Factors associated with viremia in people living with HIV on antiretroviral therapy in Guatemala

Introduction Viral suppression prevents HIV transmission and disease progression, but socio-economic and clinical factors can hinder the goal of suppression. We evaluated factors associated with viral non suppression (VNS) and persistent viremia (PV) in people living with HIV (PLHIV) receiving antiretroviral therapy (ART) in Guatemala. Methods We conducted a cross sectional analysis using data from an ongoing cohort of PLHIV attending the largest HIV clinic in Guatemala. Univariable and multivariable analyses were conducted between PLHIV with viral suppression and detectable viremia. VNS was defined as most recent HIV RNA ≥ 200 copies/ml and PV as two consecutive HIV RNA ≥ 200 copies/ml. Results Of 664 participants, 13.3% had VNS and 7.1% had PV. In univariable analysis disaggregated by gender, low income, poor education, perceived difficulty attending healthcare, and alcohol use were associated with VNS in men while low CD4 at diagnosis, multiple prior ART regimens and treatment interruptions were significant in both genders. Multiple prior ART regimens (adjusted Odds Ratio (aOR) 2.82, [95% confidence interval (CI) 1.59, 4.99], p < 0.01), treatment interruptions (aOR 4.51, [95% CI 2.13, 9.58], p < 0.01), excessive alcohol consumption (aOR 2.56, [95% CI 1.18, 5.54], p < 0.05) perceived difficulty attending healthcare (aOR 2.07, [ 95% CI 1.25, 3.42], p < 0.01) and low CD4 at diagnosis (aOR 2.34, 95% [CI 1.30, 4.20], p < 0.01) were independently associated with VNS on multivariable regression. Conclusions We conclude that socio-economic and clinical factors influence viral suppression in our cohort and vary between men and women. Gender specific approaches are necessary to achieve the 90% suppression goal. Supplementary Information The online version contains supplementary material available at 10.1186/s12981-021-00400-9.

continue to experience setbacks in the fight against HIV. Such is the case of Latin America, where no countries have reached the 90-90-90 goal, HIV new infections increased by 7% in the past eight years and only 55% of all the diagnosed PLHIV are virally suppressed [2][3][4].
Viral suppression plays a crucial role in limiting onward transmission of HIV. As HIV prevalence increases, viral suppression must be achieved in order to prevent new infections [5]. Several studies have pointed out that viral suppression decreases the likelihood of HIV transmission in the at-risk and general population and improves overall life quality and life expectancy [6][7][8]. Individuals failing to achieve viral suppression may progress to persistent viremia, which is associated with double the risk of virologic failure and with higher rates of all-cause mortality [9][10][11].
The main concern in Latin America is that progress, while steady, is uneven between regions [2]. In Guatemala, the country with the highest number of PLHIV in Central America, 62% of PLHIV know their HIV status, 69% of all linked to care are taking ART and 80% taking ART are virally suppressed. In total, the percentage of all PLHIV in Guatemala with viral suppression is estimated at 34% [2,3].
Several barriers to viral suppression have been identified and include but are not limited to stigma, difficulty in accessing healthcare because of location or financial status [12,13], alcohol and substance misuse [2], education [14], pill burden [15], and ART side effects [16]. To our knowledge, a detailed analysis on factors that influence virologic outcomes in PLHIV has not been performed in Guatemala. We aimed to identify associated factors for virologic outcomes in the treatment of PLHIV in this Central American country.

Methods
We used data from baseline interviews obtained as part of an ongoing prospective cohort study on comorbidities in PLHIV attending the Dr. Carlos Rodolfo Mejia Villatoro Infectious Diseases Clinic at Roosevelt Hospital in Guatemala City. The clinic is the largest HIV clinic in the country and provides free care for over 5000 PLHIV nationwide. PLHIV 18 years or older attending the HIV clinic are eligible for enrollment in the cohort. PLHIV were enrolled after providing verbal consent, and data was collected with in-depth interviews, clinical evaluation, and review of clinical files. The study was approved by the institutional review board of Roosevelt Hospital. The baseline interviews, containing questions on demographics, housing, medical conditions, and health risk factors, were conducted in Spanish and took place from July 2019 to February 2020. Study data were collected and managed using REDCap electronic data capture tools hosted at Washington University in Saint Louis [17].

