A Model for Reduced HIV-1 Viral Load Monitoring in Resource-Limited Settings

Abstract

Background:

Viral load monitoring of antiretroviral therapy in low-income countries is rarely used because of high costs. Reducing the frequency of monitoring may make it financially feasible.

Methods:

We modeled three testing schemes: reduced viral load monitoring (RVLM) with CD4 count at baseline and viral load testing at 6, 36, and 60 months; United States Department of Health and Human Services (US DHHS) Treatment Guidelines; and World Health Organization (WHO) Guidelines using a cohort of 313 HIV-infected patients using Kaplan-Meier analysis.

Results:

Median time to detection of antiretroviral therapy (ART) failure using RVLM was 147 days; using US DHHS, it was 115 days; and using WHO guidelines, it was 1110 days. Median time for the development of first thymidine analog mutation was 594 days. The cost of RVLM was significantly lower than US DHHS.

Conclusions:

RVLM detected failure significantly sooner than CD4 count monitoring alone at a lower cost than US DHHS monitoring. RVLM is a potentially effective method of monitoring ART in resource-limited settings.

Background

HIV type 1 (HIV-1) viral load testing has become routinely used for monitoring antiretroviral therapy (ART) in high-income countries. ^{1 –3} Reduction in HIV-1 RNA with antiretroviral (ARV) drugs has been shown to correlate with improvement in immunologic status, decreased rates of opportunistic infections, and prolonged survival. Suppression of HIV-1 RNA below the level of detection prevents the emergence of resistance to ARV drugs and extends the clinical benefit of these medications. National and international treatment bodies recommend monitoring HIV-1 RNA every 3 to 6 months and changing therapy when the viral load becomes detectable on two successive occasions.

Viral load testing has not been widely used in low-income settings because it can be expensive and technically difficult. ^4,5 The World Health Organization (WHO) recommends use of viral load testing in circumstances of virologic failure as well as in a targeted approach or in a routine approach if resources permit. ⁴ However, as resources rarely permit, clinicians rely on other clinical markers such as CD4 count, hemoglobin levels, and changes in weight to monitor response to ART. Studies in both the high- and low-income settings have demonstrated that changes in these parameters are poor surrogates for viral load rebound. ^{5 –12} Reliance on surrogate markers for virologic response can prolong the time of drug exposure in the face of viral replication and promote the emergence of resistance. ¹³ High rates of resistance have been reported in patients failing ART in Cote d’Ivoire, Uganda, and Burkina Faso. ^{14 –16}

New technologies may allow increased utilization of viral load testing in low-income settings. ^{17 –21} A recent report suggests that filter paper transfer of dried whole blood and plasma spots may provide a practical, reliable method for virologic monitoring in low-income settings. ¹⁸ Testing using either ultra sensitive P24 assays or HIV-1 reverse transcriptase enzyme-linked immunosorbent assay (RT ELISA) has compared favorably to HIV-1 RT-polymerase chain reaction (PCR) in several studies. ^{19 –21} Although the costs of these tests are substantially lower than HIV-1 RT-PCR, it remains prohibitive for routine use by many providers of HIV care in highly impacted countries.

One option for reducing the costs of ART monitoring is to employ a reduced viral load monitoring (RVLM) scheme. Monitoring HIV-1 viral loads at strategic points of the treatment course may identify those individuals who have virologic rebound while reducing the number of tests performed and therefore the cost. By sampling HIV-1 viral load on a less frequent basis, there would be an increased risk of the development of resistance. Therefore, any proposed schema must balance the risk of resistance with the benefits of decreased cost of monitoring.

To evaluate the potential of RVLM, we examined a scheme that consists of CD4 count staging at baseline and viral load testing at 6, 36, and 60 months. This RVLM scheme was compared with standard monitoring recommendations provided by the United States Department of Health and Human Services (US DHHS) and the WHO treatment guidelines. Time to detect viral replication, potential for the development of resistance, and cost of each monitoring recommendation were compared. ^3,4

Methods

HIV-1 Viral Load Sampling

Rates of virologic failure were determined from a cohort of patients with HIV in the Parkland HIV database. The Parkland database is a relational database of over 10 000 patients treated at the Parkland Health and Hospital System HIV Clinic in Dallas, Texas. Median time between measuring HIV-1 RNA levels in this cohort is 79 days.

