiotic dosing: No statistically significant differences were found (reported fully in paper).
Other analyses were reported in the review. Two studies were reported narratively.