Correlation between Cepheid GeneXpert and Abbott M2000 assays for HIV viral load measurements

Abstract

Measurement of HIV viral load (VL) is the best indicator of success of antiretroviral therapy. We investigated the correlation between results by the Cepheid GeneXpert and a standard of care VL assay (Abbott M2000). This was a prospective study of people living with HIV who attended the department for routine VL measurement with the Abbott M2000. Consenting patients agreed to provide one extra blood sample for VL measurement with the Cepheid GeneXpert assay. One hundred patients consented to participate in the study. There were 18 patients with VL ≥ 40 copies/mL and 75 patients with VL < 40 copies on both assays. The two assays had 93% agreement, with a kappa of 0.79 (p < 0.001). Treating VL as a continuous variable found measurements to be significantly higher on the Cepheid GeneXpert assay than the Abbott (p = 0.002). Analysis of samples with VL ≥ 40 copies/mL on either assay (n = 25) found the mean difference between the two assays to be 0.31 log₁₀ copies/mL (95% limits of agreement: −0.63, 1.25). Whilst the measurements on the two assays are relatively highly correlated, there is a clear bias, with the Cepheid GeneXpert tending to give higher VL values.

Keywords

HIV viral load assays correlation GeneXpert Abbott

Background

HIV viral load (VL) measurement is the best prognostic method for monitoring success of combination of antiretroviral therapy (cART) for suppression of HIV replication.¹ Because of the lifelong requirement for daily cART intake, VL should be measured regularly to ensure full viral suppression with the cART regime.^2,3

As a standard practice, VL measurements have been carried out in laboratories. In order to reduce material waste and cost, laboratories generally run VL assays in batches. This requirement generally adds to the turnaround times of the VL test. Many currently available HIV VL assays require at least six hours to complete the process of VL measurement. It is therefore unlikely that VL results will be available on the same day of patients’ phlebotomy.

Cepheid’s GeneXpert (GeneXpert) is a novel VL assay that uses reverse transcription polymerase chain reaction technology to generate quantitative detection of HIV-1 RNA in human plasma within 90 min. The use of the assay should enable HIV clinicians to deliver their consultations on the same day of patients’ phlebotomy for VL measurement.

We aimed to assess the performance of the GeneXpert’s platform and investigate the correlation between its VL measurements with those of the Abbott M2000, an established laboratory-based real-time HIV-1 VL assay in a clinical setting.

Methods

This was a prospective blinded study on adult HIV-1 positive individuals attending HIV clinics for routine blood investigations in the University Hospitals Birmingham NHS Foundation Trust (UHB) between second and 20^th February 2015.

Consecutive consenting patients booked for their routine HIV follow up blood investigations, provided an extra whole blood venous specimen. These specimens were labelled with a unique study code, before being sent to the microbiology laboratory at the UHB for testing.

Study specimens were spun in the centrifuge before transferring 1 mL of plasma into the GeneXpert assay, according to the manufacturer’s instructions.⁴ The study’s unique codes were used to register patients’ specimens in the GeneXpert assay. Laboratory staff were blinded to the identity of participating patients.

HIV-1 VL measurements on the GeneXpert were recorded and linked with those generated by the Abbott M2000 at the end of the study.

Routine VL samples were spun before the use of 0.6 mL of plasma on the Abbott M2000 assay according to the manufacturer’s instructions.⁵ Those results were used for clinical management of the patients as before.

Study sample size

At the time of the study, no preliminary data comparing the performance of the two assays were available. Our study design assumed that 60% the patients would have undetectable (less than 40 copies/mL) VL counts. The remaining 40% of patients were powered using a sign test as an indicator of bias.

With a total sample size of 100 patients, the minimal detectable bias between the assays was +20 percentage points at 80% power and 5% alpha. In other words, if one assay gave higher measurements in 70% of the cases where the assays differed, then significant bias would be detected.

