Thrombosis in COVID-19: clinical outcomes, biochemical and pathological changes, and treatments

June 4, 2020

Robin E Ferner*†, Marcel Levi, Reecha Sofat, Jeffrey K Aronson

On behalf of the Oxford COVID-19 Evidence Service Team
Centre for Evidence-Based Medicine, Nuffield Department of Primary Care Health Sciences
University of Oxford

*University of Birmingham
University College London

Correspondence to

Reports that COVID-19 is associated with venous and arterial thrombosis and pulmonary embolism and increased rates of thrombosis in cannulae and extracorporeal circuits for renal replacement or membrane oxygenation, have drawn attention to the coagulant effects of the disease.

Clinical evidence of thrombosis in patients with COVID-19
Case reports and small case series of thromboembolic complications in patients with COVID-19 have included pulmonary embolism [1 2 3], femoral vein thrombosis [1], phlegmasia caerulea dolens [4], cerebral venous sinus thrombosis [5 6], aortic thrombosis [7 8], aorto-iliac thrombosis [9], humeral artery thrombosis [9], acro-ischaemic presentations [10], and strokes in young patients [11]. Thrombotic events can occur during COVID-19; they may be a presenting feature [3 11]; and they may occur during convalescence [12].

Thromboembolism has been common in larger case series (Table 1). While many patients with thrombosis have been in intensive care units (ICUs), less seriously ill patients are also at significant risk of thromboembolic complications.

Table 1. Case series reporting thromboembolic complications in patients with COVID-19

Number of patients Presentation of thrombotic events (number) Reference
      81 Lower limb venous thrombosis (20); 8 died [13]
    184 Pulmonary emboli (25), 18 involving at least segmental pulmonary arteries; venous thromboses (3); ischaemic strokes (3); all had received ‘at least standard doses’ of prophylactic nadroparin. [14]
    150 Pulmonary emboli (25); deep venous thromboses (3); limb ischaemia (1); mesenteric ischaemia (1); 70% had received prophylactic treatment with low molecular weight heparin or unfractionated heparin, and 30% had received therapeutic treatment; compared with 233 matched patients in the intensive care unit with acute respiratory distress syndrome not due to COVID-19, the odds ratio for pulmonary embolism was 9.3 (95% CI 2.2–40) and for any thromboembolic complication 2.7 (95% CI 1.1–6.6) [15]
    107 Pulmonary emboli (22, of whom 20 had received prophylactic heparin and two were taking therapeutic anticoagulants for other reasons); the rate of pulmonary embolism in 196 patients admitted to the intensive care unit the previous year had been 6.1%, and in 40 patients with influenza pneumonia 7.5%. [16]
    106 Thirty-two scans (30%) of 106 patients with COVID-19 who had CT pulmonary angiograms in Strasbourg, France, demonstrated pulmonary embolism. Those with positive scans had a higher median plasma D-dimer concentration than those with negative scans (but the precise values are unclear from the published pre-print). [17]
      75 A more recent study from a single centre in Amsterdam observed the cumulative incidence of symptomatic deep venous thrombosis in 75 patients positive for SARS-CoV-2, admitted to ICU, and receiving routine thromboprophylaxis. The incidence increased from 10% (95% CI, 5.8–16) at seven days to 21% (95% CI, 14–30) at 14 days and 25% (95% CI 16–36) at 21 days. [18]
      25 In a cohort of 25 patients in Switzerland, ultrasound screening was positive for lower limb deep venous thrombosis at 5–10 days after admission to intensive care in 8 (32%), of whom 6 had symptomatic deep vein thrombosis. Most of the symptomatic cases were evident by 3 days from admission [19]
    100 280 patients were admitted to hospital with COVID-19 in Besançon, and 100 underwent CT pulmonary angiography. The incidence of pulmonary emboli was 17/23 (74%) among those admitted to ICU, and 22/77 (29%) among those not admitted to ICU. [20]
    135 In a retrospective study of 135 patients who underwent CT pulmonary angiography in Paris, 32/135 (24%) had radiological evidence of pulmonary embolism. Pulmonary embolism was more likely in those from ICU (50% of scans) than in others (18%). [21]
      30 Compression ultrasound scans were positive in 16 of 30 non-ICU patients with COVID-19. [22]


Among 24 patients with COVID-19 and deep venous thrombosis or pulmonary embolism only two (8.3%) had weakly positive anticardiolipin IgM and anti–β2-glycoprotein I IgM, but anticardiolipin IgG and anti–β2-glycoprotein I IgG were negative in all 24 [23].

