key: cord-0938477-9yl6lg4y
authors: Silva, Beatriz Valente; Jorge, Cláudia; Plácido, Rui; Mendonça, Carlos; Urbano, Maria Luísa; Rodrigues, Tiago; Brito, Joana; da Silva, Pedro Alves; Rigueira, Joana; Pinto, Fausto J.
title: Pulmonary embolism and COVID-19: A comparative analysis of different diagnostic models performance
date: 2021-09-08
journal: Am J Emerg Med
DOI: 10.1016/j.ajem.2021.09.004
sha: 0e6dbe1d09b900fc5a51bb880403db751d541e69
doc_id: 938477
cord_uid: 9yl6lg4y

OBJECTIVE: Pulmonary embolism (PE) is a common complication of SARS-CoV-2 infection. Several diagnostic prediction rules based on pretest probability and D-dimer have been validated in non-COVID patients, but it remains unclear if they can be safely applied in COVID-19 patients. We aimed to compare the diagnostic accuracy of the standard approach based on Wells and Geneva scores combined with a standard D-dimer cut-off of 500 ng/ml with three alternative strategies (age-adjusted, YEARS and PEGeD algorithms) in COVID-19 patients. METHODS: This retrospective study included all COVID-19 patients admitted to the Emergency Department (ED) who underwent computed tomography pulmonary angiography (CTPA) due to PE suspicion. The diagnostic prediction rules for PE were compared between patients with and without PE. RESULTS: We included 300 patients and PE was confirmed in 15%. No differences were found regarding comorbidities, traditional risk factors for PE and signs and symptoms between patients with and without PE. Wells and Geneva scores showed no predictive value for PE occurrence, whether a standard or an age-adjusted cut-off was considered. YEARS and PEGeD algorithms were associated with increased specificity (19% CTPA reduction) but raising non-diagnosed PE. Despite elevated in all patients, those with PE had higher D-dimer levels. However, incrementing thresholds to select patients for CTPA was also associated with a substantial decrease in sensitivity. CONCLUSION: None of the diagnostic prediction rules are reliable predictors of PE in COVID-19. Our data favour the use of a D-dimer threshold of 500 ng/ml, considering that higher thresholds increase specificity but limits this strategy as a screening test.

Coronavirus disease 2019 (COVID-19) caused by Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) emerged at the end of 2019 and spread rapidly, becoming a major public health problem worldwide 1 . SARS-CoV-2 infection has been recognized as a hyperinflammatory and prothrombotic state, with pulmonary embolism (PE) being one of the most common complications of COVID-19 [2] [3] [4] [5] [6] . PE was considered equally likely based on the attendant physician's impression recorded in the medical chart. Patients were categorized as having low (< 4.0 points), moderate (4.5 -6.0 points) or high ( 6.5 points) pretest probability of PE.

The revised Geneva score (GENEVA) considers the following: previous objectively diagnosed PE or deep vein thrombosis (3 points), unilateral lower limb pain (3 points), heart rate > 95 or between 75-94 beats per minute (5 and 3 points, respectively), active malignant condition (2 points), haemoptysis (2 points), age > 65 years (1 point) and pain on limb palpation (4 points) 9 . Patients were categorized as having low (0-3 points), moderate (4-10 points) or high (11 points) clinical probability of PE.

In the standard approach, patients classified as high clinical probability on Wells or Geneva scores are selected to perform CTPA. In contrast, patients with low to moderate clinical probability perform CTPA if they have a D-dimer value above 500 ng/mL or above their individual cut-off if an age-adjusted approach was considered. The age-adjusted D-dimer threshold was defined by multiplying the patients' age by ten in patients above 50 years old 10 .

The YEARS algorithm is based on D-dimer levels and three clinical variables: haemoptysis, signs or symptoms of deep vein thrombosis and whether PE is the most likely diagnosis 7 . PE is excluded in patients with 0 YEARS items and a D-dimer level less than 1000 ng/mL or in patients with one or more YEARS items and a D-dimer level less than 500 ng/mL. All other patients should perform CTPA.

In the PEGeD algorithm, PE is ruled out in patients with low pretest probability and a D-dimer level of less than 1000 ng/mL or with a moderate pretest probability and a D-dimer level of less than 500 ng/mL 8 . All other patients, including those with high clinical probability, should perform CTPA. Pretest clinical probability in this algorithm is based on Wells score described above.

