{"paper_id":"0e7d2d66-e0a6-4c87-aec1-6677017e4d63","body_text":"1\n1 Adherence Monitoring Methods to Measure Virological \n2 Failure in People Living with HIV on Long-Term \n3 Antiretroviral Therapy in Uganda\n4\n5 Stephen Okoboi 1, Joseph Musaazi 1, Rachel King 2,3, Sheri A. Lippman 4, Andrew \n6 Kambugu1, Andrew Mujugira 1,3, Jonathan Izudi 1, Rosalind Parkes-Ratanshi1,5, Agnes N. \n7 Kiragga1, and Barbara Castelnuovo1\n8\n9 Institutional affiliation \n10 1) Infectious Diseases Institute, Department of Medicine, Makerere University \n11 College of Health Sciences. \n12 2) Department of Global Health, University of California, San Francisco, San \n13 Francisco, CA, United States.\n14 3) School of Public Health, Makerere University, Kampala, Uganda.\n15 4) Division of Prevention Science, Department of Medicine, University of California, \n16 San Francisco, San Francisco, CA, United States.\n17 5) Clinical School, University of Cambridge, United Kingdom\n18\n19 Corresponding author\n20 Stephen Okoboi\n21 Infectious Diseases Institute, College of Health Sciences, Makerere University\n22 okoboi25@gmail.com / sokoboi@idi.co.ug \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n2\n23 Co-authors\n24\n25 JM: Joseph Musaazi; RK: Rachel King; SL: Sheri Lippman; AK: Andrew Kambugu; AM: \n26 Andrew Mujugira; JI: Jonathan Izudi; RP: Rosalind Parkes-Ratanshi; ANK: Agnes N \n27 Kiragga, and BC:  Barbara Castelnuovo\n28\n29\n30\n31\n32\n33\n34\n35\n36\n37\n38\n39\n40\n41\n42\n43\n44\n45\n46\n47\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n3\n48 Abstract\n49 Appointment keeping and self-report within 7-day or and 30-days recall periods are non-\n50 objective measures of antiretroviral treatment (ART) adherence. We assessed incidence \n51 of virological failure (VF), predictive performance and associations of these adherence \n52 measures with VF among adults on long-term ART. Data for persons initiated on ART \n53 between April 2004 and April 2005, enrolled in a long-term ART cohort at 10-years on \n54 ART (baseline) and followed until December 2021 was analyzed. VF was defined as two \n55 consecutives viral loads ≥1000 copies/ml at least within 3-months after enhanced \n56 adherence counselling. We estimated VF incidence using Kaplan-Meier and Cox-\n57 proportional hazards regression for associations between each adherence measure \n58 (analyzed as time-dependent annual values) and VF. The predictive performance of \n59 appointment keeping and self-reporting for identifying VF was assessed using receiver \n60 operating characteristic curves and reported as area under the curve (AUC).  We included \n61 900 of 1,000 participants without VF at baseline: median age was 47 years (Interquartile \n62 range: 41-51), 60% were women and 88% were virally suppressed. ART adherence was \n63 ≥95% for all three adherence measures. Twenty-one VF cases were observed with an \n64 incidence rate of 4.37 per 1000 person-years and incidence risk of 2.4% (95% CI: 1.6%-\n65 3.7%) over the 5-years of follow-up. Only 30-day self-report measure was associated with \n66 lower risk of VF, adjusted hazard ratio (aHR)=0.14, 95% CI:0.05–0.37). Baseline CD4 \n67 count ≥200cells/ml was associated with lower VF for all adherence measures. The 30-\n68 day self-report measure demonstrated the highest predictive performance for VF \n69 (AUC=0.751) compared to appointment keeping (AUC=0.674), and 7-day self-report \n70 (AUC=0.687). The incidence of virological failure in this study cohort was low. Whilst 30- \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n4\n71 day self-report was predictive, appointment keeping and 7-day self-reported adherence \n72 measures had low predictive performance in identifying VF. Viral load monitoring remains \n73 the gold standard for adherence monitoring and confirming HIV treatment response.\n74\n75 Keywords: Anti-retroviral Therapy, Long-term ART, Adherence, Self-report, Virological \n76 failure \n77 Word counts: 298 abstracts \n78\n79\n80\n81\n82\n83\n84\n85\n86\n87\n88\n89\n90\n91\n92\n93\n94\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n5\n95 Introduction \n96 The WHO public health approach for scaling up access to  anti-retroviral therapy  (ART)   \n97 expanded following the availability of highly active anti-retroviral medication through \n98 standardized regimes and decentralized care in Low and Middle Income Countries (LMIC) \n99 (1). In this approach, standardized simplified ART regimens and decentralized treatment \n100 delivery enabled large numbers of people with HIV (PHW) to be initiated and followed-up \n101 on treatment through public and private sector health facilities. The approach is centred \n102 on “four Ss”, an acronym for when to Start drug treatment, Substitute for toxicity, Switch \n103 after treatment failure, and Stop to enable lower-level healthcare workers to deliver \n104 appropriate care (1). \n105 In 2020, an estimated 17 million people were on ART in sub-Saharan Africa (SSA) (2);  \n106 in Uganda, 1,275,000 million persons were estimated to be on ART in 2019 (3).  