BackgroundEffective treatment of sepsis due to carbapenem-resistant Gram-negative bacteria (CR-GNB) remains a challenge for clinicians worldwide. In recent years, the combination of antibiotics has become the preferred treatment strategy for CR-GNB infection. However, robust evidence to support this approach is lacking. This systematic review aimed at critically evaluating all available antibiotic options for CR-GNB sepsis with particular focus on combination.MethodsWe systematically searched published literature from January 1945 until December 2018 for observational comparative and non-comparative studies and randomized trials examining any antibiotic option for CR-GNB. Studies were included if reporting microbiologically-confirmed infection caused by Acinetobacter baumannii, Enterobacteriaceae/Klebsiella spp., or Pseudomonas aeruginosa, reporting at least one of the study outcomes, and definitive antibiotic treatment. Carbapenem-resistance was defined as phenotypically-detected in vitro resistance to at least one of the following carbapenems: doripenem, ertapenem, imipenem, meropenem. Each antibiotic regimen was classified as “defined” when at least the molecular class(es) composing the regimen was detailed. Primary outcomes were 30-day and attributable mortality. Bayesian network meta-analysis (NMA) approach was selected for quantitative synthesis to explore feasibility of pooling data on antibiotic regimens.ResultsA total of 6306 records were retrieved and 134 studies including 11,546 patients were included: 54 studies were on Acinetobacter, 52 on Enterobacteriaceae/Klebsiella, 21 on mixed Gram-negative, and 7 on Pseudomonas. Nine (7%) were RCTs; 19 prospective cohorts (14%), 89 (66%) retrospective, and 17 (13%) case series. Forty-one studies (31%) were multicentric. Qualitative synthesis showed an heterogeneous and scattered reporting of key-clinical and microbiological variables across studies. Ninety-two distinct antibiotic regimens were identified with 47 of them (51%, 5863 patients) not reporting any details on numbers, type, dosage and in vitro activity of the included antibiotic molecules. The NMAs could not be performed for any of the selected outcome given the presence of too many disconnected components.ConclusionThe existing evidence is insufficient to allowing for the formulation of any evidence-based therapeutic recommendation for CR-GNB sepsis. Future studies must provide a standardized definition of antibiotic regimen to drive recommendations for using combination of antibiotics that can be reliably applied to clinical practice.