Acral ischaemia
A 13-year-old boy developed red-violet lesions, described as ischaemic, on the toes of both feet. He became febrile and unwell and then tested positive for SARS-CoV-2. The lesions became darker, some blistered, and they regressed over a few days [24]. Similar acral lesions occurred in 22 children in Madrid, all of whom had normal coagulation tests and negative tests for lupus anticoagulant [25]; six had skin biopsies, which showed a dense lymphocytic vascular reaction, with small vessel endothelial swelling, and sometimes focal thrombosis; only one had a positive test for SARS-CoV-2 at the time the lesions were apparent.

A Chinese report of 7 critically ill patients with COVID-19, available in English abstract, described acro-ischaemic presentations, “including finger/toe cyanosis, skin bulla[e] and dry gangrene” [26]. A further report described a patient with COVID-19 who had “evidence of ischemia in the lower limbs bilaterally as well as in digits two and three of the left hand”, multiple cerebral infarcts, and positive antiphospholipid antibodies [27]; however, antiphospholipid antibodies can be transiently positive around times of acute infections and acute thrombosis, but may be of no pathophysiological relevance [28 29].

In contrast, there is little clinical evidence that patients with COVID-19 are at increased risk of spontaneous bleeding, although two patients with COVID-19 pneumonia receiving continuous positive airways pressure (CPAP) ventilation bled from epigastric arteries [30]. One had a femoral vein thrombosis and was already being treated with low molecular weight heparin; the other received prophylactic doses of low molecular weight heparin. In one intensive care unit study, only 4/150 patients (2.7%) developed haemorrhagic complications [15].

Relevant laboratory findings
D dimers
Fibrin degradation results in an increase in circulating D-dimers and is an expected feature of disseminated intravascular coagulation (DIC, consumptive coagulopathy) but also a marker of infection. In patients with COVID-19, plasma D-dimer concentration correlated with in-hospital mortality [31 32 33 34]. In a French series of 150 patients, the median D-dimer concentration was 2.27 mg/L (interquartile range 1.16–20; reference value < 0.5). In 30 Italian patients, D-dimers were uniformly increased, with a mean of 4.88 mg/L [35]. In another French study, 22/107 patients developed pulmonary emboli, and the hazard ratio for pulmonary embolus increased with increasing D-dimer concentration [16].

Fibrinogen concentrations were increased to a median of 6.99 g/L (interquartile range 6.08–7.73, reference range 2–4) in 150 French patients [15] and to 4.7 g /L (4.4–6.6, reference range 1.9–3.5) in 83 Dublin patients [33].

Prothrombin time
Mean prothrombin time was prolonged to 15.2 ± 5.0 seconds (reference range 11.5–14.5) in 449 patients with COVID-19, and multivariable analysis showed that the odds of dying within 28 days correlated with prothrombin time (odds ratio = 1.107; 95% confidence interval 1.008–1.215; P= 0.033) [34]. The median prothrombin time was prolonged to 15.5 seconds (interquartile range 14.4–16.3, reference range 11.5–14.5) in 21 non-survivors in a study of 183 Chinese patients, but normal in the survivors [32]. In the Dublin series of 83 patients, mean prothrombin time on admission was normal in both those who survived and those who died [33]. The mean value was also normal in 81 Chinese patients, whether or not they developed venous thromboembolism [13], and in 30 Italian patients [35].

Activated partial thromboplastin time
Average activated partial thromboplastin time was normal on admission in two Chinese series [13], a Dublin series [33], a French series [15], and two groups of Italian patients [32 36]. It was at the upper limit of normal in another Chinese series of 183 patients and non-significantly increased in 21/183 patients who died [13].

Platelet count
The platelet count is expected to be reduced in patients with significant DIC. However, it was within the reference range on admission in all of a cohort of 150 French ICU patients [15], in 40 Italian patients with COVID-19 [36], and in 83 patients admitted to hospital in Dublin [33]. The mean platelet count was also normal in 449 Chinese patients with COVID-19, although it was significantly lower in 134 of the patients who subsequently died than in the 315 who survived [34]. A meta-analysis of nine studies described lower platelet counts in more severely ill patients, with the lowest counts in those who died [37] .  The results were heterogeneous, and the findings were dominated by those of the largest study, which included over 1000 Chinese patients [38].

All this suggests that the rise in D-dimers in COVID-19 is primarily due to the infection and not a consumptive coagulopathy.

Coagulation factors
In 107 French patients, the hazard ratio for pulmonary embolism depended not only on D-dimer concentration, but also on plasma Factor VIII activity and von Willebrand Factor (vWF) antigen [16].

