**2. Results**

#### *2.1. Characteristics of the Study Participants*

Between January 2017 and December 2020, 8128 blood specimens were collected at the Condesa Specialized Clinic, the largest HIV primary care center in Mexico City. From these, 6785 (83.5%) were successfully sequenced. After curation, removal of duplicates (the first sequence of each individual in this case was kept), and sequence quality filtering (see Methods), 6661 unique individuals with protease-reverse transcriptase (PR.RT) as well as integrase (IN) sequence available were included in the database. Taking into account different amounts of missing information for different variables across the data set (Table 1), 95.1% (6234/6555) of the participants were cisgender men, 4.1% (267/6555) cisgender women, 0.8% (51/6555) transgender women, and 0.05% (3/6555) transgender men; 80.0% (4223/5280) lived in Mexico City and 18.3% (967/5280) in the surrounding municipalities of the State of Mexico; and 39.7% (2323/5853) arrived with <200 CD4+ T cells/mm<sup>3</sup> to clinical care. Subtype B largely predominated in the study population with only 1.3% (84/6661) of participants having non-B subtypes. The most frequent non-B subtypes observed were circulating recombinant forms (CRF02\_AG: 0.29%; CRF01\_AE: 0.26%) and unique recombinant forms (BG: 0.20%; BF: 0.17%).

Considering a subgroup of 1348 participants enrolled from June to December 2020, for whom more detailed metadata were collected (Table 2), and accounting for missing data, an estimated 11.0% (141/1286) reported previous exposure to ART, 59.9% (717/1196) self-identified as belonging to the middle social class, 40.3% (484/1201) had at least an undergraduate degree, and 3.3% (40/1196) reported speaking any indigenous language. Regarding additional risk variables, 40.2% (426/1059) reported having a sexually transmitted infection in the previous 6 months, 78.1% (793/1016) reported a receptive role in anal

sex, 77.1% (807/1047) were not circumcised, and 9.2% (93/1013) reported using venues for sex (Table 2).


**Table 1.** Baseline demographic and clinical characteristics of study participants, 2017–2020 a.

a Data for 6661 participants enrolled from January 2017 to December 2020. b Column percentages are shown. c Using the WHO definition for PDR (see Methods). d Row percentages are shown. e Data missing for 114 participants (*n* = 6547). f Data missing for 557 participants (*n* = 6104). g Data missing for 808 participants (*n* = 5853). h Data missing for 2234 participants (*n* = 4427). \* *p* < 0.05. PDR; pretreatment drug resistance; EFV, efavirenz; NVP, nevirapine; IQR, interquartile range.

**Table 2.** Baseline demographic, clinical, and risk characteristics of study participants, June–December 2020 a.



**Table 2.** *Cont.*


**Table 2.** *Cont.*

a Data for 1348 participants enrolled from June to December 2020. b Column percentages are shown. c Using the WHO definition for PDR (see Methods). d Row percentages are shown. \* *p* < 0.05. ARV, antiretroviral; EFV, efavirenz; NVP, nevirapine; IQR, interquartile range; MSM, men who have sex with men.

#### *2.2. Overall Estimations of Pretreatment Drug Resistance*

After removal of sequences with quality control issues, 6688 PR-RT and 6709 IN sequences were available for HIVDR analysis (a total of 6661 participants had both PR-RT and IN sequences available). Considering the complete study period, overall PDR to any drug class in Mexico City was 14.4% (95% CI: 13.6–15.3%) (Figure 1a). PDR to NNRTI was higher than to any other drug class (12.0%, 95% CI: 11.3–12.8%, *p* < 0.0001), and exceeded the 10% threshold recommended by the WHO for public health action. PDR to NRTI was 4.2% (3.7–4.7%), and to PI, 3.1% (2.7–3.6%). PDR to any INSTI was the lowest (1.0%, 0.8–1.3%, *p* < 0.0001). In agreemen<sup>t</sup> with the preferential use of efavirenz-based first-line regimens in Mexico during most of the analysis period, PDR to efavirenz/nevirapine was 10.4% (9.7–11.2%). Nevertheless, PDR to efavirenz/nevirapine + any NRTI remained low (1.1%, 0.9–1.7%). Importantly, given the current preferential use of INSTI-based first-line regimens, PDR to first-generation INSTI was 1.0% (0.8–1.3%) and to second-generation INSTI 0.3% (0.2–0.4%, *p* < 0.0001).

