Abstract
Background
After scale-up of antiretroviral therapy (ART), routine annual viral load monitoring has been adopted by most countries, but reduced frequency of viral load monitoring may offer cost savings in resource-limited settings. We investigated if viral load monitoring frequency could be reduced while maintaining detection of treatment failure.
Methods
The Rakai Health Sciences Program performed routine, biannual viral load monitoring on 2489 people living with human immunodeficiency virus (age ≥15 years). On the basis of these data, we built a 2-stage simulation model to compare different viral load monitoring schemes. We fit Weibull regression models for time to viral load >1000 copies/mL (treatment failure), and simulated data for 10 000 individuals over 5 years to compare 5 monitoring schemes to the current viral load testing every 6 months and every 12 months.
Results
Among 7 monitoring schemes tested, monitoring every 6 months for all subjects had the fewest months of undetected failure but also had the highest number of viral load tests. Adaptive schemes using previous viral load measurements to inform future monitoring significantly decreased the number of viral load tests without markedly increasing the number of months of undetected failure. The best adaptive monitoring scheme resulted in a 67% reduction in viral load measurements, while increasing the months of undetected failure by <20%.
Conclusions
Adaptive viral load monitoring based on previous viral load measurements may be optimal for maintaining patient care while reducing costs, allowing more patients to be treated and monitored. Future empirical studies to evaluate differentiated monitoring are warranted.
Keywords: HIV, viral load monitoring, differentiated care, antiretroviral therapy, modeling
Adaptive viral load monitoring based on previous viral load measurements may be optimal for maintaining patient care while reducing costs, allowing more patients to be treated and monitored. Future empirical studies to evaluate differentiated monitoring are warranted.
Over the past decade there has been a marked increase in antiretroviral therapy (ART), with 21 million people living with human immunodeficiency virus (HIV) receiving ART by 2017 [1–3]. Despite this scale-up, viral load (VL) monitoring of individuals on ART has been limited due to the cost and complexity of testing. Evidence that immunologic monitoring cannot appropriately identify individuals with treatment failure has shifted guidelines toward expanded access to routine VL monitoring [4–6]. This, coupled with reductions in the cost of VL testing, has resulted in increased access and scale-up of routine VL monitoring in many countries [7], but there is limited evidence to guide the optimal frequency of VL monitoring. Studies support the use of early VL monitoring to identify early adherence problems [4] and to maintain adherence and prevent viral escape in the longer term [8–10]. Reduced frequency of CD4 and/or VL monitoring may offer cost savings to countries with limited resources by allowing for differentiated VL monitoring to reduce patient and program burden [11, 12].
Several studies have shown that reduction in or elimination of CD4 monitoring is a safe, cost-saving strategy [4, 13]. As VL monitoring becomes less expensive and newer technologies such as point-of-care VL enter the market [14–17], opportunities exist to refine and improve current monitoring strategies to maximize both patient benefit and resource utilization. We modeled the frequency of VL monitoring to minimize the duration of time prior to detection of treatment failure while reducing the number of HIV VL tests performed [13].
METHODS
HIV Population and ART Treatment
The Rakai Health Sciences Program has provided free ART since June 2004. Starting in 2005, routine biannual VL monitoring was introduced for all patients on ART. Our study included 2489 initially ART-naive participants, aged 15 years and older, who initiated ART between June 2005 and June 2011. Of these, 2229 had at least 1 measured VL within 6 months of ART initiation and at least 1 more VL at least 6 months after ART initiation; these 2229 participants were included in our simulation models.
Initial treatment regimens consisted of 2 nucleoside reverse transcriptase inhibitors (zidovudine or stavudine plus lamivudine) and nevirapine or efavirenz. Participants were seen weekly for the first month and then biweekly for 2 months followed by monthly follow-up with adherence and HIV risk reduction counseling at all visits. HIV type 1 (HIV-1) VL testing was performed on plasma using the Roche Amplicor 1.5 Monitor assay (Roche Diagonistics, Indianapolis, Indiana) until September 2010 when the Abbott Real-time m2000 assay was used (Abbott Laboratories, Lake Forest Illinois).