The 2018 \n107 and 2020 HIV treatment guidelines in Uganda recommend ART adherence monitoring \n108 using non-objective measures including pill counts, appointment keeping, visual analogue \n109 scales, and self-reported pill use, used either individually or in combination (4,5). The use \n110 of these adherence measures encourages ART adherence discussions with patients and \n111 providing information about the risk of virological failure  or to support daily tablet-taking \n112 behavior in settings where viral load testing is limited (6–8). However, in 2016, the Uganda \n113 Ministry of Health (MoH) HIV treatment guidelines recommended annual plasma HIV viral \n114 load for people on ART to monitor treatment effectiveness and identify individuals with \n115 detectable viral load (9). Annual viral load monitoring is recommended due to scarcity of \n116 resources  in LMIC (10,11).\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n6\n117 Sustained optimal adherence to ART ensures virological suppression, reduction in HIV-\n118 related morbidity and mortality, and prevents onward transmission (6,12,13), popularized \n119 by the Joint United Nations Programme on HIV/AIDS (UNAIDS) as “Undetectable equals \n120 Untransmittable (U=U) (14,15). However, previous studies report discrepancies in ART \n121 adherence thresholds used. And adherence measured as a categorical or a continuous \n122 constructs from patients or clinic reports affecting  association and predictive performance \n123 between ART adherence measures and virological failure among PWH on ART (6,7,16–\n124 18). The performance of ART adherence measures in predicting virological failure among \n125 adult PWH on long-term ART (i.e., ≥10 consecutive years of ART use), including the \n126 predictors for virological failure are not well described across ART programs in LMIC. \n127 Despite self-reporting being routinely used as an adherence proxy in clinical care, few \n128 studies have evaluated the incidence of virological failure, predictive performance, and \n129 associations of appointment keeping, self-report within 7-day or and 30-days recall \n130 periods with VF among adults on long-term ART.\n131\n132 Thus, this study aimed to describe the incidence of virological failure, compare the \n133 predictive performance of three ART adherence measures (7-days and 30-days self-\n134 reported pill use, and appointment keeping) and assess factors associated with virological \n135 failure among PWH on long-term first-line ART. \n136\n137\n138\n139\n140\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n7\n141 Methods \n142 Study design and setting\n143 This study was conducted at the HIV Centre of Excellence at the Infectious Diseases \n144 Institute (IDI) located in the Mulago Teaching Hospital in Kampala, the capital city of \n145 Uganda. The IDI clinic is a large out-patient clinic that currently serves over 8,000 patients \n146 living with HIV in five municipalities in Kampala. \n147 This was a secondary analysis of a longitudinal cohort data of patients enrolled in the \n148 Long-Term ART cohort. The ART Long-Term cohort is an observational cohort of 1,000 \n149 patients who had been on ART for at least 10 years and were enrolled between May 2014 \n150 and September 2015 to be followed up for an additional 10 years (19). Patients were \n151 eligible and enrolled in the cohort if they were ≥18-years, were willing to participate in the \n152 cohort visits and comply with the study procedures, and were in their 10th consecutive \n153 year of WHO standard ART at IDI regardless of the combination of drugs for first-line \n154 ART. Ten-year consecutive ART use was determined using data collected in the IDI \n155 electronic database, known as the Integrated Clinic Enterprise Application (ICEA). This is \n156 an in-house built system based on Microsoft.NET technologies (19). This interim analysis \n157 describes the first five years of follow-up.