We performed a systematic search from 1990 to 2011, using MEDLINE, EMBASE, and the Cochrane Library, including studies evaluating capnography as a diagnostic tool alone or in conjunction with other tests. After study quality evaluation, we calculated the pooled sensitivity, specificity, likelihood ratios, and diagnostic odds ratios.
We included 14 trials with 2,291 total subjects, with a 20% overall prevalence of pulmonary embolism. The pooled diagnostic accuracy for capnography was sensitivity 0.80 (95% confidence interval [CI] 0.76 to 0.83), specificity 0.49 (95% CI 0.47 to 0.51), negative likelihood ratio 0.32 (95% CI 0.23 to 0.45), positive likelihood ratio 2.43 (95% CI 1.70 to 3.46), and diagnostic odds ratio 10.4 (95% CI 6.33 to 17.1). The area under the summary receiver operating characteristic curve was 0.84. To reach pulmonary embolism posttest probabilities less than 1%, 2%, or 5%, pulmonary embolism prevalence or pretest probability had to be less than 3%, 5%, or 10% respectively. Because of interstudy differences in dead space measurements methodologies, the best cutoff in alveolar dead space or end tidal CO2 conferring the best negative likelihood ratio could not be evaluated.
Pooled data suggest a potential diagnostic role for capnography when the pulmonary embolism pretest probability is 10% or less, perhaps after a positive D-dimer test result.