Categorical variables were presented as frequency rates and percentages and continuous variables as median with interquartile range. Categorial and continuous variables were compared using Pearson chisquare and Mann-Whitney tests, respectively. Sensitivity, specificity, positive and negative predictive values, likelihood ratios and diagnostic odds ratio were calculated and compared among the different diagnostic prediction rules. The discriminative power of each score to predict PE was determined by receiver operating characteristic (ROC) curve analysis and the area under the curve (AUC) was calculated.

Statistical significance was defined as a P value < 0.05. The statistical software used to analyse the data was SPSSv.26 (IBM).

A total of 1346 CTPAs were performed due to PE suspicion during the study period, 315 of them in patients with confirmed SARS-CoV-2 infection. The study flowchart is summarized in Figure 1 .

The demographic, clinical and laboratory features of patients with and without PE are shown in Table   1 . CTPA confirmed PE in 46 patients (15%). The vascular allocation of emboli showed a predominantly central distribution (59%), affecting main and lobar arteries (15% and 44%, respectively). Most PE had bilateral involvement (57%) and 22% of patients had evidence of right heart strain. Thrombolytic therapy was J o u r n a l P r e -p r o o f Journal Pre-proof not performed in any patient. Notably, PE was documented in three patients under anticoagulation (1 patient with lobar and 2 with subsegmental involvement), two of those died during hospitalization.

Despite being older, PE patients did not differ from non-PE patients regarding comorbidities, traditional risk factors for venous thromboembolism (VTE) and signs and symptoms at the ED presentation.

Although elevated in all patients, those with PE had higher D-dimer levels compared with non-PE patients. Furthermore, PE patients also showed significantly higher levels of cTnT-hs and NT-proBNP.

In univariate analysis, only age (OR: 1.024, 95%CI 1.002-1.047, p=0.036), D-dimer levels (OR: 1.018, 95%CI 1.004-1.031, p=0.010) and cTnT-hs (OR: 1.007, 95%CI 1.002-1.012, p=0.011) were identified as predictors of PE occurrence. None of the comorbidities or traditional risk factors for VTE were identified as PE predictors in this cohort.

Considering adverse clinical outcomes, PE occurrence was an independent predictor of hospitalization Table 2 shows the diagnostic performance and number of CTPAs correctly avoided in our cohort when progressively higher D-dimer cut-offs were applied. The use of a cut-off of 1000 ng/mL was associated with a significant increase in specificity and a non-significant decrease in sensitivity compared with a cut-off of 500 ng/mL (p<0.001 and p=0.063, respectively), although it was associated with five more missed PE diagnosis.

On the other hand, the cut-offs of 2590 ng/mL, 2669 ng/mL and 2903 ng/mL, proposed in previous studies, were all associated with a significant decrease in sensitivity and increase in specificity compared to the cutoff of 500 ng/mL (p<0.001 for all). Table 3 illustrates the accuracy of the different diagnostic prediction rules for PE.

The prevalence of risk factors for VTE in patients with or without PE is represented in Table 4 . No differences were found in Wells and Geneva scores and traditional risk factors between groups. Regarding patients with PE, 95.7% was considered as having low probability by Wells score. In addition, according to Wells or Geneva score, none of the patients was classified as having a high probability of PE. Thus, these scores shown no predictive value for PE occurrence (OR: 1.084, 95%CI 0.841-1.396, p=0.533; OR: 1.023, 95%CI 0.869-1.205, p=0.784, respectively). Twenty-one patients had a D-dimer value higher than the fixed D-dimer threshold of 500 ng/mL but lower than their age-adjusted cut-off. Among those patients, 3 had PE (2 patients with lobar and one patient with subsegmental involvement). When pretest clinical probability was combined with an ageadjusted cut-off, it resulted in lower sensitivity than the standard threshold of 500 ng/mL, although it was not statically significant (89.13% vs 95.65%, p=0.250). Regarding specificity, an age-adjusted cut-off resulted in a substantial increase in specificity compared to the standard cut-off (15.35% vs 8.27%, respectively, p<0.001). The AUC for both Wells and Geneva scores combined with a cut-off of 500 ng/mL and an age-adjusted cut-off (Figure 2A and 2B) suggests nearly no discriminative power. adjusted cut-off (p=0.150 and p=1.0, respectively). However, even though the decrease in sensitivity was not statistically significant, the YEARS algorithm was associated with four more missed PE diagnoses than the D-dimer cut-off of 500 ng/mL.

