Ethics
The study was approved as low-risk research by the Human Research Ethics Committee of the hospital, with a waiver of individual consent. Formal trial registration was not mandated.
Results
D-dimer levels were obtained in a total of 2165 individual patients. Hospital records were not available for 37 patients (1.7%); these were excluded from further analysis. A further 103 patients had D-dimer testing for reasons other than suspected PE and 10 were less than 16 years of age. These patients were all excluded. The remaining 2015 patients met the inclusion criteria (Figure 1).
Characteristics of the study population are summarised in Table 2. The mean age of patients was 50.5 years (range 16.0 to 98.4) with 1188 females (59.0%). The mean D-dimer level was 0.93 mg/L (range 0.01 to 20.00). 63% of the study population did not have imaging for PE, while 37% did (V/Q in 14%, CTPA in 23% or both in just under 1%).
The patients were classified according to their clinical probability in Table 3. Approximately 62% were considered to be low probability; intermediate 37.4%; high, less than 1%. The total number of patients who had imaging was tabulated. Those with a final diagnosis of PE was also tabulated. In the low probability cohort (according to the RGS score), just over 32% had imaging. Of those, 7.5 % were diagnosed with PE. In the intermediate probability cohort, 44% had imaging. This yielded a diagnosis of PE in 13.3% of patients. For the high probability group, the incidence of PE imaging and diagnosis were 66.7% and 30%, respectively. In total, of approximately 36.8% of patients who underwent D-dimer testing, 10.4% were subsequently diagnosed with PE.
The number of patients imaged – and the number who were found to have PE – were gauged according to whether the D-dimer level was positive (as shown in Table 4). These values were also calculated for age-adjusted D-dimer levels (as shown in Table 5).
The clinical risk calculated for each patient (as summarised in Table 3) was derived from the individual components of RGS. The incidence of each component is listed in Table 6.
Prevalence of each component of RGS in the entire study population as well as when PE was diagnosed on imaging, when imaging did not show PE and when imaging was not performed was also calculated and shown in table 6. Correlation (coefficient) with a diagnosis of PE on imaging was calculated for each component of RGS, the overall RGS risk category and a positive D-dimer (absolute or age-adjusted) and summarised in table 7. The correlation of positive D-dimer with positive imaging for PE was 0.7033 and the correlation for a positive D-dimer by age-adjusted criteria was 0.5928. The correlation coefficient (CC) between low, intermediate and high risk group on their own and a final diagnosis of PE on imaging was calculated at 0.1332, 0.1278 and 0.0817, respectively. On the other hand, the CC between a positive absolute D-dimer and a final diagnosis of PE on imaging in patients with low, intermediate and high risk group was calculated at 0.7527, 0.6256 and 0.4195, respectively. For the age-adjusted D-dimer, the correlations were calculated at 0.6490, 0.4987 and 0.3550, respectively. The overall CC for all risk categories on their own was calculated at 0.1107 and for a positive absolute D-dimer at 0.7033 (table 8).
Discussion
While previous studies were unable to account adequately for patients without imaging, this obstacle was overcome with a novel statistical approach. Previous studies have evaluated the sensitivity and specificity of D-dimer for diagnosing or excluding PE - often in comparison to clinical risk scores. This was well reviewed by Weitz et al16. However, such methods could only be applied (and sensitivity, specificity and predictive values calculated) if every single patient in the study had V/Q or CTPA imaging subsequently. It would not account for patients who did not have subsequent imaging. To accommodate for this unknown factor, the choice of statistical method had to observe this fact.
Thus, Cramer’s V method was used as its major advantage is that it accounts for many patients not having further imaging. Fisher’s Exact Test was considered as an alternative, however was felt to be less appropriate as some categories are very large. Thus, using Fisher’s Exact Test would be less reliable. We therefore used Cramer’s V despite the potential inadequacies of having some small groups - such as those with haemoptysis who eventually diagnosed with PE (just two cases amongst the entire cohort).
Conversely, it can be construed as a drawback of this study design to have results expressed in statistical terms with which some clinicians may not be familiar. The outcomes are not expressed in terms of sensitivity, specificity and predictive values which are used much more commonly in clinical practice and guidelines. However, CC may be a more useful metric in this situation.
