Introduction
Endothelial dysfunction and thrombotic conditions play a paramount role in COVID-19 pathogenesis. Therefore, anticoagulant therapy is recommended as an essential therapy component [1-6]. Hemorrhagic complications of anticoagulant therapy are reported in 7.8–21.4% of patients with COVID-19 [8, 9]. By year 3 of the pandemic, there were dozens of reports of a specific hemorrhagic complication — massive spontaneous bleeding into the abdominal and chest walls [10-13]. The blood loss can reach several liters; it is associated with hemorrhagic shock, which aggravates the severity of the underlying disease and increases the risk of an adverse outcome [14, 15].
Contrast-enhanced multislice computed tomography (MSCT) is the primary method of diagnosing bleeding in the abdominal wall and retroperitoneal tissues. Its accuracy for diagnosing ongoing bleeding is up to 87%, which makes this method a priority for determining surgical strategy [16]. The main CT sign of ongoing bleeding is the contrast agent extravasation; its correlation with the scan phase allows us to indirectly identify the source of bleeding [17].
Surgical and non-surgical treatment methods are used to control ongoing bleeding. In ongoing arterial bleeding, endovascular hemostasis has been suggested as the method of choice, but the number of observations and reports on the treatment outcomes is scarce [10-12].
The objective of the study was to review the results of curative and preventive endovascular hemostasis in spontaneous soft tissue bleeding in the abdominal and chest wall and retroperitoneal space in COVID-19 patients.
Materials and methods
Data of 232 patients with confirmed soft tissue hemorrhages in the chest and abdominal wall and retroperitoneal space, treated at Moscow Multidisciplinary Clinical Center "Kommunarka" from 01.03.20 to 01.03.22, were reviewed for compliance with inclusion criteria.
Inclusion criteria: patients of both sexes with verified COVID-19 infection (PCR test) and signs of ongoing bleeding according to MSCT who underwent contrast-enhanced angiography.
Thirty-five (15.1%) patients met these criteria and were included in a retrospective cohort study.
Parameters studied:
1) medical history;
2) laboratory tests;
3) patients' assessment according to NEWS-2, APACHE-2 scales;
4) results of visualization studies (ultrasound, MSCT, contrast-enhanced angiography).
The severity of infectious disease was determined according to the interim guidelines for preventing, diagnosing, and treating the new coronavirus infection (COVID-19) [7].
The NEWS2 scale was used for a comparative assessment of patient status. Critically patients (n=24) were assessed using APACHE-2.
A standard complex of laboratory (complete blood count, blood chemistry, acid-base status, coagulation test) and visualization (ultrasound and MSCT) tests were performed in all patients. The four-phase MSCT included a native and bolus contrast study with an arterial, venous, and delayed phases assessment.
Angiography was performed with local or general anesthesia by femoral (ipsilateral or contralateral) (n=23), radial (n=5), or brachial (n=7) accesses. Guiding sheaths of 5 or 6 Fr and JR, IM, or Cobra 1-3 catheters (Boston Scientific, Cordis, USA) were used.
According to the angiography results, the patients were divided into three clinical groups by the type of bleeding and the treatment used:
- group 1 (n=12) — bleeding was detected by angiography; endovascular hemostasis was performed;
- group 2 (n=13) — no bleeding was detected at angiography; preventive hemostasis was performed;
- group 3 (n=10) — no bleeding was detected; hemostasis was not performed.
The groups differed in the localization of hematomas. Groups 1 and 2 included patients with predominantly abdominal wall hematomas. Group 3 also included patients with retroperitoneal and chest wall hematomas.
Complications were assessed at discharge. In-hospital mortality in the groups was analyzed. Analysis of mortality factors was performed based on the comparison of deceased and surviving patients.