Considering individual drugs, PDR to nevirapine (10.4%, 9.7–11.6%) and to efavirenz (9.3%, 8.6–10.0%) was the highest across the study period, followed by PDR to rilpivirine (5.1%, 4.6–5.7%) (Figure 1b). Emtricitabine and tenofovir showed the lowest PDR among NRTIs (0.5%, 0.4–0.8%, and 1.1%, 0.9–1.4%, respectively). Among the currently used drugs as third component, PDR to boosted darunavir (0.3%, 0.2–0.5%), dolutegravir (0.2%, 0.1–0.4%) and bictegravir (0.1%, 0.0–0.2%) was low.

**Figure 1.** Overall pretreatment drug resistance levels in Mexico City, 2017–2020. Classified by (**a**) drug class, (**b**) individual drug. PDR was estimated from 6688 protease-reverse transcriptase and 6709 integrase sequences using the Stanford HIVdb tool (v9.0). Individuals with HIV PDR were defined as those having a score ≥15 to any of the drugs in the corresponding category, as defined in the Methods. Error bars represent 95% confidence intervals. Comparisons were made with Fisher's exact test. PDR, pretreatment drug resistance; PI, protease inhibitor; RTI, reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; INSTI, integrase strand-transfer inhibitor; EFV, efavirenz; NVP, nevirapine; ATV/r, atazanavir/ritonavir; DRV/r, darunavir/ritonavir; LPV/r, lopinavir/ritonavir; EVG, elvitegravir; RAL, raltegravir; BIC, bictegravir; CAB, cabotegravir; DTG, dolutegravir.

The most frequent surveillance drug resistance mutations (SDRMs) were K103NS (6.7%) to NNRTI; M41L (1.2%) and T215CDEF (2.1%) to NRTI; M46IL (1.7%) to PI; and E138AKT (0.2%), Q148HKR (0.1%), and S230R (0.1%) to INSTI (Figure 2a). However, other polymorphic non-SDRMs were also frequent (V106I (3.6%), V179DE (7.1%) in RT) (Figure 2b). Given the low prevalence of non-B subtypes in the study population, associations between specific viral subtypes and the presence of polymorphic mutations were not particularly relevant in the context of the study population, e.g., V106I was present in all F1 subtypes; however, only 6/6661 F1 viruses were observed, while V106IM (generally combined with other NNRTI mutations) was observed in 231/6577 (3.5%) subtype B viruses.

**Figure 2.** Drug resistance mutation frequency in Mexico City, 2017–2020. Frequencies of surveillance drug resistance mutations (SDRMs) (**a**) and non-SDRMs (**b**) for the complete study period are shown, including a total of 6688 proteasereverse transcriptase and 6709 integrase sequences from unique individuals. Only resistance-associated mutations observed in the sequences are shown. All SDRMs and non-SDRMs defined in the Stanford HIV drug resistance database were analyzed. PI, protease inhibitors; NRTI, nucleoside reverse transcriptase inhibitors; NNRTI, non-nucleoside reverse transcriptase inhibitors; INSTI, integrase strand-transfer inhibitors.

We explored associations of PDR with demographic, clinical, and risk variables collected for a subgroup of participants starting from June to December 2020. As expected, persons with prior exposure to ARV had overall significantly higher prevalence of PDR (*p* < 0.01) (Table 2, Figure 3). This was true for NNRTI and NRTI, but not for boosted PI or INSTI (Figure 3). Persons who inject drugs also showed higher PDR level (*p* = 0.04). Interestingly, persons with incomplete metadata collection in general (who failed to answer several questions of the questionnaire) showed higher PDR (*p* < 0.05). Additionally, persons living in a domestic partnership, with a high school level education, who identified as heterosexual cisgender men, who were circumcised, and who engaged more frequently in insertive sexual relations showed lower PDR to efavirenz/nevirapine (all *p* < 0.05) (Table 1). In general, persons with PDR had slightly lower viral load (*p* < 0.05) (Table 2).

