Among 1,385 LTs performed between July 2005 and December 2021 at Severance Hospital, Korea, where living donor LT is predominant [11]. Patients aged <18 years (n = 120), those with a history of seizures before LT (n = 15), and those with missing data (n = 7) were excluded from the study. From the 1,243 eligible patients who underwent LT, those diagnosed with seizures at 1-year post-transplantation upon consultation with a neurologist were recruited. The causes of seizures were identified using blood tests and brain imaging studies and categorized according to the presence and type of abnormalities in imaging studies.

Patients diagnosed with seizures (seizure group) were matched to controls (no-seizure group) at a 1:5 ratio using a nested case-control design (Figure 1). Possible control individuals who had not experienced seizures (regardless of the possibility of seizures in the future) were randomly sampled at the corresponding time point (index postoperative day [POD]) when seizures developed in the seizure group. The year of LT was matched during the sampling process to ensure a comparable follow-up duration. Patients selected for the no-seizure group at certain time points were reused as potential controls at the subsequent sampling time for the seizure group unless seizures had not occurred before then. These procedures were conducted with ccwc function of Epi package (version 2.47.1) in R. The resulting left-truncated data were followed from the index POD until death, retransplantation, 5 years after sampling, or June 2022, whichever came first. If the sampled controls experienced seizures thereafter, they were censored during seizure development for the survival analyses.

Baseline characteristics of the recipient and donor and transplant factors were retrieved from the institutional LT database. In addition, data on pretransplant cerebrovascular accident (CVA), pretransplant dialysis, and intraoperative continuous renal replacement therapy (CRRT) among the sampled cohort were collected from electronic medical records. Pretransplant sodium (Na) level was categorized according to severity, and delta Na level was calculated as the difference between Na levels measured closest to the time of surgery among those recorded before and after the LT.

After applying a nested case-control design, various laboratory results at the index POD of each patient were merged with those of a matched cohort. In addition, the use of each immunosuppressant was defined as prescriptions over 50% of post-transplant days before the index POD. To analyze the association between tacrolimus exposure and seizures, tacrolimus trough levels before the index POD were estimated. Data regarding surgical complications, rejection, and sepsis before the index POD were also collected. Graft loss was defined as patient death or retransplantation.

According to their normality, data were presented as a number (percentage) for categorical variables and as a median (interquartile range [IQR]) for continuous variables. The chi-square test or Wilcoxon rank-sum test was used, as necessary, to compare the seizure group with the control group. Graft survival following the index POD was compared between the two groups using the Kaplan–Meier curve and log-rank test to examine the clinical impact of seizures. The relationship between seizures and graft survival was assessed using univariate and multivariate Cox regression analyses. For more robustness, graft survival was compared between matched population upon deciles of propensity score (PS) which was calculated using all baseline variables [12]. The matching was considered adequately balanced when the standardized mean differences between the groups were below 0.1 [13].

Risk factors for posttransplant seizures were examined using logistic regression analysis. Significant continuous variables were entered into the model after categorization with cutoff values determined by the Yuden Index, which were analyzed using the pROC package of R software [14, 15]. Considering the relatively small number of events compared to the number of variables we intended to evaluate, multivariable logistic and Cox regression models were created using the backward stepwise method. All analyses were performed using the R statistical package, version 4.2.0 for macOS, ¹ with the threshold for significance set at p < 0.05.

This study was performed in accordance with the Declaration of Helsinki and Declaration of Istanbul and was approved by the Institutional Review Board at Severance Hospital, Yonsei University Health System (IRB No. 4-2023-1567). Informed consent was not required because of the study’s retrospective design.

As shown in Table 1, the age and sex were similar in the seizure and non-seizure groups. The seizure group showed lower BMI than did the control group (22.6 [IQR: 20.7–24.5] kg/m² vs. 23.9 [IQR: 22.0–26.3] kg/m², p = 0.007). There were no differences in the incidence of hypertension, diabetes mellitus, or cardiovascular disease between the groups. The frequency of the underlying liver diseases was not statistically different between the groups; however, the seizure group had a higher proportion of patients with alcoholic liver disease than did the control group (39.3% vs. 24.6%). Pretransplant model for end-stage liver disease (MELD) score (23 [IQR: 15–32] vs. 15 [IQR: 10–24], p < 0.001) was higher, and pretransplant intensive care unit (ICU) stay (18.0% vs. 4.6%, p < 0.001) was more frequent in the seizure group than in the control group. Encephalopathy before LT was more severe in the seizure group than in the control group (p = 0.001). The seizure group showed a higher proportion of deceased donor LT (49.2% vs. 31.5%, p = 0.005) and an advanced donor age (47 [IQR: 32–54] years vs. 37 [IQR: 27–47] years, p = 0.005) than did the control group. The seizure group received more red blood cell transfusion (2.7 [IQR: 1.2–4.5] vs. 1.2 [IQR: 0.6–2.4] L, p < 0.001), had more pretransplant CVAs (6.6% vs. 1.0%, p = 0.017), and received more pretransplant dialysis (23.0% vs. 7.5%, p = 0.001) and intraoperative CRRT (36.1% vs. 7.5%, p < 0.001) than did the control group. Severe pretransplant hyponatremia was observed in the seizure group (9.8% vs. 2.6%, p = 0.007), and delta Na (5 [IQR: 4–7] vs. 3 [IQR: 2–6], p < 0.001) was also higher in the seizure group than in the control group.