Statistical Methods
We defined unsuppressed viral load using World Health Organization guidelines of a viral load >1000 copies/mL. We limited our outcomes as follows: stage 1, VL ≤1000 copies/mL; stage 2, consecutive VL >1000 copies/mL; and death. Subjects starting in stage 1 can stay in stage 1 or progress to stage 2. Subjects starting in stage 2 can similarly remain in that stage or revert to stage 1. Subjects may die while in either stage, although the risk of dying in stage 1 is much lower. The possible stages and pathways are displayed in Figure 1. We used data 6 months after ART initiation from the cohort to fit 4 Weibull regression models for time to stage change (stage 1 → stage 2, or stage 2 → stage 1, stage 1 → death, stage 2 → death), including age at stage entry as a predictor. Using these models, we simulated monthly data for 10 000 subjects over 5 years after 6 months on ART. We randomly generated ART initiation age and the stage at study entry using nonparametric bootstrap sampling from the cohort. We then simulated the stage for each subject at each time point, and the estimated time to stage change based on the Weibull model fits. To verify the model fits, we compared the simulated data to the empirical data to ascertain whether failure rates per person-years were comparable. We summarized the performance of each monitoring scheme on bootstrap samples of 1000 simulated subjects to obtain point estimates and 95% confidence intervals (CIs), for the months of failure, number of subjects with a year or more of failure, number of viral load measurements for each scheme, and number of deaths from stage 2. We also repeated these simulations ignoring death.
Figure 1.
Simulated stages and pathways to virological failure after antiretroviral therapy initiation in people living with human immunodeficiency virus. Abbreviation: VL, viral load.
We investigated 7 different monitoring schemes to assess the minimum number of VL tests required to maintain or reduce the number of VL failures and deaths. Two schemes conformed to routine monitoring of all subjects on a fixed schedule of every 6 months (scheme 1) or 12 months (scheme 2). Five schemes had viral load monitoring at varying frequency as needed based on the subjects’ risk of virological failure after ART initiation; we call these schemes “value based” because they are informed by previous VL measurements and are potentially modified with each VL measurement. Scheme 1 was the clinical practice in this setting at the time of data collection and the most frequent nonadaptive scheme of the 7 schemes compared and defined as the benchmark (Table 1).
Table 1.
Description of the Viral Load Monitoring Schemes
| Scheme | Viral Load Testing, Copies/mL Post–ART Initiation | Notes | ||||||
|---|---|---|---|---|---|---|---|---|
| Type | No. | 6 mo | 12 mo | 15 mo | 18 mo | Monitor Frequency | Per Scheme Scenario | Per Scheme |
| Routine monitoring schemes | 1 | All | All | … | … | 6 mo | … | Benchmark |
| 2 | All | … | … | All | 12 mo | … | … | |
| Targeted monitoring schemes | 3 | ≤1000 | ≤1000 | … | … | 24 mo | … | … |
| >1000 | … | … | 12 mo | … | ||||
| >1000 | ≤1000 | … | … | 24 mo | … | |||
| >1000 | … | … | NA | Virologic failure | ||||
| 4 | ≤1000 | ≤1000 | … | … | 36 mo | … | … | |
| >1000 | ≤1000 | … | 24 mo | … | ||||
| >1000 | … | NA | Virologic failure | |||||
| >1000 | ≤1000 | … | … | 36 mo | … | |||
| >1000 | ≤1000 | … | 24 mo | … | ||||
| >1000 | … | NA | Virologic failure | |||||
| 5 | ≤1000 | ≤1000 | … | … | 36 mo | … | If a VL >1000 is observed at any point, check 3 mo later | |
| >1000 | ≤1000 | … | 24 mo | … | ||||
| >1000 | … | NA | Virologic failure | |||||
| >1000 | ≤1000 | … | … | 24 mo | … | |||
| >1000 | … | … | NA | Virologic failure | ||||
| 6 | ≤1000 | ≤1000 | … | … | 36 mo | … | … | |
| >1000 | … | ≤1000 | 24 mo | … | ||||
| … | >1000 | NA | Virologic failure | |||||
| >1000 | ≤1000 | … | ≤1000 | 36 mo | … | |||
| >1000 | … | ≤1000 | 24 mo | … | ||||
| … | >1000 | NA | Virologic failure | |||||
| 7 | ≤1000 | ≤1000 | … | … | 36 mo | … | If a VL >1000 is observed at any point, check 6 mo later | |
| >1000 | … | ≤1000 | 24 mo | … | ||||
| … | >1000 | NA | Virologic failure | |||||
| >1000 | ≤1000 | … | ≤1000 | 24 mo | … | |||
| >1000 | … | … | NA | Virologic failure | ||||
Abbreviations: ART, antiretroviral therapy; NA, not applicable; VL, viral load.