Statistical processing of the obtained data was performed using Statistica 12 software (StatSoft, Inc.). Data were presented as a mean and standard deviation for continuous normally distributed variables, as a median and interquartile range for continuous data without normal distribution, and as absolute values and percentages for categorical data. Normality analysis was performed using the Shapiro-Wilk test. Statistical comparison of quantitative variables between the three independent patient groups was performed using the nonparametric Kruskal-Wallis test. If statistically significant differences between groups were found using the Kruskal-Wallis test (p<0.05), a posteriori comparisons were performed using the Mann-Whitney test with multiple comparison adjustments and a lower critical significance level. Therefore differences between the groups can be considered statistically significant only at p<0.017. Categorical data and proportions were compared using Pearson's χ2 test and Fisher's two-sided exact test. Mann-Whitney U-test was used to compare continuous values between groups of survivors and deceased.
Results
Medical history
The mean age of the patients was 71.1±10.3 years, and most were females — 31 (88.6%).
Bleeding onset time was 15.4±4.5 days from the onset of COVID-19 symptoms. All patients developed hemorrhage in the hospital; 19 (54.4%) were transferred from other medical institutions.
All patients received therapeutic doses of anticoagulants at the time of bleeding: low molecular weight heparin — 30 (85.7%), unfractionated heparin — 1 (2.9%), new oral anticoagulants or combination of drugs — 4 (11.4%).
All patients received glucocorticoids. Biological agents were used in 15 (42.8%) patients: levilimab (8 patients), olokizumab (6 patients), and tocilizumab (1 patient).
Examinations
In 28 (80%) patients, the hematoma was located in the anterior abdominal wall (sheath of rectus abdominis muscle), 4 (11.4%) in the anterior chest wall (over the greater pectoral muscle), 3 (8.6%) in the retroperitoneal space.
Patient condition severity ranged from 2 to 12 points (mean 5.1±2.5 points), according to NEWS-2. The mean APACHE-2 score in ICU patients was 22±8.7.
Clinical characteristics of bleeding
Before the operation, hypotension (systolic blood pressure of 90 mmHg or less) was revealed in 20 (57.1%) patients; 16 required vasopressor support. Mechanical ventilation was performed in 3 (8.6%) patients. The mean hemoglobin at the time of hemorrhage was 8.18±2.57 g/dL. Hemoglobin level below 7.0 g/dL was detected in 12 (34.3%) patients.
Visualization of ongoing bleeding
The hematoma volume, according to ultrasound, was 646.6±527.4 mL. Sonographic signs of ongoing bleeding were detected in 5 (14.3%) patients.
The hematoma volume, according to MSCT, was 1193.4±706.1 mL. Contrast agent extravasation during MSCT was found in all patients, including 12 (34.3%) in all scan phases, in arterial and venous phases in 3 (8.6%), in venous and delayed phase in 1 (2.8%), only in delayed phase in 3 (8.6%). The extravasation was revealed in another 15 (42.8%) patients without specifying the scan phase.
Continued bleeding detected by ultrasound (n=5) was confirmed by MSCT: in 1 patient in all scan phases; in another 4 patients, the extravasation phase was not specified; contrast agent extravasation was detected during angiography in 2 patients.
Surgery and angiography results
The average time from the diagnosis of bleeding to surgery was 7±5.3 h; the delay is due to the transfer of some patients with bleeding from other medical institutions. Thirty-two (91.4%) patients had surgery with local anesthesia and 3 (8.6%) with general anesthesia.
Selective angiography confirmed contrast agent extravasation in 12 (34.3%) patients (group 1). When comparing with MSCT data, only 8 (66.7%) of them had contrast agent extravasation in all scan phases, 2 (16.7%) in the delayed phase, for another 2 (16.7%) the scan phase was unknown (not specified in the MSCT protocol).
In all cases, the source of bleeding, according to angiography, was the inferior epigastric artery. It was embolized with microspheres in 8 patients and microspirals in 4 patients.