Study design
For the current analysis we included all PLHIV enrolled in the cohort that had been prescribed at least six consecutive months of ART since their initial diagnosis and had a recent viral load. Recent viral load was defined as a HIV viral load taken within three months prior or one month following the enrollment interview for participants with clinic visits every 3-months and viral load taken six months from enrollment at most for participants with 6-months clinic visits. Viral non-suppression (VNS) was defined as a viral load ≥ 200 copies/ml and persistent viremia (PV) was defined as two consecutive viral loads ≥ 200 copies/ml.

Variable definitions
The continuous variables age, travel time and travel cost to HIV care, individual monthly income, years of education, number of prior ART regimens, CD4 count at diagnosis and days of treatment interruptions were recategorized into binary variables for analysis. Age 50 and below or over 50 years was used to compare younger and older adults, respectively, as defined by the Center for Disease Control and Prevention (CDC) [18]. Travel time to healthcare was defined as commutes within an hour and greater than an hour based on a prior study regarding accessibility in Guatemala [19]. Individual monthly income was evaluated by earning less than or equal to the "Canasta básica alimentaria" (CBA), or national food basket, which is the monthly cost to provide sufficient food for the average Guatemalan household. As of January 2020, the cost was $464 [20]. Education level was divided by PLHIV that achieved above primary education (first six years of formal education) from those with less or no formal education. Travel cost considered whether the round-trip cost to the clinic exceeded the median cost for our study population ($2.6). Multiple-daily dosing refers to more than once-daily dosing of ART. Treatment interruptions were defined as missing seven days or more of ART between changing regimens. Multiple prior ART regimens was defined as having three or more prior ART regimens.
Categorical variables were defined as follows. Excessive alcohol consumption was defined as binge drinking (four or more drinks in one occasion for women and five or more drinks for men) and heavy drinking (eight or more drinks in a week for women and 15 or more drinks in a week for men) within the past month, in accordance with the CDC definitions [21]. Access to basic utilities considered both running water and electricity in the participant's residence. Indigenous participants were self-identified. Perceived difficulty attending healthcare was defined according to the participant's yes/no response to the question "¿Le parece dif ícil asistir a la clínica?" Translation: "Do you find it difficult to visit the clinic?". A follow up question gave participants the following reasons to choose from: lack of income, lack of transportation, nobody to take care of children, difficulty getting time off from work, price of transportation, or other reasons (free answer). Comorbidities were reported by participants and verified in the clinical file. Prior AIDS defining illnesses were assessed according to the CDC definitions [22].