\n158\n159 Data Collection \n160 General medical history, physical examination, adherence to ART, and prescription of \n161 drugs were performed at enrolment and all study visits. Follow-up visits were scheduled \n162 once a year for 10-years. In addition to study visits, the participants attended the general \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n8\n163 clinic every 3-months to pick up their ART and concomitant medications. Antiretroviral \n164 drugs are prescribed according to the WHO guidelines; patients with two consecutive viral \n165 loads >1000 copies/ml after enhanced adherence counselling are considered for \n166 treatment switch. Enhanced adherence counselling is a targeted counselling offered to \n167 PWH on ART with non-suppressed viral load, done every month for at least 3-months \n168 before the next viral load test (5).  At each study visit, real-time data entry into ICEA is \n169 performed by the respective providers (19). Laboratory results performed in the IDI Core \n170 Laboratory are automatically downloaded daily into the ICEA database. The \n171 questionnaires administered at each visit include basic demographic and epidemiological \n172 data, clinical history, adherence to ART, quality of life, and sexual behavior. Clinical data \n173 collected at each visit included vital signs and body weight, hematological and chemistry \n174 laboratory results, medications and ART regimen, and drug toxicities. All the data \n175 collected into ICEA are validated by a quality control and assurance officer who ensures \n176 that the data are complete and consistent.\n177\n178 Adherence Measures \n179 The primary outcome was virological failure defined as two consecutive plasma HIV RNA \n180 viral load measurements ≥1000 copies/ml at least within 3-months after receiving \n181 enhanced adherence counselling following the first viral load measurement. The \n182 exposure was ART adherence assessed using 3 different measures: self-reported pill use \n183 in the last 7 days, self-reported pill use in the last 30 days, and appointment keeping. The \n184 30-day and 7-day self-report of pill use ART adherence measure was assessed on a scale \n185 of 1-100 by asking the patient to recall the numbers of missed doses in the last 30-days \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n9\n186 and 7-days and then calculating, what percentage of ART doses were taken. Good \n187 adherence was determined as having a score ≥95%. Appointment keeping was defined \n188 as returning for a scheduled cohort clinic visit appointment or within a 7-day window after \n189 a missed clinic visit. Questions were assessed throughout the entire 5-year follow-up \n190 period. Additional co-variates included age, sex, marital status, employment status, HIV \n191 disclosure status, household level of income, and body mass index.\n192\n193 We extracted cohort data from 10 to 15 years on ART follow-up (or enrollment in the \n194 cohort and five years of follow-up). When the required data were missing, patient charts \n195 were retrieved and reviewed to supplement the data in the databases. We extracted \n196 clinical data including ART start dates and regimens, socio-demographics at cohort \n197 enrolment, behavioral data, CD4 cell counts, plasma HIV viral load measurements for the \n198 follow-up period using or Roche COBAS® Ampli Prep. We also extracted data on deaths, \n199 transferred out, and lost to follow-up.\n200\n201 Statistical analysis \n202 Statistical analysis was performed using STATA 16.1 (StataCorp, College Station, Texas). \n203 We described cohort participants using frequencies and percentages for categorical \n204 variables and continuous variables using means and standard deviations and medians \n205 and interquartile ranges. Adherence measures were described using frequency and \n206 percentages across calendar year. Kaplan-Meier methods were used to estimate \n207 incidence risk and incidence rate of virological failure. Associations between virological \n208 failure and ART adherence was examined using Cox-proportional hazards regression \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n10\n209 (Cox-PH) models. ART adherence measures were entered into the model as time-\n210 dependent covariates measured at annual cohort visits. The Schoenfeld residuals test \n211 was used to assess for violation of the Cox –PH assumption. Three sensitivity analyses \n212 were performed, by refitting the model when: 1) missing values on covariates were \n213 imputed by multiple imputation using chained equations (MICE), 2) considering all \n214 censored patients i.e., deaths and losses to follow-up as virological failure (worst-case \n215 scenario), and 3) when considered as non-virological failure (best-case scenario). \n216 Performance of ART adherence measures - appointment keeping, 30-days and 7-days \n217 self-report of pill use for predicting virological failure was evaluated using receiver \n218 operating characteristic curve analysis. All hypothesis tests were performed as 2-tailed \n219 tests at a 5% significance level.