The first 2 schemes measure VL every 6 months or every 12 months, while all other schemes use previous VL levels to determine the time to next VL test based on previous VL measurements. All evaluations followed the same basic rules:
Virological failure (VF) was defined as 2 consecutive unsuppressed VL measurements >1000 copies/mL.
-
When a subject has a detected VF, they may be:
Switched to second-line ART successfully and removed from further follow-up (87% chance based on empirical data), or
Not switched to second-line ART or switched to second-line ART unsuccessfully, which in either case continues to count as a VL failure (13% chance) and at risk of death. All VL >1000 are counted as unobserved failures unless they have been detected and removed from follow-up per 2a.
VF-related deaths are those that occur in subjects that are failing per 2b but either have not yet been detected as failing or have been detected and unsuccessfully switched to second-line ART (13% as in 2b).
In addition, we applied schemes 1 and 2 to the observational cohort to assess performance. Only subjects with VL measurements at 6 and 12 months after starting therapy were used in this summary analysis.
For costing analyses, we assume a VL test to cost US$20 including reagents and technician time. This does not include the equipment costs or depreciation, which are regarded as a “sunk cost” because facilities and equipment have already been purchased.
Ethics and Participant Consent
The study used retrospective, routinely collected, de-identified clinical data and was approved by the Uganda Virus Research Institute Research and Ethics Committee, the Johns Hopkins University Institutional Review Board, and the Uganda National Council for Science and Technology. Individual written consent was obtained for treatment.
RESULTS
The 2229 participants are summarized in Table 2. The majority were female (66%) and aged 26–45 years. A total of 15 755 viral load tests were done on the 2229 participants included in the models. There were 64 deaths (2.9%) and 142 virological failures (6.4%) of which 97 (68.3%) switched to second-line ART. The Kaplan-Meier cumulative probability of virological failure was 5.4% (95% CI, 4.51%–6.57%) and 6.7% (95% CI, 5.6%–7.9%) at 2 and 3 years after ART initiation, respectively. The average time in the study was approximately 3 years. Using these data, we simulated data for 10 000 similar patients and 5 years of follow-up post–ART initiation with simulated VL measurements every 6 months.
Table 2.