Statistical analysis
Statistical analyses were performed using SPSS (Statistical Package for the Social Sciences) software version 23. Baseline characteristics were analyzed with respect to frequency and percentage. Normality was evaluated for all numeric variables using Shapiro Wilk test with kurtosis and skewness. Baseline characteristics were compared with the nonparametric Pearson x 2 and Fisher's exact test of independence for categorical variables and Mann-Whitney U test for continuous variables with interquartile range. Binary logistic regression was performed to explore the association between each variable and VNS and PV. Multivariable logistic regression models were fitted with variables at p ≤ 0.05 to evaluate independent factors associated with VNS and PV.
Comorbidities were present in 408 (61.4%) and only 21 (3.2%) participants were co-infected with hepatitis B. Median CD4 count and viral load at diagnosis were 178 cells/mm 3 (IQR 57-341) and 77,000 copies/ ml (IQR 21,248-252,252) respectively. Current median CD4 count was 473 cells/mm3 (IQR 283-689). A total of 262 (39.5%) of our participants were in their first ART regimen, 402 (60.5%) had at least one prior ART regimen and 105 (15.8%) had multiple prior ART regimens. Other characteristics arranged by virologic status are presented in Table 1. VNS was present in 88 (13.3%) participants on recent HIV viral load testing. Differences between those with viral suppression and VNS are illustrated in Table 1. Of those who were not virally suppressed, 47 (7.1%) had PV.
On univariable analysis, excessive alcohol consumption, perceived difficulty attending healthcare, having no comorbidities, treatment interruptions, multiple prior ART regimens, multiple-daily dosing, low CD4 count at diagnosis, past AIDS-defining illness and sexuality were associated with VNS (p ≤ 0.05) and were included in the multivariable analysis ( , p < 0.01), were all associated with increased risk of VNS. Among those with perceived difficulty, the most commonly reported reasons for such difficulty were difficulty getting time off from work (72, 28.6%) and price of transportation (71, 27.5%) (Additional file 3: Table S3), but no association was found between specific reasons for perceived difficulty attending healthcare and VNS (Additional file 3: Table S3).
Univariable analysis disaggregated by gender (Additional file 1: Table S1 and Additional file 2: .65], p = 0.022) were also significantly associated with VNS in men but not in women. Past AIDS defining illnesses, treatment interruptions, low CD4 count at diagnosis and having multiple prior ART regimens were still significantly associated with VNS in both genders in univariable analysis. In separate multivariable models however, only multiple prior ART regimens was significant in both men and women.
In the univariable analysis for PV as the outcome of interest, past AIDS defining illness, low CD4 count at diagnosis, treatment interruptions, multiple prior ART regimens, multiple-daily dosing, bisexual orientation, perceived difficulty attending healthcare, and having no comorbidities were significantly associated with PV. In the multivariable analysis, participants with multiple prior ART regimens ( Table 3. These variables were also significantly associated with VNS.

