Comparative Evaluation of the ExaVir Load Version 3 Reverse Transcriptase Assay for Measurement of Human Immunodeficiency Virus Type 1 Plasma Load

Abstract

In resource-limited settings, the virological monitoring of antiretroviral therapy is limited by high cost and the lack of infrastructure. The Cavidi ExaVir Load assay employs a simple and inexpensive enzyme-linked immunosorbent assay format to measure human immunodeficiency virus (HIV) reverse transcriptase activity, which correlates with plasma RNA load. The version 3 assay has been described as having improved precision and sensitivity. There are limited data on its performance relative to those of current real-time assays. Our objective was to compare HIV type 1 (HIV-1) RNA load measurement in plasma by ExaVir Load version 3 (designated ExaVir), Abbott M2000sp/M2000rt RealTime HIV-1 assay (designated RealTime), and Roche COBAS Ampliprep/COBAS TaqMan HIV-1 version 1 assay (designated TaqMan). Plasma from 119 patients (34 with subtype B infection, 85 with non-subtype B infection [A-H, CRF01, CRF02, CRF06, CRF12, CRF14, and complex]; 48 subjects were treatment experienced, 71 were naive) and serial dilutions of the second international standard (IS) were tested. Assay relationship and agreement were determined by linear regression, correlation analysis, and the Bland-Altman method. The ExaVir assay quantified 77/83 (92.8%) samples with viral loads of >2.3 log₁₀ copies/ml by the molecular assays. Results were linearly associated and strongly correlated with RealTime and TaqMan measurements (R of 0.94 and 0.92, respectively) for both subtype B (R of 0.97 and 0.95, respectively) and non-subtype B (R of 0.93 and 0.91, respectively) samples. Mean differences were 0.28 and 0.18 log₁₀ copies/ml in favor of the two molecular assays; 7/119 (5.9%) and 5/119 (4.2%) samples were outside the 95% level of agreement. ExaVir underquantified the IS by a mean of 0.2 (range, 0.0 to 0.5) log₁₀ copies/ml. The ExaVir assay showed excellent concordance with real-time molecular assays, offering a suitable option for virological monitoring in settings with limited infrastructure.

The introduction of combination antiretroviral therapy in resource-limited countries has resulted in significant reductions in morbidity and mortality (8, 13, 22). It generally is accepted that the unavailability of plasma viral load (VL) monitoring should not preclude expanded access to treatment in these settings (19). However, knowledge accrued through more than a decade of experience in high-income settings indicates that virological monitoring is required in treated patients to ensure optimal long-term outcomes (2, 11). Incomplete VL suppression during therapy leads to the emergence and evolution of drug resistance, reducing treatment options and resulting in the transmission of resistant mutants (7). Neither clinical findings nor CD4 cell counts are adequate predictors of viral suppression, and in fact management by CD4 cell counts alone can lead to unnecessary treatment changes (1). VL testing is the only reliable marker for the early detection of the failure of antiretroviral therapy (17, 20).

Molecular VL assays in routine use in high-income countries require expensive instruments and reagents, sophisticated laboratory facilities to minimize the risk of contamination, a regular and stable electricity supply, and highly skilled laboratory technicians proficient in molecular biology techniques. These factors limit the implementation of VL testing in resource-limited settings. The Cavidi ExaVir Load assay employs a modified enzyme-linked immunosorbent assay (ELISA) format to measure the viral reverse transcriptase (RT) activity, which in turn correlates with plasma RNA levels (3, 23). The assay requires simple, routinely available equipment (e.g., an incubator, ELISA plate reader, freezer, mixing table, and vortex machine) and is relatively inexpensive and simple to perform. The price per test is dependent on volumes but can be as low as $13.66 (M. Eales, personal communication). These characteristics make it suitable for use in settings with limited infrastructure. In November 2007, the manufacturer signed an agreement with the Clinton Foundation HIV/AIDS Initiative (CHAI) to provide the assay at a discounted price to members of the CHAI Procurement Consortium of more than 70 developing countries. The two previous versions of the assay have been evaluated in the literature (3, 5, 12, 14, 15, 23, 25). Relative to version 2, version 3 is described by the manufacturers as showing enhanced precision, analytical specificity and sensitivity (the lower limit of quantification was reduced from 400 to 200 copies/ml), improved turn-around time (from 72 to 48 h), reduced hands-on time (from 6 to 5 h), a reduction in the use of consumables, and increased throughput (from 120 to 180 samples per week per scientist). There is no published evidence on the performance of the version 3 assay relative to current real-time molecular methods in use in high-income countries.

