Cohort I: ambulatory patients
We first sought to determine whether changes in the activity levels of select coagulation factors track with CLD severity in a single ambulatory cohort in which no instances of PVT were identified. PPP samples of CLD patients seen at an outpatient ambulatory clinic were first assessed for the activity levels of FV, FVIII, PC, PS, and thrombomodulin (TM) along with the circulating levels of sP-selectin and asTF. As expected, FV activity progressively decreased along the spectrum of CLD severity (Figure 1). PC and PS activity levels also decreased with CLD severity (Table 1). FVIII and TM activity progressively increased with disease severity (Figure 1, Table 1). The circulating levels of D-dimer along with those of asTF protein increased with disease severity, while the levels of sP-selectin decreased (Figure 1).
Cohort II: LT candidates
We next sought to evaluate more severe CLD, i.e., patients with advanced disease awaiting LT. 43 (mostly CTP-B) LT candidates were identified and consented to participate in this study which included access to disease history, severity scoring, standard lab results, and limited outcomes. Plasma was collected from each patient and evaluated for the activity levels of FV, FVIII, PC, PS, and the levels of D-dimer, sP-selectin, and asTF along with standard CLD evaluation labs: serum sodium, albumin, creatinine, total bilirubin, and PT/INR (Table 2). As in Cohort I, FV and PC activity declined significantly with CLD progression, while the levels of D-dimer rose (Figure 2). Also in agreement with the Cohort I were the low levels of PS activity in all patients (Table 2). FVIII activity was markedly elevated across the entire Cohort II, while sP-selectin protein levels were universally low (Figure 2). In agreement with our previous study [28], the levels of asTF protein were elevated in the majority of subjects (Figure 2).
PVT is associated with low Factor VIII, Factor V, and Protein S activity levels
PVT is a major complication of CLD, and our ability to effectively detect and/or predict it is lacking. In our cohort of 43 LT patients, there were 3 instances of PVT diagnosed by imaging within three months of consent (7%). Post-hoc correlation analysis between PVT incidence and the measured parameters revealed a strongly significant inverse correlation with FVIII activity levels (p=0.010). Activity levels of two other factors, FV and PS, also displayed in-trend inverse correlations with PVT (p=0.069 and p=0.064, respectively). When FV and PS activity values were matched and graphed with each patient’s FVIII activity values and distances from the graph origin were calculated for FV:FVIII (Figure 3A), the three patients with PVT were 1st, 2nd, and 10th closest to the origin; in the graph of PS:FVIII they were 1st, 2nd, and 11th closest to the origin (Figure 3B). Thus, we opted to use the activity levels of FVIII along with FV and PS to create two compensation score equations, CS-1 and CS-2, by performing logistic regressions. The equations are as follows:
CS-1 = -6.09 x ln(FV%) – 4.66 x ln(FVIII%) + 48
CS-2 = -2.27 x ln(PS%) – 4.34 x ln(FVIII%) + 50.5
A total of 11 patients, including the 3 cases of confirmed PVT, had a CS-1 score > 0 while a total of 20 patients had a CS-2 score > 0. None of the patients with a CS-1 or CS-2 less than 0 were diagnosed with PVT. 2 of the 3 patients with a confirmed PVT had CS-1 > 5.00 and CS-2 > 8.00: one had a CS-1 of 5.06, a CS-2 of 8.20 (FV 30%; FVIII 118%; CTP-B; MELD-Na 15), was not on anticoagulation at the time of sample collection, and was diagnosed with PVT one day after enrollment while the other had a CS-1 of 5.38, CS-2 of 8.43 (FV 46%; FVIII 63%; CTP-A; MELD-Na 12), was diagnosed with PVT 3 months prior to enrollment, and was not on anticoagulation at the time of sample collection. The third patient with a confirmed PVT was not on anticoagulation at the time of sample collection, had a CS-1 of 0.19, a CS-2 of 1.05 (FV% 45; FVIII 197%; CTP-B; MELD-Na 17), and was diagnosed with PVT two weeks after sample collection. Thus, multi-factor consumption may be a possible unifying condition among the three patients with PVT.
Factor V and protein C activity levels can replace PT/INR in MELD scoring
Aside from their potential utility in the detection of PVT, we also investigated whether factor activity levels could be used to calculate a CLD severity score for LT candidates. The current gold standard for tracking CLD progression in LT candidates, MELD-Na, assigns a score based on a patient’s levels of bilirubin and creatinine (indicators of liver and kidney metabolic health, respectively) as well as PT/INR. Monitoring coagulation is necessary as the liver is the main site of synthesis for pro- and anti-hemostatic factors. Expression of these factors is increasingly deranged with worsening disease severity. Given that PT/INR only measures clot formation and not dissolution, the appropriateness of its use to assess coagulative potential in CLD patients is debatable; we hypothesized that measuring activity levels of key hemostatic factors may serve as a non-inferior option to the use of PT/INR in MELD-Na scoring. Correlation analyses of factor activity levels in LT subjects and their MELD-Na scores revealed that FV activity levels significantly correlated with MELD-Na values; interestingly, activity levels of anti-coagulant Protein C behaved analogously to those of FV levels (Figure 4A, B). Seeing that activity levels of both a pro-coagulant (FV) and an anti-coagulant (PC) protein strongly correlated with MELD, we next sought to mathematically combine the FV and PC activity levels. Thus, we performed multiple linear regression using the constants for bilirubin and creatinine established by Kamath et al [31] as the dependent variable and the natural logs of FV and PC activity as independent variables. This multivariate analysis enabled the creation of a novel scoring formula:
CLD Score = 3.78 x ln(bilirubin) + 9.57 x ln(creatinine) – 1.16 x ln(FV%) – 3.10 x ln(PC%) + 29
Assessment of our novel formula vs MELD-Na showed a strongly significant positive correlation (Figure 4C). At the time of six-month follow-up, 36 of the 43 LT patients were alive. The seven deceased patients had an average MELD-Na of 15.2 ± 6.4 and an average novel CLD score of 16.2 ± 6.9. Using the c-statistic with 6-month mortality as the endpoint, the area under the receiver operating characteristic (ROC) curves for MELD-Na and our novel scoring formula did not differ significantly. The areas under the curves were 0.615 and 0.627, respectively (Figure 4D, E), demonstrating that our novel formula appears to be non-inferior to MELD-Na at predicting 6-month mortality.