Statistical analysis

Both the GeneXpert and Abbott M2000 assays have a lower level of detection of 40 copies/mL and an upper level of detection of 10 million copies/mL. For statistical analysis, undetectable VL results (<40 copies/mL) were allocated an arbitrary value of 39 copies/mL.

Because of the large number of samples with undetectable VL levels, the two sets of measurements were initially dichotomized, based on whether a VL of ≥40 copies/mL was detected. Comparisons between the two assays were then performed using a kappa statistic to quantify the level of agreement and a McNemar’s test to test for bias. These comparisons were then repeated for VL counts greater than 200 copies/mL, which represent the threshold of significant viraemia.^3,6

Since the values followed a highly skewed distribution, assessments of correlation were performed using Spearman’s rho, with Wilcoxon’s test used to compare the two assays. A Bland-Altman analysis was then performed, to assess the consistency of the two assays. VL levels were log₁₀-transformed, to reduce the degree of skew in the distribution. The mean difference was then calculated, as well as the 95% limits of agreement (LoA: mean ± 1.96 SD). As a high proportion of samples returned undetectable VLs on both assays and, hence differences of zero, the distribution of the mean differences was highly leptokurtic. For this reason, the calculated mean and LoAs of the difference between the assays were likely not to be valid as these statistics assume normality. To address this issue, the analysis was also repeated after excluding those samples that were undetectable on both assays.

All analyses were performed using IBM SPSS Statistics 22 (IBM Corp., Armonk, NY, USA), with p < 0.05 deemed to be indicative of statistical significance.

The study was reviewed and approved by the West Midlands Research Ethics’ Committee (REC reference: 14/WM/1232).

Results

One hundred patients consented to participate in the study. They had HIV infection with subtypes A (2 cases), B (36 cases), C (29 cases), CRF02_AG (2 cases) and H (1 case). The remainder of the patients (n = 30) first attended the department with undetectable VL (<40 copies/mL) for whom HIV resistance profile and subtyping was not available.

Of the 100 samples analyzed, 75 were undetectable (VL < 40 copies/mL) on both assays, and 18 gave detectable levels (VL ≥ 40 Copies/mL) on both assays, showing reasonable consistency, with 93% agreement, and a kappa of 0.79 (p < 0.001). There was non-significant (McNemar’s test: p = 0.125) evidence of bias, with six samples returning detectable on the GeneXpert but not on the Abbott (VLs of 41, 50, 51, 98, 126 and 1090 copies/mL), compared to only 1 (VL of 77 copies/mL) in the opposite direction (Table 1).

Table 1.

Comparison between detection rates (VL ≥ 40 copies/mL) on the two assays.

Abbott	GeneXpert
Abbott	VL < 40	VL ≥ 40
VL < 40	75	6
VL ≥ 40	1	18

VL: viral load; VL units = copies/mL. Kappa = 0.79 (p < 0.001). McNemar’s test: p = 0.125.

None of the six patients with plasma VL < 40 copies/mL on the Abbott and ≥40 copies/mL on the GeneXpert has evidence of viral mutation on resistance assays. The one specimen with undetectable VL on the GeneXpert and detectable VL count on the Abbott did not amplify on testing for the viral resistance.

The analysis was also performed using VL ≥ 200 copies/mL as a cut-off value (Table 2), which showed greater concordance (kappa = 0.97), with only one disagreement between the two assays (i.e. 99% agreement).

Table 2.

Comparison between rates of VL ≥ 200 copies/mL on the two assays.

Abbott	GeneXpert
Abbott	VL < 200	VL ≥ 200
VL < 200	81	1
VL ≥ 200	0	18

VL: viral load. VL units = copies/mL. Kappa = 0.97 (p < 0.001). McNemar’s test: p = 1.000.

Treating VL as a continuous variable returned a Spearman’s rho correlation between the assays of 0.87 (p < 0.001, Figure 1). However, the VL measurements from the GeneXpert were found to be significantly higher than those of the Abbott (Wilcoxon’s test: p = 0.002). Of the 25 cases where the two assays did not return identical VL levels, the GeneXpert gave higher measurement in 22 (88%) cases.

