Low-Frequency Pre-Treatment HIV Drug Resistance

Effects on 2-Year Outcome of First-Line Efavirenz-Based Antiretroviral Therapy

Ross S. Milne; Ingrid A. Beck; Molly Levine; Isaac So; Nina Andersen; Wenjie Deng; Nuttada Panpradist; James Kingoo; Catherine Kiptinness; Nelly Yatich; James N. Kiarie; Samah R. Sakr; Michael H. Chung; Lisa M. Frenkel


AIDS. 2022;36(14):1949-1958. 

In This Article

Abstract and Introduction


Objective(s): Assess the impact of pre-treatment high-frequency and low-frequency drug-resistant HIV variants on long-term outcomes of first-line efavirenz-based antiretroviral therapy (ART).

Design: Prospective observational study.

Methods: Participants' pre-treatment plasma RNA had two sections of HIV pol encoding reverse transcriptase sequenced (Illumina, MiSeq) using unique molecular identifiers to detect wild-type (pre-treatment drug-resistant variants less than 1% of viral quasispecies), low-frequency (1–9%) or high-frequency drug-resistant variants (10–100%). Associations between pre-treatment drug resistance and virologic outcomes over 24 months of efavirenz-based ART were assessed for the number and frequency of mutations by drug class and other resistance parameters.

Results: Virologic failure was detected in 30 of 352 (9%) and pre-treatment drug-resistant variants were detected in the viral quasispecies of 31 of 352 (9%) participants prescribed efavirenz-based ART. Survival analyses revealed statistically significant associations between pre-treatment drug resistance at low (P < 0.0001) and high (P < 0.001) frequencies, at oligonucleotide ligation assay (OLA) (P < 0.00001) and non-OLA (P < 0.01) codons, to a single-antiretroviral class (P < 0.00001), and a shorter time to virologic failure of efavirenz-based ART. Regression analyses detected independent effects across resistance categories, including both low-frequency (P < 0.01) and high-frequency (P < 0.001) drug-resistant variants.

Conclusion: We observed that pre-treatment HIV drug resistance detected at low frequencies increased the risk of virologic failure over 24 months of efavirenz-based ART, but that most failures, regardless of drug-resistant variants' frequencies, were detected within a year of ART initiation. These observations suggest that when efavirenz-based ART is prescribed, screening for pre-treatment drug resistance by an assay capable of detecting low-frequency variants, including OLA, may guide clinicians to prescribe more effective ART.


The prevalence of HIV variants exhibiting resistance to antiretroviral drugs in Kenya and other low-income and middle-income countries has increased with the expansion of antiretroviral uptake.[1,2] Concerns that such resistance, especially to non-nucleoside reverse transcriptase inhibitors (NNRTIs), could compromise the effectiveness of antiretroviral therapy (ART) led the WHO to update ART treatment recommendations in 2019, designating dolutegravir-based ART as preferred in first-line and second-line regimens.[3] However, some individuals may not tolerate dolutegravir, and efavirenz combined with two NRTI remains a recommended alternative regimen for adolescents and adults living with HIV and initiating ART treatment;[3] indeed, in some countries efavirenz-based ART remains the most prevalent ART regimen.[4] As ART regimens are prescribed for an individual's lifetime, and few regimens exist without cross-resistance, understanding the effects of pre-treatment HIV variants resistant to NRTIs and NNRTIs remains relevant in the age of dolutegravir.

Drug-resistant variants detected within an individual's HIV quasispecies by consensus or Sanger sequencing (typically ≥15–25% of quasispecies) have long been associated with the failure of ART to suppress viral replication.[5] More sensitive genotypic assays can detect drug-resistant variants at frequencies below the limit of detection of consensus sequencing, yielding potential insights into the role of low-frequency variants in the success or failure of ARTs.[6–9] However, assay sensitivities for 'low-frequency' pre-treatment variants have varied across studies, significantly limiting the determination of clinically meaningful thresholds. A systematic review of 103 studies assessed the impact of 'low-frequency' pre-treatment drug-resistant variants on the efficacy of ART.[10] The frequency cutoffs varied across a 4 log10 range (0.001–10%). Pre-treatment low-frequency drug-resistant variants and virologic failure were associated in 11of 25 studies (44.0%) of populations taking NNRTI-based ART. Similarly, another study suggests a variant frequency cutoff of 2% may provide optimal specificity and sensitivity[11] while a third suggests 5%.[12] The recent 'Winnipeg Consensus' described these inconsistencies as a major challenge to large-scale implementation of next-generation sequencing-based drug resistance genotyping, warranting further examination.[13]

We previously conducted a randomized clinical trial in Kenya to assess the value of a low-cost oligonucleotide ligation point mutation assay (OLA) for identifying pre-treatment drug resistance and whether guiding the selection of first-line ART regimens using this assay's results improved outcomes.[14] Among participants randomized to the OLA-guided arm, those with pre-treatment drug-resistant variants comprising at least 10% of the individual's viral population or quasispecies received a second-line ART regimen consisting of two NRTIs and lopinavir-boosted with ritonavir, while those in the standard-of-care arm and those in the OLA arm with resistance between 0 and 10% were prescribed first-line NNRTI-based ART. As hypothesized, participants with pre-treatment drug-resistant variants in the OLA-guided arm who were prescribed lopinavir-based ART experienced virologic failure at a significantly lower rate than those in the standard-of-care arm prescribed NNRTI-based ART. Among those with pre-treatment resistance by OLA who were randomized to standard-of-care and received NNRTI-based ART, failure rates by month-12 of ART were higher in participants with drug-resistant frequencies of at least 10%, and those with frequencies of 2–9%; however, the latter group size was small and did not statistically differ from those without resistance. These findings suggested that the threshold of 10% resistance for recommending second-line ART may have been too high, and raised the question as to whether participants with pre-treatment drug-resistant variants at frequencies less than 10% across OLA and non-OLA codons might experience virologic failure at higher rates, particularly if followed for a longer duration of ART.

To evaluate the effects of low-frequency pre-treatment drug-resistant variants on longer term virologic outcomes of first-line NNRTI-based ART, study participants who had maintained virologic suppression at month-12 of ART at the largest of the three study sites were offered enrollment for a second year of follow-up. Pre-treatment plasma HIV RNA specimens from these participants and those who had already experienced virologic failure by 12 months of NNRTI-based ART underwent genotyping by next-generation sequencing to assess the contribution of a broad array of codons with drug-resistant variants at low frequencies on virologic failure over 24 months of efavirenz-based ART. (Note: Because nevirapine has been rendered largely obsolete following the introduction of efavirenz in sub-Saharan Africa, we limited our analyses to the larger group of participants prescribed efavirenz-based ART.) We aimed to determine if over 2 years of first-line-efavirenz-based ART first, low-frequency drug-resistant HIV variants comprising 1–9% of an individual's quasispecies contributed to increased rates of virologic failure compared with rates in individuals with wild-type virus; second, the time to virologic failure is longer in individuals with low-frequency variants (1–9%) compared with participants with high-frequency variants (10–100%); and third, drug-resistance mutations at codons not assessed by the OLA contributed to virologic failure.