Ethical clearance was obtained from the University of the Witwatersrand Human Research Ethics Committee (reference number: M230570). Written consent was obtained from each participant when collecting data directly from patients and adding to a database for the study. Measures taken to ensure confidentiality included obtaining informed consent, anonymising data, strict access controls to data, secure storage of data on the REDCap® electronic data capture system, clear confidentiality policies for all researchers involved and the signing of a confidentiality agreement. The necessary ethical clearance documentation has been provided.

The study was a retrospective cross-sectional observational study at Chris Hani Baragwanath Academic Hospital COVID-19 and intensive care unit wards, between November 2020 and January 2021. Chris Hani Baragwanath Academic Hospital is a tertiary hospital located in Johannesburg, Gauteng, South Africa.

Patients included in the study had reverse transcription polymerase chain reaction-confirmed SARS-CoV-2 infection, were ≥ 18 years old, and had severe SARS-CoV-2 disease. Severe disease was defined as oxygen saturation < 93% with a respiratory rate ≥ 25 breaths per min, requiring supplemental oxygen support without the need for invasive or non-invasive ventilation. Critical illness was defined as hypoxaemia and the need for additional ventilatory support, in the form of non-invasive or invasive ventilation.

Exclusion criteria for patients were: a prolonged activated partial thromboplastin time of > 50 s; patients receiving chronic anticoagulation therapy (including, but not limited to, aspirin, warfarin, clopidogrel, LMWH such as enoxaparin and direct-acting oral anticoagulants such as rivaroxiban); patients with prior VTE, currently receiving quinine or a derivative thereof; known thrombotic thrombocytopenia purpura or thrombotic thrombocytopenia purpura-like disease; known haemoglobinopathies; pregnancy; oestrogen replacement therapy; and patients not receiving anticoagulation acutely or having any contraindication to anticoagulation.

To achieve a correlation coefficient of 0.5 with 80% confidence and precision of approximately 10% between thromboelastograms and anti-Xa levels, as shown previously by Tekkesin et al.,¹⁸ a sample size of 40 was required. We collected an additional 5% (2 samples) to allow for test failures, thereby arriving at 42 samples in total.

Data collection was carried out by the primary investigator and supervisors. A database was created as part of a PhD for one of the supervisors, with data being collected from 01 November 2020 to 31 January 2021. The data were collected by both the primary investigator and the supervisors during this period and entered directly into the REDCap® electronic data capture tool.¹⁹ The data were then exported into Statistica® version 13.3 (TIBCO Software Inc., Santa Clara, California, United States) for data interpretation. Data extraction from this database was conducted from 01 March 2021 to 31 May 2021. Sociodemographic data were collected directly from patients, including age, gender and ethnicity.

Whole venous blood samples were drawn directly from patients from the antecubital vein using a tourniquet and syringe for thromboelastogram, D-dimer, and anti-Xa. These blood samples were drawn on admission (prior to enoxaparin administration), 48 h post admission (3 h after enoxaparin administration); and at clinical resolution (defined as resolution of hypoxia) or day 10 of illness.

Thromboelastograms were processed within 5 min of whole blood being taken, and the corresponding sample was taken to the laboratory for anti-Xa testing within the specified 5 min as per manufacturer’s instructions. The TEG® 6s machine (Haemonetics®, Boston, Massachusetts, United States) was then used by filling a droplet of blood and a droplet of reagent into a cartridge which is then processed. The result was displayed and printed revealing an R-time, K-time, alpha angle, MA, and TCI. Only results from the standard kaolin thromboelastograms were considered.

Blood samples for D-dimer and anti-Xa tests were stored in citrate tubes (Becton-Dickinson, Oxford, UK). D-Dimer, and anti-Xa tests were done by the Chris Hani Baragwanath Academic Hospital National Health Laboratory Service Haematology Department, accredited by the South African National Accreditation System (SANAS), ISO 15189. D-dimers were measured using the D-dimer PLUS assay (Siemens Healthcare Diagnostics Products GmbH, Marburg, Germany). The test requires a latex-agglutination test which is then processed using an automated quantitative turbidimetric D-dimer assay. Anti-Xa levels were measured using the STA®-Liquid Anti Xa assay (Diagnostica Stago PTY LTD, Doncaster, Australia). The recommended target reference ranges for adults on treatment is 0.6 IU/mL – 1.0 IU/mL, and 0.2 IU/mL – 0.6 IU/mL for prophylaxis.

Reaction time evaluates the time from coagulation cascade initiation to fibrin generation and clot propagation. It reflects activity of the coagulation cascade. A shorter R-time reflects hypercoagulability. It is calculated from the beginning of the test to the first detectable clot formation (2 mm). The normal reference range is 5 min – 10 min.²⁰

Kinetics time determines the rate at which a clot is formed and is measured from the beginning of clotting to the proper formation of a clot (20 mm). It measures fibrin deposition and cross-linking. The normal reference range is 1 min – 2 min.²⁰

The alpha angle determines the speed of clot growth and strengthening. It is the angle between R-time and a line from the time of clotting initiation to the point of maximal clot formation speed. The normal angle is 53°–72°. When decreased, it suggests a deficiency in fibrinogen.^20,21

Maximum amplitude refers to the maximal amplitude of the thromboelastograph curve; the normal range is 50 mm – 70 mm.²⁰

Lysis at 30 min depicts the speed of fibrinolysis and shows the percentage of reduction 30 min after MA. The normal range is 0% – 8%.²⁰

The TCI assesses overall coagulation status, and the formula (Equation 1) is as follows:

The normal TCI range is within −3.0 and +3.0 (3 standard deviations from the mean of zero). Hypercoagulability is a TCI greater than +3.0, and coagulopathy is a TCI less than −3.0.²⁰

Data were collected and managed using REDCap®¹⁹ electronic data capture tools hosted at the University of the Witwatersrand, and statistical analyses was performed using StatSoft, Inc. data analysis system, Statistica version 13.3 (www.statsoft.com; TIBCO Software Inc., Santa Clara, California, United States).Spearman’s correlation was used to determine if there was an association between anticoagulation test parameters. Student’s t-test and Mann Whitney U test were used to determine the relationship between the dosing groups of enoxaparin and coagulation tests.

