INTRODUCTION
Plasma viral load (VL) is a key element of the virological follow-up for patients infected with HIV to measure the intensity of viral replication in untreated patients and for assessing virological response to antiretroviral drugs. Assessment of HIV-1 VL has been performed for many years by different techniques, initially the measurement of p24 antigen, and currently using the measurement of the reverse transcriptase (RT) enzymatic activity,1 2–6
These different technologies use relatively impressive high-capacity platforms, combining a separate system of extraction and amplification of the viral genome. Automation improved the reproducibility of results, facilitated technical work components, avoided contamination, and allowed savings because of processing of batched sets of samples. These platforms are advantageous for large laboratories but may be inadequate or difficult to implement in smaller laboratories or in field conditions in low-income countries because of required technical expertise, cost, and maintenance of the instruments. In addition, traditional batched samples systems are not able to perform HIV VL in real time or for emergency cases.
The GeneXpert (Cepheid, Sunnyvale, CA) technology is based on cartridges used in individual modules that perform the extraction and quantitative RT-PCR in a unique instrument system.7 Staphylococcus aureus , for oncohaematology and other applications. A new assay, Xpert HIV-1 Viral Load (Xpert HIV-1 VL), has been developed for use on the Cepheid GeneXpert instrument system. Owing to its simplicity and ease of use, this new HIV-1 VL assay could be an alternative to the current large analytical platforms, for laboratories that do not have sufficient numbers of samples to perform VL testing on large instruments in a cost-effective manner. It also could be interesting for larger laboratories wishing to conduct their high-volume VL tests in real time, and for emergency cases if necessary.
To be successful, a new test for quantifying HIV-1 must have analytical performance and a threshold of sensitivity comparable to what is currently available (20–40 copies/mL plasma depending on the assay). Owing to the intrinsic genetic properties of HIV-1, an assay must also be able to detect and quantify the wide genetic diversity of the pandemic HIV-1 group M viruses. Moreover, quantifying divergent viruses of groups N, O, and P, as is the case with the Abbott RealTime HIV-1 and Roche CAP/CTM Cobas Taqman HIV-1 assays,8
The aim of this study was to conduct a performance evaluation of the Xpert HIV-1 VL assay relative to the Abbott RealTime HIV-1 assay on a wide panel of HIV-1 variants representative of the current genetic diversity of HIV-1, as well as its practicability.
METHODS AND SAMPLES
Cepheid Xpert Assay
The Xpert HIV-1 VL assay is performed on the Cepheid GeneXpert instrument systems. These systems automate and integrate sample preparation, nucleic acid amplification, and detection of the target sequence (3′-Long Terminal Repeat region) in samples using real-time RT-PCR. The systems require the use of single-use disposable GeneXpert cartridges that include reagents, as well as high (Internal Quantitative Standard High) and low (Internal Quantitative Standard Low) internal controls and a probe check control (PCC). The IQSH and IQSL are present to control for adequate processing of the target sample and to monitor the presence of inhibitors in the RT and PCR reactions, as well as for quantification of HIV-1. The probe check control verifies reagent rehydration, PCR tube filling in the cartridge, probe integrity, and dye stability. External controls used during the study consisted of 1 HIV-1 negative, 1 low HIV-1 positive containing the 8E5 LAV strain in EDTA-based normal human plasma matrix (expected range 2.95–4.15 Log copies/mL), and 1 high HIV-1 positive control (expected range 5.48–6.44 Log copies/mL). The negative control was made of HIV-1 negative human plasma matrix. Controls were tested with the Xpert HIV-1 VL on each day that specimens were tested before analyzing any study specimens. The run time for 1 sample is 90 minutes.
The input volume of plasma (ACD-A or EDTA) is 1 mL. The range of quantification in human plasma is from 40 to 10,000,000 copies/mL (1.6 Log copies/mL–7 Log copies/mL), and the limit of detection is determined to be 40 copies/mL for HIV-1 subtype B in EDTA and ACD plasma.
Results of INVALID (eg, failure of IQSH and/or IQSL), ERROR (test failure due to Probe Check: FAIL*; all or one of the probe check results fail or insufficient volume), or NO RESULT (insufficient data collected, eg, a test in progress was stopped) are defined as indeterminate results.
