This comprehensive meta-analysis includes 11 RCTs (5989 patients) and 17 OBs (10178 patients), which compared T-AC versus P-AC treatment in a total of 16167 COVID-19 patients and analyzed all-cause mortality, events of major bleeding, and thrombosis, and outcomes of subgroups. The results showed that: (1) in terms of RCTs, the comparison of T-AC and P-AC is not associated with a lower risk of all-cause mortality in the COVID-19 patients, and the clinical benefit to patients with COVID-19 is unclear. However, the subgroup analysis of OBs shows that the mortality risk significantly reduces in critically ill COVID-19 patients treated with T-AC compared with those with P-AC treatment. In contrast, the mortality risk significantly increases in non-critically ill COVID-19 patients treated with T-AC. (2) In RCTs and OBs analyses, T-AC treatment has a significantly higher risk of major bleeding than P-AC treatment in COVID-19 patients. (3) Compared with P-AC treatment in COVID-19 patients, patients with T-AC treatment significantly reduce the incidence of venous thromboembolism, but it is not associated with arterial thrombosis events. (4) T-AC treatment does not reduce mortality risk in COVID-19 patients with high d-dimer levels in RCTs, consistent with overall outcome measures. (5) The overall sensitivity analysis results after excluding RCTs data remain consistent with the previous results.
Previous clinical studies of patients with COVID-19 demonstrate that anticoagulant treatment reduces the risk of mortality and thrombosis compared to that without anticoagulation, which generally benefits patients . Similarly, two previous studies show benefits in patients receiving anticoagulation, which is associated with increased survival compared with patients without anticoagulation treatment [7, 45]. However, there is an open question regarding anticoagulation doses treatment in COVID-19 patients, and whether therapeutic doses of anticoagulants are more effective than low-dose anticoagulation for prophylaxis is still controversial.
In our meta-analysis, the COVID-19 patients treated with T-AC compared with P-AC treatment did not reduce mortality risk. However, in the OBs, T-AC treatment in COVID-19 patients significantly reduces the mortality in critically ill patients compared with the patients with P-AC treatment. Surprisingly, in non-critically ill COVID-19 patients, T-AC treatment can increase mortality compared with the patients with P-AC treatment. Similar results are reported in another meta-analysis of observational studies . In addition, two recently published meta-analyses show no survival benefit with higher doses of anticoagulants in COVID-19 patients, and they both increase the risk of major bleeding [46, 47]. The results demonstrate that therapeutic doses increase bleeding events, while prophylactic doses decrease the risk of bleeding in COVID-19 patients. It is well known that exposure to high doses of anticoagulants can lead to major bleeding events, often with fatal consequences [48–50].
In our meta-analysis regarding arterial and venous thrombosis events, T-AC treatment in COVID-19 patients significantly reduces the risk of venous thrombosis compared with P-AC treatment and increases the risk of major bleeding. However, the risk of arterial thrombosis did not decrease in COVID-19 patients treated with T-AC, consistent with previous studies [46, 47]. The possible mechanisms for these results may be related to different pathogenesis of arterial and venous thrombosis. Venous thrombosis can be triggered by blood stasis, hypercoagulability, and endothelial dysfunction and occurs most commonly in the valve pockets of large veins . However, arterial thrombosis, caused by atherosclerosis, is mainly formed by the aggregation of platelets . The prophylaxis of arterial thrombosis is usually benefited from antiplatelet therapy. Whereas recent studies from three larger, open-label, randomized controlled trials do not support the addition of antiplatelet treatment to prevent progressive thromboinflammatory complications in hospitalized COVID-19 patients [53–55]. D-dimer levels were identified to be associated with vascular thrombosis and a poor clinical outcome in critical illness, which might suggest a strategy of d-dimer-guided anticoagulation . Compared with P-AC treatment in COVID-19 patients, T-AC treatment does not reduce mortality risk in patients with high d-dimer levels (Additional file 7: Fig S6). Its efficacy needs to be further studied.
Moreover, two included studies used rivaroxaban and apixaban as anticoagulant agents: Lopes-2021  and Connors-2021 . Both rivaroxaban and apixaban are orally available, direct factor Xa inhibitors with a different mechanism of action than heparin. They are small molecules with a distribution volume exceeding heparin [57, 58], potentially allowing them to better access lung tissue to prevent alveolar thrombosis. Rivaroxaban and apixaban are used in outpatient settings because they are limited to non-critically ill COVID-19 patients. However, it is still unclear whether this diverse mechanism reflects a clinical difference between anticoagulant treatment with rivaroxaban or apixaban and LMWH/UFH in patients with COVID-19. In addition, due to a small sample size of two sets of data (Lemos-2020 , Marcos-Jubilar-2021 ), we excluded studies of rivaroxaban and apixaban treatment in COVID-19 patients from the subsequent sensitivity analysis to decrease type II errors. The results from the sensitivity analyses after exclusion are consistent with previous mortality and major bleeding results. Sensitivity analyses for OBs are not performed due to insufficient baseline characteristics and a lack of data integrity.
In addition, UFH prolongs clotting time by enhancing or activating antithrombin III (AT-III) and anticoagulant factor Xa. Compared with UFH, LMWH exerts an anticoagulant effect by primarily enhancing and activating anticoagulant factor Xa . There was no significant difference in efficacy between UFH and LMWH . LMWH has the advantages of a robust anticoagulant effect, low risk of major bleeding, and low probability of inducing heparin-related thrombocytopenia (HIT), making it more preferred in the clinic. However, in critically ill patients, especially those with renal insufficiency or the elderly, LMWH tends to cause drug accumulation, leading to an increased risk of bleeding in these patients [59, 60]. In current available studies, many populations were unclassified, resulting in increased heterogeneity of the study populations and possibly inconsistent results between different uses of UFH and LMWH.