\n220\n221 Ethical approval\n222 This study was approved by the Infectious Diseases Institute Research Ethics Committee \n223 (reference number; IDI REC-041/2021) and the Uganda National Council for Science and \n224 Technology (reference number; HS1896ES). The IDIREC committee granted a waiver of \n225 informed consent since secondary data were retrieved and analysed. \n226\n227\n228\n229\n230\n231\n232\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n11\n233 Results \n234 Study Profile\n235 We retrieved data for 1,000 PWH adults enrolled in the long-term ART cohort who started \n236 ART between April 2004 and April 2005 and were followed up until December 2021. Of \n237 the 1,000 participants, 100 (10%) had a viral load (VL) >1000 copies/ml documented at \n238 cohort enrolment and were therefore excluded from the study. Nine hundred participants \n239 were included in the analysis, of whom 10 had transferred to other health facilities, 45 \n240 were lost to follow-up and 41 had died before reaching 15 years on ART (Fig 1). \n241 Participant characteristics  \n242 Participants’ description\n243 Of the 900 cohort participants analyzed, at cohort enrollment: the median age was 46 \n244 years (IQR 41- 51); 59.8% were females, 82.1% were employed, 43.9% lived <1 US dollar \n245 per day, median body mass index (BMI) was 22.4 (IQR 19.8-25.4), 51.3% were married \n246 or cohabiting, and 88.4% had viral load <50 copies/ml (10 years on ART), (Table 1).\n247\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n12\n248 Table 1: Baseline characteristics (at 10 years on ART)\nVariables                                           Statistics\n                                     N=900\nSex\n      Male 362 (40.2)\n      Female 538 (59.8)\nAge (years) at cohort 2 registration\n      Median (IQR) 46.0 (41.0, 51.0)\nAge categories\n      28-44 388 (43.1%)\n      45-54 368 (40.9%)\n      ≥55 144 (16.0%)\nBMI at baseline (cohort \nregistration)\n      Median (IQR) 22.4 (19.8, 25.4)\n      BMI <18kg/m2, N (%) 108 (12.6)\nVL at cohort 2registration (c/ml)\n      Median (IQR) 20.0 (20.0, 20.0)\nBaseline VL categories (copies/ml), \nN (%)\n      <50 635 (88.4)\n      ≥50 83 (11.6)\nCD4 at cohort2registration \n(cells/ml)\n      Median (IQR) 491.0 (347.0, 662.0)\n      CD4 <200 cells/ml, N (%) 36 (4.4)\nMarital status, N (%)\n   \nSingle/Separated/Divorced/Widowed\n438 (48.7)\nMarried/Cohabiting 462 (51.3)\nEmployed, N (%)\n      No 160 (17.9)\n      Yes 735 (82.1)\nHousehold monthly income (as \n<30$ vs ≥30$), N (%)\n      <30$ per month 374 (43.9)\n      ≥30$ per month 477 (56.1)\nDisclosure status, N (%)\n      No 830 (92.2)\n      Yes 70 (7.8)\n249 Table 1 footnote: SD denotes standard deviation, IQR interquartile range, BMI body mass \n250 index. Missing values: BMI (n=42, 5%), Baseline VL (n=182, n=20%), CD4 count (n=81, 9%), \n251 employed (n=5, 0.6%), household monthly income (n=49, 5%)\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n13\n252 The virological failure incidence rate was 4.37 (95% CI: 2.85 - 6.70) per 1000 person-\n253 years and the probability of virological failure was 2.4%, (95% CI: 1.6% - 3.7%) over 15 \n254 years, (Fig 2). ART adherence was very high (≥95%) over the 7-calendar years studied \n255 on all adherence measures, except appointment keeping declined in 2019 and 2020 \n256 (42.0% and 72.7%, respectively) (Table 2). \n257 Table 2: ART Adherence measures (self-reported and appointment keeping \n258 during 2014 – 2020\n 2014 2015 2016 2017 2018 2019 2020\n (N = 137) (N = 375) (N = 968) (N = 960) (N = 940) (N = 896) (N = 343)\n7-day self-reported \nadherence       \n      No 4 (2.9%) 9 (2.4%) 38 (3.9%) 28 (2.9%) 18 (1.9%) 9 (1.0%) 5 (1.5%)\n      Yes 133 (97.1%) 366 (97.6%) 929 (96.1%) 931 (97.1%) 922 (98.1%) 885 (99.0%) 329 (98.5%)\n30-day self-reported \nadherence score ≥95%)\n      No 1 (0.7%) 1 (0.3%) 241 (26.1%) 180 (18.8%) 72 (7.7%) 38 (4.3%) 8 (2.5%)\n      Yes 135 (99.3%) 373 (99.7%) 682 (73.9%) 776 (81.2%) 864 (92.3%) 851 (95.7%) 313 (97.5%)\nAppointment keeping \nadherence measure\n      No 5 (3.6%) 23 (6.2%) 65 (6.8%) 46 (4.9%) 64 (7.0%) 403 (58.0%) 83 (27.3%)\n      Yes 132 (96.4%) 346 (93.8%) 891 (93.2%) 896 (95.1%) 849 (93.0%) 292 (42.0%) 221 (72.7%)\n259 Missing data: Self-reported (0,0,1,1,0,2,9 for 2014, 2015, 2016, 2017, 2018, 2019, 2020 respectively), Appointment \n260 keeping (0,6,12,18,27,201,39 for 2014, 2015, 2016, 2017, 2018, 2019, 2020 respectively), VAS (1,1,45,4,4,7,22 for \n261 2014, 2015, 2016, 2017, 2018, 2019, 2020 respectively)\n262\n263 Associations between virologic failure and ART adherence measures\n264 