Abbott Assay
For comparison, all specimens were tested using the Abbott RealTime HIV-1 assay, as recommended by the manufacturer. This technique is based on an automated-system of extraction of the nucleic acid (m2000 SP) and an amplification system (m2000RT) using real-time RT-PCR; the detection is performed in the integrase region and the linear range is 40 copies/mL–10,000,000 copies/mL (1.6 Log–7 Log) and the limit of detection is 40 copies/mL with the 0.6 mL sample volume procedures.
Samples for Clinical Performance
The specificity was evaluated by testing 20 plasma (EDTA) specimens from HIV-1 negative blood donors.
The clinical performance was assessed on 295 HIV-1 seropositive samples, collected during routine VL measurements for 285 patients managed at the Charles Nicolle Hospital, Rouen, France, and requiring no additional blood samples. They were obtained from 224 subjects on antiretroviral therapy (ART) and 59 untreated patients. No ART information was available for 2 patients. The samples were initially quantified with the Abbott assay used routinely in the laboratory, and selected to cover the quantification range (1.6 Log ≤ n = 71 ≤ 2 Log; 2 Log < n = 50 ≤ 3 Log; 3 Log < n = 32 ≤ 4 Log; 4 Log < n = 33 ≤ 5 Log; n = 32 ≥ 5 Log) and included RNA undetectable (n = 47) or detected (n = 30) samples; 176 plasma (EDTA) were selected from registered collection of samples frozen and stored at −80°C and 119 plasma samples were selected prospectively and tested fresh. Fresh samples were stored at 2–8°C and tested simultaneously by both techniques within 5 days of collection and separation, and frozen samples were tested simultaneously within the same freeze/thaw cycle.
The genetic diversity of the panel was determined by grouping and subtyping using molecular characterization, ie, by sequencing the Pol region (protease and reverse transcriptase) on an automated capillary CEQ 8000 DNA sequencer (Beckman, Fullerton, CA), followed by phylogenetic analyses as previously described.9
Statistical Analysis
Samples with HIV-1 RNA quantitative results within the linear range of both the Xpert HIV-1 VL and the Abbott RealTime HIV-1 assay were compared using MedCalc Statistical Software version 15.11.4 (MedCalc Software bvba, Ostend, Belgium; https://www.medcalc.org
RESULTS
Specificity
None of the 20 negative samples were found positive, thus the specificity of Xpert HIV-1 VL was 100% (95% CI: 83.9 to 100.0).
Clinical Concordance
Xpert HIV-1 VL tests were successful on the first attempt for 96.6% (285/295) of the samples. The 10 indeterminate cases reported included 3 ERROR, 6 INVALID, and 1 NO RESULT; corresponding VL obtained with the Abbott assay was undetectable (n = 2), detected at <40 copies/mL but not quantifiable (n = 3), and from 1.7 to 5.3 Log copies/mL (n = 5). None of these indeterminate cases was retested because of insufficient sample volume.
The results of the 285 specimens with valid results by both VL assays are presented in Table 1 . They are classified as follows: HIV-1 undetectable; HIV-1 detected at <40 copies/mL but not quantifiable, and HIV quantifiable at ≥40 copies/mL. In both techniques, the HIV-1 VL was undetectable for 38 samples, detected for 13 plasmas, and quantifiable for 162 plasmas, leading to overall concordance of 74.7% (213/285).
TABLE 1.: Qualitative Results for Xpert HIV-1 VL and Abbott RealTime HIV-1 Assays
Taking into account that samples were initially selected on the Abbott VL, the Cepheid assay detected or quantified HIV RNA in 222/285 (78%) samples, and the Abbott assay detected or quantified HIV RNA in 240/285 (84%) samples. Xpert yielded VLs for 162 (76%) of the 213 quantifiable samples selected with the Abbott assay; this difference corresponded to 51 samples that were quantifiable with the Abbott assay but only detected or undetectable with the Cepheid assay.
Regarding the genetic diversity of the 222 samples detected or quantified by Cepheid, 84 (38%), 135 (61%), and 3 (1%) were subtype B, non-B, and NT, respectively; among the 240 samples detected or quantified by Abbott, 94 (39%), 139 (58%), and 7 (3%) were of subtype B, non-B, and NT, respectively. Regarding the 162 samples quantified by Cepheid, 58 (36%) and 104 (64%) were subtype B and non-B, respectively; among the 213 samples quantified by Abbott, 84 (39%), 124 (58%), and 5 (3%) were subtype B, non-B and NT, respectively. No difference was thus observed in terms of distribution of subtype B vs subtype non-B clinical samples among data obtained by both techniques.