At matched index POD, the seizure group showed higher total bilirubin (3.1 [IQR: 1.1–8.3] mg/dL vs. 1.3 [IQR: 0.8–2.6] mg/dL, p < 0.001), blood urea nitrogen (27.8 [IQR: 18.6–47.3] mg/dL vs. 20.3 [IQR: 14.5–31.7] mg/dL, p < 0.001), and glucose (156 [IQR: 134–184] mg/dL vs. 138 [IQR: 113–181] mg/dL, p = 0.014, Table 2) levels than did the control group. Albumin level was lower in the seizure group than in the control group (3.1 [IQR: 2.9–34] mg/dL vs. 3.4 [IQR: 3.1–3.7] mg/dL, p < 0.001). Hemoglobin level (9.5 [IQR 8.3–10.5] g/dL vs. 10.1 [IQR 8.9–11.3] g/dL, p = 0.004) and platelet count (63 [IQR 45–126] × 10³/μL vs. 110 [67–165] × 10³/μL, p < 0.001) were also lower in the seizure group than in the control group. The use of each immunosuppressant and tacrolimus trough level were similar between the groups in terms of the mean, standard deviation, maximum variance, and coefficient of variance. The rates of post-LT complications, such as rejection, bile duct complications, vascular complications, and reoperation, prior to the index POD were also similar between the groups, except for sepsis, which was higher in the seizure group than in the control group (16.4% vs. 6.9%, p = 0.029).

Among 366 matched population, 86 patients (23.5%) experienced graft loss (85 death and 1 retransplantation) within 5 years after index POD. In the Kaplan-Meier analysis, graft survival rate after the index POD was significantly lower in the seizure group than in the control group (63.9%, 56.4%, and 50.6% at 1, 3, and 5 years, respectively, in the seizure group vs. 87.5%, 81.4%, and 78.2% at 1, 3, and 5 years, respectively, in the no-seizure group, p < 0.001, Figure 2). In uni- and multivariable Cox regression models, post-LT seizure was an independent risk factor for graft loss in the matched cohort (hazard ratio [HR]: 2.04, 95% CI: 1.24–3.33, Supplementary Table S1). When the seizure group was matched with those who did not experience seizure on PS, hazardous effect of seizure on the graft survival was also observed (Supplementary Figure S2), although all variables were balanced between two groups (Supplementary Table S2).

Infection, graft failure, and HCC recurrence were three most common causes of death in both groups (Supplementary Figure S3). Among them, infectious death was significantly higher in the seizure group than the control group (18.0% vs. 5.9%, p = 0.003, Supplementary Table S3).

In uni- and multivariable logistic regressions (Table 3), the independent risk factors for post-LT seizures were BMI <23 kg/m² (odds ratio [OR]: 2.76, 95% CI: 1.39–5.61), donor age ≥45 years (OR: 3.02, 95% CI: 1.46–6.38), intraoperative CRRT (OR: 3.81, 95% CI: 1.53–9.82), and delta Na ≥4 mmol/L (OR: 5.38, 95% CI: 2.55–12.3). Among laboratory values at index POD, total bilirubin level ≥2.5 mg/dL (OR: 2.50, 95% CI: 1.21–5.24) and albumin level <3.5 mg/dL (OR: 6.75, 95% CI: 2.13–28.4) emerged as independent risk factors for post-LT seizures. When we performed sensitivity analysis only including seizures without structural abnormality, same risk factors were observed (Supplementary Table S4).

In the subgroup without pre-LT hyponatremia (≥135 mmol/L), the incidence of seizures was significantly higher when delta Na was ≥4 mmol/L than when it was <4 mmol/L (21.2% vs. 7.4%, p = 0.002; Table 4). In the subgroup with pre-LT hyponatremia (<135 mmol/L), post-transplant seizures occurred more frequently when the delta Na was ≥4 mmol/L than when it was <4 mmol/L, although this was not significant owing to the small effect size (30.2% vs. 13.6%, p = 0.215). Delta Na ≥4 mmol/L was significantly associated with post-transplant seizure after adjustment of other risk factors in both subgroups with (OR: 5.16, 95% CI: 2.12–14.1) or without pre-LT hyponatremia (OR: 11.2, 95% CI: 1.79–120, Supplementary Table S5).