Longitudinal changes in estimated glomerular filtration rates in chronic hepatitis B patients treated with Tenofovir Alafenamide vs Entecavir

Study design

This retrospective cohort study included treatment-naïve individuals with CHB receiving TAF or ETV from July 1st, 2019 to December 31st, 2020 at Shenzhen Third People’s Hospital in Guangdong China. CHB diagnosis, treatment status, and relevant clinical and laboratory data were obtained through medical chart reviews.

Study participants

Treatment-naïve CHB individuals (18 years or older) receiving monotherapy with ETV or TAF for over 48 weeks were included if they had at least one follow-up creatine measurement per 48 weeks until August 31st, 2022. This follow-up frequency reflects the variability in clinical follow-up intervals in real-world settings and aligns with the KDIGO 2012 Clinical Practice Guideline for the evaluation and management of chronic kidney disease (CKD) (Kidney Disease: Improving Global Outcomes, KDIGO, 2013), which recommends at least one annual eGFR assessment for CKD monitoring. Additionally, the Chinese guidelines (You et al., 2023) for CHB management suggest that kidney function be tested every 6–12 months.

The criteria for exclusion included: (1) under 18 years old; (2) pregnancy at baseline or during follow-up; (3) decompensated liver disease or liver transplantation; (4) alcoholic liver disease (Varga et al., 2017); (5) co-infection of other hepatitis viruses like hepatitis C and E virus; (6) presence of chronic kidney diseases (defined as: eGFR under 60 mL/min/1.73 m², persistent ≥ 3 months before baseline (Kidney Disease: Improving Global Outcomes, KDIGO, 2013), or documented clinical diagnosis with International Statistical Classification of Diseases and Related Health Problems, 10th Revision codes of N00 to N29); patients with baseline eGFR <60 mL/min/1.73 m² with unknown persistent time are typically referred to nephrologists and receive specific therapy. To minimize confounding due to baseline kidney impairment and treatment, individuals with baseline eGFR <60 mL/min/1.73 m² were excluded; (7) co-infection of other pathogens undergoing treatment, such as human immunodeficiency virus, mycobacterium and parasites; (8) malignancy; (9) severe diseases of other systems: organ transplantation, acute and chronic failure of vital organs, acute events necessitating emergent intervention (e.g., acute ischemic stroke, acute abdomen); (10) death during follow-up; (11) unknown baseline eGFR; (12) missing data on follow-up creatinine: cases with no creatinine measurements during any 48-week interval; (13) irregular treatment of ETV or TAF (Fig. 1).

The institutional review board at Shenzhen Third People’s Hospital approved this research study (Ethical Reference number: 2022-142-02, 2022-142-03). In view of the retrospective design, informed consent was not required and was therefore waived by the institutional review board.

Data collection

Demographics, medical prescriptions, and laboratory information at the initiation of ETV or TAF treatment were extracted from electronic health records, and considered as baseline measurements. All subsequent measurements post-treatment initiation were considered as follow-up data.

Accompanying disease diagnosis: (1) diabetes: fasting serum glucose level of at least 6.1 mmol/L (Society Chinese Diabetes, 2021) or current use of hypoglycemic agents. (2) Hypertension: systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or current usage of antihypertensive medications. (3) Dyslipidemia: low-density lipoprotein cholesterol level of at least 3.4 mmol/L, total cholesterol level of at least 5.2 mmol/L, triglyceride level of at least 1.7 mmol/L (Joint committee issued Chinese guideline for the management of dyslipidemia in adults, 2018), or current use of hypolipidemic agents. (4) Metabolic dysfunction-associated steatotic liver disease (MASLD) was determined by various reliable methods, including imaging techniques (magnetic resonance imaging, computed tomography, and ultrasonography), histological evidence, controlled attenuation parameter, and/or serum diagnostic markers (hepatic steatosis index or fatty liver index).

The diagnosis of cirrhosis was diagnosed through a comprehensive assessment that includes histological evidence, clinical history, imaging findings, endoscopic evaluations, laboratory findings, and signs of portal hypertension. These indicators may encompass a nodular appearance on imaging, splenomegaly, thrombocytopenia (platelets count lower than 100 × 10⁹/L), hypoalbuminemia(serum albumin (ALB) level lower than 35 g/L), a prolonged prothrombin time of more than 3 s over control, presence of varices, and hepatic decompensation symptoms such as ascites, variceal hemorrhage and hepatic encephalopathy (You et al., 2023).

Kidney outcomes

Kidney outcomes were evaluated using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (Levey et al., 2009) and categorized according to the criteria specified by the Kidney Disease: Improving Global Outcomes (KDIGO) (Kidney Disease: Improving Global Outcomes, KDIGO, 2013). CKD progression was characterized by a sustained increase in the CKD stage lasting for a minimum of three consecutive months.

Statistical analysis

Given that the proportion of missing data was less than 10%, missing data were imputed with median values (Fig. S1).