Demographic Characteristics of Observed Cohort Clients Initiated in Antiretroviral Therapy
| Category | No. | (%) |
|---|---|---|
| Sex | ||
| Male | 760 | (34) |
| Female | 1469 | (66) |
| Age, y | ||
| 15–25 | 261 | (12) |
| 26–35 | 1014 | (45) |
| 36–45 | 699 | (31) |
| 46–55 | 207 | (9) |
| ≥56 | 48 | (2) |
| Baseline WHO stage | ||
| NA | 8 | (<1) |
| 1 | 797 | (36) |
| 2 | 825 | (37) |
| 3 | 440 | (20) |
| 4 | 159 | (7) |
| Baseline VL category, copies/mL | ||
| <1000 at 6 mo; <1000 at 12 mo | 1816 | (81) |
| <1000 at 6 mo; ≥1000 at 12 mo | 96 | (4) |
| ≥1000 at 6 mo; <1000 at 12 mo | 103 | (5) |
| ≥1000 at 6 mo; ≥1000 at 12 mo | 64 | (3) |
| Missing 6 mo and/or 12 mo | 150 | (7) |
| Baseline CD4 count, cells/μL | ||
| 0–100 | 545 | (24) |
| 101–200 | 780 | (35) |
| 201–300 | 830 | (37) |
| 301–400 | 37 | (2) |
| ≥401 | 35 | (2) |
| Missing | 2 | (<1) |
| Adverse outcomes | ||
| Death during follow-up | 64 | (3) |
| Failure during follow-up | 142 | (6) |
| Switch to second-line ART during follow-up | 97 | (4) |
Abbreviations: ART, antiretroviral therapy; NA, not applicable; VL, viral load; WHO, World Health Organization.
In the simulated data, routine VL monitoring every 6 months (scheme 1) minimized months of failure but required the most VL tests. Schemes 3–7 increased failure time by 18%–25% compared to testing every 6 months. Scheme 2, which measures VL every 12 months, had the highest number of subjects with 12 or more months of failure time, and the highest total number of months of failure. Nonetheless, scheme 2 required more VL measurements than any of the other schemes except the gold standard of every 6 months (scheme 1). While the total months of failure were similar among all schemes, scheme 1 (VL every 6 months) had a significantly lower number of subjects with 12 or more months of failure time than all other schemes (Table 3).
Table 3.
Results of Monitoring Schemas Applied to Simulated Data
| Schemea | Mean Failure, mo (95% CI) | Mean Subjects With ≥12 mo of Failure (95% CI) | Mean VL Measurement, copies/mL (95% CI) | Mean Cost of Monitoringb, US$ (95% CI) | Mean Failure Deaths, no. (95% CI) |
|---|---|---|---|---|---|
| No monitoring | 3560 (3088–4092) | 124 (104–144) | … | … | 24 (15–34) |
| 1 (every 6 mo) | 1988 (1747–2251) | 33 (22–44) | 9055 (8914–9192) | 181 100 (178 280–183 840) | 12 (5–19) |
| 2 (at 6 mo, then every 12 mo) | 2456 (2135–2779) | 83 (65–101) | 5603 (5536–5669) | 112 060 (110 720–113 380) | 15 (8–24) |
| 3 | 2691 (2313–3097) | 83 (65–101) | 3855 (3817–3893) | 77 100 (76 340–77 860) | 17 (9–26) |
| 4 | 2381 (2060–2741) | 63 (48–78) | 2984 (2955–3012) | 59 680 (59 100–60 240) | 15 (8–23) |
| 5 | 2382 (2066–2747) | 63 (48–78) | 3005 (2975–3033) | 60 100 (59 500–60 660) | 15 (8–23) |
| 6 | 2501 (2158–2873) | 71 (54–87) | 2971 (2946–2997) | 59 420 (58 920–59 940) | 16 (8–25) |
| 7 | 2503 (2176–2873) | 71 (54–87) | 3001 (2973–3029) | 60 020 (59 460–60 580) | 16 (8–25) |
Means and 95% CIs based on 1000 samples of 1000 subjects selected with replacement from the 10 000-subject simulated population.
Abbreviations: CI, confidence interval; VL, viral load.
aSee Table 1 for details on monitoring schema.
bCost of viral load testing was assumed to be US$20 including reagents and technician time.
Scheme 6 uses the least amount of resources, on average <3000 VL tests, as compared to the >9000 VL tests needed with scheme 1 monitoring every 6 months. However, schemes 4 and 5 equivalently had the best balance between lowest number of VL measurements and lowest number with 12 or more months of failure or the fewest months of failure, with the second lowest number of tests. When we ignored death, these results were similar (Supplementary Table 1).