Direct angiography showed no signs of contrast agent extravasation in the remaining 23 (65.7%) patients. Preventive embolization was performed in 13 (37.1%) patients (group 2). In the remaining 10 (28.6%) patients, embolization was not performed (group 3). The results of the examination and treatment of the clinical groups are presented in Table 1.
Table 1. Characteristics of clinical groups
Parameter |
Group 1 (n=12) |
Group 2 (n=13) |
Group 3 (n=10) |
p |
Age, years (M±SD) |
71.4±11.5 |
69.7±8.3 |
72.5±10.7 |
0.6477 |
Sex (F/M) |
10/2 |
12/1 |
9/1 |
0.77 |
Anticoagulant therapy |
||||
UFH |
1 |
0 |
0 |
>0.05 |
LMWH |
10 |
12 |
9 |
>0.05 |
NOAC |
1 |
0 |
0 |
>0.05 |
combination |
0 |
1 |
1 |
>0.05 |
Days from the illness onset to the hematoma appearance |
15.1±5.6 |
14.1±3.4 |
17.4±3.5 |
0.1371 |
Days from anticoagulant therapy initiation to the hematoma appearance |
9.1±5.9 |
7.9±3.1 |
12.3±5.1 |
0.1388 |
Transfer from another hospital (n=19) |
7 |
9 |
3 |
0.164 |
Surgery and anesthesia risk (ASA) |
3.7±0.7 |
3.7±0.6 |
4.1±0.7 |
0.3245 |
NEWS-2, points |
||||
On admission |
5.6±3.2 |
4.5±1.8 |
5.1±2.3 |
0.7176 |
Before the procedure |
4.75±3.1 |
4.4±1.5 |
6.6±2.7 |
0.1575 |
COVID-19 severity |
||||
Moderate |
6 |
8 |
0 |
0.037 |
Severe |
5 |
3 |
7 |
0.037 |
Very severe |
1 |
2 |
3 |
0.037 |
Localization of the hematoma |
||||
Anterior abdominal wall |
11 |
13 |
4 |
<0.004 |
Retroperitoneal space |
1 |
0 |
2 |
<0.004 |
Chest wall |
0 |
0 |
4 |
<0.004 |
Hematoma volume, mL (according to ultrasound) |
795.1±628.9 |
603.5±383.8 |
646.4±621.3 |
0.5314 |
Hematoma volume, mL (according to CT data) |
1370±716.9 |
1332.1±804.1 |
993±504.2 |
0.3686 |
APACHE-2 (for ICU patients), points (n — number of patients) |
n=8 21±10.1 |
n=6 24.8±9.4 |
n=8 20.9±5.9 |
0.9037 |
INR |
||||
On admission |
1.19±0.19 |
1.06±0.13 |
1.12±0.1 |
0.1182 |
Before the appearance of the hematoma |
1.15±0.19 |
1.05±0.1 |
1.21±0.13 |
0.0281 |
aPTT |
||||
On admission |
32.2±6.2 |
33.1±5.3 |
33.6±3.1 |
0.7145 |
Before the appearance of the hematoma |
61.7±72.1 |
29.8±5.5 |
44.6±32.9 |
0.4182 |
Platelets, 109/L |
||||
On admission |
259.4±104.3 |
192.3±73.7 |
165.9±69.5 |
0.0388 |
Before the appearance of the hematoma |
343±145.2 |
257.5±79.1 |
218±80.6 |
0.0919 |
Hemoglobin before angiography, g/dL |
77.7±21.8 |
90.9±27.1 |
73.3±23.5 |
0.206 |
Vasopressor support before surgery (n=16) |
7 |
3 |
6 |
0.118 |
Time from diagnosis to the procedure, min |
260±218.9 |
394.6±244.9 |
622.5±399.4 |
0.0666 |
Manipulation duration, min |
87.1±76.7 |
57.7±21.5 |
30±16.9 |
0.0045 |
Re-operation |
0 |
1 (7.7%)* |
1 (10%)* |
0.56 |
Arterial dissection |
2 (16.7%) |
0 |
0 |
0.131 |
Note. Here and in Table 3: UFH — unfractionated heparin; LMWH — low molecular weight heparin; NOAC — new oral anticoagulants; * — in open surgery performed because of suspected recurrence, there was no ongoing bleeding.
