High incidence of thrombin generation resistance to thrombomodulin in patients with unprovoked venous thromboembolism Open Access

Venous thromboembolism (VTE), comprising both deep vein thrombosis and pulmonary embolism, affects nearly 10 million people every year worldwide.¹ Risk factors are classified as strong, weak, transient, or persistent,² but most events are unprovoked. Growing evidence suggests causal interactions between the coagulation and immune systems, innate immune cells contributing to thrombus formation (immunothrombosis).³ The contribution of particularities of the coagulation system predisposing to thrombosis to the risk of VTE is not fully identified.

Several global coagulation assays to assess the global coagulation potential have been developed over the past 50 years. Among them, thrombin generation (TG) testing (TG assay [TGA]) investigates TG and the process of thrombin inhibition as a whole,⁴ giving insights to potential coagulation hyperactivity. Adding purified thrombomodulin (TM) to the system reveals thrombin-mediated activation of the protein C system, limiting TG and allowing certain thrombotic phenotypes to be recognized.⁵ 

We report the incidence of TG resistance to TM in patients with unprovoked VTE and study the impact of clinical and coagulation parameters on that resistance.

We retrospectively screened the files of adult outpatients referred to our Department of Hematology between 1 January 2021 and 1 January 2025 to test for thrombophilia following a first, objectively diagnosed, unprovoked VTE event.

After 6 months of anticoagulant treatment, we performed a complete clinical evaluation backed by laboratory tests following a 15-day treatment washout. None of the women included were taking hormone treatment. VTE events were classified as unprovoked if the patient had none of the currently described clinical risk factors for VTE.¹ 

A control group of 200 sex- and age-matched healthy individuals were recruited among individuals coming to our institution for a systematic laboratory check-up and were tested for automated TGA (aTGA) after giving informed consent.

Thrombophilia testing included the F5 rs6025 and F2 rs1799963 polymorphisms, D-dimers (Vidas D-dimers Exclusion II; bioMérieux), fibrinogen, factors XI, IX, VIII, and II, antithrombin, protein C, and protein S, all by functional assays using reagents and the STA-R Max² analyzer from Stago, Asnières, France. Antiphospholipid antibodies (aPLAbs) were tested according to the guidelines of the International Society on Thrombosis and Haemostasis:^5,6 lupus anticoagulant (LA) using an activated partial thromboplastin time (results generating the Rosner index) and a dilute Russell’s viper venom time (generating the ratio dilute Russell’s viper venom time patient time:control time; the PTT-LA and STA-Staclot-DRVV screen reagents [Stago], respectively); anticardiolipin immunoglobulin G (IgG) and anticardiolipin IgM antibodies, and anti-β2-glycoprotein I IgG and anti-β2-glycoprotein I IgM antibodies using commercially available kits from Orgentec Diagnostika GmbH, Mainz, Germany. Positive solid-phase aPLAbs were >99th percentile of values from 200 normal plasma samples.

aTGA was performed using the STG-ThromboScreen reagent, in which the activation of TG is triggered by 5 pM tissue factor, and the ST-Genesia analyzer, Stago, as described.⁷ 

The study was approved by the institutional review board and ethics committee (reference: Local/2025/JCG-01) at Nîmes University Hospital.

The primary objective was to compare the variables describing the aTGA profiles, both in the absence and in the presence of TM, between patients and controls: lag time, velocity index, thrombin peak height, time to peak, endogenous thrombin potential (ETP), start tail, both in the absence of TM (_TM^–) and in the presence of TM (_TM⁺). The TM effect on TG was defined by ETP inhibition as follows: (1 – [ETP_TM⁺/ETP_TM^–]) × 100.

Quantitative data, handled as continuous variables, are given as medians and interquartile ranges, and comparisons were made using nonparametric tests (Wilcoxon-Mann-Whitney). Qualitative data are given as numbers (%) and compared using a χ² test or Fisher exact test.