Although not statistically significant, the PEGeD algorithm had the lowest sensitivity of all diagnostic prediction rules considered (p=0.063 for a D-dimer cut-off of 500 ng/mL; p=0.50 for an age-adjusted cut-off; and p=1.00 for YEARS algorithm). PEGeD algorithm was associated with five more missed PE diagnoses than the D-dimer cut-off of 500 ng/mL. The PEGeD algorithm had a significantly higher specificity than a fixed and an age-adjusted D-dimer cut-off (p<0.001 for both). There was no difference regarding sensitivity and specificity between YEARS and PEGeD algorithms (p=1.00 for both).

In summary, none of the diagnostic strategies was significantly superior regarding sensitivity compared to each other (p>0.06 for all). However, an age-adjusted approach had a significantly higher specificity than a fixed D-dimer threshold of 500 ng/mL, whether clinical probability was evaluated by Wells or Geneva scores (p<0.001 for all). YEARS and PEGeD algorithms had higher specificity compared to an age-adjusted approach (p<0.001 for both). Lastly, there was no difference regarding specificity between YEARS and PEGeD algorithms (p=1.0).

Since the first reports, PE was described as one of the most common complications of SARS-COV-2 infection 4, 5 . This study showed a diagnostic yield of 15% in patients with SARS-CoV-2 infection that underwent a CTPA due to PE suspicion. Except for older age, we found no differences regarding comorbidities, traditional risk factors for VTE and signs and symptoms at presentation between patients with and without PE.

Wells and Geneva scores have been used for decades to predict PE. Currently, the need for CTPA is determined by combining the clinical probability with D-dimer levels 9 . However, considering that COVID-19 patients have a different thrombotic and inflammatory milieu, the usefulness of these prediction rules for PE in this condition is a matter of debate. Previous studies demonstrated no differences between patients with and without PE regarding traditional risk factors for VTE 11, [14] [15] [16] . Furthermore, Whyte M.B. et al. showed no difference in Wells score between those with and without PE in a cohort of 214 patients that performed CTPA due to PE suspicion 14 . Consistent with these reports, we found that neither the Wells nor Geneva scores were higher in COVID-19 patients with PE than those without PE. This finding highlights the importance of the association between the induced COVID-19 hyperinflammatory status and a thrombotic outcome 17 .

Pulmonary embolism was detected in 15% and 13% of patients considered as having low probability by Wells and Geneva scores, respectively, which is much higher than the 1 to 3% rate expected from the literature in non-COVID patients 18, 19 . Moreover, 95.7% of PE patients were considered as having low probability by Wells score (58.7% of those with a Wells score of 0 pts). We demonstrated that even when clinical pretest probability, evaluated by these scores, was combined with the D-dimer measurement, the discriminative power to predict PE in COVID-19 patients remains low.

According to the international guidelines for diagnosing and managing PE, the YEARS and PEGeD algorithms should be considered as an alternative to the standard approach to rule out PE 9 . The main advantage of these scores is their ability to safely reduce CTPA requests by adjusting the D-dimer cut-off to the clinical probability. To the best of our knowledge, our study was the first to compare these new proposed algorithms' diagnostic performance with the standard approach based on Wells or Geneva scores to predict J o u r n a l P r e -p r o o f Journal Pre-proof PE in COVID-19 patients. These algorithms were associated with an absolute reduction of 19.3% in the number of CTPA. However, they were also associated with decreased sensitivity that, although not statistically significant, is clinically relevant.

In agreement with previous studies, we found that PE patients had significantly higher D-dimer levels than non-PE patients 20, 12 . Considering that COVID-19 patients tend to have higher D-dimer levels even in the absence of PE, several authors proposed higher D-dimer thresholds to select patients for CTPA, based on maximizing the Youden index [11] [12] [13] . However, considering that the D-dimer is used as a triage test, the safety of this diagnostic approach is questionable. Mouhat B et al. proposed a D-dimer cut-off value of 2590 ng/mL to best predict PE occurrence in COVID-19 patients, which is slightly lower than the suggested by Ventura-Díaz et al. (2903 ng/mL) 11, 13 . With the cut-offs above mentioned, the authors reported a sensitivity of 83% and 81%, respectively, being far from the pretended negative predictive value of 100% for a triage test. As represented in Table 2 , when these cut-offs were applied to our cohort, we found an unacceptably low sensitivity, despite allowing a reduction of up to 68% in CTPAs. Considering the clinical implications of a missed PE diagnosis, defining a D-dimer threshold according to a higher negative predictive value might be more acceptable than adjusting the D-dimer threshold to increase specificity. Our results suggest that the 500 ng/mL cut-off point might be the safer strategy in COVID-19 patients. We found no difference in clinical presentation between patients with or without PE, which is in line with the previous reported 14, 15, 21, 22 .

