OBJECTIVES Infections caused by Gram-negative bacteria, including nosocomial pneumonia (NP), complicated urinary tract infections (cUTI), and complicated intra-abdominal infections (cIAI), have been increasing. Although microbiological eradication is on the presumed causal path from antibiotic susceptibility to clinical success, other factors impact clinical success rates as well. This study assessed the relationship between microbiological eradication and clinical outcomes for NP, cUTI, and cIAI based on randomized controlled trial (RCT) evidence. METHODS A systematic literature search identified RCTs (25 NP trials, 10 cUTI, 28 cIAI, and 1 cUTI & cIAI trial) that met the following criteria: Adult patients with cUTI, cIAI, or NP; Gram-negative bacteria present in at least a fraction of the population; treatment including coverage of Gram-negative bacteria; any measure of microbiological eradication and either clinical response, cure or mortality. RCTs with information on both eradication and clinical outcome were selected to estimate their relationship to treatment effects using multivariate meta-analyses. RESULTS Given the variation in analysis populations and timing of outcome assessment across RCTs, a limited number of studies were considered sufficiently similar for meta-analyses. For NP, a positive relationship between microbiological eradication and cure (correlation coefficient of 0.84; 95%CI 0.07, 0.98; 5 studies) and a negative relationship between eradication and mortality (-0.86; 95%CI -0.97, -0.34; 7 studies) were observed. For cIAI, clinical outcome was used as a proxy measure for microbiological eradication, but no correlation with mortality was identified. No relationship was observed for cUTI either. CONCLUSIONS Relationships between treatment effects in terms of eradication and clinical cure and between eradication and mortality were identified for NP. For cIAI and cUTI, the relationship between microbiologic eradication and treatment effects is unclear based on available study level RCT evidence. Given the great variation between studies and several uncertain findings, evaluations using patient level data are recommended.