Among the 51 samples not quantifiable with the Cepheid assay, 40 were detected with Cepheid (ie, detected at <40 copies/mL but not quantifiable). These samples had a low VL with the Abbott assay, with a range of 40 copies/mL (1.6 Log)–170 copies/mL (2.23 Log), a median of 56 copies/mL (1.75), and an average of 61 copies/mL (1.76 Log). There were 11 samples quantifiable with the Abbott assay that was undetectable with Cepheid Xpert HIV-1 VL. These samples also had a low VL with a range of 41 copies/mL (1.61 Log)–77 copies/mL (1.89 Log), a median of 55 copies/mL (1.74 Log), and an average of 54 copies/mL (1.73 Log). Regarding the genetic diversity of these 51 different samples, 26 were subtype B, 10 were CRF02_AG, 5 were URFs, 5 were NT, and there was 1 for D, F1, CRF01, CRF06, CRF11, each.
There were 14 samples detected with Abbott that were undetectable with the Cepheid kit. Of these 7 were subtype B, 2 were CRF02, 2 were NT, and there was 1 for A, CRF06, CRF11 each. Conversely, there were 7 samples detected with Cepheid that were undetectable with Abbott. Of these samples, 3 were subtype B, 2 were subtype A, 1 was CRF02, and 1 was an URF.
Quantitative analysis was performed for 162 clinical samples with quantifiable results by both systems (104 subtype non-B and 58 B). These samples had VLs ranging from 40 copies/mL (1.6 Log) to 6,800,000 copies/mL (6.83 Log). Correlation between the Cepheid Xpert HIV-1 VL and the Abbott RealTime HIV-1 Assay was performed using a Passing–Bablok regression on the 162 samples. Figure 1 shows very good correlation between both assays with a slope for the regression line of 1.0355 (1.0114–1.0597) and an intercept of −0.1184 (−0.2151 to −0.03756), showing a slightly higher quantification in the lowest VL for Abbott and in the highest VL for Cepheid, with a close quantification among the middle ranges of VL.
FIGURE 1.: Scatter diagram and the regression line according to Passing and Bablok on the 162 samples quantifiable by both the Cepheid and Abbott assays. In addition to the regression line (solid line), the CI for the regression line (dashed lines) is displayed. Samples are differentiated between subtype B (●, n = 58), subtype CRF02_AG (▼, n = 19), and other non-B (○, n = 85).
Using a Bland–Altman analysis (Fig. 2 ), we observed an extremely close quantification of all samples, because the mean of differences was −0.01. Values were close along the VL level, with slightly higher values for the Abbott assay in the lowest levels (between 1.6 and 2.5 Log, Figs. 1 and 2 ). No dispersion or heterogeneous distribution was observed, with SD of +0.4 and −0.4, values less than 0.5 considered as the maximum difference between 2 techniques. Only 2 of the 162 (1.2%) samples had a difference of quantification of more than 0.5 Log. Among them, 1 (CRF02_AG) had a difference of −0.53 (in the 2–2.5 Log range) in favor of Abbott. The other sample (subtype B) had a difference of +0.78 (in the 4–4.5 Log range) in favor of Cepheid.
FIGURE 2.: Degree of agreement in Log copies/mL for the 162 samples quantifiable by both the Cepheid and Abbott assays. In the graphical method of Bland and Altman, the differences between the 2 techniques are plotted against the averages of the 2 techniques. Samples are differentiated between subtype B (●, n = 58), subtype CRF02_AG (▼, n = 19), and other non-B (○, n = 85).
Performance for detection and quantification of the HIV-1 group's non-M samples (N, O, and P) was analyzed for the 3 dilutions of the 18 culture supernatants (Fig. 3 ). Results showed a median with Cepheid and Abbott of 2.80 and 2.74 for the “2–3 Log” dilution, of 3.80 and 3.69 for the “3–4 Log” dilution, and of 4.81 and 4.74 for the “4–5 Log” dilution, respectively. Globally, these values were linear along the dilutions for the same assay and very close between the assays (slightly higher for Xpert). For 4 supernatants, we observed significantly higher values with Cepheid than with Abbott. Of those found, 2 were group N samples (difference of +0.87 Log and +0.74 Log each at the 3–4 Log dilutions; similar differences were found for the 2 other dilutions) and 2 of the 15 group O samples (difference of +0.87 Log and +1.83 Log at the 3–4 Log dilutions; similar differences were found for the 2 other dilutions). No difference (regardless of dilutions) was observed for the unique group P sample.