The objective of this study was to evaluate the performance of the ExaVir Load version 3 assay (referred to hereafter as the ExaVir assay) in comparison to two real-time PCR assays widely used in high-income countries: the Abbott M2000sp/M2000rt RealTime human immunodeficiency virus type 1 (HIV-1) assay (referred to as the RealTime assay) and the Roche COBAS-Ampliprep/COBAS-TaqMan HIV-1 version 1 assay (referred to as the TaqMan assay).

(This work was presented at the 16th Conference on Retroviruses and Opportunistic Infections, Montreal, Canada, 8 to 11 February 2009.)

MATERIALS AND METHODS

Patients and samples.

Blood samples anticoagulated with EDTA were collected from 119 patients attending the Royal Free Hampstead NHS Trust for routine HIV care. The study population was infected with diverse HIV-1 subtypes and comprised 71 antiretroviral treatment-naive patients and 48 patients receiving antiretroviral therapy. Plasma was separated within 6 h of collection and stored at −80°C in three separate aliquots prior to parallel testing in the three assays. Serial dilutions (n = 10) ranging from 25,000 to 40 (4.4 to 1.6 log₁₀) IU/ml of the WHO subtype B second international standard for HIV-1 RNA (IS) (product code 97/650; National Institute for Biological Standards and Control, United Kingdom) also were tested in parallel. Plasma samples from 10 HIV antibody-negative patients were used as negative controls to assess ExaVir assay specificity.

ExaVir assay.

The ExaVir assay (Cavidi, Sweden) measures the viral RT enzymatic activity in an ELISA format. Following the separation of virus particles from 1 ml of plasma using a solid-phase extraction manifold, virus is lysed to obtain the RT enzyme and the lysate is added to an RNA template bound to the solid phase in the presence of primer and RT substrate. In the presence of RT, the enzyme synthesizes a DNA strand, which is detected by anti-bromodeoxyuridine monoclonal antibody conjugated to alkaline phosphatase (AP). The product is quantified by the addition of a colorimetric AP substrate. The RT activity in the sample is determined by the ExaVir Load Analyzer software through a standard curve generated by an 11-point serial dilution of a known amount of recombinant HIV-1 RT. The range of quantification, as reported by the manufacturers, is from approximately 200 (2.3 log₁₀) to 600,000 (5.8 log₁₀) copies/ml. The upper limit varies with the reading range of the ELISA plate reader.

RealTime and TaqMan assays.

The RealTime and TaqMan assays employ high-throughput, automated real-time PCR methodologies targeting conserved regions in HIV-1 integrase and gag, respectively. The RNA is extracted from 1 ml of plasma and concentrated using magnetic particle technology in the automated Abbott M2000sp instrument and the Roche COBAS Ampliprep instrument. The detection and quantification of the amplified PCR product is accomplished within hours by monitoring the emission intensity of fluorescent reporter dyes released during the amplification process. The reported range of quantification is 40 to 10,000,000 (1.6 to 7.0 log₁₀) copies/ml.

HIV-1 subtyping.

