Patients
A total of 158 patients who had a CTA were included. 66 (41.8%) patients had a positive CTA and 92 (58.2%) had a negative CTA. The median age of CTA+ was 59 years, of whom 61% were male. The median age of CTA- was 64.5 years, of whom 52% were male. CTA+ comprised more of White/Caucasians (33% vs 13%), and less Asians (0% vs 10%) compared to CTA- (p=0.002). CTA+ had a significantly fewer patients with chronic obstructive pulmonary disease (COPD)/asthma (9% vs 24%, p=0.03) compared to CTA-. The baseline characteristics of both groups are summarized in Table 1.
Table 1. Demographic Data of Patients with and without a PE
Characteristics
|
Positive chest CTA consistent with a PE
|
Negative chest CTA ruling out a PE
|
Total
|
P-value
|
Demographics
|
|
|
|
|
Age (yr), median (IQR)
Males
Females
BMI (kg/m2), median (IQR)
|
59 (49-71)
40 (61)
26 (39)
28.6 (24.6-35.0)
|
64.5 (52.8-71)
48 (52)
44 (48)
27.6 (23.4-33.2)
|
62.5 (49.3-71.0)
88 (56)
70 (44)
28 (23.7-33.9)
|
0.40
0.37
0.32
|
Race
|
|
|
|
|
Black or African American
Hispanic
White
Asian
Others
|
18 (27)
15 (23)
22 (33)
0 (0)
11 (17)
|
24 (26)
22 (24)
12 (13)
9 (10)
25 (27)
|
42 (27)
37 (23)
34 (22)
9 (6)
36 (23)
|
0.002
|
Comorbidities
|
|
|
|
|
Hypertension
Diabetes
Hyperlipidemia
Active malignancy
COPD/asthma
Coronary artery disease
Congestive heart failure
Atrial fibrillation
Chronic kidney disease
Chronic liver disease
Stroke
Autoimmune disease
Previous gastrointestinal bleed
Prior DVT/PE
Oral contraceptive pill use
Smoking
No
Active/prior
Family history of DVT/PE
History suggestive of hyper-coagulable state
Recent surgery / immobilization
Chronic home anticoagulation
None
Lovenox/DOAC/coumadin
Antiplatelet
DAPT
|
25 (38)
17 (26)
19 (29)
6 (9)
6 (9)
8 (12)
2 (3)
3 (5)
4 (6)
1 (2)
5 (8)
7 (11)
0 (0)
7 (11)
0 (0)
46 (70)
20 (30)
2 (3)
2 (3)
8 (12)
63 (96)
3 (5)
9 (14)
1 (2)
|
50 (54)
24 (26)
31 (34)
16 (17)
22 (24)
17 (19)
9 (10)
8 (9)
10 (11)
4 (4)
9 (10)
8 (9)
4 (4)
12 (13)
1 (1)
62 (67)
30 (33)
0 (0)
7 (8)
4 (4)
83 (90)
9 (10)
20 (22)
3 (3)
|
75 (48)
41 (26)
50 (32)
22 (14)
28 (18)
25 (16)
11 (7)
11 (7)
14 (9)
5 (3)
14 (9)
15 (10)
4 (3)
19 (12)
1 (1)
108 (68)
50 (32)
2 (1)
9 (6)
12 (8)
146 (92)
12 (8)
29 (18)
4 (3)
|
0.05
1.00
0.63
0.21
0.03
0.39
0.18
0.36
0.40
0.40
0.84
0.90
0.14
0.83
1.00
0.89
0.17
0.31
0.12
0.36
0.28
0.64
|
Clinical data: laboratory values, vital signs
Differences in laboratory values and vital signs at admission, at peak, and at the time of CTA are shown in Table 2. Compared to CTA-, CTA+ had a longer time from symptom onset to admission (7 days vs 4 days, p=0.05), higher admission D-dimer (6.87 vs 1.59, p<0.0001), peak D-dimer (9.26 vs 3.8, p=0.0008), admission white blood cells (WBC) (10.2 vs 7.45, p=0.005), admission platelets (272.5 vs 199.5, p=0.001), admission total bilirubin (0.8 vs 0.5, p=0.003), admission direct bilirubin (0.4 vs 0.3, p=0.021), and admission troponin (0.015 vs 0.01, p=0.01). The median PESI score at the time of CTA was significantly higher for CTA+ than CTA- (125.5 vs. 117.5, p=0.006). No statistically significant difference was seen in lactate, BNP (b-type natriuretic peptide), troponin, blood pressure, oxygen saturations, or heart rate at the time of CTA between the two groups.