A multiple linear regression model was performed (Online Supplementary Table 1) using the thromboelastogram parameters (R-time, K-time, alpha angle, MA and A30) to predict anti-Xa levels (Table 1). Only R-time, K-time and A30 were included in the final model, as p-values were ≤ 0.2 (a 20% probability that the correlation was a result of chance was considered acceptable). Alpha angle was omitted, as this is dependent on K-time. Only R-time was significantly associated with anti-Xa level, with a p-value of < 0.05.

Forty-two participants were enrolled into the study (Figure 1). A total of 22 patients required a prophylactic dose (≤ 40 mg/day) of enoxaparin (52.38%), while 20 patients were treated with a therapeutic dose (> 40 mg/day) of enoxaparin (47.61%) (Table 2). Fifteen men and 27 women, with ages ranging from 33 years to 74 years, met the entry and exit criteria, enrolling them into the study. Patients with prior VTE, bleeding disorders, or on any anticoagulants, were excluded from the study to ensure a population which would not be biased towards bleeding or thrombosis.

There was a statistically significant low to moderate correlation between anti-Xa levels and TCI, using Spearman’s correlation = 0.43 (p = 0.014) among patients admitted for suspected COVID-19 pneumonia (Figure 2).

There was a statistically significant moderate correlation between anti-Xa levels and R-time, with Spearman’s correlation = 0.52, p = 0.002 (Figure 3).

There was a statistically significant low correlation between anti-Xa levels and K-time. with Spearman’s correlation = 0.35, p = 0.049 (Figure 4).

Although the K-time and alpha angle are mathematically correlated, we did not find a statistical correlation between anti-Xa levels and alpha angle.

There were also no significant correlations between D-dimer and thromboelastogram parameters.

Spearman Rank Order Correlations revealed no marked correlations between thromboelastogram and D-dimers, at a significance of p < 0.05.

Only anti-Xa levels were significantly associated with the dose of enoxaparin, while amongst the thromboelastogram parameters, lysis time at 30 min was significantly associated with the dose of enoxaparin (Online Supplementary Table 2).

We found a statistically significant low correlation between anti-Xa levels and K-time (r = 0.35), which is different from the study by Buckley et al.,²² who found no relationship between anti-Xa levels and K-time (r = 0.08). This was likely because we measured the K-time at peak enoxaparin dose effect (3 h post dose) compared to the continuous infusion, which may not have achieved the optimal anti-Xa level. Furthermore, differences in our study population, especially regarding disease profile, may be contributory as previously mentioned.²²

We failed to detect a statistically significant relationship between anti-Xa levels and MA, which is similar to the findings of Buckley et al.²² These findings are not surprising, as much as 80% of the MA is contributed to by platelet activity,²⁶ which is not affected by enoxaparin.

Our study found no statistically significant correlations between D-dimer and thromboelastogram parameters. This has also been found by Chandel et al. in a study of 24 patients with severe COVID-19 pneumonia.²⁷ Their study found a discordant or unpredictable relationship between clot burden and D-dimer levels. This may explain the lack of relationship between D-dimer, anti-Xa levels and thromboelastogram parameters.²⁷

Although we found a statistically significant difference in A30 percentages between prophylactic and therapeutic dose enoxaparin, there was no clinically significant difference. This finding is confirmed by Lloyd-Donald et al. (Australia, 2020) who failed to find any clinically significant relationship in A30.²⁸ However, our finding is an interesting one, and may require further investigation on the effect of enoxaparin on fibrinolysis.

Our study had several limitations which need to be considered. Firstly, we included a limited sample size, which is open to large variations in precision. Secondly, although we had two dosing strategies of enoxaparin, the exact dose in the therapeutic dosed group was not considered for the analysis. The data are also limited by use of a single LMWH medication (enoxaparin) and are likely not to apply exactly to other anticoagulants. The nature of a single centre study limits the generalisability of the findings. However, given the limited published data available, our study findings are not insignificant, given that they contribute the second largest published sample sized study.

We have demonstrated that the TCI (comprising R-time, K-time, and alpha angle) is statistically significantly correlated with anti-Xa levels in this study sample receiving LMWH. In particular, TCI values appear to increase as anti-Xa levels increase. It would be useful to see if this finding is replicated in other studies and, if so, why. Our findings should be interpreted with reserve and do not mean that TCI can replace anti-Xa testing. Anti-Xa levels were associated with the dose of LMWH (as expected), with lysis time at 30 min being the thromboelastogram parameter associated with LMWH dose (an interesting finding worth further investigation regarding the relationship between LMWH and fibrinolysis).

The possibility should also be considered that the high TCI values obtained in patients with higher anti-Xa levels are merely a reflection of the hypercoagulability present in these patients (who are correctly, it would seem, treated with higher doses of enoxaparin), and that these high TCI values have no direct relationship with the measured anti-Xa levels at all, but rather with the particular patient group receiving higher enoxaparin doses.

This study serves as a basis for future research with larger sample sizes, including patients with varying disease profiles and severity of illness. Further studies should focus on threshold values for R-time and other thromboelastogram indices that indicate specific anticoagulation targets.