When analyzing the groups, differences in the COVID severity were significant: the most severe disease was in group 3. However, no differences were found between the groups when prognostic score points, anemia severity, and vasopressor support were assessed. No severe coagulation abnormalities were recorded in all patients except for platelet levels at hospital admission. The INR value at hematoma detection differed among the three groups (p=0.0281). In a pairwise comparison using the Mann-Whitney test, significant differences were found only between groups 2 and 3 (p=0.005).
Surgery and intraoperative complications
The shortest duration of surgery was in group 3. Although the surgery extent in groups 1 and 2 was similar, the duration of surgery was longer in group 1, probably due to 2 (16.7%) intraoperative complications. Both patients had dissection of the common femoral artery (CFA) during the access, which required temporary balloon occlusion of the CFA in one of them (the patient was subsequently discharged from the hospital), and stent implantation in the CFA in the second patient (died 63 hours after surgery). No postoperative complications were recorded in both patients.
Postoperative care
After surgery, 23 (65.7%) patients were admitted to the ICU. The rest were treated in surgical departments.
Blood transfusion was performed in 26 (74.3%) patients (group 1 — 10 (83.3%), group 2 — 9 (69.2%), group 3 — 8 (80%)), plasma transfusion — in 17 (48.6%) patients (group 1 — 58.3%, group 2 — 5 (38.5%), group 3 — 5 (50%)).
Due to suspected recurrence of bleeding according to ultrasound, 2 (5.7%) unstable patients underwent re-operation — open hematoma evacuation, ligation of the inferior epigastric artery. However, no ongoing bleeding was revealed during the surgery in both cases.
Anticoagulant therapy was resumed 2.1±2.8 days after the manipulation with the prophylactic dose of the low molecular weight heparin and the therapeutic one — on the next day. No recurrence of bleeding was observed in any case afterward.
Fourteen (40%) patients died at 139.6±187.5 h after the surgery (median 62 [31; 143]) (Table 2).
Table 2. Analysis of mortality in all groups
Mortality factor |
Group 1 (n=12) |
Group 2 (n=13) |
Group 3 (n=10) |
p |
Mortality in groups |
4 |
3 |
7 |
0.064 |
Depending on hematoma localization (n=14) |
||||
Anterior abdominal wall |
4 |
3 |
3 |
0.232 |
Retroperitoneal space |
0 |
0 |
1 |
0.232 |
Chest wall |
0 |
0 |
3 |
0.232 |
Depending on the COVID-19 severity (n=14) |
||||
Moderate |
1 |
1 |
0 |
0.35 |
Severe |
2 |
0 |
4 |
0.35 |
Very severe |
1 |
2 |
3 |
0.35 |
Time from surgery to death, h |
110±69.3 |
265±291.1 |
102.7±150.5 |
0.6303 |
Cause of death (n=14) |
||||
Hemorrhagic shock |
0 |
1 |
1 |
0.718 |
Pneumonia |
3 |
2 |
5 |
0.718 |
Posthemorrhagic anemia |
1 |
0 |
1 |
0.718 |
No statistically significant differences between the groups were found during the mortality analysis. The leading cause of mortality was progressive respiratory failure due to pneumonia, confirmed at autopsy in 10 (71.4%) patients.
We compared deceased and surviving patients to assess mortality factors according to previously used criteria (Table 3).