Table 3 shows that after adjusting for other patient factors, ART adherence assessed \n265 using 30-day self-report was associated with lower risk of virological failure (adjusted \n266 hazard ratio [AHR] 0.14; 95% CI: 0.05 - 1.76). However, the relationship was not \n267 significant when ART adherence was measured using 7-day self-report or appointment \n268 keeping (AHR 0.36; 95% CI: 0.05 – 2.75 and AHR 2.27; 95% CI: 0.27–18.83), \n269 respectively. Among other patient factors, only baseline CD4 ≥200 cells/ml was \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n14\n270 associated with lower risk of virological failure in models including all three adherence \n271 measures: 30-day self-report (AHR 0.24; 95% CI: 0.07–0.85), 7-day self-report (AHR \n272 0.26; 95% CI: 0.08 – 0.91) and appointment keeping (AHR 0.22; 95% CI: 0.06 – 0.76). \n273 In sensitivity analyses, when imputing missing covariate data, the association between \n274 adherence and virological failure only remained for 30-day ART self-report (AHR 0.14; \n275 95% CI: 0.05 – 0.35). There was no significant association with 7-day self-reported or \n276 appointment keeping measures (Supplementary tables).\n277\n278\n279\n280\n281\n282\n283\n284\n285\n286\n287\n288\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n15\n289 Table 3: Associations of ART adherence and virologic failure, adjusting for \n290 baseline and time-dependent patient characteristics\n291\n292 Footnote: ART adherence modelled as time-dependent covariates in all models. Analysis performed on complete \n293 cases for all covariates adjusted in the model: 11% (100/900) missing in model 1 and 2, whereas 12% (105/900) missing \n294 in model 3. aHR denotes adjusted hazard ratio from Cox proportional hazard regression models, CI denotes confidence \n295 interval. Apart from adherence (the main exposure), in the adjusted models we Included only covariates with P \n296 value<0.2 at unadjusted model. Age groups, marital status and household monthly income had P values>0.2 in \n297 unadjusted Cox models, and thus were excluded from adjusted models.\n298\n299 Adherence predictive performance of virologic failure\n300 In the receiver operating characteristics curve (ROC) analysis for predictivity ability of \n301 adherence measures for virological failure, 30-day self-report best predicted virological \n302 failure (area under the curve [AUC] 0.751; 95% CI: 0.66 - 0.90) versus appointment \n303 keeping (AUC 0.674; 95% CI: 0.53 - 0. 81) and 7-day self-report (AUC 0.687; 95% CI: \n304 0.51-0.82) (Fig 3).\nFactor Model 1\n30-day self-report ART \nadherence measure \n (as main exposure) \n(n=800) \nModel 2\n7-day self-report ART \nadherence measure\n (as main exposure) (n=800) \nModel 3\nART adherence measure \nAppointment keeping (as main \nexposure)\n (n=795) \nART adherence \n(time-updated)\naHR (95% CI) P value aHR (95% CI) P value aHR (95% CI) P value\nNon-adherent 1 1 1\nAdherent 0.14 (0.05 – 0.37) <0.001 0.36 (0.05 – 2.75) 0.325 2.27 (0.27 – 18.83) 0.447\nOther factors \nadjusted\nSex\nMale 1 1 1\nFemale 2.29 (0.74 – 7.07) 0.149 2.46 (0.81 – 7.50) 0.133 2.84 (0.80 – 10.06) 0.106\nCD4 count at \ncohort registration \n(cells/mL)\n<200 1 1 1\n≥200 0.24 (0.07 – 0.85) 0.027 0.26 (0.08 – 0.91) 0.035 0.22 (0.06 – 0.76) 0.017\nEmployment \nstatus\nUnemployed 1 1 1\nEmployed 0.57 (0.21 – 1.55) 0.275 0.57 (0.22 – 1.53) 0.267 0.60 (0.21 – 1.76) 0.353\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n16\n305 Discussion\n306 Overall, we found that unlike 30-day self-report, appointment keeping and 7-day self-\n307 reported adherence measures had low predictive performance in identifying virological \n308 failure. We also found a low incidence of virological failure among person living with HIV \n309 in this long-term ART study cohort. Our finding of low incidence of virological failure is \n310 consistent with an observational cohort analysis conducted in Uganda at The AIDS \n311 Support Organization among 3,340 persons who initiated ART from 2004 - 2009 and \n312 followed-up for a median of 5.7 years (IQR, 4.1 - 7.2 years) which found a low rate of \n313 virological failure among adult HIV patients on first-line antiretroviral therapy (20). Our \n314 study reports a low incidence of virological failure comparable to 7.4% reported among \n315 participants in the first Infectious Diseases Institute cohort followed up-to 10 years  on \n316 ART (21). The lower incidence of virological failure observed in our study could be \n317 attributed to the longer duration on ART and the fixed dose once-daily therapy which was \n318 introduced around the time of cohort enrollment (19,22,23).  