Data were extracted for 313 ARV naive, HIV-1-infected patients treated with zidovudine (ZDV), lamivudine (3TC), and efavirenz (EFV). Virologic failure was defined as HIV-1 RNA above 400 copies/mL after previously having achieved viral suppression. We used Kaplan-Meier analysis to determine median time to detect virologic failure when obtaining HIV-1 RNA every 6 months. Clinically feasible intervals for obtaining 3 viral load measurements over 60 months were tested to identify the interval that minimized median time for the detection of virologic failure.

Time to Clinical Failure

Time to clinical failure using CD4 count monitoring alone was derived from published sources. ^{22 –24} Individuals in the Parkland cohort had ART changed based on virologic failure and therefore could not be used to derive time to failure using CD4 count monitoring alone.

Development of Resistance

Time for the development of resistance was derived from published reports of first-line regimens consisting of a thymidine analog, 3TC, and a nonnucleoside reverse transcriptase inhibitor (NNRTI). ^{13,22 –25} Thymidine analog mutations (TAMs) were of particular interest because of the high incidence of 3TC and NNRTI resistance mutations at virologic failures and because of the potential impact of the accumulation of TAMs on subsequent regimens.

Cost Analysis

The cost analysis was based on the 313 individuals treated with ZDV-3TC-EFV from the Parkland HIV database. Average cost per patient was determined for each sampling scheme. Costs included CD4 testing and viral load testing costs for each of the guidelines: CD4 count and HIV-1 viral load at baseline and every 6 months for US DHHS treatment guidelines; CD4 count at baseline and every 6 months for WHO guidelines; and CD4 count at baseline and viral load testing at 6, 36, and 60 months for RVLM scheme. The cost analysis included only the costs of the testing and did not include any other potential costs associated with ART or patient care. The cost analysis assumed that tests would be performed using the BD FACSCount CD4 enumeration machine and the Cavidi EvaVir Load viral load testing machine, both of which are widely available in resource-limited settings. The cost used was US$5 for each CD4 test and US$25 for each viral load measurement.

Results

Three hundred and thirteen ARV naive patients who were treated with ZDV, 3TC, and EFV were identified, and time to virologic failure was determined using Kaplan-Meier analysis. There were 84 virologic failures over 5 years. Figure 1 shows the time to virologic failure of this cohort, marked with a sampling interval of every 6 months as recommended by US DHHS treatment guidelines. Figure 2A shows the number of virologic failures detected in each 6-month interval. More than 50% of the virologic failures occurred in the first 6 months.

Figure 1.

Time to virologic failure in patients treated with zidovudine (ZDV)/lamivudine (3TC)/efavirenz (EFV). Horizontal lines represent US DHHS HIV treatment guideline VL monitoring recommendations. Bolded lines indicate time points for viral load testing using reduced viral load monitoring.

Figure 2.

A, Percentage of total virologic failures detected in each 6-month time interval using US DHHS monitoring guidelines. B, Percentage of total virologic failures detected in each time interval using reduced viral load monitoring scheme.

Using only 3 viral load tests in the first 60 months of treatment, the interval that led to the shortest median time to detect virologic failure was monitored at 6, 36, and 60 months. Bolded lines in Figure 1 indicate this reduced sampling interval. Figure 2B shows the number of virologic failures detected in each interval using RVLM.

Table 1 shows the median time to detect virologic failure from the previous test for each sampling scheme. The median time to detect virologic failure was shorter for the US DHHS guidelines than for RVLM (115 vs 147 days, P < .05). The time to detect clinical failure using CD4 count monitoring data was significantly longer than with either other method (1110 days). The minimum time for the development of TAMs was 120 days, while the median time for the development of TAMs was over 500 days.

