Carriers of multidrug-resistant bacteria are at risk of infections with these bacteria; the precise size of this risk is unclear. We aimed to quantify the effect of gut colonisation on subsequent risk of infection with multidrug-resistant bacteria.

We performed a systematic review and meta-regression analysis. We searched PubMed, Embase, Web of Science Core Collection, and Google Scholar for follow-up studies published from Jan 1, 1995, to March 17, 2022, that measured the incidence of infections with multidrug-resistant Gram-negative bacteria (MDR-GNB) and from Jan 1, 1995, to March 15, 2022, that measured the incidence of infections with vancomycin-resistant enterococci (VRE). We included original cohort studies and case-control studies that used incidence-density sampling, included 50 or more patients with enteric colonisation or positive urinary samples as a surrogate marker of colonisation, or both, and analysed infections clearly preceded by colonisation. We did not use any language restrictions. We excluded studies not reporting length of follow-up. Summary data were extracted and independently cross-verified by two authors. Carriage was defined as MDR-GNB or VRE, detected in faecal or urinary cultures. Our primary outcomes were cumulative incidence and incidence density of infection in patients colonised by multidrug-resistant bacteria. To estimate pooled incidences, general linearised mixed-effects meta-regressions were used, adjusting for varying follow-up durations. This study is registered with PROSPERO, CRD42020222415.

Of the 301 studies identified, 44 studies (26 on MDR-GNB, 14 on VRE, and four on both MDR-GNB and VRE) from 14 countries were retained for qualitative synthesis, 40 of which were analysed with meta-regression, comprising data for 14 049 patients colonised with multidrug-resistant bacteria. The pooled cumulative incidence of infection was 14% (95% CI 10–18; p<0·0001) at a median follow-up time of 30 days for MDR-GNB (845 cases of infection in 9034 patients colonised) and 8% (5–13; p<0·0001) at 30 days for VRE (229 cases of infection in 4747 patients colonised). Infection incidence density (4·26 infections per 1000 patient-days; 95% CI 1·69–6·82) and cumulative incidence of infection (19%, 95% CI 15–25; p<0·0001; 602 cases of infection in 4547 patients colonised) were highest for carbapenem-resistant Gram-negative bacteria at 30 days. Risk of bias was rated low to moderate.

The risk of infection was substantial, with the highest risk for patients colonised with carbapenem-resistant Gram-negative bacteria and the lowest in patients with VRE. These data might help to guide prophylactic and treatment decisions and form a valuable resource for planning clinical trials on targeted prevention.

The Netherlands Organization for Health Research and Development