Table 3. Clinical data of deceased and discharged patients
Criterion |
Deceased (n=14) |
Survived to discharge (n=21) |
p |
Age, years (M±SD) |
73.4±11.1 |
69.6±9.4 |
0.263 |
Sex (F/M) |
12/2 |
19/2 |
1.000 |
Anticoagulant therapy |
|||
UFH |
1 |
0 |
0.4 |
LMWH |
12 |
19 |
1.000 |
NOAC |
0 |
1 |
1.000 |
combination |
1 |
1 |
1.000 |
Days from the illness onset to the hematoma appearance |
14.8±4.4 |
15.8±4.4 |
0.752 |
Days from anticoagulant therapy initiation to the hematoma appearance |
9.3±4.7 |
9.7±5.4 |
0.987 |
Transfer from another hospital (n=19) |
6 |
13 |
0.317 |
Surgery and anesthesia risk (ASA), points |
4.1±0.7 |
3.6±0.6 |
0.048 |
NEWS-2, points |
|||
On admission |
5.9±2.6 |
4.7±2.3 |
0.095 |
Before the procedure |
6.6±2.8 |
4±2.1 |
0.011 |
COVID-19 severity (number of patients) |
|||
Moderate |
2 |
12 |
0.016 |
Severe |
6 |
9 |
1.000 |
Very severe |
6 |
0 |
0.002 |
Localization of hematoma (number of patients) |
|||
Anterior abdominal wall |
11 |
17 |
1.000 |
Retroperitoneal space |
1 |
2 |
1.000 |
Chest wall |
2 |
2 |
1.000 |
Hematoma volume, mL (according to ultrasound) |
757.1±546 |
542.2±502.4 |
0.135 |
Hematoma volume, mL (according to CT data) |
1374.6±806.7 |
1072.6±600.6 |
0.309 |
APACHE-2 (n — ICU patients), scores n1±n2 |
(n=10) 25.4±6.1 |
(n=12) 19.2±9.6 |
0.059 |
INR |
|||
On admission |
1.16±0.13 |
1.1±0.17 |
0.052 |
Before the appearance of the hematoma |
1.23±0.16 |
1.07±0.13 |
0.001 |
aPTT |
|||
On admission |
34.25±3.3 |
60.52±66.4 |
0.066 |
Before the appearance of the hematoma |
31.9±6.1 |
35.24±27.8 |
0.007 |
Platelets, 109/L |
|||
On admission |
178.3±65.1 |
243.6±125.2 |
0.154 |
Before the appearance of the hematoma |
227.6±75.1 |
307.2±129.7 |
0.118 |
Hemoglobin before angiography, g/dL |
73.7±24.6 |
86.4±24.8 |
0.066 |
Vasopressor support before surgery (n=16) |
11 |
5 |
0.002 |
Time from diagnosis to the procedure, min |
429.3±351.7 |
413.6±296.4 |
0.803 |
Manipulation duration, min |
51.8±41.9 |
54.8±21.3 |
0.249 |
Open surgery for recurrent bleeding |
2 |
0 |
0.153 |
The analysis showed more patients with extremely severe infection, high ASA scores, and those with vasopressor support and more severe INR abnormalities among the deceased. No significant differences using the APACHE-2 score were found, although a trend towards greater severity in the deceased was observed (p=0.059). Both patients with repeat surgeries died.
Factors directly related to bleeding (anticoagulant therapy, localization, and volume of hematoma, procedure duration, intraoperative complications) did not differ in survivors and deceased.
Discussion
Mortality in patients with spontaneous soft tissue bleeding associated with COVID-19 is up to 70% [18], which is significantly higher than in the "pre-COVID era," when the rate did not exceed 10% [19-21].
Bleeding into the abdominal wall is more common in females of elderly and senile age [22, 23]. Risk factors for spontaneous bleeding reported in the literature also include anticoagulant therapy, steroid use, severe cough, anterior abdominal wall injections, female gender, and advanced age [18, 24-27]; all these factors were observed in most of our patients.
The diagnosis of ongoing bleeding is a critical strategic issue. Clinical findings (hypotension requiring vasopressors, decreased hemoglobin, mechanical ventilation) did not differ between the groups and were not relevant for diagnosing ongoing bleeding. Fifteen (42.8%) patients were hemodynamically stable despite contrast agent extravasation during MSCT.
