Thijs E van Mens, Luuk JJ Scheres, Paulien G de Jong, Mariska MG Leeflang, Mathilde Nijkeuter, Saskia Middeldorp

Pulmonary embolism is a leading cause of pregnancy-related death. An accurate diagnosis in pregnant patients is crucial to prevent untreated pulmonary embolism as well as unnecessary anticoagulant treatment and future preventive measures. Applied imaging techniques might perform differently in these younger patients with less comorbidity and altered physiology, who largely have been excluded from diagnostic studies. To determine the diagnostic accuracy of computed tomography pulmonary angiography (CTPA), lung scintigraphy and magnetic resonance angiography (MRA) for the diagnosis of pulmonary embolism during pregnancy. We searched MEDLINE and Embase until July 2015. We used included studies as seeds in citations searches and in 'find similar' functions and searched reference lists. We approached experts in the field to help us identify non-indexed studies. We included consecutive series of pregnant patients suspected of pulmonary embolism who had undergone one of the index tests (computed tomography (CT) pulmonary angiography, lung scintigraphy or MRA) and clinical follow-up or pulmonary angiography as a reference test. Two review authors performed data extraction and quality assessment. We contacted investigators of potentially eligible studies to obtain missing information. In the primary analysis, we regarded inconclusive index test results as a negative reference test, and treatment for pulmonary embolism after an inconclusive index test as a positive reference test. We included 11 studies (four CTPA, five lung scintigraphy, two both) with a total of 695 CTPA and 665 lung scintigraphy results. Lung scintigraphy was applied by different techniques. No MRA studies matched our inclusion criteria.Overall, risk of bias and concerns regarding applicability were high in all studies as judged in light of the review research question, as was heterogeneity in study methods. We did not undertake meta-analysis. All studies used clinical follow-up as a reference standard, none in a manner that enabled reliable identification of false positives. Sensitivity and negative predictive value were therefore the only valid test accuracy measures.The median negative predictive value for CTPA was 100% (range 96% to 100%). Median sensitivity was 83% (range 0% to 100%).The median negative predictive value for lung scintigraphy was 100% (range 99% to 100%). Median sensitivity was 100% (range 0% to 100%).The median frequency of inconclusive results was 5.9% (range 0.9% to 36%) for CTPA and 4.0% (range 0% to 23%) for lung scintigraphy. The overall median prevalence of pulmonary embolism was 3.3% (range 0.0% to 8.7%). Both CTPA and lung scintigraphy seem appropriate for exclusion of pulmonary embolism during pregnancy. However, the quality of the evidence mandates cautious adoption of this conclusion. Important limitations included poor reference standards, necessary assumptions in the analysis regarding inconclusive test results and the inherent inability of included studies to identify false positives. It is unclear which test has the highest accuracy. There is a need for direct comparisons between diagnostic methods, including MR, in prospective randomized diagnostic studies.