The secondary objective was to describe predictors, among clinical and laboratory results generated by thrombophilia testing, of ETP resistance to TM (R-ETP/TM) defined at the fifth percentile threshold of the results observed in controls. Only data from those individuals who had provided full information were included in the analysis. We computed a logistic regression analysis, univariate (crude odds ratios and 2-sided 95% confidence intervals) then multivariate (adjusted odds ratios; backward stepwise selection beginning with all variables with a P value <.20 in univariate analysis; threshold for removal: P > .20). The goodness-of-fit of the model was assessed by the Hosmer-Lemeshow test. The estimated multivariate regression coefficients of the informative variables finally retained were used as the weighting coefficients of the parameters in a score for ETP resistance to TM, whose prediction ability was tested by receiver operating characteristics (ROC) curve analysis.

Statistical analyses were performed using StatView-Windows software v.5.0 (SAS Institute Inc, Cary, NC) and XLSTAT software v.2015.4.01.20116 (Addinsoft SARL, Paris, France). P values of ≤.05 were considered significant.

We were able to screen 1123 relevant patients’ files. Laboratory evidence of a residual circulating anticoagulant activity was evidenced in 54 patients, aTGA could be performed in 989 among the remaining 1069 (80 technical failures).

Table 1 shows the individual characteristics and aTGA results. Patients had higher TG rates than controls, and a lower inhibiting capacity for a fixed amount of TM on individual TG. The distribution of values of the TM-mediated ETP inhibition obtained in controls and patients is described in Figure 1, showing the systematic shift toward low values in patients (P = 2.2 × 10^–16). In controls, the fifth percentile of the distribution of values corresponded to the 20% threshold value, thus defining ETP resistance to TM (R-ETP/TM), which could be evidenced in 406 out of the 989 patients (41.1% [95% confidence interval, 38.1-44.2]; P < .0001). We selected 10 patients with aPLAbs-negative plasma with R-ETP/TM and performed mixing tests with the normal reference plasma used in TGA, or with a pool of normal control plasma samples. No evidence of any activity inhibiting the TM effect on ETP could be demonstrated.

The 422 individuals with full information were investigated for predictors of R-ETP/TM and found positive in 163 of them (38.6%; Table 2). We studied the global quantitative risk factors for VTE, as well as by referring to threshold values described in the literature as being associated with a first event or VTE recurrence.^8-15 

Univariate analysis showed that age, male sex, antithrombin, protein C, and protein S were protective factors, whereas D-dimers, factor VIII (FVIII) >150 IU, and results of LA-related assays were risk factors for R-ETP/TM.

Increasing D-dimer levels exposed to R-ETP/TM in a first multivariate quantitative model, protein S, and Rosner index were negative risk factors. The R² coefficient of determination was only 0.307. The score derived for R-ETP/TM had an area under the ROC curve of 0.851 (0.797-0.904); P < .0001; best threshold value (Youden index): positive predictive value = 0.646, positive likelihood ratio LR⁺ = 4.030, negative predictive value = 0.884, negative likelihood ratio LR^– = 0.290.

A second, qualitative model evidenced D-dimer higher than the age-adjusted normal upper limit, FVIII >150 IU, protein C <65 IU, and protein S <50 IU as risk factors for ETP resistance to TM, R² coefficient 0.183. The derived score for R-ETP/TM had an area under the ROC curve of 0.758 (0.681-0.835); P < .0001; best threshold value: positive predictive value = 0.585, LR⁺ = 4.235, negative predictive value = 0.834, LR^– = 0.596. Neither of the 2 models had sufficient predictive performances to diagnose ETP resistance to TM.

This is the first work to describe TG resistance to TM in patients with unprovoked VTE, and its high incidence. This is in addition to greater TG capacities. Finally, among the laboratory predictors were D-dimers, the Rosner index used to diagnose LA, high levels of FVIII, and low levels of protein C and protein S. The latter had a striking global effect on the assay, with activities highly correlated with ETP inhibition values (Spearman’s rank correlation coefficient: 0.585; by comparison, protein C: 0.286; D-dimers: −0.242, all: P < .0001).

