HIV Low-Level Viremia is Not Related to Subsequent Virological Failure Under NNRTI-Based Regimens: A Multicenter Retrospective Study in China

Introduction

The HIV plasma viral load is one of the most important indicators for monitoring responses to antiretroviral therapy (ART).^1–4 After therapy, most patients reach and maintain undetectable viral loads. However, some patients still exhibit persistent and detectable viremia above the lower limit of detection (LLOD) despite the use of potent antiretroviral regimens. This phenomenon is referred to as low-level viremia (LLV). Another type of viremia, characterized by one or more temporary and intermittent increases in the HIV viral load during ART, is known as a blip.

Although the threshold for defining viremia varies between studies owing to different local health policies, numerous studies have confirmed the association between viral loads of 200–1000 copies/mL and clinical progression.^5–7 The updated guidelines define virological failure (VF) as 200 copies/mL or more.^1,2,4 However, the clinical significance of viremia levels of < 200 copies/mL remains controversial. A retrospective analysis of a large European study indicated that both blips and LLV of 50–200 copies/mL during ART are associated with an increased risk of subsequent VF,⁸ which aligns with previous evidence^9–15 but contradicts others.^16–19 For example, a Spanish cohort of 5986 patients reported no association between LLV and AIDS events, death, or virological failure.¹⁶ There is also ongoing debate regarding the role of drug resistance mutations (DRMs) in LLV, particularly within the range of the LLOD to 200 copies/mL, where low-level viral replication makes resistance detection challenging.²⁰ Limited evidence suggests that the rate of DRMs in patients with LLV is comparable to that in virologically suppressed individuals.⁸ Moreover, several studies have indicated that the majority of LLV patients do not exhibit resistance mutations to the antiretroviral drugs they are currently receiving, prompting a reevaluation of the relationship between LLV and future virological failure.

Here, we conducted a retrospective analysis of six cohorts of people living with HIV (PLWH) in China, where non-nucleoside reverse transcriptase inhibitor (NNRTI)-based ART regimens are predominantly used, to investigate whether blip or LLV is associated with an increased risk of virological and virological failure. Additionally, we retrospectively analyzed drug resistance mutations at baseline and during LLV episodes to assess the role of resistance mutations in the occurrence of LLV.

Methods

Cohorts and Population

This retrospective, multicenter, observational study was based on six HIV cohorts, including five prospective nationwide HIV cohorts from the National Free Antiretroviral Treatment Program of the 13^th Five-Year Plan and one clinical HIV cohort of Peking Union Medical College Hospital (PUMCH). Five nationwide HIV cohorts were established from 2008 to 2017 and mainly enrolled adults aged 18 to 65 years who were ART-naïve. All patients received two nucleoside reverse transcriptase inhibitors (NRTIs) + 1NNRTI ART regimen after enrollment and were followed up for a minimum of 2 years with visits scheduled every 24 or 48 weeks, following the cohort design (details in Supplementary Material 1). The PUMCH cohort was an observational cohort that enrolled all outpatients at the HIV center of PUMCH between 2003–2022. Follow-up visits were scheduled every 12–48 weeks based on clinical needs.

We extracted data from the databases of the aforementioned cohorts, including PLWH aged 18–65 years who were ART-naïve at baseline. Patients were excluded if they received less than 24 weeks of ART, had fewer than two documented viral loads and CD4+ T cell count measurements, or initiated ART during the acute infection stage due to the significant difference in VF between immediate and deferred initiation of ART.^21,22 Follow-up began on the date of the first virological suppression or at 48 weeks after ART. Patients were also excluded if their total follow-up duration was less than 48 weeks. Follow-up ended upon virological failure, loss to follow-up for more than four years, or administrative censoring (August 20, 2022).

Definitions

All patients were classified into three groups based on their viremia pattern during follow-up: virological suppression (VS), virological blip (VB), and low-level viremia (LLV). Blips were defined as a single viral load between the LLOD (50 or 20 copies/mL, depending on the local laboratory) and 399 copies/mL, followed by a subsequent viral load below the LLOD. LLV was defined as two or more consecutive viral loads between the LLOD and 200 copies/mL. Persistent viral loads below the LLOD or results reported as “target not detected” were considered virological suppression. Additionally, a single viral load rebound that did not meet the criteria for virological failure was classified as virological suppression if no subsequent measurements were available after administrative censorship. If an individual experienced both blips and LLV during the follow-up, they were classified as having LLV to reflect the highest historical viremia exposure. Virological failure was defined as a single viral load of ≥400 copies/mL or two or more consecutive viral loads between 200 and 399 copies/mL.

