While previous studies have reported that TFPI is increased in patients with cardiovascular risk factors, to the best of our knowledge, this is one of few studies to evaluate the ability of TFPI to predict prospective cardiovascular events. Our study, importantly, demonstrates the predictive capacity of TFPI for ATE in CKD patients, which is the largest cause of morbidity and mortality in this population. While our findings were not significant in diabetes patients without CKD, the latter may be due to low event numbers and limited follow-up duration.

TFPI has previously been explored as a biomarker of atherosclerotic disease. Interestingly, TFPI is best known for its anticoagulant properties, but paradoxically, increased TFPI has been strongly associated with the presence of atherosclerosis and cardiovascular disease. The extrinsic coagulation pathway has an important role in atherosclerosis. While tissue factor (TF) is not present on the surface of healthy endothelial cells, TF is exposed to the circulation following blood vessel injury [15]. Multiple mechanisms are in place to avoid subsequent systemic coagulation including the actions of TFPI along with protein S to inhibit FXa [16].

TFPI has also been found to co-localised with TF in atherosclerotic plaques [17], with murine models having also demonstrated that TFPI deficient mice have increased atherosclerosis [18]. It is hypothesized that TFPI acts to inhibit atherogenesis through downregulation of TF/VIIa, FXa and TF [4]. It is also thought that the role of TFPI in atherosclerosis extends beyond its direct anticoagulant function, with TFPI also having antiproliferative activity including as an inhibitor of basic fibroblast growth factor induced proliferation [19].

Several mechanisms have been proposed to explain the association of increased circulating TFPI with atherothrombotic disease. Elevated TFPI has been associated with established CVD risk factors [8, 9] and these factors may act to increase TFPI with it then following that these high risk groups would be predisposed to atherothrombotic events. It is also possible that increased TFPI may reflect the degree of endothelial dysfunction and platelet activation present, or similarly, that this is a compensatory mechanism for increased thrombotic activity [16]. Despite the strong association between TFPI and CVD, there have been few studies that have demonstrated the predictive capability of TFPI for subsequent ATE. Our study specifically explored TFPI in two vulnerable groups– diabetes mellitus and CKD, in which CVD is the largest cause of morbidity and mortality and remains the key focus of treatment in these populations.

The most significant finding in this study is the ability of a TFPI cut off level to predict ATE in the vulnerable CKD population (SHR 3.23). This association appears to be independent of renal function given the lack of correlation between TFPI with urea, creatinine or creatinine clearance. As far as we are aware, this association in CKD has not been described before, and could represent an important next step in personalised cardiovascular risk prevention in this population. There have been a number of studies that have evaluated TFPI in CKD and normal controls, interestingly with mixed results. Our previous work and Malyszko et al. [20, 21] showed increased TFPI in CKD while Pawlak et al. reported no significant difference [22].

Even beyond the specific CKD population, studies looking at cardiovascular events have predominantly focused on evaluation of patients acutely post an event [4, 23, 24] rather than targeting prospective event prediction in a chronic setting. Morange et al. did report on clinical outcomes in patients presenting with acute coronary syndrome, however TFPI levels were associated with severity of myocardial damage but not clinical outcomes [25]. Hence these findings in CKD patients significantly adds to the literature demonstrating the role of TFPI as a critical predictive biomarker of subsequent cardiovascular events.

Our findings in the relatively small pilot study of diabetic patients without CKD, however, were not as clear and did not demonstrate statistical significance in predicting subsequent cardiovascular events. The wide heterogeneity in diabetes population (including duration of disease, other cardiovascular risk factors and glycyaemic control), may explain the lack of significance seen in predicting ATE. In addition, the relatively infrequent and longer time to subsequent ATE may have limited our findings during the study follow-up period. It has previously been reported by El-Hagracy et al. that TFPI levels were higher in patients with diabetes with existing cardiovascular complications [26], suggesting that TFPI may be significantly impacted in DM patients. We did demonstrate correlation between TFPI and insulin-dependence in the DM cohort, in keeping with another study by Leurs et al. who found patients with insulin-dependent diabetes mellitus had increased TFPI in relation to normal controls [27].

Our findings highlight the complexity of the pathophysiology underlying atherothrombosis. The multi-faceted role played by TFPI within both the coagulation and endothelial systems, and in response to thromboinflammation, may explain the conflicting results reported by the various studies and for differing underlying disease processes. It is well described that CKD is associated with increased tissue factor (TF) levels leading to a procoagulant state [28, 29]. It has also been postulated that a resultant imbalance between TF and TFPI as its inhibitor may be an important factor contributing to atherothrombotic complications [22, 30]. We hypothesise that elevated TFPI levels may be the result of partial compensation for other hypercoagulable changes in the CKD population such as increased TF, explaining the paradoxical association between elevated TFPI, an anticoagulant protein and ATE. Some of the mechanisms leading to elevation in TF levels in CKD are renal specific including expression of TF in kidney-derived cells including renal podocytes and tubular cells in disease states [28]. Hence, these changes may be less apparent in the diabetic population, as the vascular homeostasis or imbalance may be less disrupted compared to the CKD population and they may not display the same renal mediated TF expression.

It should also be acknowledged that established CVD risk factors has been associated with higher TFPI along with increased risk of CVD events [8, 9], and the CKD cohort had relatively higher baseline CVD risk factors than the diabetes cohort. In addition, the hypercoagulable state in diabetes could be more heavily influenced by other changes in the coagulation pathways and endothelial system, with TF pathway only a minor part of a myriad of factors contributing to thrombotic events. Regardless, given there was also a lower event rate in the DM cohort, further longer duration exploration is required to exclude follow up duration as a potential factor influencing these results.

We acknowledge the limitations of this single site study with relatively low study numbers and short follow-up periods for ATE, which is particularly relevant for the DM population. Furthermore, the wide confidence intervals of the TFPI results in both cohorts may mean a potential for error, hence it is important that larger studies are performed to validate these results. Nonetheless, these important pilot results have identified that TFPI may be predictive of ATE, particularly in CKD patients, and explore the role of TFPI in the setting of prospective prediction of ATE in chronic disease states. Further studies, particularly incorporating TFPI with other biomarkers of CVD are crucial as part of our transformation towards personalised medicine in CVD.