**Figure 3.** Pretreatment drug resistance in ART-naïve and prior ARV-exposed individuals in Mexico City, 2020. PDR levels are shown for a subset of 1348 participants enrolled from June to December 2020 (data on ARV exposure for 62 participants are missing), for whom data on prior exposure to antiretrovirals were available. Individuals with HIV PDR were defined as those having a score ≥ 15 to any of the drugs in the corresponding category, as defined in the Methods. Error bars represent 95% confidence intervals. Comparisons were made with Fisher's exact test. PDR; pretreatment drug resistance; PI, protease inhibitor; RTI, reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; INSTI, integrase strand-transfer inhibitor; EVG, elvitegravir; RAL, raltegravir; BIC, bictegravir; CAB, cabotegravir; DTG, dolutegravir.

The group of participants with prior exposure to ARVs showed significant differences in demographic, clinical, and risk characteristics in comparison to ART-naïve participants, including a higher frequency of cisgender women (12.8 vs. 5.6%), lower frequency of cisgender MSM (60.3 vs. 68.3%), higher age (median: 35 vs. 29 years), lower education (elementary or none: 12.2 vs. 4.8%), higher frequency of persons who inject drugs (8.5 vs. 3.1%), and lower use of apps for finding sexual partners (31.9 vs. 42.8%) (*p* < 0.05 in all cases; Supplementary Table S1). Considering only ART-naïve persons enrolled from June to December 2020, HIVDR reached 10.2% (CI 95%: 8.5–12.1%) to efavirenz/nevirapine, 4.0% (3.0–5.3%) to NRTI, 1.7% (1.1–2.7%) to boosted atazanavir, lopinavir or darunavir, and 1.4% (0.8–2.3%) to any INSTI (Figure 3). Of note, no cases of bictegravir or dolutegravir resistance were observed among prior ARV-exposed participants (only 1 participant (0.7%) showed cabotegravir resistance due to the N155H mutation).

#### *2.3. Pretreatment Drug Resistance Trends across the Study Period*

A significant increasing trend for NNRTI PDR was observed from 2017 to 2020 (10.3 to 13.6%, *p* = 0.02). PDR to all other drug classes remained stable: NRTI below 5%, PI below 3.5%, and INSTI below 1% (Figure 4a). Importantly, no increase in PDR to secondgeneration INSTI was observed, remaining under 0.3% across the complete study period. The observed increase in NNRTI PDR was mainly attributable to increasing PDR to efavirenz (8.0 to 10.4%; *p* = 0.05), but also to rilpivirine (4.5 to 5.9%, *p* = 0.03) (Figure 4b). No other drugs showed significant trends (Figure S1).

**Figure 4.** Pretreatment drug resistance trends in Mexico City, 2017–2020. (**a**) By drug class, (**b**) For NNRTI, (**c**) Including relevant antiretroviral drugs in the Mexican context, and (**d**) Including selected drug resistance mutations with relevance to NNRTI PDR. PDR was estimated from protease (PR), reverse transcriptase (RT) and integrase (IN) sequences using the Stanford HIVdb tool (v9.0). Individuals with HIV PDR were defined as those having a Stanford score ≥ 15 for any of the drugs in the corresponding category, as described in the Methods. DRMs included SDRM and non-SDRM. First-generation INSTIs include raltegravir and elvitegravir; second-generation INSTIs include dolutegravir, bictegravir and cabotegravir. Trends were tested using linear regression. PDR, pretreatment drug resistance; PI, protease inhibitors; RTI, reverse transcriptase inhibitors; NRTI, nucleoside reverse transcriptase inhibitors; NNRTI, non-nucleoside reverse transcriptase inhibitors; INSTI, integrase strand-transfer inhibitors; DRM, drug resistance mutation; SDRM, surveillance DRM.

In the Mexican context, PDR to all relevant drugs for currently recommended first-line ART regimens remained stable and below 2%, except for efavirenz (Figure 4c). Regarding possible NRTI backbones, PDR to abacavir, emtricitabine/lamivudine, and tenofovir, remained low (under 2%) and stable. The same was true for boosted darunavir (under 0.5%), and dolutegravir/bictegravir (under 0.3%).

