Prognostic value of von Willebrand factor and ADAMTS13 in patients with COVID-19: A systematic review and meta-analysis

A comprehensive literature search was performed using MEDLINE (via PubMed), Cochrane Library, Web of Science, and the EMBASE databases from inception to March 4th, 2022. We searched for studies that investigated the plasma levels of VWF-related variables in patients with COVID-19 according to clinical outcomes using predefined keywords: (“SARS-CoV-2” OR “COVID-19” OR “2019-nCoV” OR “Coronavirus disease 19”) AND (“von Willebrand Factor” OR “VWF” OR “von Willebrand” OR “Willebrand Protein”) AND (“ADAMTS13” OR “von Willebrand Factor cleaving proteinase”) AND (“Factor VIII” OR “FVIII”). The references listed in the retrieved articles were also reviewed to identify additional studies. Duplicate articles were removed after the initial search. Two reviewers (YJ and XZ) independently screened the title and abstract of the articles. The full-text of those that had passed the initial screening were assessed according to the eligibility criteria. Discrepancies were resolved by consensus between the two reviewers or by a third reviewer (XB) when a consensus could not be achieved.

The inclusion criteria were as follows: 1) clinical trial or observational prospective/retrospective studies on COVID-19 with clinical outcomes for mortality (survivor vs. non-survivor), need for intensive care unit (ICU) admission (non-ICU vs. ICU), and severity (non-severe vs. severe); 2) reporting data on plasma levels of VWF:Ag, VWF:Rco, VWFpp, VWF multimer distribution, ADAMTS13:Ac, VWF:Ag/ADAMTS13:Ac ratio, and FVIII; 3) results were presented as or could be converted/digitized to standardized mean difference (SMD). The following types of articles were excluded from the analysis: case reports/series, brief reports, communications, correspondence, or letters that involved less than ten patients, review articles, commentaries, non-English language articles, research articles on the pediatric population, animal or in-vitro studies, unpublished studies, and studies with irrelevant or non-extractable results.

Two reviewers (YJ and XZ) extracted data from eligible studies independently using a predefined spreadsheet containing authors, country, publication date, study design, sample size, patient demographics, including age and gender, criteria used for clinical outcome classification, VWF related variable. The raw data were compared and pooled by the third investigator (XB) to eliminate extraction errors. The pooled effect estimates were SMD in terms of these variable between patients with and without poor clinical outcomes. When the data were expressed in median and quantiles, we derived and estimated the mean and standard deviation with accepted methods [, ].

Two reviewers (YJ and XZ) also independently assessed the risk of bias (methodological quality) for each included study based on the principle of the eight-item Newcastle-Ottawa Scale (NOS; for case-control study) [] or revised NOS version (for cohort study; additional file 1) []. Studies with scores of 0–4 and 5–9 points were identified as low quality and high quality, respectively. Any disagreement was resolved by the two reviewers through discussion, if necessary, with the help of a third reviewer (XB). Funnel plots were used to assess the publication bias for outcomes that included more than ten studies.

The meta-analysis was conducted using the Review Manager software (RevMan5.3, Cochrane Collaboration, Oxford, UK). Continuous outcome variables were presented as SMD with 95 % confidence interval (CI). I-square (I²) statistics was used to evaluate the data heterogeneity. Outcomes with I² > 50 % were regarded as having a high heterogeneity. The random-effects model was used to analyze outcomes for included studies when I² > 50 %. Otherwise, a fixed-effect model was used. A p value of <0.05 was considered statistically significant in the test for overall effect. Publication bias was assessed by visualization of funnel plots.

A total of 1211 potentially relevant studies were included in the combined electronic and paper reference search. After removing 446 duplicates, 602 studies were excluded by title and abstract screening (reviews or not relevant). The final full-text review included 163 reports, from which 123 studies did not meet the inclusion criteria and were excluded. A total of 40 studies comprising of 3764 patients met the including criteria and were included in this systematic review and meta-analysis [, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ]. Fig. 1 shows the flow diagram of study selection. Table 1 presents study characteristics and patient demographics.