Analysis of antithrombin, clotting factors V and VII, and von Willebrand factor in the French cohort of 150 patients showed a marked increase in von Willebrand Factor (vWF) antigen (Table 2) [15]. 

Table 2. Coagulation factors in 150 patients with COVID-19 [15] 



All 150 patients

Median [IQR]

Reference range
Antithrombin activity (%)     91 [78; 102] 50–150
Factor V (%) 136 [115; 150] > 70
Factor VIII (%) 341 [258; 416] 60–150
vWF activity (%) 328 [212; 342]  
vWF antigen (%) 455 [350; 521] 50–150
Lupus anticoagulant n (%) 50/57 (87.7)  


Lupus anticoagulant was found in 31/34 British patients with COVID-19 (91%), in whom the aPTT was prolonged and Factor VIII activity was increased [39].

A detailed examination of haemostasis in 11 intensive care patients with COVID-19 in Italy showed low mean antithrombin concentrations, and marginally low protein S free antigen, but very high concentrations of vWF (mean VWF antigen 529 units/dL; range 210–863; reference range 40–165) and vWF ristocetin cofactor activity (mean 387 units/dL; range 195–550; reference range 41–151) [35].

A single Swiss case report described a 72-year-old man with COVID-19 whose condition deteriorated rapidly six days after admission to hospital, and who required intensive care with ventilation and renal replacement therapy. D-dimers increased over three weeks to a maximum of 20.6 mg/L, and at the same time vWF antigen was 555% (reference range 42–136%) and vWF activity 520% (reference range 42–168%), with an increase in Factor VIII activity to 369% (reference range 55–164%) [40].  The authors interpreted the findings to indicate “massive endothelial stimulation and damage with release of vWF from Weibel–Palade bodies.” [Weibel–Palade bodies are the storage granules in vascular endothelial cells that contain vWF and other proteins, including tissue plasminogen activator and P-selectin [41].]

Relevant pathological findings
Detailed pathological and histopathological studies in COVID-19 have been hampered by the potential risk of viral transmission to the pathologist [42]. 

Diffuse pulmonary congestion, haemorrhage, and haemorrhagic necrosis were observed in a man with a clinical diagnosis of COVID-19 who died and whose lungs were examined post-mortem [43]. There was interstitial pulmonary fibrosis with haemorrhagic infarcts and microvascular thrombi, and an inflammatory infiltrate contained lymphocytes, plasma cells, macrophages, and mononuclear cells. Immunohistochemistry showed a wide spectrum of immunological cells, including cells positive for CD3, CD4, CD5, CD8, CD20, CD38, and CD79a. Vascular wall thickening, lumen stenosis, and occlusion were frequent, and there were microthrombi. These offered a potential explanation for the late stage of severe hypoxia and respiratory failure seen in COVID-19.

Biopsy specimens from a man who died 14 days after presenting with COVID-19, but who declined mechanical ventilation, showed diffuse alveolar damage with hyaline membranes in both lungs, desquamating pneumocytes in the right lung, and lymphocytic infiltrates, but no obvious viral inclusion bodies [44]. Flow cytometry of peripheral blood cells showed reduced CD4 and CD8 T-lymphocyte counts, but a high proportion of lymphocytes showed evidence of activation.

Needle biopsy of lung in four fatal cases showed changes consistent with diffuse alveolar damage [45]. There were hyaline membranes in three cases, and in one case fibrinoid necrosis of small vessels.

The English abstract of a Chinese study of limited post-mortem examination in three patients reported hyaline membranes in some alveoli, with intra-alveolar macrophages and monocytes, and proliferation and desquamation of Type II alveolar lining cells [46]. Hyaline thrombi were seen in a minority of small vessels.

A man with COVID-19 and oedematous lungs had diffuse alveolar damage on light microscopy [47]. There were patchy hyaline membranes and a scanty lymphocytic infiltrate, and thrombi were noted in a few small branches of the pulmonary arteries. Immunohistochemistry identified sparse CD3-positive T-cells.

Another study examined the possible differences between the pathology of COVID-19 and of typical acute respiratory distress syndrome (ARDS) [48]. The authors examined lung and in some cases skin tissue from five patients with COVID-19. Two had died and three had rashes consistent with thrombotic microangiopathy. In the first case, the lungs were haemorrhagic and oedematous, and light microscopy showed a severe haemorrhagic pneumonitis; major changes were limited to the capillaries, which showed fibrin deposits and endothelial cell necrosis. A second patient, who died after four days of mechanical ventilation, also showed haemorrhagic pneumonitis with fibrin deposition in capillaries. Skin biopsy in the three cases with rashes showed a marked thrombogenic vasculitis, an arterial thrombus, and small venular thrombi. Capillaries in all five cases stained for components of complement, and all cases contained one or more of C4d associated with SARS-CoV-2 virions, C5b-9, and MASP2. The authors suggested that catastrophic microangiopathic damage, related to sustained complement activation via the alternative and lectin pathways, was pathogenic in at least some cases of COVID-19.

