Ventilator-associated pneumonia (VAP) is an important complication in intensive care unit (ICU) patients. Current guidelines in the United States strongly recommend that surveillance be conducted for bacterial pneumonia in ICU patients, but diagnostic criteria have been criticized for the lack of sensitivity and specificity, especially regarding multidrug-resistant (MDR) pathogens. Brusselaers and investigators from Ghent University in Belgium performed a systematic review and meta-analysis to determine the sensitivity and specificity of surveillance cultures in predicting pathogens in VAP. Their results were published in the March issue of Intensive Care Medicine.

A systematic review of the literature was conducted using several major data sources. No language restrictions were applied. All studies had to include lower respiratory tract sampling before any clinical suspicion of VAP. Studies were assessed according to current Quality Assessment of Studies of Diagnostic Accuracy Included in Systematic Reviews (QUADAS) guidelines. A mixed-effect regression model was calculated to perform a meta-analysis. A Bayes nomogram was constructed, and heterogeneity was assessed with forest plot analyses and calculation of the I2 statistic.

Five hundred twenty-five studies were retrieved; 14 studies were ultimately included in the meta-analysis, representing 791 VAP episodes in 688 patients. The incidence of VAP was only reported in three studies, and ranged from 10.4/1000 to 55/1000. In nine studies, the prevalence of VAP was reported and ranged between 7.3% and 44.6%. Eight studies used endotracheal aspirates, five used protected specimen brushes, and bronchoalveolar lavage (BAL) was used in ten studies. High accuracy was observed (area under the curve [AUC] =0.92; 95% confidence interval [CI], 0.89-0.94) when all variables (i.e., true positives/negatives, false positives/negatives, etc.) were included for the prediction of bacterial pathogens. The pooled sensitivity for all studies was 75% (95% CI, 65% to 83%) and overall specificity was 92% (95% CI, 85% to 96%). The ability to predict MDR organisms was good (AUC 0.95), but only four studies could be used to calculate the AUC, and heterogeneity was significant (I2=63%). Twice-a-week sampling resulted in better sensitivity and specificity compared to less frequent sampling.

This study is the first to provide coherent evidence for the accuracy of surveillance cultures to predict VAP. The absence of MDR organisms on surveillance cultures may be reassuring in some cases as most studies reported moderate to high specificity and accuracy.  Some studies may have been missed because the authors did not examine other databases, such as EMBASE or grey literature. Additionally, the incidence and prevalence of VAP was not reported consistently in all studies, and the range was wide. Various techniques for lower respiratory tract sampling were employed, possibly biasing the sensitivity and specificity of the cultures. Heterogeneity was high for the accuracy of VAP prediction, with I2 values ranging from 0% to 92%. These differences in case mix may affect the external generalizability of the findings for specific patient populations. The interstudy variation, limited number of studies and cited limitations preclude formulating recommendations regarding routine lower respiratory surveillance cultures for the prediction of VAP. 

Concise Critical Appraisal is a regular feature authored by SCCM member Samuel M. Galvagno Jr., DO, PhD. Each installment highlights journal articles most relevant to the critical care practitioner.