Table 2. Laboratory values and vital signs at admission, at peak, and at the time of CTA
Variables
|
Positive chest CTA consistent with a PE (n=66)
|
Negative chest CTA ruling out a PE (n=92)
|
Total
|
P-value
|
|
Median (IQR)
|
Median (IQR)
|
Median (IQR)
|
|
Time from symptom onset to admission (days)
Admission D-dimer (ug/mL)
Admission LDH (U/L)
Admission CRP
Admission WBC
Admission Platelets
Admission total bilirubin
Admission direct bilirubin
Admission troponin
Admission BNP
Peak IL-6
Peak LDH (U/L)
Peak CRP (mg/L)
Peak D-dimer
Troponin at time of CTA
Lactate at time of CTA
PESI score at time of CTA
SBP at time of CTA
DBP at time of CTA
SpO2 at time of CTA
Heart rate at time of CTA
|
7 (2-14)
6.87 (2.10-20)
476.5 (353.75-6)
105.65 (68.8-199.5)
10.2 (6.5-14.78)
272.5 (195.25-349)
0.8 (0.6-1.2)
0.4 (0.3-0.5)
0.015 (0.01-0.097)
27.08 (10.3-95.48)
74.6 (27.23-130.5)
555 (377.5-696.25)
226.5 (124.7-302.3)
9.26 (2.60-20)
0.02 (0.01-0.1)
1.7 (1.3-2.9)
125.5 (113.3-140.5)
126 (111-141)
75 (66-82)
96 (94-98)
104 (86-114)
|
4 (2-7)
1.59 (0.79-3.4)
415.5 (308-591.8)
105 (42.01-192.9)
7.45 (5.9-10.4)
199.5 (158.8-261.3)
0.5 (0.4-0.9)
0.3 (0.2-0.5)
0.01(0.00-0.03)
25.8 (10.0-63.8)
61.15 (28.03-218.6)
558 (401.5-813)
196.2 (103.1-272.6)
3.8 (1.47-8.92)
0.01 (0.01-0.04)
1.6(1.2-2.0)
117.5 (100.8-133.5)
126( 114-136.5)
76 (70-83)
95 (93-97.3)
99.5 (86-110)
|
5 (2-10)
2.32 (1.14-10.7)
440 (319.5-619.5)
105.45 (55.4-198.4)
8.1 (6.1-12.3)
216 (170-316.3)
0.7 (0.4-1)
0.3 (0.2-0.5)
0.013 (0.01-0.05)
25.9 (10.0-71.2)
72.75 (26.675-174)
558 (386-789.5)
209 (113.04-281.4)
4.7 (1.71-17.5)
0.02 (0.01-0.06)
1.7 (1.2-2.3)
122 (105-136)
126 (112.8-138)
75 (68-83)
95 (93-98)
101 (86-113)
|
0.05
<0.0001
0.29
0.22
0.005
0.001
0.003
0.021
0.01
0.34
0.67
1.00
0.17
0.0008
0.26
0.26
0.006
0.76
0.15
0.37
0.45
|
Note – Continuous variables are presented as means and standard deviations for normally distributed data or as medians and interquartile ranges for nonparametric data. Differences in distributions of characteristics of those with and those without pulmonary embolism (PE) were analyzed using Student t test or Mann-Whitney U test. Bold indicates statistical significance (p<0.05). CTA=computed tomography pulmonary angiography; LDH=lactate dehydrogenase; CRP=c-reactive protein; WBC=white blood cell; BNP=B-type natriuretic peptide; IL-6=interleukin-6; PESI= pulmonary embolism severity index; SBP=systolic blood pressure; DBP=diastolic blood pressure.