Alternatively, some studies on PE have used assessment by respiratory physician at six months after the index event as the gold standard. Obviously, this is equally problematic. For the pragmatic intents of this study, it had to be assumed that CTPA and V/Q are sufficiently and equally accurate. It is known, however, that there is a percentage of indeterminate CTPA studies – reported as up to 6%17. This may account for some of the patients listed in table 2 as having undergone both studies. The central finding was that the D-dimer level was superior to RGS in terms of its higher correlation with a diagnosis of PE on subsequent imaging. We demonstrated that the correlation between a positive D-dimer level with a diagnosis of PE was 0.7033 compared to 0.1103 for risk category; the former implying strong correlation in statistical terms, and the latter, moderate.
Unsurprisingly, the correlation between a positive D-dimer result and diagnosis of PE was well short of 100%. However, D-dimer has been shown to have a greater correlation with diagnosis of PE (on imaging). It can, thus, avoid using further tomographic imaging unnecessarily. This would be desirable clinically6. This study, as elsewhere18, clearly showed over-use of tomographic imaging in patients with lower pre-test probability.
If more patients with a high clinical pre-test probability were imaged, how would this have altered results? The group with a high clinical pre-test probability would probably have had a higher correlation with diagnosis of PE but it would be unlikely to bridge the gap from 0.0814 to 0.4195. If all clinically high-probability patients with negative imaging results had positive imaging instead, the correlation would have been 0.2716. How is it possible that the correlation was so low even if the test had relatively high sensitivity, specificity and positive predictive value? The answer is simply because there was such a large proportion (still 33%) with no imaging results. This highlights how calculating correlation better accounts for those without imaging.
If a greater number of patients with negative D-dimer were imaged – and presumed to have no evidence of PE on either CTPA or V/Q – then the correlation would be even higher than that documented. Similarly, if a greater number of patients with positive D-dimer were imaged – and presumed to have PE on either CTPA or V/Q – then the correlation would also be higher than that found.
It has been suggested that age-adjusted D-dimer reference ranges have higher sensitivity and specificity for PE19-22. This is particularly in the context of a low pre-test probability23. In this study, however, the absolute D-dimer level had a higher correlation with a diagnosis of PE than age adjusted D-dimer. Further, the correlation of D-dimer levels with a subsequent diagnosis of PE - for each category of pre-test probability - were consistently higher for the absolute value than the age-adjusted values (as shown in table 8). Our findings in this regard would be consistent with a minority of studies/publications such as the work of a Canadian group24.
Conclusion
A positive D-dimer test, absolute or age-adjusted, was found to have a higher correlation with a diagnosis of PE on subsequent V/Q SPECT or CTPA than a clinical risk score. Thus, the relevance of a D-dimer test in the context of suspected PE is no longer limited to a negative result to aid excluding PE and a positive test could be equally valuable. Given a positive D-dimer had a higher correlation with PE diagnosis than a clinical risk score system (like RGS) in our cohort, we conclude that it would be more sensible and reliable to determine if tomographic imaging is further required based on a positive D-dimer than a clinical prediction score on its own and thus, reduce resource and economic burdens in the health system. A validation study, preferably utilising the same statistical method, should be carried out in light of these findings.
References
1. Shiraev TP, Omari A, Rushworth RL. Trends in pulmonary embolism morbidity and mortality in Australia. Thromb Res 2013;132:19-25.
2. Jelinek GA, Ingarfield SL, Mountain D, Gibson NP, Jacobs IG. Emergency Department diagnosis of pulmonary embolism is associated with significantly reduced mortality: a linked data population study. Emerg Med Australas 2009;21:269-276.
3. Australian Government Department of Health. MBS Online Medicare Benefits Schedule. Available from: http://www.mbsonline.gov.au/internet/mbsonline/publishing.nsf/Content/downloads-200801. [accessed 20 April 2021]
4. Ehrman RR, Malik AN, Smith RK, Kalarikkal Z, Huang A, King RM, et al. Serial use of existing clinical decisions aids can reduce computed tomography pulmonary angiography for pulmonary embolism. Intern Emerg Med 2021. doi: 10.1007/s11739-021-02703-1.
5. Kline JA, Runyon MS, Webb WB, Jones AE, Mitchell AM. Prospective study of the diagnostic accuracy of the simplify D-dimer assay for pulmonary embolism in emergency department patients. Chest 2006;129:1417-1423.
6. Eng CW, Wansaicheong G, Goh SK, Earnest A, Sum C. Exclusion of acute pulmonary embolism: computed tomography pulmonary angiogram or D-dimer? Singapore Med J 2009;50:403–406.
7. Kline JA, Nelson RD, Jackson RE, Courtney DM. Criteria for the safe use of D-dimer testing in emergency department patients with suspected pulmonary embolism: a multicenter US study. Ann Emerg Med 2002;39:144-152.
