DISCUSSION
In the current study, the prevalence of HIV-1 subtypes and drug resistance in South Korea was assessed between 2017 and 2022. The prevalence of TDR in South Korea reportedly ranges from 2.5% to 12.0%; however, it varies considerably in different studies10-13. For instance, Jung et al. showed that TDR increased between 2011 and 2019 (5.1%‒7.2%)12, while Park et al. reported no significant increase in TDR between 1999 and 2012 11. These discrepancies might be due to the analysis of a small fraction of patients with HIV-1 in South Korea, with a maximum of 141 patients being diagnosed yearly (range: 27‒141). To date, the current study is the largest-scale study, comprising 2,107 patients, to estimate TDRM in South Korea. More than half of the patients with HIV-1 diagnosed nationwide between 2017 and 2022 were included in this analysis. Moreover, while previous DRM studies only reported the resistance trends of NRTIs, NNRTIs, and PIs11,12, the present study analyzed INSTI mutations and is the first to report the trends in INSTI resistance mutations in South Korea.
In a previous analysis of the HIV-1 subtypes reported between 2005 and 2009, approximately 90% were subtype B; CRF02_AG was the most common among the non-B subtypes, followed by CRF01_AE and subtype A114,19. Meanwhile, a recent study analyzed HIV-1 subtypes between 1999 and 2018 and showed that subtype B infections were more predominant (78.7%) than non-B subtypes (21.3%), which increased to 27.4% between 2015 and 2019 in Busan, Korea 16. Lee et al. also reported CRF01_AE (52.7%) as the most common non-B subtype, followed by CRF02_AG, subtype A1 16. Similarly, in the present analysis between 2017 and 2022, subtype B infections were predominant (75.7%). Non-B subtype strains were identified in 24.3% of patients. CRF01_AE (64.8%) was the most prevalent in non-B subtype infections, followed by CRF02_AG (10.9%) and A1 (10.4%). Although the overall prevalence of subtype B was similar to that reported by previous studies 14,16,19, when stratified by year, the current analysis showed a rapid increase in non-B subtypes from 20% to 33.3% between 2019 and 2022, thus increasing by 4.4% each year. This increase in non-B subtypes is driven primarily by the rapid expansion of the CRF01_AE population, which increased from 58.4% to 73.7% in the non-B subgroup between 2017 and 2022. However, CRF02_AG decreased from 11.7% to 3.0%, whereas subtype A1 remained relatively consistent during the same period. These results are related to the rapid increase in the number of Chinese immigrants, primarily carrying CRF01_AE (around 50%) 20,21, from 0.1 million to 1.1 million between 2000 and 2019, accounting for approximately 50% of all immigrants 22. Moreover, subgroup analysis showed that patients of Korean ethnicity largely carried subtype B compared with non-Korean patients (79.9% vs. 25.6%). Women were also more likely to harbor non-B subtypes than men (51.5% vs. 22.3%). These results are consistent with those of a previous study16.
The current study DRM data showed that 34.7% of the HIV-1 sequences harbored DRMs (low-, intermediate-, or high-level). However, only high-level DRMs, which are clinically significant, were detected in 12.3% of patients. These results showed a higher prevalence of DRMs than previous reports in South Korea (range: 2.5% to 12.0%)10-13. The criteria for defining DRMs can lead to differences in estimation of the prevalence of DRM by studies that include only high-level resistance or all types of DRM. Indeed, the prevalence of high-level DRMs in the current study is higher than that of a previous global study 23, which conducted a meta-analysis of 149 studies, including five from upper-income Asian countries (12.3% vs. 8.7%). However, the prevalence results of this study may be exaggerated as INSTI mutations were also included, unlike previous studies.
Certain mutations exhibited significantly different percentages in different sub-types. For instance, the NRTI mutation A62V was observed in 32.1% and 31.4% of patients harboring subtype A and A6, respectively 24; meanwhile, the PI mutation M46I/L was observed only in 0.9% and 0.6% of patients harboring subtype B and CRF01_AE, respectively. These results are consistent with a previous Chinese HIV-1 subgroup analysis 21. Moreover, the prevalence of DRM declined over the 5 years of the study period. All mutation types declined from 39.4% to 31.6%. High-level resistance mutations also decreased from 16.7% to 7.7%, and the prevalence of NRTI and INSTI DRMs decreased. However, no significant change was observed in NNRTI or PI. These declining trends in drug resistance are not common globally. In fact, most previous reports have demonstrated an overall increase in drug resistance 1,21,23 with only a few noting decreases in TDRM prevalence. That is, Tostevin et al. observed a decrease in DRMs between 2010 and 2013. However, it was confined to the men who have sex with men population, and the rates remained stable in those with heterosexually acquired HIV infections25. Weng et al. observed a stable resistance prevalence between 2009 and 2015 (slope = -0.086) under a fixed regimen with zidovudine/lamivudine + efavirenz or nevirapine as the first-line therapy 26. These TDRM trends are affected by environmental and economic factors, including the COVID-19 pandemic. In fact, a recent study conducted in Canada revealed that drop-in services and outreach work by commercial sex workers were reduced due to COVID-related restrictions 27. Access to healthcare also declined significantly during the COVID-19 pandemic.28. These unusual decreases in TDRM were observed in countries with national health insurance (NHI) systems, which fully cover HIV-related medical expenses for NHI subscribers to ensure the treatment of all patients. These results show that accessibility to ART can play an essential role in changing TDRM trends. Additionally, the adoption of pre-exposure prophylaxis treatment since 2018 in South Korea may have affected the TDRM trends.
Additionally, NNRTI resistance was detected in 26.1% and 42.3% of patients in the ART-naïve and ART-treated groups, respectively. The prevalence of DRM was higher for NRTIs (36.5% vs. 6.3%), PIs (9.6% vs. 0.9%), and INSTIs (26.9% vs. 2.7%) in ART-treated patients. These differences were more significant in dual-class resistance: NRTI + NNRTI (19.2% vs. 3.6%), NRTI + PI (5.8% vs. 0.5%), NNRTI + PI (9.6% vs. 0.5%), NRTI + INSTI (21.2% vs. 1.4%), and NNRTI + INSTI (15.4% vs. 1.4%). Similar results have been demonstrated in previous studies21,29. These findings underscore the need to implement surveillance programs for HIV drug resistance in clinical management.
The current study has certain limitations. First, this was a nationwide multi-hospital-based, retrospective, observational study. The collection of clinical and epidemiological information was limited due to insufficient information provided to the centralized laboratory from hospitals that requested sample analysis. Second, the analysis comparing ART-naïve and ART-treated groups was performed on a limited number of patients. However, this was accounted for by performing a subtype distribution analysis and confirming that patients enrolled at Severance Hospital were representative of the whole cohort to some extent (Supplementary Table 4). Third, despite efforts to clearly classify medical records, clinical information for some patients in the ART-treated group before 2018 may be inaccurate. GART was adopted as a routine surveillance test for HIV-1 treatment in 2018 in South Korea9. Therefore, the GART results before 2018 were primarily weighted toward ART-treated patients, especially those who were ART-resistant. This could have exaggerated the prevalence of DRM before 2018. Fourth, it is difficult to directly compare DRM prevalence between studies as the algorithm version used in each study differs. That is, each study used a different mutation analysis program or version, even if the Stanford University HIV Drug Resistance Database was applied. In particular, the Stanford University HIV Drug Resistance Database accumulates more mutation data with each version update. Therefore, if researchers analyze a previous sequence with a newer version, the results could be different.