Distinct virologic trajectories in chronic hepatitis B identify heterogeneity in response to nucleos(t)ide analogue therapy

Abstract

Background & aims: The dynamics of HBV viral load (VL) in patients with chronic hepatitis B (CHB) on nucleos(t)ide analogue (NA) treatment and its relationship with liver disease are poorly understood. We aimed to study longitudinal VL patterns and their associations with CHB clinical outcomes.

Methods: Utilising large scale, routinely collected electronic health records from six centres in England, collated by the National Institute for Health and Care Research Health Informatics Collaborative (NIHR HIC), we applied latent class mixed models to investigate VL trajectory patterns in adults receiving NA treatment. We assessed associations of VL trajectory with alanine transaminase, and with liver fibrosis/cirrhosis.

Results: We retrieved data from 1,885 adults on NA treatment (median follow-up 6.2 years, IQR 3.7-9.3 years), with 21,691 VL measurements (median 10 per patient, IQR 5-17). Five VL classes were identified from the derivation cohort (n = 1,367, discrimination: 0.93, entropy: 0.90): class 1 'long term suppression' (n = 827, 60.5%), class 2 'timely virological suppression' (n = 254, 18.6%), class 3 'persistent moderate viraemia' (n = 140, 10.2%), class 4 'persistent high-level viraemia' (n = 44, 3.2%), and class 5 'slow virological suppression' (n = 102, 7.5%). The model demonstrated a discrimination of 0.93 and entropy of 0.88 for the validation cohort (n = 518). Alanine transaminase decreased variably over time in VL-suppressed groups (classes 1, 2, 5; all p <0.001), but did not significantly improve in those with persistent viraemia (classes 3, 4). Patients in class 5 had twofold increased hazards of fibrosis/cirrhosis compared with class 1 (adjusted hazard ratio, 2.00; 95% CI, 1.33-3.02).

Conclusions: Heterogeneity exists in virological response to NA therapy in CHB patients, with over 20% showing potentially suboptimal responses. Slow virological suppression is associated with liver disease progression.

Impact and implications: Treatment recommendations for people living with chronic hepatitis B virus (HBV) infection are becoming less stringent, meaning that more of the population will be eligible to receive therapy with nucleos(t)ide analogue agents. We explored outcomes of HBV treatment in a large UK dataset, describing different responses to treatment, and showing that the viral load is not completely suppressed after 1 year in about one in five cases, associated with an increased risk of liver complications. As treatment is rolled out more widely, patients and clinicians need to be aware of the potential for incomplete virologic responses. The findings can support the identification of high-risk individuals, improve early fibrosis and cirrhosis prediction, guide monitoring and preventive interventions, and support public health elimination goals.

Keywords: Antiviral treatment; Cirrhosis; HBV; Health Informatics Collaborative (HIC); Latent class mixed models; Liver fibrosis; Longitudinal; Nucleotide analogue; Viral load.

Graphical abstract

Fig. 1

Individual trajectories of HBV VL for adults with chronic HBV infection on NA treatment divided into five classes using latent class mixed modelling. Data presented for derivation cohort (n = 1,367 individuals). Dots represent the raw values of VL linked by coloured lines for each individual, and solid black lines with shading area represent the predicted VL trajectory patterns with 95% CIs, which were computed by a Monte Carlo approximation of the posterior distribution of the predicted values. HBV, Hepatitis B Virus; NA, nucleos(t)ide analogue; VL, viral load.

Fig. 2

Schematic view of division of adult population with chronic HBV infection treated with NA therapy into five viral load classes. Data presented for 1,885 individuals by pooling the derivation and validation cohorts. Proportion of the population in each class and characteristics of each class are shown. Age in years is given as median [IQR]. Anti-HBe, Anti-Hepatitis B e-antigen; HBV, hepatitis B virus; HBeAg, Hepatitis B e-Antigen; NA, nucleos(t)ide analogue; VL, viral load.

Fig. 3

Multivariate Cox proportional-hazards model investigating associations of different VL trajectory patterns with liver disease progression to fibrosis and/or cirrhosis among adults with chronic HBV infection on NA therapy. Forest plot showing the hazard ratios and 95% CIs for the development of liver fibrosis and cirrhosis, which were calculated with Cox proportional-hazards model. ADV, adefovir; ALP, alkaline phosphatase; ALT, alanine transaminase; Anti-HBe, Anti-Hepatitis B e-antigen; AST, aspartate aminotransferase; eGFR, estimated glomerular filtration rate; ETV, entecavir; HBV, hepatitis B virus; HBeAg, Hepatitis B e-Antigen; HR, hazard ratio; IMD, Index of Multiple Deprivation; LAM, lamivudine; NA, nucleos(t)ide analogue; TDF, tenofovir disoproxil fumarate; VL, viral load.

Fig. 4

ROC curves of HBV VL trajectory and other identified predictors for liver fibrosis and cirrhosis. (A) Single identified predictor; (B) combination of different identified predictors. All models were adjusted for other parameters that were included in multivariate analysis. In panel B, ∗∗p <0.01 using the roc.test function with the DeLong’s test method. ALB, albumin; ALP, alkaline phosphatase; AST, aspartate aminotransferase; HBV, hepatitis B virus; PLT, platelets; NA, nucleos(t)ide analogue; ROC, receiver operating characteristic curve; VL, viral load.