The application of schemes 1 and 2 to the empirical data are given in Supplementary Table 2. Numbers of failures are also given in the Supplementary Appendix under each scheme, as well as months of “missed” failures, but these are difficult to interpret because the patterns of both viral load suppression and of missingness in our observed data are likely related to the actual monitoring and might be expected to change with different monitoring and feedback to the participants.
DISCUSSION
In our simulated data we demonstrate that adaptive VL monitoring can reduce the number of VL tests while maintaining an adequate detection of treatment failure. Adaptive VL monitoring may be approached to providing ART roll-out and monitoring for all patients in resource limited countries where VL monitoring is challenging. Although the months of undetected ART failure were higher in all the adaptive schemes compared to monitoring every 6 months, the 6-month monitoring may not be feasible in resource-limited settings. A more realistic standard of comparison might be 12-month monitoring (scheme 2), in which case all of our proposed adaptive methods improve both monitoring outcomes and minimize frequency of monitoring, especially after the first 12-month viral load test [18, 19].
As the number of people living with HIV started on ART has increased and quality of care becomes a concern, differentiated care delivery that adapts HIV services to patient needs has been proposed, including differentiated VL monitoring frequency [12]. The adaptive VL monitoring schemes proposed in our simulation are consistent with differentiated care service delivery strategies of monitoring people living with HIV based on need, since they allow for a VL monitoring frequency based on the individual patient’s viral suppression and stability of health [11, 17].
Our methods are limited; we used a 3-stage agent-based model. Due to the limitation of these data, we were not able to reliably fit models for more stages or model beyond 5 years after ART initiation Thus, these results only hold under our assumed data generation. While our model suggests schemes 4 and 5 as the best of the “value-based” strategies evaluated, we acknowledge that this is still less than an ideal goal of minimizing months of undetected viremia due to the potential for HIV transmission and acquired HIV drug resistance. Our data also are drawn from a time where nonnucleoside reverse transcriptase inhibitor–based regimens were routine, which is currently shifting with integrase inhibitors for first-line regimens. Also, generalizability of our findings is limited to settings with similar underlying HIV suppression patterns; therefore, other schemes, tailored to the population and resources, should be tested empirically to determine if they are viable options for maintaining patient care while reducing costs.
CONCLUSIONS
Adaptive VL monitoring could improve the efficiency of resource allocation while maintaining clinical outcomes by focusing on patients most in need of more intensive monitoring strategies while reducing resource requirements for those on stable ART. Future studies empirically evaluating alternative frequency monitoring to facilitate differentiated care are warranted. Innovative methods such as point-of-care VL assays could dramatically change the monitoring landscape, but currently Uganda still supports a centralized VL monitoring service.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Notes
Author contributions. S. J. R. was a senior clinician on the study, conceived the research idea, wrote the manuscript, and supported interpretation and discussion of results. M. N. and E. G. performed the analysis, interpreted results, and contributed in writing the original draft of the manuscript. G. N., R. G., M. W., T. C. Q., and D. S. were senior clinicians on the study and contributed in interpretation of analyses and writing. A. N. and V. S. supported data management and statistical analysis of results. L. W. C. supported interpretation of results and writing.
Acknowledgments. The authors thank the Rakai Health Sciences Program clinical cohort participants and study staff, whose contributions made this work possible.
Disclaimer. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US government.
Financial support. This project was supported by the Division of Intramural Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH) (grant number AI001040). It has also been funded in whole or in part with federal funds from the National Cancer Institute, NIH (contract number HHSN261200800001E). Support for treatment services was provided by the President’s Emergency Plan for AIDS Relief through the Centers for Disease Control and Prevention Uganda.
Potential conflicts of interest. The authors report no potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1. World Health Organization. Scaling up priority HIV/AIDS interventions in the health sector: progress report 2010. Available at: http://www.who.int/hiv/pub/2010progressreport/report/en. Accessed 15 June 2018.
- 2. Joint United Nations Programme on HIV/AIDS. An ambitious treatment target to help end the AIDS epidemic. 2014. Available at: http://www.unaids.org. Accessed 15 June 2018.