We analyzed reported plasma levels of VWF:Ag, VWF:Rco, ADAMTS13:Ac, VWF:Ag/ADAMTS13:Ac ratio, and FVIII in patients with COVID-19. Since there were 2 studies that met the inclusion criteria having plasma VWFpp values [, ] and 2 studies measured VWF multimeric pattern [, ], these variables were not included in this meta-analysis.

All included studies were assessed with 5 to 9 points (Table 1). The funnel plot was used to assess the publication bias for endpoint measures which pooled more than ten studies. No apparent publication bias was found in all the eligible endpoints. The funnel plots were listed in the supplement Fig. S6.

CAC was first described by Tang et al. in a single-center study in Wuhan, China []. Autopsy results from COVID-19 deaths have consistently highlighted the marked endotheliopathy, which is characterized with EC desquamation, cytoplasmic vacuolization, swelling, tight junction disruption, and loss of contact with the basilar membrane in the lung, heart, brain, mesentery, and kidney [, ]. Recent evidence suggests that endotheliopathy contributes to hyperinflammatory and hypercoagulable states that predispose COVID-19 patients to thrombosis and microvascular events [, , ]. Won et al. [] reported the upregulated pro-coagulants, downregulated anti-coagulants and severe thrombosis, and infiltration of activated macrophages, monocytes, and T cells within autopsy lungs of COVID-19 patients. Ranucci et al. [] demonstrated a closely correlation between interleukin (IL)-6 and fibrinogen in patients with COVID-19. A correlation between inflammatory cytokines and VWF was also reported by Dupont and colleagues []. The coordinated activation of the inflammatory and thrombotic responses is now termed immunothrombosis (or thromboinflammation), whereby COVID-19 induces endothelial injury, immune dysregulation, and inflammation, with resultant thrombosis and further propagation of inflammation [, ]. Ciceri et al. [] thus recommended the use of MicroCLOTS (microvascular COVID-19 lung vessels obstructive thromboinflammatory syndrome) to define thromboinflammatory response to COVID-19. A prevalent view holds that SARS-CoV-2 enters ECs directly by binding to the transmembrane angiotensin-converting enzyme 2 (ACE2) receptor [, , ]. But more recent studies have disputed the expression of ACE2 on ECs in postmortem COVID-19 tissues and in cultured ECs, and they suggest that the endotheliopathy is induced by pro-inflammatory cytokine milieu, complement activation, or tissue hypoxia [, , , ]. Ma et al. [] recently reviewed and summarized that ROS, VEGFA/VEGFR2, and HMGB1/RAGE/TLR4 are potential signaling pathways that involved in the COVID-19-associated endotheliopathy. These findings explain why patients with conditions of systemic endotheliopathy, such as cardiovascular disease, diabetes mellitus, chronic kidney disease, and cancer, are at a higher risk for severe COVID-19. It is therefore also not surprisingly that endotheliopathy and resultant CAC are more severe in the elderly with preexisting comorbidities [, , , , ]. An interesting question is as whether individuals on anti-platelet or anti-coagulant regimens for other conditions are less vulnerable for more severe COVID-19. Although D-dimer and fibrinogen are recognized to be related to thrombotic risk in COVID-19 early during the pandemic []. Circulating soluble biomarkers associated with endotheliopathy, including syndecan-1, VWF, selectins P and E, and intercellular adhesion molecule-1 (ICAM-1) have emerged as more clinically relevant biomarkers of CAC, multi-organ failure, and disease severity [, , , ]. However, whether these variables can be used for prospective clinical assessments may require further studies [].