All 21 patients with SARS-CoV-2 infection who underwent post-mortem examination in a study in Basel had congested lungs, with histopathological changes of severe capillary congestion, hyaline membranes, and diffuse alveolar damage [49]. Ten had superimposed bronchopneumonia, four had pulmonary emboli, and one had evidence of vasculitis. In 3/18 patients, histological examination of the kidneys showed evidence of fibrin thrombi consistent with DIC. The authors concluded that microthrombi in the lung were consistent with complement-mediated microvascular damage.

In an Austrian post-mortem study of 11 COVID-19 patients, all had thrombus in small and middle-sized pulmonary arteries [50]. All but one had received prophylactic anticoagulant therapy, and all but one had been “selected at random for autopsy”. Other pathological changes in the lung, signifying diffuse alveolar disease, included oedema (10/11 cases), hyaline membrane formation (10/11), and emphysema (11/11). The authors considered that the thrombi were “unexpected” and most likely to arise secondary to endothelial damage.

Ackermann et al compared findings in the lungs of seven patients who died from COVID-19 with those from seven patients who had died from acute respiratory distress syndrome caused by influenza A H1N1 [51]. All showed diffuse alveolar damage with perivascular T-cell infiltration. However, the patients who had died from COVID-19 showed distinctive distortion of the lung microvasculature associated with severe endothelial injury. Alveolar capillary microthrombi were much more likely in patients with COVID-19 than those with influenza. There was, unexpectedly, significant new vessel growth in the alveolar microcirculation, but its significance is unclear.

A renal transplant patient who died from COVID-19 had viral inclusion bodies in the transplanted kidney on electron microscopy [52]. On light microscopy, clumps of inflammatory cells were found associated with endothelium in the heart, lungs, and small bowel. A second COVID-19 patient, who had mesenteric ischaemia during her final illness, had lymphocytic inflammatory changes affecting the vascular endothelium in the lung, heart, kidney, liver, and small intestine. She also had evidence of myocardial infarction. In a third patient, who developed mesenteric ischaemia and survived small bowel resection, the resected gut showed endothelial inflammatory changes. The authors concluded that “our findings show the presence of viral elements within endothelial cells and an accumulation of inflammatory cells, with evidence of endothelial and inflammatory cell death. These findings suggest that SARS-CoV-2 infection facilitates the induction of endotheliitis in several organs as a direct consequence of viral involvement.”

A series of 26 patients who died from COVID-19 and who suffered acute kidney injury underwent post-mortem examination [53]. A common finding was “erythrocyte stagnation in the lumen of glomerular and peritubular capillaries without platelets, red blood cell fragments, fibrin thrombi, or fibrinoid necrosis.” Endothelial injury was seen even in patent glomerular capillary loops. Three cases showed fibrin thrombi in the glomeruli, associated with severe endothelial injury. In a subset examined by electron microscopy, coronavirus-like particles were identified in 7/9 cases.

Evidence is accumulating that infection with SARS-CoV-2 can be associated with widespread vascular endothelial damage. Presence of the virus in the endothelium appears to cause an inflammatory reaction, and the microthrombi are a consequence. In the lung there is evidence of alveolar capillary angiogenesis, but its significance remains to be established.           

Practical therapeutics
Systematic reviews favour low molecular weight heparin to prevent deep venous thrombosis in acutely ill patients [54], and this is standard practice. There is no evidence from clinical trials in COVID-19 showing efficacy of any prophylaxis or treatment for the thrombotic complications [55].

The pathology is uncertain, with evidence that predictors of poor prognosis include an increased concentration of D-dimers, a reduced lymphocyte count, and paradoxically high fibrinogen concentrations; and antiphospholipid antibodies are present in some cases. The severe form of COVID-19 has been likened to cytokine release syndromes and to hyperferritinaemic states, such as macrophage activation syndrome [56]. It may be that, in addition to the mundane prevention of deep venous thrombosis in acutely ill medical patients, treatments should be directed at the procoagulant state and the inflammatory endothelial damage that is associated with microvascular thrombi. As one group observed, “the optimal preventive strategy warrants further investigation” [57].

Several suggestions, consensus guidelines, and research protocols for treatment strategies have been published [58 59 60 61]. Recommendations include the provision of prophylactic doses of low molecular weight heparin to all patients admitted with COVID-19, unless there are contraindications [62]. This seems eminently sensible, and in keeping with standard UK practice. 