Discussion
To our knowledge, a detailed analysis on factors that influence virologic outcomes in PLHIV has not been performed in Guatemala. Our study found factors associated with VNS and PV among PLHIV in Guatemala, complementing prior findings about the modifiable and unmodifiable factors associated with detectable viremia in PLHIV on ART [23][24][25][26][27][28]. Several studies have demonstrated that the risk for negative outcomes such as mortality, virologic failure and AIDS increases for PLHIV at low viremic levels, starting from 50 to 200 copies/ml [24,29]. Therefore, 200 copies/ml was used in this study as a threshold to identify people with detectable viremia.
Perceived difficulty attending healthcare was one of the variables associated with VNS. Several socio-economic factors have been shown to have a negative impact on perception of HIV care such as quality of services, financial fairness, transportation convenience, perceived negative attitude, and stigma [30]. However, none of the reasons for perceived difficulty in our study was significantly associated with VNS, suggesting that each participant may have individualized difficulties that influence their viral load suppression. We found that paradoxically, having no comorbidities increases the risk of VNS and PV. Several studies have found no association or a similar association to our study [31][32][33]. This may be explained by more frequent interactions with healthcare for people with comorbidities, due to their chronic conditions, where they may be encouraged to seek HIV care [34,35].
In this study, indigenous ethnicity and age did not have an impact on VNS or PV. This differs from other studies where ethnic minorities [36,37] and younger populations [10,27,37,38] are more likely to experience worse outcomes in HIV care. Our findings may have been influenced by the low frequency of older PLHIV and those who self-identify as indigenous in our cohort, and further studies are needed in these populations.
Gender was also not associated with VNS in the overall population, in keeping with previous research on gender and detectable viremia [39,40]. However, analysis disaggregated by gender showed important differences (Additional file 1: Table S1 and Additional file 2: Table S2). Lower individual income, low educational attainment, engaging in risk behaviors, bisexual orientation and perceived difficulty attending healthcare exclusively had an impact on viremia in men. Individual income may be less relevant in women due to differences in workforce participation in the cohort (23.7% in women vs 42.6% in men) [41]. Other studies have examined the interplay between gender, HIV and risk behaviors and have observed that the prevalence of heavy alcohol consumption is higher in men with HIV than in women [42], at-risk drinking in men is predictive of virologic failure [43], and MSM who engage in higher risk drinking are more likely to engage in HIV risk behaviors [44]. We emphasize these gender Table 2 Univariable and multivariable logistic regression of viral non-suppression in 647 PLHIV on ART in Guatemala MSM men who have sex with men, ART antiretroviral therapy, aOR adjusted odds ratio, CI confidence interval Significant variables (p < 0.05) in univariable analysis were entered into multivariable analysis a Monthly income ≤ Canasta Básica Alimentaria, the cost to feed an average Guatemalan household per month b $2.6 was the median cost of transport reported in the cohort differences due to our concern that men are getting left behind in the fight against the HIV epidemic [45] which calls to revise policies regarding men's sexual health services. Unlike VNS, we found no social or economic variables associated with PV. However, the same variables related to HIV background were significantly associated with both VNS and PV. Having multiple prior ART regimens was independently associated with both outcomes with the highest odds ratio of all relevant variables. Similar findings are reported in South Africa [46], Uganda [47], and the United States [48] where having multiple prior ART regimens or being on 2nd line therapy was significantly associated with PV. We agree with other hypotheses that multiple ART changes are more likely associated with poor adherence, acquired drug resistance mutations, and long-term drug toxicities that influence treatment efficacy and viral suppression [46][47][48]. Not all studies support this link between the viral suppression and ART history [49], but this finding should encourage further research on reasons for multiple ART changes and HIV outcomes in Guatemala, especially since the HIV treatment guidelines were updated in December 2019 to include integrase inhibitors as first line of therapy [50,51].
Low CD4 count at diagnosis was another variable significantly associated with VNS and PV, similar to what was found in several cohort studies [23,[52][53][54]. Progressing to advanced disease before treatment may put patients at higher risk of persistent viremia once treatment is started as a consequence of increased viral reservoir size, but further research is needed to understand this association.
We also demonstrate that treatment interruptions are associated with VNS as well as PV. This result is similar to what is reported from multiple studies where missing any  [55,56]. These treatment interruptions may occur due to lack of HIV care facilities or treatment support programs in Guatemala, in addition to individual challenges in attending care. This could be a key area of intervention to impact both the percentage of PLHIV taking ART and the proportion of those that are virally suppressed.
There are important limitations in our study. Although all participants were selected randomly, a small number of people declined participation mainly due to insufficient time to conduct an interview. Those that were temporarily lost to follow up in HIV care, hospitalized in-patient, or attending prenatal care were not captured in this cohort. Adolescents were not considered due to health policies regarding age and sexual health consent. Transgender population and sex workers are underrepresented in our cohort due to the lack of public outreach and gender specific programs [2]. We were unable to include HIV subtypes and clades for our participants because this data is not available as standard of care. However, one cohort study previously published from our clinic showed that 98.9% of sequences performed belong to B Pandemic clade [57]. Treatment interruptions were only captured in participants lost to follow up for more than three months or with ART that differed in at least one drug or differed in doses, but not in those that did not change regimen or dosing. During treatment interruptions some participants may share treatment with a partner or split pills. For prior ART switches, we were able to accurately capture reasons for treatment switch due to lack of prior information for some of the files. Household income, which was not available in this data, may be a better measurement of economic status than individual income.

Conclusions
This is the first study, to our knowledge, demonstrating potential social and clinical influences on VNS and PV in PLHIV in Guatemala. In our analysis, we found that social factors such as excessive alcohol consumption, perceived difficulty attending healthcare, and bisexual orientation were significantly associated with VNS in PLHIV in Guatemala. In this cohort low CD4 count at diagnosis, treatment interruptions, multiple prior ART regimens, and multiple-daily dosing were significantly associated with both VNS and PV. This highlights potential areas of improvement to advance towards achieving the goal of 90% suppression. Simplifying regimens when possible, addressing personal barriers, improving access to ART outside of the clinic, and concentrating efforts on groups at greater risk are key interventions to continue improving viral suppression. The strong link between perceived difficulty attending healthcare and VNS warrants further exploration, as PLHIV often need to overcome barriers to optimal care.