Figure 1.

Scatter plot of viral load measurements from the two assays.

The two assays compared using a Bland–Altman approach. When all samples were considered in the analysis (Figure 2(a)), the mean difference between the two assays was found to be 0.08 log₁₀ copies/mL (95% LoA: −0.46, 0.61), with the tendency for measurements to be higher on the GeneXpert. This analysis assumed that the difference in VL was zero for samples that were undetectable on both assays.

Figure 2.

Bland–Altman analysis of the GeneXpert versus the Abbott for: (a) the full set of 100 samples and (b) the 25 samples that had a detectable viral loads on either assay.

The analysis was repeated for the subgroup samples with VLs that were detectable on at least one of the assays (N = 25, Figure 2(b)). This increased the mean difference to 0.31 log₁₀ copies/mL (95% LoA: −0.63, 1.25).

Discussion

Our study used prospectively collected paired specimens to compare the Abbott M2000 and GeneXpert VL assays. We found a good correlation between plasma VL counts measured by the two assays. This was particularly reassuring as the study specimens included different HIV subtypes. This suggests that switching from the Abbott M2000 to GeneXpert should not produce discordant results for the majority of patients. Our results are consistent with those reported previously.⁷

When categorizing plasma VL results into less than and greater than 40 copies/mL, 93% of the results by the two assays were in agreement in our study. This is consistent with other published studies that have reported 84–93% agreement between the GeneXpert and Abbott M2000.^6–9

Low-level viraemia is a recognized observation close to the lowest level of quantification of most assays.^10–13 Current guidelines recommend low-level viraemia with values less than 200 copies/mL can be considered as clinically suppressed without the need of further investigations.^2,3 When using a cut-off of 200 copies/mL in our study, the consistency between the two assays was greater. Hence, the two assays are highly consistent for identifying those patients where further investigations are required.

Treating the VL measurements as continuous variables did find a mean difference of 0.08 log₁₀ copies/mL between the two assays. The difference is consistent with those of between 0.08 and 0.13 log₁₀ copies/mL reported previously.^6–9 This can potentially impact on the choice of cART for the management of HIV infection as abacavir and rilpivirine would not be suitable for those with VL values of over 100,000 copies/mL.^2,3

The GeneXpert has a number of advantages for clinical use. Its modular design permits flexibility with the volume of the specimens that can be processed at any time. It also generates VL measurement results within 90 min that makes it ideal for a model of HIV care delivery where the patients receive their results on the same day of undergoing phlebotomy. The use of the assay is easy and does not require special laboratory skills.

However, strict adherence to the manufacturer’s instructions is required, in order to reliably produce VL measurements. For example, inadequate volume of specimens in the cartridge has been associated with failure to produce results, which occurred in 9% (26/300) of specimens in one study.⁶ This did not occur in any of the samples from our study, which may be explained by the previous experience of our laboratory staff with this assay. One procedural drawback with the GeneXpert is that it requires a higher volume of blood samples than the Abbott M2000, leaving less residual specimen for viral resistance profiling when required.

There is currently no single “gold standard” plasma VL assay available. The Abbott M2000 has been reported to have more consistent results at lower plasma VL measurements with robust inter or intra reproducibility.^10,11 It may also report lower values for plasma VL counts close to its lowest level of detection.^12,13

The main advantages of this study were that it used paired samples obtained prospectively in a routine clinical setting. The study population was representative of HIV patients in the United Kingdom where 88% of patients have a plasma VL count of less than 50 copies/mL.¹⁴The study participants were infected with different HIV subtypes that should improve the generalizability of our findings.

The study also had some limitations. Because of the limited quantities of blood specimens obtained, we did not repeat testing of samples with discordant results between the GeneXpert and Abbott M2000 on the same assay, a recommendation of the current guidelines.^3,4 Also, whilst the proportion of patients with detectable VL were representative of the UK HIV populations, the fact that it was relatively small caused issues in the statistical analysis. Consequently, whilst the findings are generalizable to the UK HIV population, this may not be the case for populations with vastly different rates of detectable VL.