Participants on long-term \n319 ART have received many ART adherence counselling sessions  that should increase \n320 awareness about the importance of ART adherence and possibly drug side effects (21, \n321 22). Furthermore, this is a survivor cohort of persons who have been on ART for at least \n322 ten years; there is evidence that shorter ART duration is conversely associated with \n323 increased risk of virological failure (23). This is due to the recent policy of universal test \n324 and treat, which increase the likelihood that patients with a new diagnosis of HIV tend to \n325 be unprepared to start ART as they have had limited psychosocial support due to rapid \n326 initiation (23,24). The issues of stigma, sero-status disclosure to people close to them \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n17\n327 and discrimination is another concern among people on ART as it increases non-\n328 adherence to treatment (23,24).  \n329\n330 We found that baseline CD4 cells ≥200 cells/ml was associated with lower risk of \n331 developing virological failure in models using each of the three adherences measures. \n332 Our finding is similar to several studies both in developed and LMIC that have reported \n333 low baseline CD4 count have increased risk of developing virological failure (26). This \n334 finding is supported by other studies which recommended closer monitoring and ART \n335 adherence counselling for persons who commence ART with low CD4 count (26,27). \n336  \n337 We also found that all the three adherence measures had low predictive performance in \n338 identifying virological failure. However, the 30-day self-report adherence measure was \n339 most able to predict virological failure. Our finding that 30-day self-report predicted \n340 virological failure is consistent with a study by Minyi et al., 2008 (28) who found that 1-\n341 month self-report ART adherence was more accurate in measuring ART adherence and \n342 predicating virological failure than 3-day or 7-day self-reported ART adherence (28–30). \n343 Other studies conducted in sub-Saharan Africa have found that self-report adherence \n344 measures have low predictive performance in detecting virological failure among \n345 participants on long and short-term ART (29). This could be because each of these \n346 adherence measures has inherent weakness such us their accuracy and precision due \n347 to recall and social desirability in different settings (25).  Therefore, viral load monitoring \n348 as per WHO, remains the gold standard for identifying virological failure, monitoring ART \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n18\n349 adherence, and confirming treatment failure among people on ART even if its scale-up in \n350 resource-limited setting is hindered by financial and technical constraints (31).\n351   \n352 To improve ART adherence, HIV care programs in LMIC should continue to educate \n353 people living with HIV on the importance of reporting accurate and consistent ART \n354 adherence, keeping dosing schedules, and explaining adverse effects. National ART \n355 guidelines should pay particular attention to monitoring virological failure and supporting \n356 ART adherence among persons who initiate ART with lower CD4 count.  Monitoring viral \n357 load helps identify PWH on ART who have sustained long-term viral load suppression \n358 and is crucial in HIV prevention efforts given that national programs are promoting the \n359 UNAIDS slogan of undetectable equals to untransmissible (U=U). As we disseminate and \n360 implement the UNAIDS policy, programs should integrate objective methods of \n361 measuring adherence like medication event monitoring systems, and biologic measures \n362 like point of care tenofovir testing that best predict virological failure.\n363 The key strength of our study is the prospective data collection design among long-term \n364 ART persons, large sample size, long duration of follow-up, and objective ascertainment \n365 of virological failure. However, our study has limitations. These findings may not be \n366 generalizable because persons were from an HIV centre of excellence, which may not be \n367 representative for smaller centres or primary care settings, social desirability bias, \n368 immortal time bias could have affected the study findings. Also, this is a non-randomized \n369 comparison and is subject to unmeasured confounding. Twenty percent of viral load data \n370 were missing but we used multiple imputation in the sensitivity analysis.