The present MR study found further evidence for TM as the risk factors for stroke and IS. These results were consistent in multivariable MR adjusted for BMI, SBP, LDL, T2D, and smoking. In the reverse MR analyses, no evidence shown significant association of genetic liability to stroke and its phenotypes with TM.
Previous clinical studies had suggested that circulating thrombomodulin levels were increased in IS patients, and increased plasma TM levels at baseline were associated with decreased risks of adverse clinical outcomes at 3 months after IS 27. However, to date, there are few studies investigating the association between plasma TM levels and the risk of stroke. A population-based study about vascular injury biomarkers and stroke risk performed by Laura et al. found only a weak borderline significant trend between TM and overall stroke risk (hazard ratio:1.47 [0.99–2.19]; P = 0.05) 28. Therefore, still little is known about associations between plasma TM levels and stroke risk.
The current MR study extend the literature on the issue of TM in stroke in several ways. Firstly, in the context of the MR framework, our study provided compelling evidence of the causal relationship between TM and an increased risk of any stroke, as well as IS. In particular, evidence for causal inference was supported by the consistent direction and magnitude of effect estimates across various MR methods, including IVW, MR-Egger regression, weighted median, simple mode, and weighted mode. These results suggest that measuring serum TM levels could serve as a valuable screening tool for predicting the passible occurrence of any stroke and IS. Endothelial cells play an important role in the prevention of intravascular thrombus formation29, and ECD is critically involved in development of atherosclerosis and stroke30,31. The anticoagulant properties of the endothelium are mediated primarily by TM, a multidomain type-1 transmembrane glycoprotein constitutively expressed on the luminal surface of endothelial cells32. Binding of thrombin to the high affinity TM receptor transforms its procoagulant activity into an anticoagulant potential, by activating protein C33. TM detaches from the cell membrane surface and releases when the vascular endothelial cells are damaged; thus, the plasma TM levels are increased. However, at this time, the binding between thrombin and TM weakens, making it impossible to activate protein C. Eventually, the anticoagulant effect of body fluids is weakened, resulting in a procoagulant effect that increase the risk of stroke. Secondly, we found no causal evidence that the occurrence of IS, or indeed any other type of stroke, leads to increased serum TM levels. Although, there have been observational studies showing that circulating thrombomodulin levels were elevated in IS patients, as is known, results from such studies are limited by small sample size, residual confounding and reverse causality13. Alternatively, we prefer to believe that the reason why circulating TM levels were increased in IS patients may be due to the fact that TM has already been released into the blood due to the occurrence of ECD before the onset of IS. Nevertheless, further studies are needed to support this hypothesis.
The present study had several strengths. It was the first MR study with a comprehensive assessment of TM in relation to stroke risk. Compared to previous observational studies, MR analysis could effectively reduce potential bias including confounders and reverse causation, thus enhancing the causal inference. Another strength is that the bidirectional analysis guaranteed the inference of causality between TM and stroke in both directions, avoiding misleading causal effect14. In addition, we performed multivariate MR analysis to determine whether the observed association was driven by confounding factors.
Our study has several limitations that need to be considered. Firstly, the study only included a single population, and the representativeness of the results remains to be further verified in the whole population. Secondly, it is possible that our findings might have been affected by weak instrument bias, which is influenced by the selection of the genetic instrument through the relatively lenient threshold of P = 1 × 10− 5 for IS phenotypes (including LAS, CEA, and SVS), although the F statistics did not indicate that our instruments were weak.