HIV-1 subtypes were determined from pol gene sequences using the ViroSeq system (Celera Diagnostics). Briefly, after the reverse transcription of plasma RNA, a 1.8-kb amplicon comprising the whole of protease and codons 1 to 335 of RT underwent population sequencing in an ABI PRISM 3100 genetic analyzer. The sequences were submitted to the NCBI and Rega HIV-1 subtyping tools, and the assignment was confirmed by phylogenetic analysis with Mega 4.0 using reference sequences from the Los Alamos database (www.lanl.gov).

Statistical analysis.

VL measurements were log₁₀ transformed before analysis, and the value of the assay lower limit of quantification (LLQ) was assigned to samples with VL below this level. Pair-wise Pearson's correlation coefficients were used to assess whether VL values determined using different assays were correlated. Since correlation coefficients do not account for the fact that one assay may provide consistently higher values than another assay, pair-wise Bland-Altman plots were used to further assess the level of agreement. These plots compare two measurement techniques by plotting the difference in VL measurements between any two assays against the average of the two assays. These differences then were tested using paired t tests for each pair-wise comparison. The Pitman's test, based on calculating the correlation between the difference and the mean, was used to test for a null hypothesis of equal variances given bivariate normality. The t test also was repeated assuming unequal variances (and unpaired data), with similar results (not shown).

RESULTS

Patient samples.

The 119 samples comprised HIV-1 group M strains representing 34 subtype B and 85 diverse non-subtype B strains (Table 1). At the time of sampling, 71 patients were antiretroviral treatment naive and 48 were receiving antiretroviral therapy.

TABLE 1.

Comparison of HIV-1 plasma VL levels measured by the ExaVir, RealTime, and TaqMan assays by antiretroviral (ARV) treatment status and HIV-1 subtype^a

Sample type	No. of samples	Mean viral load in log10 copies/ml (SD)
		ExaVir	RealTime	TaqMan
All	119	3.3 (1.4)	3.6 (1.4)	3.5 (1.4)
ARV status
Naïve	71	4.2 (0.9)	4.5 (0.9)	4.3 (0.9)
Experienced	48	2.1 (0.9)	2.3 (1.0)	2.3 (1.0)
Subtype
A	19	3.0 (1.4)	3.2 (1.5)	3.2 (1.4)
B	34	3.5 (1.3)	3.8 (1.3)	3.7 (1.2)
C	30	3.0 (1.3)	3.3 (1.4)	3.3 (1.3)
D	9	3.7 (1.2)	3.8 (1.3)	3.8 (1.2)
F	2	2.2 (0.6)	2.5 (0.9)	2.5 (0.9)
G	1	3.3	3.8	3.0
H	1	5.1	5.6	5.5
CRF01	3	2.4 (1.1)	2.2 (0.7)	1.8 (0.2)
CRF02	13	4.2 (1.2)	5.0 (0.9)	4.6 (0.9)
CRF06	1	1.6	2.0	1.8
CRF12	1	1.6	2.0	1.9
CRF14	1	5.5	5.7	4.0
Cpx	4	2.4 (1.3)	2.7 (1.3)	2.7 (1.4)

^aThe means and standard deviations are shown and were calculated after conversion into log₁₀ copies/milliliter. Undetectable VL levels were scored as 40 (1.6 log₁₀) copies/ml, corresponding to the lower limit of quantification of the two molecular assays. Cpx, complex mosaic pol sequence.

Comparison of ExaVir and RealTime assays.