HIV-1 Drug Resistance Testing

Cryopreserved plasma samples were thawed and concentrated by ultracentrifugation at 20,000 g for 2h at 4°C. Nucleic acids were extracted from the resulting pellet using a Viral RNA Extraction Kit (Guangzhou Life Technologies, Daan Diagnostics Co., Ltd). Sequences spanning the reverse transcriptase and protease genes were amplified by Sanger sequencing, and the resulting products were analyzed on a 3500XL DX Genetic Analyzer (Applied Biosystems, USA). The experimental procedures were described in detail as previously reported.²⁰

HIV-1 drug resistance mutations (DRMs) and resistance profiles were interpreted using the Stanford HIV Drug Resistance Database algorithm (version 9.0). The genotypic sensitivity score (GSS) was calculated for each patient according to the latest interpretation guidelines provided by the Stanford HIV Drug Resistance Database (https://hivdb.stanford.edu/). For each antiretroviral drug, a GSS of 1 was assigned if there was no resistance or only potential low-level resistance. A score of 0.5 was assigned for drugs showing low-level or intermediate resistance and a score of 0 was assigned in cases of high-level resistance. The total GSS for each treatment regimen was obtained by summing the scores of all included drugs.

HIV-1 subtypes and circulating recombinant forms (CRFs) were identified using the COMET tool (http://comet.retrovirology.lu). Phylogenetic analysis was conducted using the MEGA software (version 7.0, USA), and the robustness of the phylogenetic trees was evaluated using bootstrap analysis.

Statistical Analysis

The baseline characteristics of the participants are summarized as means with standard deviations for normally distributed data and as medians with interquartile ranges for non-normally distributed data. Pearson’s χ²-test was used for categorical variables, whereas the Kruskal–Wallis test was used for continuous variables. Survival analyses assessing the risk of virological failure across different viremia categories were conducted using Kaplan-Meier curves and Cox proportional hazards regression models. The adjusted Cox models accounted for potential confounders, including age (years, modeled as a continuous linear variable), sex (male/female), baseline CD4+ T cell count (cells/μL, modeled linearly), and baseline viral load (log₁₀ copies/mL, modeled linearly). Schoenfeld and Martingale residuals were examined to verify the proportional hazards assumption and assess model adequacy. All statistical analyses were performed using R programming language (version 4.4.2; R Foundation for Statistical Computing, Vienna, Austria).²³

Results

Baseline Characteristics

Our study database included 2655 participants from five nationwide cohorts and 1394 participants from the PUMCH cohort (Figure 1). Of the 4049 individuals extracted from the database, 1532 (37.8%) met inclusion criteria. Of these, 992 (64.8%), 374 (24.4%), and 166 (10.8%) were categorized into VS, VB, and LLV groups, respectively (Table 1). The majority of participants in all three groups were male, with a median age of 33 years, with no significant differences between the groups. The median baseline CD4+ T cell counts were significantly lower in the VB group (237 cells/μL) and LLV group (218 cells/μL) groups than in the VS group (266 cells/μL; p < 0.001). Median baseline viral loads also differed significantly across the three groups: 4.6 log₁₀ copies/mL in the VS group, 4.7 log₁₀ copies/mL in the VB group, and 4.9 log₁₀ copies/mL in the LLV group (p < 0.001). Additionally, participants were compared based on whether ART was initiated before or after 2012, as China’s National Free Antiretroviral Therapy Program provided only NNRTI-based first-line regimens for PLWH with CD4⁺ T cell counts < 350 cells/μL prior to 2012. The details regarding the co-infections and initial ART regimens are summarized in Table 1. Most participants had their baseline plasma samples stored for drug resistance testing; however, ART initiation did not necessarily delay the test results.

Table 1 Clinical Characteristics at Baseline

Figure 1 Enrollment flowchart of study population.

