**3. Discussion**

The present study provides evidence that ART re-starters, as well as persons with pretreatment NNRTI resistance, in general have worse clinical outcomes than persons without previous exposure to ART and persons without NNRTI resistance in a cohort of Mexican individuals followed for two years. In the adjusted model with ART-regimen change variable, prior exposure to ART was strongly associated with poor clinical outcomes, and with ART-experienced participants, having nearly 40% lower probability of remaining in care with viral suppression. Previous exposure to ART was more common in women with low education arriving late to clinical care. From a total of 1823 Mexican persons living with HIV with baseline drug resistance testing and followed by a median of almost 2 years of follow-up time, we classified 1% as experienced + resistant, 9% as naïve + resistant, 11% as experienced + non-resistant, and 78% as naïve + non-resistant. At the end of follow-up, the experienced + resistant group had the lowest proportion of participants remaining in care compared to other groups. Participants in the experienced + resistant group also showed higher mortality and LTFU, as well as lower levels of viral suppression, even after switching to NNRTI-sparing ART regimens.

Participants in the experienced + resistant group were more frequently women, with lower education level and more frequently unemployed, compared to other groups. Higher social and economic vulnerability among women living with HIV in Mexico has been observed in previous studies [10–15], including in circumstances associated with HIV diagnosis during pregnancy [15]. Additionally, a higher risk for ART discontinuation has been observed among Mexican women living with HIV compared to men [16], with a recent study reporting as high as 20% ART interruption rate [16], with reasons including access to care, depression, and ART adverse effects. Other common reasons observed for ART interruption, and higher viral failure rate in women include a general lower education level, economic dependence on other members of the household, responsibility to provide care and sustenance for children, long distance to clinics and difficulty to pay for transport [12,15]. In general, we found that MSM had better outcomes than heterosexual participants. This could reflect both significant structural differences in the MSM population compared to the heterosexual population living with HIV in Mexico [12], as well as a traditional focus on MSM as a priority group for HIV care and focus of most national HIV prevention efforts.

Changes in ART regimen were more commonly observed among ART-experienced than naïve patients, regardless of their resistance status. However, from this study, we cannot deduce whether clinicians were more proactive in changing ART regimens for ARTexperienced and/or NNRTI resistant patients. We also do not know which information clinicians had available to them at the time of ART-regimen decision-making, as resistance test results. Thus, it is not possible to assess whether the results influenced selection of first-line ART regimens.

In Mexico, as of 2019, bictegravir, an integrase inhibitor drug with high genetic barrier to resistance [8], is recommended as the basis for first line ART-regimens. However, the use of bictegravir is contraindicated in patients using rifampicin or rifabutin; and in pregnan<sup>t</sup> women. In 2020, 30% of patients were still on EFV-based regimens in Mexico [16]. Thus, the implementation of measures to improve adherence and prevent failures due to resistance are important and still needed. HIV drug resistance testing is recommended to guide the choice of second-line ART line, and adherence issues and potential drug interactions need to be addressed.

Our study provides an evaluation of clinical outcomes in real-life setting evaluating clinical outcomes among ART-naive and experienced persons with and without resistance to NNRTI, using surveillance data cross-referenced with the official ministry of health database analyzed with robust statistical techniques. We evaluated virological success among resistant patients, in a context of competing risk with lost to follow-up and death; including the effect of ART change and multiple imputation analyses to address possible bias due to missing data. However, our study also has limitations. First, the original surveillance study was not designed to follow-up participants or to evaluate their clinical outcomes in longer periods, which may reduce availability of information. Using the national database SALVAR allowed us to improve completeness of the data, but we acknowledge that some quality issues could exist with a possible impact on our results. The SALVAR database was designed to record viral load and CD4 T cell count studies, as well as antiretroviral drug dispensation practices for administrative purposes, and record of visits to the clinic or vital status are not among its main objectives. We used the information available as a proxy to inform retention in care, lost to follow-up and viral suppression. Second, not all participants of the original surveillance study were included due to lack of follow-up information, representing a possible selection bias. This situation is frequent in the Mexican setting due to the fragmentation of the health system, obligating persons with formal employment to seek clinical care and ART in social security clinics and persons without formal employment in ministry of health clinics. The lack of a unified national database and migration of persons between health systems, due to changes in employment status has been previously reported as an important reason for ART defaulting and LTFU [17]. The percentage of participants in our study in this situation was 9%, and we found some sociodemographic differences between the persons excluded due to missing data and the persons with data available in the SALVAR database, and thus included in the current study, raising representativeness issues. Nevertheless, these participants with

missing data were mainly men, with a slightly higher median CD4+ count, with higher education, higher employment rates, and lower prevalence of re-starters. Interestingly, a Mexican study including outpatients of a large Ministry of Health clinic in Mexico City observed that persons arriving to care with CD4+ T cell counts < 100 cells/mm<sup>3</sup> were more frequently classified as intermittent ART users and 43% came from social security clinics [18], possibly suggesting a return to care in Ministry of Health clinics, in a worse condition after employment loss, ART defaulting, and LTFU. A program linking the different health care systems in Mexico and a unified HIV database are urgently needed to improve follow-up and care of patients repeatedly changing employment status. Finally, although our study leverages access to SALVAR, we recognize that the information available in this national database is limited. The variables collected for sociodemographic description of the study cohort only describe education level and employment, which may also be variable over time. Inclusion of more adequate variables to describe economic class or poverty level could help to better define socioeconomic status and risk associations with the evaluated outcomes.

Our study provides evidence on associations between pretreatment drug resistance and prior exposure to antiretrovirals in persons starting ART and deleterious clinical outcomes in the Mexican context. However, these observations may also be generally true elsewhere, mainly in other LMICs, and help in public health decision making.