The clinical trial of “Therapeutic Anticoagulation versus Standard Care as a Rapid Response to the COVID-19 Pandemic” (RAPID) was to determine if therapeutic heparin is superior to prophylactic heparin in moderately ill patients with COVID-19 . Their results showed no significant reduction in the primary outcome of death, mechanical ventilation, or length of ICU stay with therapeutic heparin. However, therapeutic heparin was associated with a significant decrease in all-cause mortality and a reduced risk of major bleeding in the patients with COVID-19. The results from the RAPID trial suggest that therapeutic heparin was beneficial to moderately ill patients with COVID-19 who were admitted to hospital wards . A large observational study of 3,119 patients with COVID-19 showed that both prophylactic and therapeutic doses reduced mortality in COVID-19 patients with hypercoagulable states . In addition, COVID-19 patients who received the therapeutic dose had a higher survival probability than those with the prophylactic dose treatment. But in critically ill patients, there was an increased probability of major bleeding . Notably, the benefit of high-dose anticoagulants is unclear due to active or potential bleeding complications, low baseline hemoglobin or platelet counts, and physician practice choices.
Based on the above interpretation and analysis, compared without anticoagulants treatment, COVID-19 patients prone to hypercoagulability have increased clinical benefits of anticoagulation treatment. However, the choice of dose between T-AC and P-AC is still controversial. In our integrated analysis of included RCTs and OBs, meta-analysis evidence supports the necessity for T-AC treatment in critically ill COVID-19 patients. Still, close monitoring of the associated bleeding risk is also required. A previous study showed a hypercoagulable state in the COVID-19 presents during the early stage of infection . Therefore, prompt anticoagulation treatment may prevent the disease from progressing to severe forms in COVID-19 patients who may present with a disseminated intravascular coagulopathy (DIC)-like state , thus preventing an increased risk of major bleeding.
Meanwhile, prophylactic anticoagulants may be the most beneficial option in non-critically ill patients, reducing the incidence of major bleeding and fatal events. Routine use of therapeutic doses of anticoagulants beyond the standard prophylactic dose is not suggested in non-critically ill patients with COVID-19. Even though thromboprophylaxis does not reduce mortality in patients with acute illness , it can be used to prevent venous thrombotic events in patients without indications for anticoagulation as a guideline-recommended . Thromboprophylaxis reduces the risk of thrombosis in patients with risk factors but not the risk of death. Therefore, for patients without anticoagulation indications, the benefits of venous thrombosis prophylaxis outweigh the risks, which may be clinically beneficial. In COVID-19 patients with a hypercoagulable state, the prophylactic dose may be less effective than the therapeutic dose for preventing venous thrombosis. In addition, another concern is the lack of guidance for the anticoagulation treatment window timing and symptom onset time, even though the average time window for symptoms of COVID-19 patients is about one-week . Therefore, further clinical trials are needed to confirm the effectiveness of the study on the time window of coagulation to prevent thrombosis in patients with COVID-19.
Strengths and limitations
The current comprehensive meta-analysis study incorporated all relevant RCTs and OBs and analyzed more studies and patients with COVID-19 than previous studies, which only included RCTs . In RCTs, we did not find a statistical difference between the mortality of T-AC and P-AC treatment in patients with COVID-19. However, data analysis from OBs demonstrated that T-AC significantly increased the survival of critically ill patients with COVID-19, but not the non-critically ill patients. In both RCTs and OBs, T-AC treatment in COVID-19 patients showed a decreased risk of venous thromboembolism but increased the risk of bleeding compared to P-AC treatment.
Contrasting with merely pooling RCTs, we discovered some controversies about the conclusion of the meta-analysis. Observational studies play an essential role in guiding patient care, making medical decisions, and providing evidence among representative patients with varying severity. Nowadays, policymakers increasingly require much real-world data such as electronic medical records to conduct observational studies on the medical evidence and clinical practice . RCTs provide the primary evidence for regulatory decisions. Still, as the methods of real-world research mature, there will be more opportunities to supplement the limited aspects of RCTs using real-world research .
However, several limitations to our study should be noted. First, due to the limitations of baseline characteristics, we did not perform detailed subgroup analyses, and all analyses were classified into severe and non-severe cases according to clinical conditions. In addition, the quality of the included RCTs varied, and all but two trials had an open-label design, which may lead to bias in the determination of thrombotic and bleeding events. Detailed sensitivity analyses for OBs were not performed due to a lack of information. Second, the sample sizes of two studies were too small with inadequate representation, and two used rivaroxaban and apixaban as anticoagulation treatment in patients with COVID-19. This may lead to increased heterogeneity and statistical error in type II. But we performed a sensitivity analysis later and excluded them. Thirdly, the population definitions used in the studies are different, such as critically ill and non-critically ill patients, which cannot be summarized uniformly, resulting in biased population classification. In addition, there was no standardization of prophylactic and therapeutic treatment strategies, and definitions of primary and secondary outcomes were inconsistent. Therefore, our results should be considered carefully, as possible confounding cannot be completely ruled out. Fourth, the observation methods and time of outcome indicators, such as major bleeding, were inconsistent in different studies. There was heterogeneity among study populations, settings, experimental designs, interventions, and detection methods. Last but not least, there was one study used direct oral anticoagulants  as its therapeutic anticoagulation, and two other studies used rivaroxaban or bemisaban.