Overall, 78/119 (65.5%) samples were quantified by both assays, with a median VL of 4.4 (range, 2.4 to 5.8) log₁₀ copies/ml by ExaVir assay and 4.6 (range, 2.4 to 6.8) log₁₀ copies/ml by RealTime assay. There were 25/119 (21.0%) samples with undetectable VL by both assays, all from treated patients. A further 15/119 (12.6%) samples with a median VL of 2.3 (range, 1.6 to 3.3) log₁₀ copies/ml by RealTime assay showed an undetectable VL by ExaVir assay, including seven samples that were quantified by RealTime at levels above the expected LLQ of ExaVir (2.3 log₁₀; 200 copies/ml). One sample (1/119; 0.8%) (subtype D, on antiretroviral therapy) showed a VL of 1,259 (3.1 log₁₀) copies/ml by ExaVir but an undetectable VL by RealTime assay. The coefficient of correlation (R) between the assays was 0.94 overall and was 0.97 and 0.93 for subtype B and non-subtype B, respectively (Fig. 1a). The VL measurements differed on average by 0.28 (95% confidence interval [CI], 0.19 to 0.37) log₁₀ copies/ml in favor of the RealTime assay (P < 0.0001). Pitman's test of the difference in variance (r) was 0.107 (P = 0.248). In the Bland-Altman comparison, the limits of agreement (the reference range for difference) were −0.72 to 1.27 log₁₀ copies/ml (Fig. 2a). Overall, 7/119 (5.9%) samples (three CRF02, two subtype D, one subtype A, and one subtype B) fell outside the reference range, including six samples underquantified by ExaVir assay and the subtype D sample quantified by ExaVir but not by RealTime assay (Table 2).

FIG. 1.

Correlation between ExaVir and RealTime assays (a), ExaVir and TaqMan HIV-1 assays (b), and RealTime and TaqMan assays (c), as determined by the parallel testing of 119 samples. All VL values are in log₁₀ copies/milliliter. Undetectable VL levels were scored as 40 (1.6 log₁₀) copies/ml, corresponding to the lower limit of the quantification of the two molecular assays. The linear regression line is shown.

FIG. 2.

Bland-Altman analysis of the agreement between ExaVir and RealTime assays (a), ExaVir and TaqMan assays (b), and RealTime and TaqMan assays (c). The labels show the subtype of samples outside the 95% level of agreement.

TABLE 2.

Samples showing HIV-1 plasma VL levels outside the 95% level of agreement between two assays when tested by ExaVir, RealTime, and TaqMan assays

Subtype	Viral load (log10 copies/ml)
	ExaVir	RealTime	TaqMan
A	Undetectable	3.3	2.4
B	3.8	5.2	5.0
B	4.8	5.2	4.2
D	Undetectable	3.0	3.0
D	3.1	<1.6	<1.6
D	4.4	4.8	3.7
D	3.6	3.0	4.0
CRF01	3.9	3.2	<1.6
CRF02	3.3	5.5	5.2
CRF02	2.7	4.3	3.7
CRF02	Undetectable	3.0	3.2
CRF02	5.4	5.7	4.7
CRF14	5.5	5.7	4.0

Comparison of ExaVir and TaqMan assays.

Overall, 77/119 (64.7%) samples were quantified by both assays, with a median VL of 4.4 (range, 2.4 to 5.8) log₁₀ copies/ml by ExaVir assay and 4.5 (range, 2.8 to 6.8) log₁₀ copies/ml by TaqMan assay. There were 25/119 (21.0%) samples with undetectable VL by both assays, all from treated patients. A further 15/119 (12.6%) samples with a median VL of 2.3 (range, 1.8 to 3.2) log₁₀ copies/ml by TaqMan assay showed an undetectable VL by ExaVir assay, including seven samples that were quantified by the TaqMan assay at levels above the expected LLQ of the ExaVir assay. Two samples (2/119; 1.7%) (subtype D and CRF01) showed a VL of 3.1 and 3.9 log₁₀ copies/ml, respectively, by ExaVir assay but an undetectable VL by TaqMan assay. By RealTime assay, the subtype D sample from a treated patient also showed an undetectable VL, whereas the CRF01 sample showed a VL of 3.2 log₁₀ copies/ml. The coefficient of correlation (R) between the two assays was 0.92 overall and was 0.95 and 0.91 for subtype B and non-subtype B samples, respectively (Fig. 1b). The VL measurements differed on average by 0.18 (95% CI, 0.08 to 0.29) log₁₀ copies/ml in favor of the TaqMan assay (P = 0.0005). Pitman's test of the difference in variance (r) was −0.043 (P = 0.647). In the Bland-Altman comparison, the limits of agreement (the reference range for difference) were −0.94 and 1.31 log₁₀ copies/ml (Fig. 2b). Overall, 5/119 (4.2%) samples (two CRF02, one subtype D, one CRF01, and one CRF14) fell outside the reference range, including two samples (both CRF02) underquantified by the ExaVir assay and three samples (subtype D, CRF01, and CRF14) underquantified by the TaqMan assay (Table 2).