Table 1.

Time to Detection of Antiretroviral Failure Based on HIV-1 Viral Load Monitoring Using Different Testing Schemes.^a

Monitoring Method	Median Time to Detect Failure, Days	95% CI, Days
US DHHS treatment guidelines	115	100-129
Reduced viral load monitoring	147	137-156
WHO guidelines	1110	987-1263

Abbreviations: CI, confidence interval; US DHHS, United States Department of Health and Human Services.

^aMinimum time to thymidine analog mutations (TAMS) is 120 days; medium time to TAMS is 594 days.

The costs of performing each of the testing schemes are shown in Table 2. The US DHHS was the most expensive (US$48 per patient/y) followed by the reduced viral load testing scheme (US$18 per patient/y) with the lowest cost being the WHO HIV treatment guidelines (US$8 per patient/y).

Table 2.

Cost of Various Testing Schemes.

Monitoring Method	Total 60-Month Cost per Patient, US $	Total Cost per Patient per Year, US$
US DHHS treatment guidelines	239	48
Reduced viral load monitoring	89	18
WHO treatment guidelines	40	8

Abbreviations: US DHHS, United States Department of Health and Human Services; WHO, World Health Organization.

Discussion

Recent studies have focused largely on using computer modeling that uses a 6-month interval for viral load testing. ^{26 –28} Our intent is to propose that we should also consider models of reduced viral load monitoring that add value to clinical care with minimal added costs.

To highlight this possibility, we developed a model to test the feasibility of performing viral load testing in low-income settings using a reduced monitoring scheme. We found that the reduced monitoring scheme detected virologic failure almost as frequently as using standard 6-month monitoring. In addition, the risk of developing resistant virus, particularly TAMs, was not appreciably increased. In our model, the median time of exposure to replicate virus was 147 days, which is well below the median time for the development of TAMs in clinical trials and only 37 days longer than was seen using US DHHS treatment guidelines (while a statistically significant difference, not necessarily clinically significant). In contrast, monitoring ART using CD4 counts only results in a median time to detect clinical treatment failure of 1110 days, making TAMs quite possible. The cost of RVLM would be significantly less than standard viral load monitoring. While more expensive than CD4 count monitoring, it provides a distinct benefit by reducing the probability of resistance and thus increasing the potential effectiveness of second-line regimens.

There are caveats to this model. Other models of viral load testing in resource-limited settings have found that its usage is not cost effective. ^26,27 As part of their analysis, the authors used standard 6-month viral load sampling to monitor ART and were therefore unable to realize the cost advantages associated with reduced monitoring schema. Both our and other models reported in the literature rely on clinical cohorts based on high-income countries and thus these data may not be fully generalizable to the resource-limited setting. In our study, the cost analysis evaluated only the actual costs of performing the tests and did not determine the impact of a reduced monitoring scheme on other economic factors or on the quality of life years gained nor did it assess the costs that may be associated with increased morbidity expected when clinical and or immunological criteria are used as recommended by WHO.

Despite these caveats, we believe that a reduced sampling scheme for the monitoring of viral loads in low-income settings deserves further evaluation. Using a cohort in Uganda, Meya et al developed a model for predicting who would benefit from viral load testing. ²⁹ They found that using a combination of an individual's adherence to ART and CD4 count decline was predictive of virologic failure. Using defined criteria such as ART adherence and CD4 counts to direct viral load testing could improve the utility of viral load testing and further reduce its cost. The best way to further determine the utility of viral load testing in resource-limited settings would be through a strategy trial. In such a trial, patients would be randomized to receive standard viral load testing every 6 months or to one or more reduced viral load-monitoring schemes. Clinical end points such as CD4 count decline, opportunistic infection, and death could be evaluated in each group. The development of resistance and response to second-line regimens could also be compared. The ultimate goal would be to determine whether the use of a reduced sampling scheme in low-income settings would be equally effective at a lower cost than standard testing.