Association of Virological and Immunological Failure with Viremia

Over a total follow-up period of 5452 person-years, 48 virological failure (VF) events were documented: 31 in the VS group, 10 in the VB (Blip) group, and 7 in the LLV group. The median follow-up duration were 2.3 years. Among participants in the LLV group, the median viral load during the LLV period was 57 copies/mL (interquartile range [IQR]: 38–105), lasting a median of 168 and occurring a median of 591 days after ART initiation. For those in the VB group, the median viral load at the time of blipping was 41 copies/mL [IQR: 27–74 copies/mL], with a median of 498 days after ART initiation.

After adjusting for age and sex, neither the VB group nor the LLV group demonstrated a statistically significant increase in the risk of VF compared to the VS group (Figure 2A; VB, adjusted hazard ratio [aHR] = 0.6, 95% confidence interval [CI]: 0.3–1.2, p = 0.15; LLV, aHR = 0.9, 95% CI: 0.4–2.0, p = 0.760). This association remained nonsignificant after further adjustment for baseline CD4+ T cell count and baseline viral load (Supplementary Table 1; VB, aHR = 0.5, 95% CI: 0.2–1.2, p = 0.116; LLV, aHR = 0.7, 95% CI: 0.3–1.9, p = 0.474). After excluding 272 participants with missing baseline data, the number of VF events reduced to 37. Given the potential impact of shorter follow-up intervals on the detection of viremia and VF events, we compared the frequency of follow-up visits among the three groups. The analysis showed an average annual visit rate of 2.2 visits per patient, with no significant differences between the groups (Figure 2B; p = 0.436).

Figure 2 Survival curves of patients in virological failure. (A) Survival curves for virological failure defined as one viral load of ≥400 copies/mL or ≥2 consecutive viral loads of 200–399 copies/mL. (B) Frequency of visit in VS, blip and LLV groups. Abbreviations: aHR, adjusted hazard ratio; VS, virological suppression; LLV, low-level viremia.

Drug Resistance Mutation in LLV

Among 166 patients with LLV, paired plasma samples from both baseline and LLV episodes were retrieved from 70 individuals from the specimen repository. An ultracentrifugation-based method was used to process plasma collected at both time points. All baseline samples underwent successful PCR amplification and sequencing, whereas only 17 LLV episode specimens yielded detectable DRMs and valid genotypic results, likely due to the low viral load during these episodes. The baseline characteristics of these 17 LLV patients were comparable to those of the remaining LLV cohort (Supplementary Table 2). All patients received an 2NRTIs+1NNRTI initial ART regimen. Among the 17 patients with successful LLV genotyping, 11 (64.7%) were classified as CRF01_AE (Table 2). Notably, three patients (P04, P15, and P17) developed newly emerging DRMs; however, none of them experienced VF. All patients maintained a GSS score of 3 during LLV episodes. By contrast, P11, the only patient who experienced VF, showed no detectable DRMs during LLV. Phylogenetic analysis of 34 sequences from the baseline and LLV episodes revealed close genetic relatedness between the paired sequences in all patients, except for P15, P16, and P17. These results were consistent with the corresponding DRM profiles (Figure 3). Overall, the association between emergent DRMs and VF during LLV remains unclear.

Table 2 Genotypic Drug Resistance and GSS at Baseline and LLV Episode

Figure 3 Phylogenetic tree of pol sequences. P01B represented pol sequence at baseline from patient 01 and P01V represented pol sequence at LLV duration from patient 01. The numbers on the line represent genetic distance. The numerical values at the nodes of the phylogenetic tree represent the bootstrap confidence levels following 100 replicates. Four reference sequences were utilized to denote distinct subtypes: CRF01_AE (GenBank: U54771), CRF_BC (GenBank: AF286226), B subtype (GenBank: K03455), and C subtype (GenBank: AF067154). The four color-coded blocks correspond to the subtypes classified based on the pol sequences. The dashed lines indicate the three patients with newly emerged drug resistance mutations.

