Risk factors and clinical prediction models for low-level viremia in people living with HIV receiving antiretroviral therapy: an 11-year retrospective study

Abstract

Objective: This study explores the risk factors for low-level viremia (LLV) occurrence after ART and develops a risk prediction model.

Method: Clinical data and laboratory indicators of people living with HIV (PLWH) at Hangzhou Xixi Hospital from 5 April 2011 to 29 December 2022 were collected. LASSO Cox regression and multivariate Cox regression analysis were performed to identify laboratory indicators and establish a nomogram for predicting LLV occurrence. The nomogram's discrimination and calibration were assessed via ROC curve and calibration plots. The concordance index (C-index) and decision curve analysis (DCA) were used to evaluate its effectiveness.

Result: Predictive factors, namely, age, ART delay time, white blood cell (WBC) count, baseline CD4⁺ T-cell count (baseline CD4), baseline viral load (baseline VL), and total bilirubin (TBIL), were incorporated into the nomogram to develop a risk prediction model. The optimal model (which includes 6 variables) had an AUC for LLV after 1-year, 3-year, and 5-year of listing of 0.68 (95% CI, 0.61-0.69), 0.69 (95% CI, 0.65-0.70), and 0.70 (95% CI, 0.66-0.71), respectively. The calibration curve showed high consistency between predicted and actual observations. The C-index and DCA indicated superior prediction performance of the nomogram. There was a significant difference in CD4 levels between LLV and non-LLV groups during the follow-up time. The dynamic SCR, ALT, TG and BG levels and occurrence of complications differed significantly between the high- and low-risk groups.

Conclusion: A simple-to-use nomogram containing 6 routinely detected variables was developed for predicting LLV occurrence in PLWH after ART.

Keywords: low-level viremia; nomogram; people living with HIV; prediction model; viral load.

Figure 1

Flowchart of all participants in the study. LASSO, the least absolute shrinkage and selection operator; ART, antiretroviral therapy; DCA, decision curve analysis; C-index, concordance index.

Figure 2

LASSO Cox regression plot. (A) Plot of partial likelihood deviance; (B) plot of LASSO coefficient profiles. Each color curve represents the LASSO coefficient profile of a feature against the Log (λ) sequence. The values above the figure represent the numbers of variables included in the model, with the corresponding λ shown on the x-axis. λ, lambda.

Figure 3

Multivariate Cox regression and ROC curves analysis was used to evaluate the risk factors associated with LLV. (A) Forrest plot of multivariate Cox regression analysis. (B) Time-independent ROC curves of the nomogram for 1-, 3- and 5-year overall survival prediction in the PLWH cohort.

Figure 4

Construction of the risk score model with respect to PLWH with LLV. (A) Kaplan–Meier curves of the overall survival of the two LLV risk groups of PLWH with LLV. (B,C) Risk score curves and scatter plots of the risk of LLV among the PLWH cohorts.

Figure 5

Nomogram for evaluation of LLV in PLWH. To use the nomogram, a line was first drawn from each parameter value to the score axis for the score; then the points for all the parameters were added. Finally, a line from the total score axis was drawn to determine the occurrence of LLV on the lower line of the nomogram.

Figure 6

Calibration plots, DCA curves, C-index for the predictive nomogram. (A) Calibration curves for 1-, 3- and 5-year survival depict the calibration of nomogram in terms of the agreement between the predicted probabilities and observed outcomes of PLWH cohort. (B) The results of DCA analysis for the nomogram. (C) Time-dependent C-index of the nomogram model.

Figure 7

Comparisons of the differences in clinical indexes between different groups after ART. The levels of CD4⁺ T cells (A), white blood cells (B), total bilirubin (C) at different follow-up time points in LLV and non-LLV groups. Serum levels of creatinine (D), alanine aminotransferase (E), aspartate aminotransferase (F), total triglycerides (G), total cholesterol (H), and blood glucose (I) at various different follow-up time points in high-risk and low-risk groups. Wilcoxon matched-pairs signed-rank test with two-tailed p-value was used for comparison between groups. ns: p > 0.05, *: p < 0.05, **: p < 0.01, ***: p < 0.001.