Predictors of PE
The PE predictor model for patients hospitalized with COVID-19 was developed by initially performing univariate logistic regression analysis on all sixty-nine variables individually, which identified variables that were significantly associated with PE. Nine variables were selected to be included in the final model and are shown in Table 3. Multivariable logistic regression analysis was used to build the model to predict PE, and stepwise selection was used for variable selection based on AIC. Out of these, two statistically significant variables were identified: time from symptom onset to admission (OR=1.11, 95% CI 1.03-1.20, p=0.008), and PESI score at the time of CTA (OR= 1.02, 95% CI 1.01-1.04 (p= 0.008). In contrast, hypertension (OR=0.34, 95% CI 0.13-0.85, p=0.02) and COPD/asthma (OR=0.22, 95% CI 0.06-0.68, p=0.01) significantly predicted the absence of a PE.
Table 3. Regression analysis of predictors of PE in patients hospitalized with COVID-19 with and without a PE
Variable
|
Odds ratio [95% CI]
|
P-value
|
Peak D-dimer
COPD/bronchial asthma
Time from symptom onset to admission (days)
PESI score at time of CTA
Hypertension
Admission total bilirubin
Admission platelets
Admission troponin
Admission AST
|
1.06 (0.998-1.119)
0.22 (0.06-0.68)
1.11 (1.03-1.20)
1.02 (1.01-1.04)
0.34 (0.13-0.85)
2.27 (0.97-5.85)
1.003 (0.999-1.008)
3.67 (0.91-62.72)
0.99 (0.98-1.00)
|
0.06
0.01
0.008
0.008
0.02
0.07
0.1
0.24
0.17
|
Note – Multivariable logistic regression analysis was used to build the model to predict pulmonary embolism and stepwise selection was used for variable selection. Bold indicates statistical significance (p<0.05). COPD=chronic obstructive pulmonary disease; PESI= pulmonary embolism severity index; CTA=computed tomography pulmonary angiography; AST=aspartate transaminase.
Predictors of mortality in patients with PE
Eighty-one percent of patients were discharged, and overall mortality was 18.9%, including 22.8% (n=21) in CTA- and 13.6% in CTA+ (n=9). CTA- was sicker than CTA+. This was demonstrated by 12/21(57.1%) of CTA- requiring ICU admission vs 4/9 (44.4%) of CTA+, of whom 9/21 (42.9%) of CTA- were intubated vs 3/9 (33.3%) of CTA+, and 10/21 (47.6%) of CTA- required pressors vs 1/9 (11.1%) of CTA+ . Finally, 3/21 (14.2%) of CTA- underwent renal replacement therapies vs 0 of CTA+. In addition, 8/21 (38.1%) of CTA- had a diagnosis of active malignancy vs 1/9 (11.1%) of CTA+
Variables for predictors of mortality in COVID-19 patients with PE were identified performing univariate Cox proportional hazard regression on all sixty-nine variables individually to select the variables that were significantly associated with mortality. Age, chronic home anticoagulation (AC), time from symptom onset to CTA, admission ferritin, C-reactive protein (CRP), blood urea nitrogen (BUN) were selected to be included in the multivariable Cox proportional hazard model and stepwise selection was used for variable section based on AIC, as shown in Table 4. Three statistically significant variables were identified: age (HR 1.13, 95% CI 1.04-1.22, p=0.006), chronic home AC (13.81, 95% CI 1.24-154, p=0.03), and admission ferritin (1.001, 95% CI 1-1.001, p=0.01).
Table 4. Predictors of mortality among COVID-19 patients (univariate regression)
Variable
|
Hazard ratio [95% CI]
|
P-value
|
Age
Chronic home AC
Time from symptom onset to CTA (days)
Admission ferritin
Admission CRP
Admission BUN
|
1.13 (1.04, 1.22)
13.81 (1.24, 154)
0.94 (0.88, 1.01)
1.001 (1, 1.001)
1.01 (0.99, 1.01)
1.03 (0.97, 107)
|
0.006
0.03
0.07
0.01
0.07
0.07
|
Note – Cox proportional hazard model was implemented to predict mortality in COVID-19 patients with PE and stepwise selection was used for variable selection. Bold indicates statistical significance (p<0.05). AC=anticoagulation; CTA=computed tomography pulmonary angiography; CRP=c-reactive protein; BUN=blood urea nitrogen.
Diagnostic performance of D-dimer for PE in patients with COVID-19
With respect to the admission D-dimer and the traditional cut-off level of 0.5, the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 98.46%, 11.11%, 44.44% and 90.91%, respectively. For a cut-off level of 1.0, they were 90.77%, 30.0%, 48.36% and 81.82%, respectively.