Patients with PE had significantly higher cTnT-hs and NTpro-BNP levels than those without PE.

Although this finding is consistent with the reported in a cohort of 162 hospitalized patients with COVID-19

that performed CTPA 23 , other studies found no difference concerning cardiac biomarkers between patients with or without PE 15, 16, 24 . A higher incidence of central PE can explain our findings compared to those studies.

In our cohort, PE was not associated with higher mortality. This finding is consistent with a French multicentre cohort of 1204 patients, although other studies documented otherwise 25, 26 . This could be related to the heterogeneity of the severity of the concomitant COVID-19 disease.

According to the European Society of Cardiology recommendations in COVID-19, PE diagnosis in these patients should be based on algorithms combining pretest clinical probability with D-dimer 27 . These recommendations suggest that an unexpected respiratory worsening, signs of deep vein thrombosis, hypotension not attributable to other causes and a new or unexpected tachycardia may trigger the suspicion of PE. Although this is true for the general population, COVID-19 population seems to break the rule.

We demonstrated a remarkable absence of the usual comorbidities, typical VTE risk factors and overlap of signs and symptoms between PE and non-PE COVID-19 patients, which acknowledge recent published data. These findings reinforce the difficulty of PE diagnosis in COVID-19 patients and claims in favour of a different screening strategy. However, the use of higher D-dimer thresholds does not seem to be the solution to improve the diagnostic accuracy for PE in these patients, as it limits this strategy as a screening test.

Our study has some limitations. It was a retrospective single-center and chart review study. One of 

A Novel Coronavirus from Patients with Pneumonia in China

Wells score to predict pulmonary embolism in patients with coronavirus disease-2019

Abnormal coagulation parameters are associated with poor prognosis in patients with novel coronavirus pneumonia

Pulmonary Embolism and Deep Vein Thrombosis in COVID-19: A Systematic Review and Meta-analysis

Incidence of thrombotic complications in critically ill ICU patients with COVID-19

Cardiovascular Complications of COVID -19

Simplified diagnostic management of suspected pulmonary embolism (the YEARS study): a prospective, multicentre, cohort study

Diagnosis of Pulmonary Embolism with d -Dimer Adjusted to Clinical Probability

ESC Guidelines for the diagnosis and management of acute pulmonary embolism developed in collaboration with the European respiratory society (ERS)

Age-adjusted D-dimer cutoff levels to rule out pulmonary embolism: The ADJUST-PE study

Elevated D-dimers and lack of anticoagulation predict PE in severe COVID-19 patients

Acute pulmonary embolism in patients with 14

Pulmonary embolism in hospitalised patients with COVID-19

Prevalence of pulmonary embolism in patients with COVID-19 pneumonia and high D-dimer values: A prospective study

Pulmonary embolism in COVID-19 patients: A French multicentre cohort study

Incidence, risk factors, and thrombotic load of pulmonary embolism in patients hospitalized for COVID-19 infection

Clinical prediction rules for pulmonary embolism: A systematic review and meta-analysis

Clinical experience and pre-test probability scores in the diagnosis of pulmonary embolism

Pulmonary embolism in patients with COVID-19 pneumonia

Acute pulmonary embolism in non-hospitalized COVID-19 patients referred to CTPA by emergency department

Association Between Pulmonary Embolism and COVID-19 in Emergency Department Patients Undergoing Computed Tomography Pulmonary Angiogram: The PEPCOV International Retrospective Study

Elevated D-dimers and lack of anticoagulation predict PE in severe COVID-19 patients

Characteristics of Acute Pulmonary Embolism in Patients With COVID-19 Associated Pneumonia From the City of Wuhan

Pulmonary embolism in COVID-19 patients: Prevalence, predictors and clinical outcome

Thromboembolism risk of COVID-19 is high and associated with a higher risk of mortality: A systematic review and meta-analysis

EClinicalMedicine Elsevier Ltd

European Society of cardiology. ESC Guidance for the Diagnosis and Management of CV Disease during the COVID-19 Pandemic

Diagnostic accuracy of Wells and Geneva scores combined with a fixed and an age-adjusted cut-off, YEARS algorithm and PEGeD algorithm to predict pulmonary embolism in COVID-19 patients

AUC, area under the curve