A French consensus statement suggests that patients be stratified by risk based on D-dimer concentration and prothrombin time [63].

There is a caution that antithrombin deficiency may impair the efficacy of heparins, although the suggestion that bivalirudin may be more effective in patients receiving extracorporeal oxygenation or renal support [64] is problematic, because bivalirudin is contraindicated in patients with an estimated glomerular filtration rate below 30 mL/minute, and it is not removed by continuous veno-venous haemofiltration.

On the basis of the early results of an analysis of a retrospective non-randomized open study in which different forms of (unspecified) “treatment-dose” anticoagulants were given at different stages of COVID-19 to patients in New York hospitals, it was suggested that patients who received anticoagulants and were mechanically ventilated were more likely to survive than ventilated patients who did not receive anticoagulants; but the overall survival was the same, and the patients who received anticoagulants were more likely to be ventilated [65].

The authors of the case report in which vWF and Factor VIII activity were high also suggested using higher doses of anticoagulants [40].  The demonstration of antiphospholipid antibodies adds further complexity [26 39], and suggests that at least in those with potentially pathogenic antiphospholipid antibodies, oral anticoagulation during recovery should be with warfarin rather than direct-acting anticoagulants (DOACs) [66].

The effect of low molecular weight heparin on D-dimers was examined in a retrospective analysis of 42 patients, half of whom had received the treatment. D-dimer concentrations were significantly higher than those in the control group before treatment (3·75 ± 4·04 versus 1·23 ± 1·15; P = 0·009). There was no significant difference between the groups after treatment with a low-molecular-weight heparin (0·90 ± 0·44 versus 1·00 ± 1·06: P = 0·368) [67]. The authors interpreted the result to mean that heparin may be useful in the treatment of SARS-CoV-2 infection, but the bias introduced by the non-randomized retrospective design, and the failure to demonstrate any difference in outcome make any conclusion speculative.

There is increasingly strong evidence that venous thrombosis and pulmonary embolism are common in severely ill patients with COVID-19 and that the incidence is probably higher than in other conditions. This has encouraged recommendations to use treatment doses of an anticoagulant in seriously ill patients, especially those with raised D-dimer concentrations [68], which is probably primarily due to the infection and not a consumptive coagulopathy. However, as yet there is no proof that this strategy is helpful. A sensible recommendation, again without controlled evidence, is to prefer low-molecular-weight heparin and to monitor antiFactor Xa activity [69], although assays are not always available. 

The practical problems of recurrent thrombosis of extracorporeal circuits for renal replacement or membrane oxygenation has led to suggestions that alternatives, such as bivalirudin or prostacyclin analogues, be used. There is no evidence to support these suggestions.

There is also evidence that in many fatal cases, there is microvascular thrombosis, and that this is associated with endothelial damage associated with the presence of viral particles. The association of endothelial damage and viral infection is already established, for example in dengue haemorrhagic fever and hantavirus pulmonary syndrome [70]. The roles played by disruption of the vascular endothelial barrier, by immune cell activation, and by chemokine and cytokine dysregulation are unclear. However, the insights have led to speculation about potential treatments to supplement anticoagulation. For example, the demonstration that macrophage activation occurs, and that the pattern of interleukin (IL) expression is unusual, has led to the suggestion that drugs targeting macrophages or IL-6 may help [71]. A retrospective analysis of treatment with the IL-1 antagonist anakinra suggested that it may be effective, but concluded that “confirmation of efficacy will require controlled trials” [72]. Trials of axatilimab, a macrophage-colony stimulating factor receptor inhibitor, tocilizumab (an IL-6 receptor antagonist), and situximab and clazakizumab (IL-6 antagonists) are underway or proposed.

Given the evidence for complement activation, based on a small number of histological reports and the finding of high vWF activity, it may be that inhibitors of the complement pathway also have a role in treating COVID-19; eculizumab, a monoclonal antibody directed against C5, has been suggested as a potential therapy [73]. We have not identified protocols for trials of the anti-vWF monoclonal antibody caplacizumab in the treatment of COVID-19 [74].

The best current strategies for confronting large vessel thrombosis in COVID-19 are prophylaxis with low-molecular-weight heparin and treatment with full-dose low-molecular-weight heparin with monitoring of anti-Factor Xa. There are no strong hypotheses regarding the pathogenesis of the coagulant effect of COVID-19 to guide therapy. Until the results of masked randomized controlled trials are available, treatments directed against components of putative pathogenic pathways, such as interleukin and complement, should be regarded as experimental.