- 3. Joint United Nations Programme on HIV/AIDS. Fact sheet—latest statistics on the status of the AIDS epidemic 2017. Geneva, Switzerland: UNAIDS, 2017. [Google Scholar]
- 4. Reynolds SJ, Nakigozi G, Newell K, et al. Failure of immunologic criteria to appropriately identify antiretroviral treatment failure in Uganda. AIDS 2009; 23:697–700. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. World Health Organization. Consolidated guidelines on the use of antiretroviral drugs for treating and preventing HIV infection: recommendations for a public health approach. 2016. Available at: http://www.who.int/hiv/pub/arv/arv-2016/en/. Accessed 15 June 2018. [PubMed]
- 6. World Health Organization.. What’s new in treatment monitoring: viral load and CD4 testing. WHO/HIV/201722. Geneva, Switzerland: WHO, 2017. [Google Scholar]
- 7. Médicins Sans Frontières. Achieving undetectable: what questions remain in scaling-up HIV virologic treatment monitoring? 2014. Available at: https://www.msfaccess.org/sites/default/files/MSF_IssueBrief_undetectable6.pdf. Accessed 15 June 2018.
- 8. Calmy A, Ford N, Hirschel B, et al. HIV viral load monitoring in resource-limited regions: optional or necessary? Clin Infect Dis 2007; 44:128–34. [DOI] [PubMed] [Google Scholar]
- 9. Bonner K, Mezochow A, Roberts T, Ford N, Cohn J. Viral load monitoring as a tool to reinforce adherence: a systematic review. J Acquir Immune Defic Syndr 2013; 64:74–8. [DOI] [PubMed] [Google Scholar]
- 10. Jourdain G, Le Cœur S, Ngo-Giang-Huong N, et al. PHPT-3 Study Team Switching HIV treatment in adults based on CD4 count versus viral load monitoring: a randomized, non-inferiority trial in Thailand. PLoS Med 2013; 10:e1001494. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Grimsrud A, Barnabas RV, Ehrenkranz P, Ford N. Evidence for scale up: the differentiated care research agenda. J Int AIDS Soc 2017; 20:22024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Grimsrud A, Bygrave H, Doherty M, et al. Reimagining HIV service delivery: the role of differentiated care from prevention to suppression. J Int AIDS Soc 2016; 19:21484. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Schneider K, Puthanakit T, Kerr S, et al. Economic evaluation of monitoring virologic responses to antiretroviral therapy in HIV-infected children in resource-limited settings. AIDS 2011; 25:1143–51. [DOI] [PubMed] [Google Scholar]
- 14. Ritchie AV, Ushiro-Lumb I, Edemaga D, et al. SAMBA HIV semiquantitative test, a new point-of-care viral-load-monitoring assay for resource-limited settings. J Clin Microbiol 2014; 52:3377–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Rowley CF. Developments in CD4 and viral load monitoring in resource-limited settings. Clin Infect Dis 2014; 58:407–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Estill J, Egger M, Blaser N, et al. IeDEA Southern Africa Cost-effectiveness of point-of-care viral load monitoring of antiretroviral therapy in resource-limited settings: mathematical modelling study. AIDS 2013; 27:1483–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Dorward J, Drain PK, Garrett N. Point-of-care viral load testing and differentiated HIV care. Lancet HIV 2018; 5:e8–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Shet A, Neogi U, Kumarasamy N, DeCosta A, Shastri S, Rewari BB. Virological efficacy with first-line antiretroviral treatment in India: predictors of viral failure and evidence of viral resuppression. Trop Med Int Health 2015; 20:1462–72. [DOI] [PubMed] [Google Scholar]
- 19. Alvarez-Uria G, Naik PK, Pakam R, Midde M. Early HIV viral load determination after initiating first-line antiretroviral therapy for indentifying patients with high risk of developing virological failure: data from a cohort study in a resource-limited setting. Trop Med Int Health 2012; 17:1152–5. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.