A significant increasing trend was observed in the prevalence of any RT SDRM across the study period (10.6 to 12.5%, *p* = 0.04), particularly to any NNRTI SDRM (7.3 to 10.0%, *p* = 0.04). The prevalence of K103NS varied from 5.2 to 7.7%, but this increase was not significant (*p* = 0.08) (Figure 4d). No other DRM (including both SDRM and non-SDRM) showed significant prevalence changes during the period (Figure S2).

#### *2.4. Associations of Pretreatment Drug Resistance with Epidemiological Variables*

We analyzed possible associations between the presence of PDR and demographic and clinical variables. This analysis encompassed only individuals enrolled from June 2020 to December 2020, a period in which variables were collected through a computer-based questionnaire and the completeness of the metadata was acceptable. After multivariable adjustment, the presence of PDR to any drug remained strongly associated with prior exposure to ARV drugs (adjusted odds ratio, aOR: 2.7, 95% CI: 1.7–4.1, *p* = 0.001). The odds of PDR were higher in persons presenting to care with <200 CD4+ T cells/mm<sup>3</sup> compared to 200–499 CD4+ T cells/mm<sup>3</sup> (aOR: 1.6, 95% CI: 1.2–2.2, *p* = 0.005) (Table 3).


**Table 3.** Risk factors for HIV pretreatment drug resistance in Mexico City, June–December 2020.

a *n* = 1348; b Analyzed as a continuous variable; c Sexual risk category was assessed as a composite variable including sex at birth, gender identity and sexual practices. aOR, adjusted odds ratio; CI, confidence interval; ARV, antiretroviral; NNRTI, non-nucleoside reverse transcriptase inhibitors; MSM, men who have sex with men; Ref., reference category. \* Statistically significant.

#### *2.5. Analysis of HIV PDR Transmission within Mexico City's HIV Genetic Network*

We identified clusters of individuals with PDR within Mexico City's HIV transmission network. The network was inferred from 6688 PR-RT sequences from individuals arriving to care from 2017 to 2020, from which 2960 (44.3%) were found to belong to 820 clusters, ranging in size from 2 to 41 nodes (Figure 5). No difference was observed between

clustering and non-clustering individuals in the proportion of viruses with PDR to any drug (13.9 vs. 14.7%, *p* = 0.65) or to efavirenz/nevirapine (9.7 vs. 11.0%, *p* = 0.15). A total of 19.4% (159/820) of all clusters included at least one person with PDR and 14.8% (121/820) included at least one person with resistance to efavirenz/nevirapine. Resistance to efavirenz/nevirapine was associated with transmission of K103NS alone in 7.9% (65/820) of clusters, K103NS plus other NNRTI mutations in 3.4% (28/820) clusters and with other NNRTI mutations in 8.0% (66/820) of clusters (Figure 5). In 41.3% (50/121) of the clusters with efavirenz/nevirapine PDR, resistance was shared by 100% of the nodes, and in 32.4% (38/121) by 50 to 99% of the nodes. Some examples of the larger clusters evidencing NNRTI PDR transmission included cluster NNRTI-1, with 14 nodes, all men living in Mexico City with median age 28 years (IQR 23–31), all of them with K103NS; cluster NNRTI-2, with 27 nodes formed by men with median age 25 (23–30) enrolled across the complete study period, 11% (3/27) with K103NS and 74% (20/27) with other mutations; NNRTI-3, with 21 nodes, including men living in Mexico City and the State of Mexico and median age 24 (20–29), 62% (13/21) with K103NS; and cluster NNRTI-4, with 12 nodes, all men with median age 28 (25–32), all sharing E138A (Figure 5).

A total of 2.7% (22/820) of the clusters included at least one person with INSTI PDR and only 0.7% (6/820) with at least one person with bictegravir/dolutegravir resistance. The largest clusters with INSTI PDR transmission were cluster INSTI-1, with 5 nodes sharing the S230R mutation, formed by 4 cisgender men and 1 transgender woman, with median age 23 (20–32); and cluster INSTI-2, with 4 nodes, all men sharing the E157Q and G163K mutations and median age 28 (26–34).