8. Dunn KL, Wolf JP, Dorfman DM, Fitzpatrick P, Baker JL, Goldhaber SZ. Normal D-dimer levels in emergency department patients suspected of acute pulmonary embolism. J Am Coll Cardiol 2002;40:1475-1478.
9. Wong DD, Ramaseshan G, Mendelson RM. Comparison of the Wells and Revised Geneva Scores for the diagnosis of pulmonary embolism: an Australian experience. Intern Med J 2011;41:258-263.
10. Yap KS, Kalff V, Turlakow A, Kelly MJ. A prospective reassessment of the utility of the Wells score in identifying pulmonary embolism. Med J Aust 2007;187:333-336.
11. Douma RA, Mos IC, Erkens PM, Nizet TA, Durian MF, Hovens MM, van Houten AA, et al.; Prometheus Study Group. Performance of 4 clinical decision rules in the diagnostic management of acute pulmonary embolism: a prospective cohort study. Ann Intern Med. 2011;154:709-718.
12. Ghanima W, Abdelnoor M, Mowinckel MC, Sandset PM. The performance of STA-Liatest D-dimer assay in out-patients with suspected pulmonary embolism. Br J Haematol 2006;132:210-215.
13. Douma RA, le Gal G, Söhne M, Righini M, Kamphuisen PW, Perrier A, et al. Potential of an age adjusted D-dimer cut-off value to improve the exclusion of pulmonary embolism in older patients: a retrospective analysis of three large cohorts. BMJ 2010;340:c1475.
14. Bajc M, Neilly JB, Miniati M, Schuemichen C, Meignan M, Jonson B. EANM guidelines for ventilation/perfusion scintigraphy : Part 2. Algorithms and clinical considerations for diagnosis of pulmonary emboli with V/P(SPECT) and MDCT. Eur J Nucl Med Mol Imaging. 2009;36:1528-1538.
15. Akoglu H. User’s guide to correlation coefficients. Turk J Emerg Med 2018;18:91-93.
16. Weitz JI, Fredenburgh JC, Eikelboom JW. A Test in Context: D-Dimer. J Am Coll Cardiol 2017;70:2411-2420.
17. Yeo JH, Zhou L, Lim R. Indeterminate CT pulmonary angiogram: Why and does it matter? J Med Imaging Radiat Oncol 2017;6:18-23.
18. Kauppi JM, Airaksinen KEJ, Saha J, Bondfolk A, Pouru JP, Purola P, Jaakkola S, Lehtonen J, Vasankari T, Juonala M, Kiviniemi T. Adherence to risk-assessment protocols to guide computed tomography pulmonary angiography in patients with suspected pulmonary embolism. Eur Heart J Qual Care Clin Outcomes. 2021 Mar 16:qcab020. doi: 10.1093/ehjqcco/qcab020.
19. Righini M, Van Es J, Den Exter PL, Roy PM, Verschuren F, Ghuysen A, et al. Age-adjusted D-dimer cutoff levels to rule out pulmonary embolism: the ADJUST-PE study. JAMA 2014;311:1117-1124.
20. Jaconelli T, Crane S. Towards evidence based emergency medicine: best BETs from the Manchester Royal Infirmary. BET 2: Should we use an age adjusted D-dimer threshold in managing low risk patients with suspected pulmonary embolism? Emerg Med J 2015;32:335-337.
21. Fuchs E, Asakly S, Karban A, Tzoran I. Age-Adjusted Cutoff D-Dimer Level to Rule Out Acute Pulmonary Embolism: A Validation Cohort Study. Am J Med 2016;129:872-878.
22. Sharp AL, Vinson DR, Alamshaw F, Handler J, Gould MK. An Age-Adjusted D-dimer Threshold for Emergency Department Patients With Suspected Pulmonary Embolus: Accuracy and Clinical Implications. Ann Emerg Med 2016;67:249-257.
23. Farm M, Siddiqui AJ, Onelöv L, Järnberg I, Eintrei J, Maskovic F, et al. Age-adjusted D-dimer cut-off leads to more efficient diagnosis of venous thromboembolism in the emergency department: a comparison of four assays. J Thromb Haemost 2018;16:866-875.
24. Takach Lapner S, Julian JA, Linkins LA, Bates SM, Kearon C. Questioning the use of an age-adjusted D-dimer threshold to exclude venous thromboembolism: analysis of individual patient data from two diagnostic studies. J Thromb Haemost 2016;14:1953-1959.