Modelling the impact of protein-kinase R allelic variant on HIV biomarkers trajectories by means of latent class mixed models

Abstract

This paper is based on a retrospective longitudinal study on people living with HIV under antiretroviral treatment (ART) where allelic variants (either heterozygous CT genotype or homozygous CC genotype) have been evaluated at position −168 of the promoter region of the protein kinase R (−168/PKR). In general, antiviral effects of interferon are partially mediated by a RNA-dependent protein kinase (PKR) that, once activated, inhibits protein synthesis. Indeed, activation of PKR response can inhibit HIV replication. To explore the role of allelic variants in shaping dynamics of commonly monitored HIV biomarkers, CD4 cells, CD8 cells and HIV-load were modelled within a latent class mixed model (LCMM) to account for participants’ heterogeneity over time. The estimated models identified two sub-groups from CD4 and HIV-load dynamics, revealing better outcomes for subgroups of participants with the heterozygous CT genotype. Heterozygous CT subjects in one of the two identified subgroups exhibited higher increase of CD4 cells and more marked decrease of HIV-load, over time, with respect to the homozygous CC subjects assigned to the same group.

Classification of virologic trajectories during nucleos/tide analogue treatment of hepatitis B virus (HBV) infection

“…Most patients in class 3 (VL persistent at moderate levels) were anti-HBe positive (84.6%), while 94.7% in class 4 were HBeAg positive at baseline, which may account for differences in VL levels. Patients in class 2 had a significantly higher median ALT level (median 64, [IQR 37, 125] IU/L) at baseline, followed by class 5 (median 41, [32, 69] IU/L), compared to other classes (29 [20,47] IU/L for class 1, 32 [21,41] IU/L for class 3, 35 [22,50] IU/L for class 4), p<0.001.…”

“…Latent class mixed modeling, a contemporary unsupervised approach, has already demonstrated clinical relevance in various disease areas, including cardiovascular disease (18), chronic kidney disease (19), human immunodeficiency virus (HIV) (20), and coronavirus disease (caused by SARS-CoV-2) (21), by disaggregating sub-phenotypes. We used this approach to avoid any prior assumptions about HBV VL trajectories and to take an unbiased approach to identify distinct on-treatment groups.…”