Other interesting clusters evidencing PDR transmission within the network included cluster PI-1, with 29 nodes, all men with PI resistance and median age 24 (22–28) enrolled across all years; cluster NRTI-1, with 23 nodes, 22 cisgender men and 1 cisgender woman, with median age 22 (10–27) and constant growth across all years; cluster complex-1, with 10 nodes, all men enrolled from 2018 to 2020 and median age 23 (20–30), 60% (6/10) with PI + NRTI + NNRTI resistance and 40% (4/10) with PI + NRTI resistance; and complex-2, with 7 nodes, all men with median age 26 (21–31) sharing PR M46I, L90M, RT M41L, D67N, T69D, L210W, T215D and RT K103N, Y181C (Figure 5).

#### *2.6. Characteristics and Prevalence of Acquired Drug Resistance in Mexico City, 2020*

In order to describe acquired drug resistance in the context of ART optimization in Mexico City, starting in 2019, we analyzed all clinically indicated HIV genotypes for persons cared for at the Condesa clinics in 2020, who had two consecutive viral load estimations > 1000 copies/mL. A total of 143 individuals were eligible (see Methods). From these, we obtained 142 successful PR-RT sequences and 137 IN sequences. After quality filtering, 133 PR-RT and 128 IN sequences were used for the analysis. Overall, 39.8% (95% CI: 31.5–48.7%) individuals had ADR to any drug class (Figure 6a). Considering individual drug classes, 34.6% (26.6–43.3%) of individuals showed ADR to NNRTI, 26.3% (19.1–34.7%) to NRTI, 1.5% (0.2–5.3%) to boosted PI, and 2.3% (0.5–6.7%) to INSTI. A total of 21.1% (14.5–29.0%) showed resistance to NRTI and efavirenz/nevirapine. Considering individual drugs, 33.8% individuals were resistant to efavirenz, 23.3% to emtricitabine/lamivudine, and 8.3% to tenofovir. Resistance to dolutegravir/bictegravir was 2.3% and to darunavir 0%. Regarding other drugs commonly used in second or third line, ADR to etravirine was 14.3%, 18.8% to rilpivirine, 20.3% to doravirine, and 2.3% to zidovudine (Figure 6b). The most commonly observed DRMs were K103NS (24.1%) and M184VI (23.3%) in RT. Resistance in IN was mainly due to R263K (1.6%) and E138AKT, G140ACS, Q148HKR (0.8% each) (Figure 6c). The relatively high resistance to doravirine is noteworthy, and was mainly associated with mutations Y188L (4.5%, in most cases combined with V106I), L100IV (4.5%), K103EP (3.8%), P225H (3.0%).

**Figure 5.** PDR transmission within Mexico City's HIV genetic network, 2017–2020. The network was inferred from 6688 protease-reverse transcriptase sequences from individuals arriving to care from 2017 to 2020, using a locally adapted version of the HIV-TRACE tool. Each circle represents a cluster. The size of the circle reflects the size of the cluster. Clusters are colored according to the presence of pretreatment drug resistance by drug class. Specific clusters mentioned in the text are identified with letters on the lower-right side. Red circles surrounding clusters show the presence of viruses with the K103NS mutation.

**Figure 6.** Acquired drug resistance in Mexico City, 2020. (**a**) By drug class, (**b**) by individual drugs, (**c**) SDRM and non-SDRM. ADR was estimated from protease (PR), reverse transcriptase (RT) and integrase (IN) sequences using the Stanford HIVdb tool (v9.0). Individuals with HIV ADR were defined as those having a Stanford score ≥ 15 for any of the drugs in the corresponding category, as described in the Methods. ADR, acquired drug resistance; PI, protease inhibitors; RTI, reverse transcriptase inhibitors; NRTI, nucleoside reverse transcriptase inhibitors; NNRTI, non-nucleoside reverse transcriptase inhibitors; INSTI, integrase strand-transfer inhibitors; EFV, efavirenz; NVP, nevirapine; ATV/r, atazanavir/ritonavir; DRV/r, darunavir/ritonavir; LPV/r, lopinavir/ritonavir; EVG, elvitegravir; RAL, raltegravir; BIC, bictegravir; CAB, cabotegravir; DTG, dolutegravir; DRM, drug resistance mutation; SDRM, surveillance DRM.