A recent study has established the low sensitivity of ultrasound in the diagnosis of ongoing bleeding [28]. Our findings were consistent with these observations. Due to the overdiagnosis of recurrent bleeding, ultrasound-guided recurrent surgeries were performed on two unstable patients.
The CT scan has the highest diagnostic accuracy for ongoing bleeding to determine the localization and volume of the hematoma. The main CT sign of ongoing bleeding is the extravasation of the contrast agent [29]. On angiography, ongoing bleeding was found in only 12 (34.3%) patients. The group of Italian researchers showed that CT signs of ongoing bleeding include only extravasation of contrast agent in the arterial scan phase and subsequent enhanced accumulation in the images of the venous and late (delayed) phases [16]; these signs should be considered an indication for emergency surgery. Fifteen patients met these criteria, but the results of MSCT and angiography were consistent only in 8 (67%) of them. However, during angiography, ongoing bleeding was found in two patients with extravasation in the delayed phase of MSCT only. Thus, we did not find an absolute association between extravasation on MSCT and bleeding on angiography, especially if extravasation was in the late phases of MSCT. However, the absence of contrast agent extravasation on direct angiography does not exclude ongoing bleeding previously identified on MSCT. According to A. Dohan et al., MSCT can have greater diagnostic accuracy than direct angiography due to several limitations, such as target artery spasm, thrombosis, and hypotension [17].
The role of endovascular embolization as a treatment for spontaneous soft tissue bleeding is unclear. Proponents of the method recommend its use in ongoing bleeding [19]. The benefits of endovascular embolization have not been shown in a recent study [18]. According to our data, X-ray endovascular embolization effectively controlled ongoing arterial bleeding, but the effect of the technique on patient survival is unknown. Despite the technical success of embolization in 100% of patients, mortality remained very high in group 1. The advantages of preventive embolization performed in group 2 when compared with the results of the treatment without embolization (group 3) were also not established, especially considering the disease severity of group 3 patients. Perhaps the reduction of arterial blood flow in the target artery circulation, the most likely source of the hematoma, may help to reduce ongoing non-arterial bleeding. Despite a clear trend to reduced mortality in the preventive embolization group, the difference with patients without embolization did not reach statistical significance.
When comparing the characteristics of deceased and surviving patients, the severity of the condition due to the underlying disease was likely to be more important as an adverse outcome factor than factors related to bleeding, its source, and blood loss.
An important interim result of the study can be considered our success in resuming anticoagulant therapy two days after stopping bleeding, which was not accompanied by bleeding recurrence and is relevant for patients with severe COVID-19.
From the practical aspects, we would like to note our preference for an embolization agent in favor of embospheres due to the ease of their delivery to the distal arterial circulation. Spiral insertion is often difficult because of spasm or small arterial caliber (10).
Our study had a number of limitations. First, the results are based on radiological, clinical, and laboratory retrospective data from a single center. The clinical groups differed in the localization of hematomas and the COVID-19 severity. In addition, some logistical issues in treating these patients have been identified. First of all, the long transportation of patients to the specialized hospital could affect the treatment outcomes.
Thus, MSCT is the main method of diagnosing spontaneous soft tissue hemorrhages of the abdominal and chest wall. In contrast agent extravasation on MSCT, the arterial bleeding on angiography was confirmed in 33.4% of patients. Contrast agent extravasation in all phases of MSCT corresponded to ongoing bleeding on angiography in 67% of patients. Contrast agent extravasation only in the late phases of MSCT did not completely rule out arterial bleeding on angiography. Embolization has been effective as a method of stopping ongoing arterial bleeding; however, more research is needed to evaluate the effect of the technique on patient survival compared with non-surgical treatment. Endovascular embolization was associated with potentially life-threatening complications.
The authors declare no conflict of interest.