Comparison of RealTime and TaqMan assays.

Overall, 91/119 (76.5%) samples were quantified by both assays, with a median VL of 4.5 (range, 1.6 to 6.8) log₁₀ copies/ml by RealTime assay and 4.4 (range, 1.8 to 6.8) log₁₀ copies/ml by TaqMan assay. There were 25/119 (21.0%) samples with undetectable VL by both assays, all from treated patients. In addition, one sample (1/120; 0.8%) (subtype A) showed a VL of 2.2 log₁₀ copies/ml by TaqMan assay but an undetectable VL by RealTime assay. Conversely, two samples (2/119; 1.7%) (subtype C and CRF01) showed a VL of 2.3 and 3.2 log₁₀ copies/ml, respectively, by RealTime assay but an undetectable VL by TaqMan assay. The coefficient of correlation (R) was 0.96 overall and was 0.97 and 0.96 for subtype B and non-subtype B samples, respectively (Fig. 1c). The VL measurements differed on average by 0.09 (95% CI, 0.02 to 0.17) log₁₀ copies/ml in favor of the RealTime assay (P = 0.01). Pitman's test of the difference in variance (r) was 0.194 (P = 0.035). In the Bland-Altman comparison, the limits of agreement (the reference range for difference) were −0.69 and 0.88 log₁₀ copies/ml (Fig. 2c). Overall, 6/119 (5.0%) samples fell outside the reference range, comprising five samples (one subtype B, one subtype D, one CRF01, one CRF02, and one CRF14) underquantified by TaqMan assay and one sample (subtype D) underquantified by RealTime assay (Table 2).

Assay performance with the WHO HIV-1 RNA IS.

The ExaVir assay consistently underquantified the subtype B IS, whereas the TaqMan assay consistently overquantified the IS (Fig. 3). With the ExaVir assay, across seven quantified dilutions ranging from 4.4 to 2.6 log₁₀ IU/ml, the mean difference in VL was 0.3 (range, 0.0 to 0.5) log₁₀ copies/ml. With the molecular assays, across 10 dilutions ranging from 4.4 to 1.6 log₁₀ IU/ml, the average difference was 0.0 (range, 0.1 to −0.3) log₁₀ copies/ml with the RealTime assay and −0.2 (range −0.1 to −0.3) log₁₀ copies/ml with the TaqMan assay.

FIG. 3.

Comparison of VL measurements obtained by the ExaVir, RealTime, and TaqMan assays with serial dilutions of the WHO second IS for HIV-1 RNA, ranging from 4.4 to 1.6 log₁₀ (25,000 to 40) IU/ml.

Reproducibility and specificity of the ExaVir assay.

To assess the intraassay reproducibility of the ExaVir assay, 10 samples were tested in duplicate, of which 7 showed a detectable VL. Overall, replicate values differed by a mean of 0.04 (standard error, 0.07) log₁₀ copies/ml. Specificity was assessed with 10 HIV-negative plasma samples, all of which showed an undetectable VL.

DISCUSSION

Molecular VL assays widely used in high-income countries for the virological monitoring of antiretroviral therapy are difficult to implement in resource-limited settings due to both financial and practical constraints. The ExaVir assay offers a cheaper and simpler methodology for VL measurement in these settings. In this study, the version 3 assay showed an excellent correlation and a high degree of concordance with two widely used commercial real-time PCR assays, a comparative performance that was consistent with the relationship that the two molecular assays showed with each other.