Discussion

In this study, based on data from large-scale, long-term, multicenter cohorts in China, we found no significant association between LLV or blips and increased risk of VF. A retrospective analysis of resistance mutations at the baseline and during LLV episodes in 17 patients revealed no substantial emergence of new DRMs. Moreover, the viral strains in these patients remained susceptible to the ART regimens used during LLV episodes. These findings suggest that in China, a resource-limited setting where NNRTI-based ART remains the predominant first-line regimen, LLV ranging from the lower limit of detection to 200 copies/mL does not appear to be definitively associated with VF.

Our findings differ from those reported by other centers in developed area,^8–15 some of which included larger sample sizes and longer follow-up periods. However, we believe that our study is the first real-world study in a resource-limited setting, where > 80% of PLHW received NNRTI-based regimens as the initial treatment based on the National Free Antiretroviral Treatment Program. The difference between the initial ART regimens may be attributed to the lack of an association between LLV and VF.

Previous studies have shown that among individuals with baseline viral loads exceeding 5 log₁₀ copies/mL, NNRTI-based regimens take longer to achieve virological suppression and are associated with a higher risk of VF.^24,25 To minimize the misclassification of patients whose VL was still declining as LLV cases, we extended the observation window for LLV onset to 48 weeks after ART initiation rather than 24 weeks. Nevertheless, this adjustment does not fully eliminate a key limitation. Our study found that over 60% of patients with LLV were infected with the CRF01_AE subtype. Several epidemiological studies conducted in China have indicated an association between CRF01_AE and pretreatment resistance to NNRTI-based antiretroviral therapy.^26,27 This suggests that in regions where NNRTI-based regimens are commonly used as first-line treatment, individuals harboring CRF01_AE may experience rapid virological rebound once a secondary resistance mutation develops. Consequently, the transient LLV phase may go undetected under routine monitoring, potentially leading to an underestimation of the true prevalence of LLV.

We observed that patients with blips or LLV in our cohort did not undergo follow-up more frequently than those with sustained virological suppression, which reflects the relatively insufficient frequency of follow-up for patients with LLV. As of 2019, more than 80% of HIV-infected individuals in China had initiated ART with NNRTI-based regimens.²⁸ For this population, enhancing the frequency of clinical follow-up and reducing the interval between VL monitoring sessions may be particularly beneficial.

It is still controversial whether low-level viremia originates from cyclical viral replication, the release of latent viruses, or intermittent viral escape from sanctuary.^29–38 The emergence of drug-resistant mutations is a potential marker for cyclical viral replication. Despite numerous studies indicating a relationship between LLV and virological failure, ambiguous and inconsistent definitions of LLV have undermined the clarity and interpretability of their findings. Substantial causal evidence between LLV and drug resistance mutations is still lacking,^8,39,40 as echoed in our study. These findings collectively suggest that, although LLV is associated with VF, the emergence of drug-resistant mutations may not be the major driver.

Our findings suggest that maintaining previous ART regimens and intensifying viral load monitoring is currently the best management strategy for LLV, especially in middle- or low-income regions using NNRTI-based ART regimens. Consideration must be given to further drug resistance testing, which is dependent on the baseline viral load of patients, ART regimen employed, and success rate of resistance testing. The necessity of drug resistance testing should be further discussed in regions where drug resistance testing is less accessible and relatively expensive.

This study had some limitations. First, most patients in this study were enrolled from five prospective clinical trials, which limited modifications to ART regimens unless virological failure occurred. Second, our database lacks specific quantitative measures of medication adherence, a known factor associated with the occurrence of LLV. Owing to data limitations, we failed to use all-cause death and AIDS-related events as the direct study outcomes. Finally, not all cohorts were designed to routinely assess drug resistance mutations, except when requested by physicians, regardless of the viremia status. The low success rate of drug resistance mutation detection may have introduced selection bias and limited the representativeness of the findings for the overall LLV population. Notwithstanding these problems, our retrospective study provides a high-quality reference for developing countries to better understand and manage HIV viremia.

Conclusion

This study showed that the occurrence of blips or LLV during routine ART had no clear correlation with virological failure in China, where NNRTI-based ART regimens are mainstream. Emerging drug resistance during LLV is rare and not associated with subsequent virological failure. Maintaining previous ART regimens and intensifying viral load monitoring are currently the best management strategies for LLV, and further drug resistance testing should be considered based on baseline viral loads and ART regimens.