Categorical variables were presented as numbers (%) and compared by Chi-squared test or Fisher exact test. Continuous variables were expressed as means (standard deviations) and compared by t test if they followed a normal distribution; otherwise, they were presented as medians (interquartile ranges) and analyzed by Kruskal-Wallis test.

Factors influencing CKD development over time were explored using univariable Cox analysis. Collinearity among variables was tested using the variance inflation factor (VIF), and any variables with high collinearity (VIF > 10) were excluded from the Cox regression analysis. Creatinine (Cr) and total bilirubin (TB) demonstrated high VIF and were consequently excluded from the Cox analysis. Multivariable Cox analysis was conducted on variables that showed statistical significance (p < 0.5) in the univariable Cox analysis. Proportional hazard assumptions were validated by conducting the proportional hazard test based on Schoenfeld residual using the Cox.zph function in the R package of survival.

To investigate temporal changes in eGFR and the association between change rates in follow-up eGFR and treatment groups, generalized additive mixed model (GAMM) was employed, considering repeated measurements of follow-up eGFR and irregular measurement intervals (Lin & Zhang, 1999). The model fitted nonlinear smooth functions of time for overall and treatment groups, using cubic regression splines, to capture nonlinear eGFR trajectories. To assess potential effect modification by sex and age, interaction terms between treatment groups (TAF vs. ETV) and sex (male vs. female) or age group (≥35 vs. <35 years) were included in the GAMM models.

To quantify temporal changes in eGFR for specific intervals (0–12, 12–24, 24–48, and 48–86 weeks. 86 was the median follow-up duration), average slopes of eGFR (mL/min/1.73 m²/week) were calculated for the treatment groups, using the fitted smooth functions. Then GAMM with smooth curve fitting was used to assess the overall differences in eGFR change between treatment groups, both with and without covariates. Sex- and age-stratified analyses were conducted to investigate variations in eGFR between the two treatments across diverse demographic subgroups.

Sensitivity analyses were conducted to evaluate the stability of the primary results as follows: (1) a piecewise linear mixed model (LMM) was fitted to confirm the 12-week decline rate. This model included fixed effects for group and time (0–12 weeks, 12–24 weeks, 24–48 weeks, 48–68 weeks), as well as random intercepts and slopes for each subject. (2) Multivariable Cox analysis was conducted in the subgroup of eGFR ≥ 90 mL/min/1.73 m² without missing baseline data. (3) GAMM analyses were conducted on the following groups of individuals: those without any missing baseline data; those with eGFR measurements within 12 and 30 weeks, respectively; and those who had at least three eGFR measurements during the periods of 0 to 12 weeks, 12 to 24 weeks, and 24 to 48 weeks.

Statistical significance was determined as p value < 0.05. The analyses were carried out in R software (version 4.1.3).

Baseline characteristics

According to the screening flow (Fig. 1), 466 treatment-naïve individuals with CHB were included with 296 receiving ETV and 170 receiving TAF for statistical analysis (Table 1).

At baseline, the ETV group demonstrated a higher median age (39.5 vs. 36.5, p = 0.019) and a greater proportion of male participants (79.1% vs. 67.6%, p = 0.009) compared to the TAF group. Both groups had statistically similar proportions of comorbidities, including hypertension, diabetes mellitus, dyslipidemia, non-alcoholic fatty liver disease (NAFLD), and compensated cirrhosis (all p > 0.05). Medication usage for hypertension, dyslipidemia, diuretics, and propranolol did not vary between the two groups (p > 0.05), although the ETV group had a higher percentage of hypoglycemic agent usage (6.4% vs. 1.8%, p = 0.040).

Moreover, when compared to the TAF group, the ETV group demonstrated elevated median levels of aspartate aminotransferase (AST) (45.0 vs. 41.0 U/L), TB (15.4 vs. 14.1 µmol/L), direct bilirubin (DB) (4.8 vs. 4.3 µmol/L), and gamma-glutamyl transferase (GGT) (41.0 vs. 34.0 U/L) (all p < 0.05). Conversely, there were no differences in ALT, ALB, Cr, and baseline eGFR (all p > 0.05). The proportion of detectable HBV DNA was similar between the two agents (p = 0.068), while a higher percentage of positive HBeAg was noted in the TAF group (61.8% vs. 48.6%, p = 0.008).

395 patients were in the subgroup of baseline eGFR ≥ 90 mL/min/1.73 m², and 71 were in the subgroup of baseline eGFR between 60 and 90 mL/min/1.73 m² (Table 1). The characteristics of the former subgroup were largely comparable to those of the overall group, with the exception of a non-significant difference in median age (p = 0.074) and the proportion of hypoglycemic drugs (p = 0.171). Within the subgroup of individuals with a baseline eGFR between 60 and 90 mL/min/1.73 m², the average age was significantly lower in the TAF group (44.4 vs. 51.3, p = 0.015), while other variables were found to be statistically similar.