Previous studies analyzed the performance of the ExaVir assay versions 1 and 2 and found good detection rates for samples with VL above 10,000 and 400 copies/ml, respectively, and a good overall correlation with molecular assays, most commonly the Roche Amplicor HIV-1 Monitor test version 1.5 (3, 5, 12, 14, 15, 23, 25). An evaluation of the performance of the version 3 assay compared to that of version 2 (and the Roche Amplicor HIV-1 Monitor test version 1.5) was presented in abstract form in 2008 (9). It demonstrated increased sensitivity with version 3 compared to that of version 2, with a mean difference of 0.19 log₁₀ copies/ml. In this study, VL measurements with clinical samples generally were underquantified by the ExaVir assay compared to quantification by the molecular assays. With the IS, we also observed underquantification by the ExaVir assay, while we detected good performance for the RealTime assay and a small but consistent overquantification by the TaqMan assay. The ExaVir assay quantified 93% of samples with VL of >200 (2.3 log₁₀) copies/ml by both molecular assays and 100% of samples with VL of >1,585 (3.2 log₁₀) to 1,995 (3.3 log₁₀) copies/ml. Thus, performance was in agreement overall with the range of quantification reported by the manufacturers. These findings indicate that the ExaVir assay can reliably identify significant viremia in treated patients.

The significance of low-level viremia continues to be debated in high-income countries (16, 18). Interpretation and management likely will be even more challenging in developing countries, where drug options are limited. Previous studies reported stable CD4 cell counts and a low risk of disease progression at VL levels below 4.0 log₁₀ copies/ml (4, 18). However, it would be of importance to consider additional outcome data, including the emergence of drug resistance and the exhaustion of treatment options, in order to establish appropriate VL cutoffs that should trigger a treatment change where resources are limited. Meanwhile, an assay with a lower limit of quantification of around 200 copies/ml would be of immediate practical use.

A few samples showed significant differences in VL measurements between assays. ExaVir generally underquantified these discrepant samples compared to the quantification of the molecular assays, consistently with the reduced sensitivity of the assay. While a problem with false-positive results was apparent in a study of the ExaVir version 2 assay (24), there was no evidence of a significant problem with assay specificity in this study. Among 26 samples with an undetectable VL by the two molecular assays, all from treated patients, only one showed a detectable VL by ExaVir assay, at 1,259 (3.1 log₁₀) copies/ml, while HIV-negative samples all showed an undetectable VL by ExaVir assay. One additional sample, from a patient infected with CRF01, was quantified by ExaVir assay as well as by RealTime assay but not by TaqMan assay, suggesting a problem with quantification by the latter. Although RealTime and TaqMan assays showed a high degree of concordance, four other samples, comprising subtype B, subtype D, CRF02, and CRF14, were significantly underquantified by TaqMan assay, whereas one subtype D sample was underquantified by RealTime assay. Previous studies suggested an impaired performance of the TaqMan version 1 assay for the quantification of non-subtype B samples (10, 21), and the recently launched version 2 assay promises to address this problem. We previously reported the good overall performance of the RealTime assay (6). In this study, the underperformance of either real-time PCR assay was more common with non-subtype B samples but not consistent with specific subtypes. Taken together, these findings indicate that, despite the significant improvements introduced in recent years, HIV sequence variability continues to challenge molecular VL assays. Testing with a second method is recommended when VL results are not consistent with the patient's history, and in these circumstances the use of a nonmolecular assay like the ExaVir assay also could be considered.

In summary, we found an excellent correlation and a high degree of concordance between the ExaVir assay and current real-time molecular assays. The increased throughput and reduced turnaround time, hands-on time, and use of consumables with version 3 relative to that of version 2 make the assay an attractive option for the virological monitoring of treated patients where infrastructure is limited. Some of the previously recognized limitations (25) remain, including the large sample volume required for analysis and the lack of automation.