\n371\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n19\n372 Conclusion \n373 The incidence of virological failure among PWA on long-term ART in this cohort study \n374 was low. Unlike the 30-day self-report, appointment keeping and 7-day self-reported ART \n375 adherence measures had low predictive performance in identifying virological failure. \n376 Routine plasma viral load monitoring remains the gold standard for adherence monitoring \n377 and confirming HIV treatment response.\n378\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n20\n379 Acknowledgements\n380 This project was supported by the Fogarty International Center of the National Institutes \n381 of Health (NIH) under Award Number D43TW009343 and the University of California \n382 Global Health Institute (UCGHI). The content is solely the responsibility of the authors \n383 and does not necessarily represent the official views of the NIH or UCGHI. BC was partly \n384 supported by the Fogarty International Centre, National Institute of Health (grant# \n385 2D43TW009771-06 “HIV and co-infections in Uganda\"). \n386\n387 Declaration of conflicts of interest \n388 The authors declare no conflict of interest\n389 Author Contributions\n390 Conceptualization: Stephen Okoboi, Barbara Castelnuovo, Rachel King\n391 Data curation: Stephen Okoboi and Joseph Musaazi\n392 Formal analysis: Stephen Okoboi, Joseph Musaazi, Izudi Jonathan, Sheri A. Lippman\n393 Methodology: Stephen Okoboi, Joseph Musaazi, Rachel King, Sheri A. Lippman, \n394 Andrew Kambugu, Andrew Mujugira, Jonathan Izudi, Rosalind Parkes-Ratanshi, Agnes \n395 N. Kiragga, and Barbara Castelnuovo\n396 Project administration: Stephen Okoboi\n397 Supervision: Rachel King, Sheri A. Lippman and Barbara Castelnuovo\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n21\n398 Review and approval: Stephen Okoboi, Joseph Musaazi, Rachel King, Sheri A. \n399 Lippman, Andrew Kambugu, Andrew Mujugira, Jonathan Izudi, Rosalind Parkes-\n400 Ratanshi, Agnes N. Kiragga, and Barbara Castelnuovo\n401\n402 References \n403 1. Ford N, Ball A, Baggaley R, Vitoria M, Low-Beer D, Penazzato M, Vojnov L, \n404 Bertagnolio S, Habiyambere V, Doherty M, Hirnschall G. The WHO public health \n405 approach to HIV treatment and care: looking back and looking ahead. Lancet \n406 Infect Dis. 2018 Mar;18(3):e76-e86. doi: 10.1016/S1473-3099(17)30482-6. Epub \n407 2017 Oct 20. PMID: 29066132 \n408 2. UNAIDS. FACT SHEET 2021 Global HIV Statistics. End AIDS epidemic. \n409 2021;(June):1–3. \n410 3. Ministry of Health Uganda Population - Based HIV I Mpact a Ssessment. 2019;0–\n411 252. \n412 4. Ministry of Health-Uganda. Consolidated Guidelines for Prevention. \n413 2016;(December). \n414 5. Uganda Ministry of Health. Consolidated Guidelines on the prevention and \n415 treatment of HIV. 2018;(September). \n416 6. Sangeda RZ, Mosha F, Prosperi M, Aboud S, Vercauteren J, Camacho RJ, et al. \n417 Pharmacy refill adherence outperforms self-reported methods in predicting HIV \n418 therapy outcome in resource-limited settings. BMC Public Health. 2014;14(1):1–\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n22\n419 11. \n420 7. Aye WL, Puckpinyo A, Peltzer K. Non-adherence to anti-retroviral therapy among \n421 HIV infected adults in Mon State of Myanmar. BMC Public Health. 2017;17(1):1–\n422 10. \n423 8. Mekuria LA, Nieuwkerk PT, Yalew AW, Sprangers MAG, Prins JM. High level of \n424 virological suppression among HIVinfected adults receiving combination \n425 antiretroviral therapy in Addis Ababa, Ethiopia. Antivir Ther. 2016;21(5):385–96. \n426 9. Ministry of Health-Uganda. Consolidated Guidelines for the Prevention and \n427 Treatment of HIV and AIDS in Uganda. Minist Heal Uganda [Internet]. \n428 2020;(February):142–70. Available from: \n429 https://uac.go.ug/sites/default/files/Consolidated HIV Guidelines 2020.pdf\n430 10. Médecins Sans Frontières. How Low Can We Go? Pricing for HIV Viral Load \n431 Testing in Low- and Middle-Income Countries. 2013;41(0):1–8. \n432 11. Rouet F, Rouzioux C. HIV-1 viral load testing cost in developing countries: What’s \n433 new? Expert Rev Mol Diagn. 2007;7(6):703–7. \n434 12. Wu P, Johnson B, Nachega J, Wu B, Ordonez C, Hare A, et al. The Combination \n435 of Pill Count and Self-Reported Adherence is a Strong Predictor of First-Line ART \n436 Failure for Adults in South Africa. Curr HIV Res. 2014;12(5):366–75. \n437 13. Kabore L, Muntner P, Chamot E, Zinski A, Burkholder G, Mugavero MJ. Self-\n438 report measures in the assessment of antiretroviral medication adherence: \n439 Comparison with medication possession ratio and HIV viral load. J Int Assoc \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n23\n440 Provid AIDS Care. 2015;14(2):156–62. \n441 14. Okoli C, Van De Velde N, Richman B, Allan B, Castellanos E, Young B, et al. \n442 Undetectable equals untransmittable (U = U): Awareness and associations with \n443 health outcomes among people living with HIV in 25 countries. Sex Transm Infect. \n444 2021;97(1):18–26. \n445 15. Bor J, Fischer C, Modi M, Richman B, Kinker C, King R, et al. Changing \n446 Knowledge and Attitudes Towards HIV Treatment-as-Prevention and \n447 “Undetectable = Untransmittable”: A Systematic Review. AIDS Behav [Internet]. \n448 2021;(0123456789). : AIDS Behav. 