Because of the availability of plasma VL counts on the day of clinical attendance, the use of the GeneXpert HIV VL assay may provide an opportunity for delivering a new model of HIV care with better patients’ acceptance whilst maintaining good quality of clinical care.

Footnotes

Acknowledgements

The authors would like to acknowledge the patients’ contribution to the study. The study was funded by Cepheid Inc. Cepheid Inc. did not contribute to the design, execution, analysis or write up of the manuscript.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

ORCID iD

Kaveh Manavi

References

Mellors

Rinaldo

Jr Gupta

, et al. Prognosis in HIV-1 infection predicted by the quantity of virus in plasma. Science 1996; 72: 1167–1170.

British HIV Association . BHIVA guidelines for the treatment of HIV-1-positive adults with antiretroviral therapy 2015. https://www.bhiva.org/file/RVYKzFwyxpgiI/treatment-guidelines-2016-interim-update.pdf (2016 interim update, accessed 16 April 2020).

Department of Health and Human Services . Guidelines for the use of antiretroviral agents in HIV-1 infected adults and adolescents. https://aidsinfo.nih.gov/guidelines/html/1/adult-and-adolescent-arv/0 (2019, accessed 16 April 2020).

Cepheid . Xpert HIV-1 viral load. https://www.fda.gov/media/136314/download (accessed August 09, 2020).

Abbott . Abbott RealTime HIV-1. https://www.fda.gov/media/73278/download (accessed August 09, 2020).

Caffa

Luhanga

Andreotti

, et al. Comparison of the Cepheid GeneXpert and Abbott M2000 HIV-1 real time molecular assays for monitoring HIV-1 viral load and detecting HIV-1 infection. J Virol Methods 2016; 229: 35–39.

Agutu

Ngetsa

Price

, et al. Systematic review of the performance and clinical utility of point of care HIV-1 RNA testing for diagnosis and care. PLoS One 2019;14(6): e0218369. DOI: 10.1371/journal.pone.0218369.

Mor

Gozlan

Wax

, et al. Evaluation of HIV-1 RealTime, Xpert HIV-1 and Aptima HIV-1 Quant Dx assays in comparison to the NucliSens EasyQ v2.0 HIV-1 assay for quantification of HIV-1 viral load. J Clin Microbiol 2015; 53(11): 3458–3465.

Gous

Scott

Berrie

, et al. Options to expand HIV viral load testing in South Africa: evaluation of the GeneXpert HIV-1 viral load assay. PLoS One 2016; 11(12): e0168244. DOI: 10.1371/journal.pone.0168244.

10.

Bourlet

Signori-Schmuck

Roche

, et al. HIV-1 Load comparison using four commercial real-time assays. J Clin Microboil 2011; 49: 292–297.

11.

Naeth

Ehret

Wiesmann

, et al. Comparison of HIV-1 viral load assay performance in immunological stable patients with low or undetectable viremia. Med Microbiol Immunol 2013; 202(1): 67–75.

12.

Amendola

Marsella

Bloisi

, et al. Ability of two commercially available assays (Abbott Real time HIV-1 and Roche Cobas AmpliPrep/Cobas Taqman HIV-1 version 2.0) to quantify low HIV-1 RNA levels (<1000 copies/mililiter): comparison with clinical samples and NIBSC working reagent for nucleic acid testing assays. J Clin Microbiol 2014; 52(6): 2019–2026.

13.

Yan

Hanafi

Kelleher

, et al. Lack of correlation between three commercial platforms for the evaluation of human immunodeficiency virus type 1 (HIV-1) viral load at the clinically critical lower limit of quantification. J Clin Virol. 2010; 49(4): 249–253.

14.

Public Health England . HIV New Diagnoses, Treatment and Care in the UK 2019 report. https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/469405/HIV_new_diagnoses_treatment_and_care_2015_report20102015.pdf (accessed 16 April 2020).