Selecting pregnant or postpartum women with suspected pulmonary embolism for diagnostic imaging: the DiPEP diagnostic study with decision-analysis modelling

Abstract

Background Pulmonary embolism (PE) is a leading cause of death in pregnancy and post partum, but the symptoms of PE are common in normal pregnancy. Simple diagnostic tests are needed to select women for diagnostic imaging. Objective To estimate the accuracy, effectiveness and cost-effectiveness of clinical features, decision rules and biomarkers for selecting pregnant or postpartum women with a suspected PE for imaging. Design An expert consensus study to develop new clinical decision rules, a case–control study of women with a diagnosed PE or a suspected PE, a biomarker study of women with a suspected PE or diagnosed deep-vein thrombosis (DVT) and decision-analysis modelling. Setting Emergency departments and consultant-led maternity units. Participants Pregnant/postpartum women with a diagnosed PE from any hospital reporting to the UK Obstetric Surveillance System research platform and pregnant/postpartum women with a suspected PE or diagnosed DVT at 11 prospectively recruiting sites. Interventions Clinical features, decision rules and biomarkers. Main outcome measures Sensitivity, specificity, area under receiver operating characteristic (AUROC) curve, quality-adjusted life-years (QALYs) and health-care costs. Results The primary analysis involved 181 women with PE and 259 women without PE in the case–control study and 18 women with DVT, 18 with PE and 247 women without either in the biomarker study. Most clinical features showed no association with PE. The AUROC curves for the clinical decision rules were as follows: primary consensus, 0.626; sensitive consensus, 0.620; specific consensus, 0.589; PE rule-out criteria, 0.621; simplified Geneva score, 0.579; Wells’s PE criteria (permissive), 0.577; and Wells’s PE criteria (strict), 0.732. The sensitivities and specificities of the D-dimer measurement were 88.4% and 8.8%, respectively, using a standard threshold, and 69.8% and 32.8%, respectively, using a pregnancy-specific threshold. Previous venous thromboembolism, long-haul travel, multiple pregnancy, oxygen saturation, recent surgery, temperature and PE-related chest radiograph abnormality were predictors of PE on multivariable analysis. We were unable to derive a rule through multivariable analysis or recursive partitioning with adequate accuracy. The AUROC curves for the biomarkers were as follows: activated partial thromboplastin time – 0.669, B-type natriuretic peptide – 0.549, C-reactive protein – 0.542, Clauss fibrinogen – 0.589, enzyme-linked immunosorbent assay D-dimer – 0.668, Innovance D-dimer (Siemens Healthcare Diagnostics Products GmbH, distributed by Sysmex UK Ltd, Milton Keynes, UK) – 0.651, mid-regional pro-atrial natriuretic peptide (MRproANP) – 0.524, prothrombin fragment 1 + 2 – 0.562, plasmin-antiplasmin – 0.639, Prothombin time – 0.613, thrombin generation lag time – 0.702, thrombin generation endogenous potential – 0.559, thrombin generation peak – 0.596, thrombin generation time to peak – 0.655, tissue factor – 0.531 and troponin – 0.597. The repeat analysis excluding women who had received anticoagulation was limited by the small number of women with PE (n = 4). The health economic analysis showed that a strategy of scanning all women with a suspected PE accrued more QALYs and incurred fewer costs than any selective strategy based on a clinical decision rule and was therefore the dominant strategy. Limitations The findings apply specifically to the diagnostic assessment of women with a suspected PE in secondary care. Conclusions Clinical features, decision rules and biomarkers do not accurately, effectively or cost-effectively select pregnant or postpartum women with a suspected PE for diagnostic imaging. Future work New diagnostic technologies need to be developed to detect PE in pregnancy. Trial registration Current Controlled Trials ISRCTN21245595. Funding details This project was funded by the National Institute for Health Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 22, No. 47. See the NIHR Journals Library website for further project information.