With respect to the Peak D-dimer cut-off level of 0.5, the sensitivity, specificity, PPV, and NPV were 98.46%, 4.44%, 42.67%, 80%. For a D-dimer cut-off of 1.0, they were 96.92%, 15.56%, 45.32% and 87.5%, respectively. These are summarized in Table 5 and 6.
The summary receiver operating characteristic (ROC) curves yielded an area of 0.729 (Figure 1) for admission D-dimer, and 0.662 for peak D-dimer (Figure 2), suggesting cut-offs of D-dimer levels for PE diagnosis from 0 to maximum admission D-dimer in small incremental steps for PE diagnosis. For each single cut-off, sensitivity, specificity, PPV, NPV were calculated. Cut-off values that provided the best trade-off between sensitivity and specificity were selected.
For instance, using an admission D-dimer cut-off of 1.710, the sensitivity was 83.1% and specificity was 53.5%, whereas for an admission cut-off value of 4.370, the sensitivity of D-dimer tests for PE decreased (56.9%), but the specificity increased (84.4%). When admission D-dimer cut-off was raised, there was a reduction in sensitivity but an increase in specificity and PPV, as demonstrated in Table 5 and Figure 1.
Using a peak D-dimer cut-off of 3.44 (Figure 2), the sensitivity was 64.6%, specificity was 46.7%, and PPV was 46.7%. Raised cut-off value to 11.1 was associated with decreased sensitivity (49.2%), but an increased specificity (78.9%) and PPV (61.5%), as demonstrated in Table 6 and Figure 2.
Table 5. Diagnostic performance of different cut-offs for admission D-dimer values for detecting pulmonary embolism in patients with COVID-19 (n=155)
D-dimer cut-off (ng/ml)
|
Sensitivity
|
Specificity
|
PPV
|
NPV
|
0.50
|
98.46%
|
11.11%
|
44.44%
|
90.91%
|
1.00
|
90.77%
|
30%
|
48.36%
|
81.82%
|
1.25
|
86.15%
|
38.89%
|
50.45%
|
79.55%
|
2.50
|
61.54%
|
65.56%
|
56.34%
|
70.24%
|
5.00
|
50.77%
|
84.44%
|
70.21%
|
70.37%
|
7.50
|
49.23%
|
86.67%
|
72.73%
|
70.27%
|
10.00
|
46.15%
|
88.89%
|
75%
|
69.57%
|
15.00
|
35.39%
|
92.22%
|
76.67%
|
66.40%
|
20.00
|
30.77%
|
93.33%
|
76.92%
|
65.11%
|
PPV=positive predictive value; NPV=Negative predictive value.
Table 6. Diagnostic performance of different cut-offs for peak D-dimer values for detecting pulmonary embolism in patients with COVID-19 (n=155)
D-dimer cut-off (ng/ml)
|
Sensitivity
|
Specificity
|
PPV
|
NPV
|
0.50
|
98.46%
|
4.44%
|
42.67%
|
80%
|
1.00
|
96.92%
|
15.56%
|
45.32%
|
87.5%
|
1.25
|
93.85%
|
18.89%
|
45.52%
|
80.95%
|
2.50
|
76.92%
|
42.22%
|
49.02%
|
71.69%
|
5.00
|
56.92%
|
56.67%
|
48.68%
|
64.56%
|
7.50
|
55.39%
|
68.89%
|
56.25%
|
68.13%
|
10.00
|
49.23%
|
76.67%
|
60.38%
|
67.65%
|
15.00
|
44.62%
|
83.33%
|
65.91%
|
67.57%
|
20.00
|
35.39%
|
86.67%
|
65.71%
|
65%
|
PPV=positive predictive value; NPV=Negative predictive value.
Prophylactic Anticoagulation
Out of 92 PE negative patients, 34 patients were on prophylactic low molecular weight heparin (LMWH) (37%), 4 patients were on therapeutic LMWH (4.3%), 19 patients were on unfractionated heparin (UFH) (20.7%), 8 were on DOAC (8.7%), 27 patients were not on any anticoagulant (29.3%) prior to CTA. In contrast, of 66 PE positive patients, 14 patients were on prophylactic LMWH (21.2%), 6 patients were on therapeutic LMWH (9.1%), 12 patients were on UFH (18.2%), 4 were on DOAC (6.1%), 30 patients were not on any anticoagulant (45.5%) prior to CTA.