Sources of heterogeneity in studies reporting the venous thrombotic risk in patients admitted with COVID-19: A meta-analysis and explorative meta-regression

ABSTRACT

Background: Meta-analyses on venous thromboembolism (VTE) risk in admitted patients with COVID-19 have shown substantial heterogeneity among included studies. Aim(s) To explore sources of heterogeneity among studies estimating VTE risk in COVID-19 patients. Method(s) A systematic review using the databases PubMed and Embase was performed searching for studies reporting VTE risk in patients admitted to the ward or intensive care unit (ICU). Analysis was performed with studies from which a incidence proportion could be retrieved. The pooled incidence of VTE and heterogeneity (I2) were estimated in a random-effects model stratified by ICU or ward. Incidence estimates were logit-transformed. The effect of 12 pre-selected clinical and methodological variables were explored in a univariable linear meta-regression model. Because of limited degrees of freedom, a multivariable model was constructed in a stepwise fashion. The outcome measure was regression coefficient with a 95% confidence interval (CI). Result(s) Fifty-three and 26 studies reported on incidence of VTE at the ICU and ward respectively. The pooled incidence at the ICU was 20.3% (95% CI 16.4-24.8%, I2 95.4%) and at the ward was 4.7% (95% CI 3.0-7.3%, I2 96.2%). In studies concerning ICU patients, asian region, VTE-screening, standard error and date of publication were significantly associated with incidence of VTE in the univariable model. (Table Presented) Two multivariable models were fitted (Table 1). In the multivariable model only standard error remained significantly associated. In studies concerning ward patients, average age, VTE-screening, standard error were significantly associated in the univariable model. Two multivariable models were fitted (Table 2). Again, only standard error remained significantly associated. Conclusion(s) Studies with a high standard error yielded a higher VTE incidence in COVID-19, strongly suggesting publication bias. Remaining heterogeneity was not completely explained, but may be due to differences in setting/outcome definition and clinical practice.