VWF is a large multimeric glycoprotein encoded by the VWF gene on the short arm of chromosome 12, and is synthesized primarily in ECs and megakaryocytes [, ]. The architecture of pro-VWF composes four types of domains that are constructed as repeats in the following order: D1-D2-D′-D3-A1-A2-A3-D4-C1-C2-C3-C4-C5-C6-CK. The D1-D2 domains represent VWFpp that is cleaved off by furin, and the D′-CK represents mature VWF monomer. Mature VWF and VWFpp are secreted in equimolar amounts from ECs into plasma []. Upon synthesis, VWF is either secreted constitutively as smaller multimers (basal secretion), or stored in Weibel-Palade bodies (WPBs) of ECs and α-granules of platelets where multimerization continues [large and ultra-large VWF (ULVWF)] and release occurs upon various pathological stimuli (regulated release) [, ]. The majority of circulating VWF (80–90 %) in blood is derived from ECs, and it is therefore frequently used as a biomarker for endothelial activation or injury [, ]. Autopsy results reveal that VWF expression was increased in ECs of the lung, heart, and kidney from COVID-19 patients []. Babkina et.al [] found more intensive VWF immunostaining in pulmonary ECs of COVID-19 patients with thrombotic complications than those without thrombotic complications. Moreover, platelet-VWF plug formation was present within the pulmonary microcirculation of patients with COVID-19 [, ]. Several studies have reported elevated plasma levels of VWF:Ag in patients with COVID-19 compared with healthy controls, which confirmed a severe inflammatory state and fulminant endotheliopathy [, , , , , , , , , ]. In addition, several studies found increased levels of VWFpp in patients with COVID-19, however, this increase was less pronounced than that of VWF:Ag. They assumed that the decreased ratio of VWFpp to VWF:Ag indicated a diminished clearance of VWF. Although the mechanism is unclear, this may further contribute to the markedly increased VWF:Ag [, , , ]. However, whether the increase in VWF is the result of increased production, WPB exocytosis, or decreased clearance remains to be determined. Here, we indicated that levels of VWF:Ag were higher in patients with poor prognosis, consistent with those of two previous meta-analyses that included 10 studies/996 patients and 7 studies/695 patients, respectively [, ]. Notably, Ward et al. [] reported that elevated VWF:Ag was persisted during 3-week ICU stay, suggesting the sustained synthesis and release of VWF. The “post-acute COVID-19 syndrome”, which is defined as having dyspnea, fatigue, sleep disorder, and exercise intolerance following acute COVID-19 resolution, has also been attributed to persistent endotheliopathy [, ]. Levels of plasma VWF:Ag and VWFpp remained persistently high in some convalescent COVID-19 patients [, ], and were correlated with elevated D-dimer levels []. The mechanisms underlying these persistent endotheliopathy after recovery from COVID-19 remain poorly understood. It is worth noting that ECs of pulmonary small vessels and microvessels express 5–50 times higher concentrations of VWF mRNA than similar sized vessels in the kidney and liver []. Another interesting observation is that a greater absolute increase in VWF:Ag than FVIII (mainly synthesized in hepatocytes []) in COVID-19 patients [, ]. These may highlight the predominantly pulmonary-centric nature of COVID-19 and the critical roles of VWF in the pathophysiology of COVID-19.

Upon vascular injury, VWF undergo conformational changes that expose the functional A1 domain, which is hidden in a globular structure by forming a complex with the adjacent A2 domain, facilitating 1) binding of platelets to subendothelium under blood flow through its interactions with platelet receptor glycoprotein (GP)Ibα (via A1 domain) and subendothelial collagen (via A3 domain); 2) platelet-platelet interaction (platelet aggregation) by binding to platelet receptor GPIIb/IIIa, thereby inducing hemostatic plug formation (primary haemostasis) [, ]. VWF also serves as chaperone for FVIII (via D′-D3 domain), 1) protecting FVIII from enzymatic degradation and extending its half-life in blood; more importantly; 2) directing its localization to the site of vascular injury and promoting coagulation cascades (secondary haemostasis) [, ]. The ULVWF freshly released from activated ECs, either circulating in plasma or locating on endothelial surface, are intrinsically active and hyperadhesive because the functional A1 domain are continuously exposed [, ]. Activated VWF and freshly released ULVWF can also mediate leukocyte recruitment to facilitate inflammatory endotheliopathy at the site of injury and elsewhere either directly or indirectly after binding platelets [, ]. Consistent with this notion, neutrophil activation and the formation of neutrophil extracellular traps (NETs) have been found to be involved in thromboinflammatory response to COVID-19 [, , ]. VWF interacts with NETs to provide a scaffold for platelet/leukocyte adhesion thus promoting thrombus formation and inflammation [, ], which also observed in patients with COVID-19 [, ]. In addition, Ackermann et.al [] reported an unexpected new vessel growth within lungs of COVID-19 patients through a mechanism of intussusceptive angiogenesis, and speculated that it is because of endothelialitis and thrombosis in the lungs. Similarly, we have recently shown that VWF served as a coupling factor that tethered platelet-derived extracellular vesicles (EVs) to ECs, thus locally concentrating vascular endothelial growth factor (VEGF) for aberrant angiogenesis in patients with left ventricular assist device implantation []. We also found that hyperadhesive VWF released during acute traumatic brain injury mediated EVs to active ECs and platelets, thus responsible for consumptive coagulopathy and vascular leakage in the brain and the lung [, ]. Increased circulating EVs, containing procoagulant, proinflammatory, and prothrombotic factors in their cargo, have been reported to be predictive biomarkers and likely to enhance immunothrombosis in patients with COVID-19 [, , , , , , ]. Hence, VWF tethered EVs (VWF⁺EVs) may also be involved in COVID-19-associated endotheliopathy and coagulopathy, and a promising diagnostic or prognostic marker.