2021 Dec;25(12):4209-4224 Available from: \n449 https://doi.org/10.1007/s10461-021-03296-8\n450 16. Haberer JE, Bwana BM, Orrell C, Asiimwe S, Amanyire G, Musinguzi N, et al. \n451 Adherence in early versus late ART initiation in sub-Saharan Africa. Top Antivir \n452 Med. 2018;26 (Supple:214s-215s. \n453 17. Gare J, Kelly-Hanku A, Ryan CE, David M, Kaima P, Imara U, et al. Factors \n454 influencing antiretroviral adherence and virological outcomes in people living with \n455 HIV in the Highlands of Papua New Guinea. PLoS One. 2015;10(8). \n456 18. Wu P, Johnson B, Nachega J, Wu B, Ordonez C, Hare A, et al. The Combination \n457 of Pill Count and Self-Reported Adherence is a Strong Predictor of First-Line ART \n458 Failure for Adults in South Africa. Curr HIV Res. 2014;12(5):366–75. \n459 19. Castelnuovo B, Mubiru F, Kiragga AN, Musomba R, Mbabazi O, Gonza P, et al. \n460 Antiretroviral treatment Long-Term (ALT) cohort: A prospective cohort of 10 years \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n24\n461 of ART-experienced patients in Uganda. BMJ Open. 2018;8(2):1–8. \n462 20. Okoboi S, Ding E, Persuad S, Wangisi J, Birungi J, Shurgold S, et al. Community-\n463 based ART distribution system can effectively facilitate long-term program \n464 retention and low-rates of death and virologic failure in rural Uganda. AIDS Res \n465 Ther. 2015;12(1):37. \n466 21. Castelnuovo B, Kiragga A, Musaazi J, Sempa J, Mubiru F, Wanyama J, et al. \n467 Outcomes in a cohort of patients started on antiretroviral treatment and followed \n468 up for a decade in an urban clinic in Uganda. PLoS One. 2015;10(12):1–11. \n469 22. Bukenya D, Mayanja BN, Nakamanya S, Muhumuza R, Seeley J. What causes \n470 non-adherence among some individuals on long term antiretroviral therapy? \n471 Experiences of individuals with poor viral suppression in Uganda. AIDS Res Ther \n472 [Internet]. 2019;16(1):1–9. Available from: https://doi.org/10.1186/s12981-018-\n473 0214-y\n474 23. Lippman SA, El Ayadi AM, Grignon JS, Puren A, Liegler T, Venter WDF, et al. \n475 Improvements in the South African HIV care cascade: findings on 90-90-90 \n476 targets from successive population-representative surveys in North West \n477 Province. J Int AIDS Soc. 2019;22(6). \n478 24. Damulak PP, Ismail S, Manaf RA, Said SM, Agbaji O. Interventions to improve \n479 adherence to antiretroviral therapy (Art) in sub-saharan africa: An updated \n480 systematic review. Int J Environ Res Public Health. 2021;18(5):1–18. \n481 25. Orrell C, Cohen K, Leisegang R, Bangsberg DR, Wood R, Maartens G. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n25\n482 Comparison of six methods to estimate adherence in an ART-naïve cohort in a \n483 resource-poor setting: Which best predicts virological and resistance outcomes? \n484 AIDS Res Ther. 2017;14(1):1–11. \n485 26. Leierer G, Grabmeier-Pfistershammer K, Steuer A, Sarcletti M, Geit M, Haas B, et \n486 al. A single quantifiable viral load is predictive of virological failure in human \n487 immunodeficiency virus (HIV)-infected patients on combination antiretroviral \n488 therapy: The austrian HIV cohort study. Open Forum Infect Dis. 2016;3(2):1–9. \n489 27. Doyle T, Smith C, Vitiello P, Cambiano V, Johnson M, Owen A, et al. Plasma HIV-\n490 1 RNA detection below 50 copies/mL and risk of virologic rebound in patients \n491 receiving highly active antiretroviral therapy. Clin Infect Dis. 2012;54(5):724–32. \n492 28. Minyi, Safren SA, Skolnik PR, Rogers WH, Coady W, Hardy H, et al. Optimal \n493 recall period and response task for self-reported HIV medication adherence. AIDS \n494 Behav. 2008;12(1):86–94. \n495 29. Finitsis DJ, Pellowski JA, Huedo-Medina TB, Fox MC, Kalichman SC. Visual \n496 analogue scale (VAS) measurement of antiretroviral adherence in people living \n497 with HIV (PLWH): a meta-analysis. J Behav Med. 2016 Dec;39(6):1043-1055. doi: \n498 10.1007/s10865-016-9770-6. \n499 31. World Health Organization. Updated recommendations on HIV prevention, infant \n500 diagnosis, antiretroviral initiation and monitoring. 2021. 67 p. \n501\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted May 17, 2022. ; https://doi.org/10.1101/2022.05.11.22274977doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}