The same test as the one we used was the best indicator of activated protein C pathway resistance in women using combined oral contraceptives.¹⁶ It combines with the hypercoagulant phenotype in pregnant women.¹⁷ It also indicated a later transfer of patients with severe COVID-19 disease to the intensive care unit,⁷ and was discriminatory for the risk of VTE recurrence in older adults.¹⁸ 

Using TM supplementation in the TGA makes it possible to assess the whole activated protein C pathway, including its activation by the thrombin-TM complex, the level of protein C in the plasma, and the subsequent response of inhibiting activated factor V and FVIII by the activated protein C generated.¹⁶ We found that protein S activity was a strong modulator in that assay, and that also the activation level of the coagulation and fibrinolytic pathways (as reflected by D-dimer levels) had an impact.

The mechanisms underlying TM resistance in patients with VTE remain unclear. The question is how increased D-dimer or reduction in protein S would contribute to the TM resistance is not known. Our data are based on association studies, and increased D-dimer or low protein S activities can be the consequence, the partial cause, or both, of resistance to TM. Increased D-dimer suggests increased TG, thus high basal ETP values and a relatively lower response to a fixed dose of TM, but also can be the consequence of a high residual TG after a TM challenge in the circulation. Low protein S may impair the efficiency of TM-induced activated protein C generation, thus impairing the inhibition of TG, but can also be the consequence of a partial protein S consumption due to a chronic excess of TG after a TM challenge in the circulation. Further studies are warranted.

Multivariate models could only partly predict ETP resistance to TM, suggesting that some determinants are still unknown. Genome-wide association studies might highlight some new actors of ETP resistance to TM. Of note, the family history of VTE did not predict ETP resistance to TM, which might argue against a strong genetic determinant to this resistance, unlike what is observed in severe monogenic thrombophilia. Hypotheses range from the association of an inherited genetic determinant with low thrombotic risk with an acquired modification of hemostasis, for example modulated by lifestyle or inflammation, to the association of several determinants with low thrombotic risk inherited from both parents without either of them being at high thrombotic risk.

We also noted that high TG measured in the presence of TM was associated with a greater risk of recurrent VTE¹⁹ in a subgroup of 254 patients prospectively enrolled in the Prolong clinical trial, which included patients with a first, unprovoked episode of VTE.

The main strength of our pilot study is a substantial number of cases tested using an automated, commercially available reagent-based technique within a single laboratory. The retrospective nature and single-center design of our study are, however, associated with certain limitations. A specific prospective multicenter study is quickly warranted.

Fully understanding the determinants of R-ETP/TM may help to categorize the risk of recurrence after a first unprovoked episode of VTE. Most of the observed differences in TG parameters are small but highly significant, raising the question whether this difference is biologically or clinically significant or not. Future multicenter prospective studies will have to see if TM resistance is associated with disease severity and treatment outcomes.

The authors thank Sophie de Boüard for administrative help and data management and Teresa Sawyers, medical writer at the BESPIM, Nîmes University Hospital, Nîmes, France, for expert editorial assistance.

Contribution: J.-C.G. conceived and designed the study; J.-C.G., C.B., J.L., and A.P.-M. managed the patients; M.C., E.N., M.F., and S.B. supervised and performed all laboratory tests; J.-C.G., M.C., and R.D.J. were in charge of data collection and statistical analysis; J.-C.G. drafted the first version of the manuscript; and all the authors interpreted the data, and reviewed and approved the final manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Jean-Christophe Gris, Consultations et laboratoire d’Hématologie, CHU, GHU Carémeau, Place du Pr. Robert Debré, 30029 Nîmes cedex 9, France; email: jean.christophe.gris@chu-nimes.fr.