Kidney outcomes during follow-up

The follow-up durations were similar between the TAF and ETV groups in both the entire cohort and subgroups (all p > 0.05) (Table 1). Specifically, in the subgroup of baseline eGFR ≥ 90 mL/min/1.73 m², 14 (9.8%) patients in the TAF group and 35 (13.9%) in the ETV group developed CKD stage 2, with no statistically significant difference observed (p = 0.304). Furthermore, no patients in either treatment group experienced progression beyond CKD stage 2. Univariable Cox analyses revealed that age, sex, AST, ALB, and baseline eGFR were related to CKD development during the follow-up. However, only the male sex (hazard ratio (HR) 2.72; 95% confidence interval (CI) [1.02–7.25]; p = 0.045) and baseline eGFR (HR 0.86; 95% CI [0.82–0.90]; p < 0.001) were found to be independently associated with CKD development in multivariable Cox analysis (Table 2).

Within the subgroup of individuals with baseline eGFR between 60 to 90 mL/min/1.73 m², one patient in the TAF group and two patients in the ETV group experienced an eGFR lower than 60 mL/min/1.73 m² during the follow-up, but none had their eGFR decline for more than 12 weeks or to lower than 30 mL/min/1.73 m².

eGFR dynamic change during follow-up

Within the subgroup of individuals with a baseline eGFR ≥ 90 mL/min/1.73 m², GAMM analysis demonstrated a pattern in which the eGFR initially decreased before stabilizing around week 40 (Fig. 2). The eGFR changes in TAF and ETV groups are presented in Fig. 3. Specifically, the ETV group, exhibited an initial decline in eGFR followed by stabilization, whereas the TAF group experienced a decrease in eGFR initially, followed by a gradual increase after approximately week 30. The eGFR trajectories of the two groups intersected at approximately week 45, indicating differing rates of change between them (Fig. 3). In the initial 0 to 12 weeks, eGFR in the TAF group showed a rapid decline compared to the ETV group (slopes: TAF vs. ETV: −0.38 vs. −0.15, p = 0.015), with a significant difference. From 12 to 24 weeks, the decline in eGFR within the TAF group slowed and became comparable to that of the ETV group (TAF vs. ETV: −0.18 vs. −0.14, p = 0.627). Between 24 and 48 weeks, the eGFR in the TAF group began to recover, while in the ETV group, the eGFR continued to decline (TAF vs. ETV: 0.06 vs. −0.10, p = 0.005). Beyond 48 weeks (from 48 to 86 weeks), the eGFR in the TAF group continued to increase slightly, while it stabilized in the ETV group (TAF vs. ETV: 0.04 vs. −0.03, p = 0.151) (Table 3).

Further analysis revealed that on average, the magnitude of decrease in eGFR from the baseline was lower in the TAF group (β: 0.43, 95% CI [0.16–0.69], p = 0.002) every 12 weeks. The observed disparity remained statistically significant even after adjusting for all variables (model 2, β: 0.40, 95% CI [0.13–0.67], p = 0.004) as well as the variables with p < 0.05 (model 3, β: 0.38, 95% CI [0.11–0.65], p = 0.006) (Table 4). Subsequent sex-stratified analysis revealed significant distinctions between treatment groups among males (all p < 0.05 in three models), while no disparities were evident among females (all p > 0.05 in three models). The age-stratified analysis further demonstrated that individuals aged 35 to 65 years exhibited significant variances between TAF and ETV groups (p < 0.05 in three models), whereas those under 35 years did not display any differences (all p > 0.05 in three models).

The interaction between treatment groups and sex (male vs. female) was not statistically significant (all p > 0.05 in three models) (Table S1). The interaction between treatment groups and age (≥ 35 vs. <35 years) demonstrated a significant result in the unadjusted model (β: 3.81, 95% CI [0.10–7.53], p = 0.045). However, this effect showed no significance after adjustment (p > 0.05 in adjusted models) (Table S1).

In the subgroup of baseline eGFR between 60 to 90 mL/min/1.73 m², GAMM analysis demonstrated a linear change in eGFR (Fig. S2). The rate of eGFR decline varied between the two treatment groups (Fig. S3), but it did not reach a statistically significant difference (β: 0.34, 95% CI [−0.31–1.00], p = 0.303) (Table 4).

Sensitivity analyses

In the sensitive analyses, the piecewise LMM showed similar findings as the GAMM (Table S2). Among the individuals with complete baseline data, the multivariable Cox regression indicated similar results that sex and baseline eGFR were independently associated with CKD development in the subgroup of baseline eGFR ≥ 90 mL/min/1.73 m² (Table S3).

GAMM analyses also demonstrated consistent patterns and overall differences of eGFR changes between TAF and ETV groups in the following scenarios: (1) individuals with complete baseline data (Figs. S4–S7, Table S4); (2) individuals who had no missing eGFR measurements in the first 12 (Figs. S8–S9, Table S5) and 30 weeks (Figs. S10–S11, Table S6); (3) individuals with at least three eGFR assessments over the periods of 0 to12, 12 to 24 and 24 to 48 weeks (Figs. S12–13, Table S7). However, among individuals without missing eGFR in the first 30 weeks, no significant difference was observed in the male subgroup.