Persistent low-level viremia (pLLV), commonly defined here as two consecutive plasma viral load (VL) measurements of 50–999 copies/mL, occurs in a subset of people living with HIV (PWH) despite antiretroviral therapy (ART). pLLV has been associated with adverse outcomes in prior work, including increased reservoir complexity, impaired immune recovery, and elevated risks of virologic failure and non-AIDS events. Definitions and management recommendations vary among guidelines, and resistance testing is often unsuccessful at low VL, complicating regimen change decisions. Evidence on whether switching ART after pLLV improves outcomes is mixed, and real-world data from China are limited. This multicenter retrospective cohort study compared virologic outcomes for patients whose treating clinicians either maintained or switched ART after confirmed pLLV.
This was a multicenter retrospective observational study using an LLV database from seven medical institutions in China. Records from June 1, 2018, to December 31, 2023, were screened. Eligible adults were ≥18 years old, had received ART for ≥6 months, and had two consecutive VLs between 50 and 999 copies/mL separated by at least 7 days. Exclusion criteria included pregnancy or lactation, recent ART interruption, severe hepatic or renal dysfunction, severe opportunistic infections or malignancies at baseline, or lack of follow-up VLs within prespecified windows. The ethics committee of Beijing Youan Hospital approved the protocol and waived informed consent because data were deidentified.
Participants were classified into a control group (no regimen modification after pLLV confirmation) or a switch group (ART modified after pLLV confirmation during routine care). The index date was pLLV confirmation for controls and the date of switching for the switch group. Follow-up VL and CD4 measurements were assessed at 24, 48, and 96 weeks after the index date using a ±4-week visit window; the measurement closest to each target visit was used when multiple values were available. Adherence counseling was reinforced for all patients, but objective adherence measures were inconsistently recorded and therefore not included as covariates.
The primary outcome was viral suppression (VL <50 copies/mL) at week 48. Secondary outcomes included suppression at weeks 24 and 96, longitudinal VL (log10-transformed) and CD4 trajectories, and selected laboratory safety indices at 12 and 24 months when available.
Sample-size calculations, based on effect sizes from a prior study, targeted at least 49 participants per group after accounting for missingness. Categorical variables were summarized as counts and percentages; continuous variables as mean (SD) or median (IQR) as appropriate. Group comparisons used χ² or Fisher’s exact tests for categorical variables and t tests or Wilcoxon rank-sum tests for continuous variables. Logistic regression estimated odds ratios (ORs) and risk differences (RDs) for switching versus maintenance, with adjusted models including age, sex, baseline VL, baseline CD4 count, and current ART regimen category. Linear regression was used for continuous outcomes. Primary analyses used available data; multiple imputation by chained equations (MICE) was performed as a sensitivity analysis for missing data. An inverse probability of treatment weighting (IPTW) analysis was conducted to address potential treatment-selection bias; propensity scores included demographic and baseline clinical covariates and ART history.
A total of 162 participants met inclusion criteria: 97 in the control group and 65 in the switch group. At week 48—the primary outcome time point—viral suppression (VL <50 copies/mL) was observed in 64.37% of the control group and 68.52% of the switch group. The adjusted odds ratio (aOR) for regimen switching was 1.40 (P = 0.427), and the adjusted risk difference (aRD) was 0.07 (P = 0.423), indicating no statistically significant association between switching and improved suppression at week 48.
At week 24, suppression rates were 40.98% in controls and 50.00% in switch patients. The aOR was 1.34 (P = 0.521) and aRD 0.07 (P = 0.518), also not statistically significant. Results at week 96 and for laboratory safety indices were reported in the manuscript; longitudinal summaries showed modest changes over time without significant between-group differences.
Longitudinal analyses found that HIV RNA levels decreased modestly and CD4 cell counts increased modestly over time in both groups. Rank-based nonparametric longitudinal models were used to assess these trends. No significant differences between the switch and control groups were observed in the trajectories of viral load (log10-transformed) or CD4 counts.
To assess robustness to missingness, multiple imputation by chained equations (MICE) was performed under a missing-at-random assumption. To address potential treatment-selection bias, an IPTW sensitivity analysis was conducted using propensity scores that included age, sex, baseline HIV RNA log10, baseline CD4 count, current ART class, history of PI- or INSTI-based regimens, and prior resistance testing. Stabilized weights were applied and covariate balance before and after weighting was assessed. The manuscript reports these sensitivity analyses but does not adjust for medical center in primary models because some centers had few participants.
In this retrospective real-world cohort from seven Chinese centers, regimen switching after confirmation of pLLV was not associated with improved short-term virologic suppression at 24 or 48 weeks. Both groups experienced modest viral declines and modest CD4 gains over time, and no between-group differences in these longitudinal trends were observed. The study highlights challenges in managing pLLV, including inconsistent availability of objective adherence measures and limited ability to perform resistance testing at low VL levels. Previous studies have reported mixed findings regarding the benefit of switching for LLV, and this study provides real-world evidence from China where data had been limited.
The study is retrospective and observational, limiting causal inference. Objective adherence measures were inconsistently recorded and therefore not included as covariates. Some centers contributed few participants and center-level effects were not adjusted for in primary models. Missing follow-up measurements required imputation and sensitivity analyses. Resistance testing at low VL was often not feasible and therefore generally unavailable to guide switching decisions.
Among adults with pLLV in this multicenter Chinese cohort, switching ART after pLLV confirmation was not associated with improved short-term virologic suppression. HIV RNA and CD4 measures changed modestly over time in both the switch and control groups without significant between-group differences. The data and supporting information are available in the manuscript and its supplemental files.