FIGURE 3.: HIV-1 VL results obtained for the 3 dilutions of supernatants of groups N, O, and P. In the Box-and-Whisker plot, the central box represents the values from the lower to upper quartile (25–75 percentiles) and the middle line represents the median. For each dilution, groups are differentiated between group O (○, n = 15), group N (●, n = 2), and group P (▼, n = 1).
DISCUSSION
The aim of this study was to evaluate the performance of the new HIV-1 VL test from Cepheid. Given the important genetic diversity of HIV-1, we wanted to evaluate performance on a large panel of samples representative of the current HIV-1 diversity. We thus tested 295 clinical samples, including 120 subtype B and 150 subtype non-B of HIV-1 group M, as well as 18 culture supernatants derived from the diversity of the rare variants of groups N, O, and P.
This wide panel of specimens tested demonstrated that the performance of Xpert HIV-1 VL assay was globally close to that of the Abbott HIV-1 RealTime assay. Comparison of the quantifiable samples (ie, >40 copies/mL) by both techniques showed a remarkable correlation between the 2 systems regardless of the VL level because the mean difference was almost zero, what we did not observe in previous comparisons of VL assays.4,6 10 genetic diversity from their samples; in particular, involvement of a particular subtype that will not be well represented in our panel. Distribution of VL values is also highly correlated with very little scattering of points in a variance of ±0.4 Log. This very strong correlation is even more interesting given that the technologies and regions of amplification are different between the 2 assays. For the rarest divergent variants group N, O, and P circulating in Central Africa and in the countries associated with this region, the results showed a good correlation between the 2 assays with a more significant quantification for some samples with the Xpert HIV-1 VL kit. In particular, an underquantification of group N samples by the Abbott test was observed, confirming data of a previous evaluation.4 11
Concerning the concordance of assay results near the limit of quantification, data showed that 11 quantifiable samples above the threshold of 40 copies/mL with the Abbott technique were undetectable with the Cepheid assay. These data seem concordant with those reported by Mor et al,10 12 13 14,15 16 genetic diversity of these samples showed that these results were not related to a particular subtype, because it involved samples of subtype B (n = 26) and non-B subtypes (n = 20) or NT (n = 5).
It is important to note that a low indeterminate rate of approximately 3% was observed with the Cepheid assay. This rate, rarely shown in studies, should be considered, because it is important for questions of cost and time for results; a low indeterminate rate is an indirect indicator of quality and robustness of the test. Unfortunately, insufficient samples were available to allow for assessment of whether invalid results were sample specific or incidental.
The use of this assay and its unique technology in a cartridge has proved extremely simple compared with the Abbott platform. The volume requirement of 1 mL of plasma to add to the single-use disposable cartridge greatly simplifies the technical task and can be performed by any staff member from reference laboratories to point of care sites. This is an extremely positive and favorable element for implementation in small laboratories or in difficult field conditions. In addition, the achievement of extraction and amplification in a single cartridge within a relatively short period (90 minutes) is also extremely important to be able to respond in emergency or “on demand” situations, as compared with the longer and heavier workflow of the series with the Abbott platform.
Other positive elements of this technology are the absence of mechanical requirements (ie, for extraction) and the ease of transport (for a module of 4 cartridges), allowing the technology to be implemented anywhere, especially in low-income countries. The GeneXpert can also be used in larger laboratories for high-volume samples with up to 80 modules or even for emergency use because the technology is not burdened by the series process. The sample addition to the cartridge remains manual compared with the automated Abbott platform. This can be a disadvantage but if the test is performed in “real time,” this manual step is equivalent to a stage of manual decantation of the blood samples for freezing before testing on the platform.
In conclusion, this study, using a wide panel in terms of number of samples, genetic diversity , and VL ranges, allowed us to evaluate this new assay in contexts representative of diverse clinical situations and genetic background. Our data showed that the Xpert HIV-1 VL assay offered very good performance for detection and quantification of the current HIV-1 genetic diversity . The samples quantifiable by the Abbott assay that were below 40 copies/mL by the Cepheid technique are an issue and requires further studies with nonselected samples near the limit of quantification. It is also necessary now to compare this new test to other assays, such as the Roche Cobas Taqman HIV-1, also widely used, to confirm its global performances. In addition, the practicability of this new assay makes it a useful and interesting tool both for high-income countries and in low-income countries, where it could facilitate and improve follow-up of patients.