A key determinant of VWF functional capacity is that larger VWF size is more active due to more monomeric subunits and higher sensitivity for shear forces [, ]. ADAMTS13, mainly synthesized in hepatic stellate cells, is primarily responsible for specifically cleaving the ULVWF (>10,000 kDa; Tyr1605-Met1606 peptide bond within A2 domain) to smaller and less active multimers (<10,000 kDa), thus preventing the spontaneous intravascular platelet aggregation and resultant thromboembolism, as is seen in patients with thrombotic thrombocytopenic purpura (TTP), while maintaining the basic hemostatic activity of VWF [, ]. Roh et al. [] recently performed a case-control plasma proteomics study, and demonstrated that, of the 4996 protein analytes assessed, ADAMTS13 was the most significantly decreased in severe COVID-19, and displayed the strongest inverse association with myocardial injury. Abnormalities in plasma VWF multimeric pattern have been reported in patients with COVID-19. Several groups reported a relative decreased high molecular weight multimer (HMWM) of VWF in COVID-19 patients [, , ], which could be explained by an early increase in VWF proteolysis by ADAMTS13. This may also be attributable in part to the formation of ULVWF-platelet aggregates and the corresponding VWF and platelet consumption. However, thrombocytopenia is relatively rare in COVID-19 patients [, ]. But other studies of severe COVID-19 have observed evidence of increased HMWM [, ]. The differences likely due to the study cohort recruitment, time of sample collection, and the methodologies of VWF multimer distribution []. The exact dynamic changes of VWF multimeric pattern after COVID-19 still need further study. The majority of studies reported normal or mildly to moderately decreased ADAMTS13:Ac levels [, , , , , ], and a strongly elevated VWF:Ag/ADAMTS13:Ac ratio in patients with COVID-19 when compared to healthy controls, especially in those with worse illness or in non-survivors [, , ]. We confirmed that lower levels of ADAMTS13:Ac and higher VWF:Ag/ADAMTS13:Ac ratio and VWF:Rco were related to poor clinical outcomes. Several lines of evidence indicated that this is because massive increase of VWF with relative deficiency of ADAMTS13: 1) inflammatory cytokines and/or tissue hypoxia induce massively increased WPB exocytosis of VWF multimers by activating ECs [, , , ]; 2) there is no alteration in ADAMTS13 gene expression after stimulation of liver cells with pro-inflammatory stimuli []; 3) there is no intracellular storage pool of ADAMTS13 []. The latter two suggest AMAMTS13 may not increase as rapidly after COVID-19 as VWF does. Other possible mechanisms contributing this imbalanced VWF-ADAMTS13 axis include: 1) inflammatory (including complement and NET products) [, ] and oxidative [, ] mediators enhance VWF self-association to increase VWF reactivity, make VWF resistant to cleavage, and reduce ADAMTS13 activity in cleaving VWF; 2) reduced ADAMTS13 synthesis because of COVID-19-induced pathologies of the liver [, , ]; 3) Considering high-density lipoprotein (HDL) prevents VWF self-association, while low-density lipoprotein (LDL) has the opposite effect []. HDL reduction and LDL elevation observed in patients with COVID-19 may also contribute to the imbalanced VWF-ADAMTS13 axis [, ].