Toward a tailored diagnostic standard for future diagnostic studies in pulmonary embolism: Communication from the SSC of the ISTH

Abstract

Dear Editor, Over the past 20 years, there has been a shift in diagnostic strategies for diagnosing venous thromboembolism. Diagnostic tests for suspected acute pulmonary embolism (PE) have evolved from pulmonary angiography, which was the only available diagnostic test up to 30 years ago, to ventilation perfusion scintigraphy, computed tomography pulmonary angiography and combinations of a D‐dimer test and clinical decision rules to determine the pretest probability of having PE.1.Huisman M.V. Klok F.A. Diagnostic management of acute deep vein thrombosis and pulmonary embolism.J Thromb Haemost. 2013; 11: 412-422Crossref PubMed Scopus (125) Google Scholar The current diagnostic standard is an integrated algorithm starting with pretest probability assessment and a D‐dimer test, only followed by radiological imaging in case of high pretest probability and/or a D‐dimer test result above the applicable threshold.2.Huisman M.V. Barco S. Cannegieter S.C. et al.Pulmonary embolism.Nat Rev Dis Primers. 2018; 4Crossref PubMed Scopus (152) Google Scholar Because of the less invasive character and more widespread availability of diagnostic tests for acute PE, the threshold for testing for PE has considerably decreased, which translated to a lower disease prevalence in study populations.3.Dronkers C.E.A. Ende‐Verhaar Y.M. Kyrle P.A. et al.Disease prevalence dependent failure rate in diagnostic management studies on suspected deep vein thrombosis: communication from the SSC of the ISTH.J Thromb Haemost. 2017; 15: 2270-2273Crossref PubMed Scopus (15) Google Scholar, 4.Dronkers C.E.A. van der Hulle T. Le Gal G. et al.Towards a tailored diagnostic standard for future diagnostic studies in pulmonary embolism: communication from the SSC of the ISTH.J Thromb Haemost. 2017; 15: 1040-1043Crossref PubMed Scopus (67) Google Scholar Nonetheless, the accepted failure rate (i.e., the 3‐month incidence of symptomatic venous thromboembolism in patients in whom PE was considered absent and were left untreated) in diagnostic management studies remained the upper limit of the 95% confidence interval (95% CI) of the failure rate of conventional pulmonary angiography (2.7%).5.van Beek E.J. Brouwerst E.M. Song B. et al.Clinical validity of a normal pulmonary angiogram in patients with suspected pulmonary embolism–a critical review.Clin Radiol. 2001; 56: 838-842Abstract Full Text PDF PubMed Scopus (150) Google Scholar Because, according to the theorem of Bayes the diagnostic failure rate (posttest probability) is associated with the disease prevalence (pretest probability) in a study population, the SSC proposed a new safety threshold dependent on disease prevalence to prevent that unsafe strategies would get accepted as accurate. To provide guidance on how to evaluate new diagnostic strategies, or existing strategies in various health care settings with corresponding higher or lower disease prevalence, we proposed a varying safety threshold, modeled on a linear regression analysis of disease prevalence and failure rate in pooled data from published high‐quality diagnostic management studies. Application of this new safety threshold will lead to a stricter safety threshold in study populations with a sporadic disease prevalence. Freund et al.6.Freund Y. Roussel M. Kline J. et al.Towards a tailored diagnostic standard for future diagnostic studies in pulmonary embolism: communication from the SSC of the ISTH: COMMENT.J Thromb Haemost. 2021; 15: 1040-1043Google Scholar raise an interesting comment to the SSC of the ISTH communication paper published in 2017, which proposed a new diagnostic safety threshold for future diagnostic studies in patients with suspected acute PE.4.Dronkers C.E.A. van der Hulle T. Le Gal G. et al.Towards a tailored diagnostic standard for future diagnostic studies in pulmonary embolism: communication from the SSC of the ISTH.J Thromb Haemost. 2017; 15: 1040-1043Crossref PubMed Scopus (67) Google Scholar They discuss that the definition of failure rate should be calculated only in patients in whom the new diagnostic strategy actually affected the diagnostic workup (i.e., in whom imaging was avoided) rather than the complete population. We agree with this comment. However, in our view, this latter group are all patients in whom PE is ruled out without imaging test and not the small subgroup of patients where the study strategy was effectively applied (i.e., in whom the D‐dimer threshold was changed compared to the “standard” algorithm). We agree that reporting the failure rate in the subgroup of patients affected by the change in algorithm is important to enable a comprehensive appraisal of the new algorithm. However, determining an acceptable failure rate in such a subgroup was outside the scope of this SSC guidance. In their interpretation of the SSC communication, Freund and colleagues6.Freund Y. Roussel M. Kline J. et al.Towards a tailored diagnostic standard for future diagnostic studies in pulmonary embolism: communication from the SSC of the ISTH: COMMENT.J Thromb Haemost. 2021; 15: 1040-1043Google Scholar slightly misinterpreted the application of the formula to calculate an appropriate safety threshold (1.82 + 0.0053 × prevalence). This formula does not determine the upper limit of the CI of the acceptable safety threshold, but the maximum acceptable “point estimate” of the failure rate. Considering this, the point estimate of the failure rate of the pregnancy‐adjusted YEARS algorithm observed in the Artemis study (0.51; 95% CI, 0.09–2.9) is well below the post hoc calculated safety threshold (1.82 + 0.0053 × 4 = 1.84%) for avoiding computed tomography pulmonary angiography.7.van der Pol L.M. Tromeur C. Bistervels I.M. et al.Pregnancy‐adapted YEARS algorithm for diagnosis of suspected pulmonary embolism.N Engl J Med. 2019; 380: 1139-1149Crossref PubMed Scopus (0) Google Scholar Hence, according to the SSC communication, the diagnostic algorithm should be considered safe, even despite the low baseline PE prevalence. In the second example discussed, a “heads and tails” strategy is applied to a virtual study population with a disease prevalence of 2.5%. The failure rate with “the toss of a coin” in a population of 4000 patients, of whom 2000 would have a negative test result, would be 50/2000 = 2.5%. The accepted failure rate according to the SSC formula would be 1.83%, however. Besides the question if it is reasonable to perform any test in a study population with such a low disease prevalence, the heads and tails strategy does not match the safety threshold as proposed by the SSC, confirming its relevance. We therefore conclude that the proposed new way of determining the safety threshold for PE and DVT diagnostic studies as proposed by the SSC remains valid and appropriate.3.Dronkers C.E.A. Ende‐Verhaar Y.M. Kyrle P.A. et al.Disease prevalence dependent failure rate in diagnostic management studies on suspected deep vein thrombosis: communication from the SSC of the ISTH.J Thromb Haemost. 2017; 15: 2270-2273Crossref PubMed Scopus (15) Google Scholar, 4.Dronkers C.E.A. van der Hulle T. Le Gal G. et al.Towards a tailored diagnostic standard for future diagnostic studies in pulmonary embolism: communication from the SSC of the ISTH.J Thromb Haemost. 2017; 15: 1040-1043Crossref PubMed Scopus (67) Google Scholar None. Charlotte E. A. Dronkers and Frederikus A. Klok wrote the first draft of the manuscript. Tom van der Hulle, Grégoire Le Gal, Paul A. Kyrle, Menno V. Huisman, and Suzanne C. Cannegieter critically revised the paper for important intellectual content.