ACKNOWLEDGMENTS
The authors also thank the patients for the COREVIH Haute-Normandie and all the staff of the laboratory, particularly Fabienne De Oliveira, Marie Leoz and Anne Savalle, for characterizing the collection samples.
REFERENCES
1. Labbett W, Garcia-Diaz A, Fox Z, et al. Comparative evaluation of the exavir load version 3 reverse transcriptase assay for measurement of human immunodeficiency virus type 1 plasma load. J Clin Microbiol. 2009;47:3266–3270.
2. Bourlet T, Signori-Schmuck A, Roche L, et al. HIV-1 load comparison using four commercial real-time assays. J Clin Microbiol. 2011;49:292–297.
3. Garcia-Diaz A, Labbett W, Clewley GS, et al. Comparative evaluation of the artus HIV-1 QS-RGQ assay and the abbott realtime HIV-1 assay for the quantification of HIV-1 RNA in plasma. J Clin Virol. 2013;57:66–69.
4. Mourez T, Delaugerre C, Vray M, et al. Comparison of the bioMerieux nucliSENS easyQ HIV-1 v2.0-HIV-1 RNA quantification assay versus abbott realtime HIV-1 and Roche Cobas TaqMan HIV-1 v2.0 on current epidemic HIV-1 variants. J Clin Virol. 2015;71:76–81.
5. Hopkins M, Hau S, Tiernan C, et al. Comparative performance of the new Aptima HIV-1 quant dx assay with three commercial PCR-based HIV-1 RNA quantitation assays. J Clin Virol. 2015;69:56–62.
6. Sire JM, Vray M, Merzouk M, et al. Comparative RNA quantification of HIV-1 group M and non-m with the roche cobas ampliPrep/cobas taqMan HIV-1 v2.0 and abbott real-time HIV-1 PCR assays. J Acquir Immune Defic Syndr. 2011;56:239–243.
7. Raja S, Ching J, Xi L, et al. Technology for automated, rapid, and quantitative PCR or reverse transcription-PCR clinical testing. Clin Chem. 2005;51:882–890.
8. Mourez T, Simon F, Plantier JC. Non-M variants of human immunodeficiency virus type 1. Clin Microbiol Rev. 2013;26:448–461.
9. Plantier JC, Dachraoui R, Lemee V, et al. HIV-1 resistance genotyping on dried serum spots. AIDS. 2005;19:391–397.
10. Mor O, Gozlan Y, Wax M, et al. Evaluation of the realTime HIV-1, xpert HIV-1, and aptima HIV-1 quant dx assays in comparison to the nucliSens easyQ HIV-1 v2.0 assay for quantification of HIV-1 viral load. J Clin Microbiol. 2015;53:3458–3465.
11. Gueudin M, Leoz M, Lemee V, et al. A new real-time quantitative PCR for diagnosis and monitoring of HIV-1 group O infection. J Clin Microbiol. 2012;50:831–836.
12. WHO. Consolidated Guidelines on the Use of Antiretroviral Drugs for Treating and Preventing HIV Infection: Recommendations for a Public Health Approach. World Health Organization, Geneva, Switzerland. 2013.
13. Essajee S, Kumarasamy N. Commentary: the monitoring of adults and children on antiretroviral therapy in the 2013 WHO consolidated ARV guidelines. AIDS. 2014;28(suppl 2):S147–S149.
14. European AIDS Clinical Society (EACS) Guidelines Version. Available at:
www.eacsociety.org/guidelines/eacs-guidelines/eacs-guidelines.html . Accessed on October 8, 2015.
15. Recommandations du groupe d'experts. Prise en charge médicale des personnes vivant avec le VIH. Sous la direction du Pr Morlat; La documentation Française, Paris, France. 2013.
16. Marcelin AG, Delaugerre C, Beaudoux C, et al. A cohort study of treatment-experienced HIV-1-infected patients treated with raltegravir: factors associated with virological response and mutations selected at failure. Int J Antimicrob Agents. 2013;42:42–47.
