Comparing HIV viral load assays and frequency of low level virological rebound in clinical practice

Introduction

The efficacy of antiretroviral therapy (ART) for HIV infection has traditionally been measured by the achievement of an ‘undetectable’ plasma viral load (VL), where the VL is less than the lower limit of quantification (LLOQ) of the specific assay utilised. However, with the advent of real-time nucleic amplification tests (NAATs) for the measurement of VLs in clinical practice, the LLOQ has progressively decreased from a previous 400–500 copies/ml to currently under 20 copies/ml, making an undetectable level a more difficult target to achieve. In addition, it is now well recognised that these VL assays have a degree of variability at their LLOQ, making results around this level less reliable.^1,2

Low-level virological rebound (LLVR) in patients stable on ART is a well-recognised and studied phenomenon, usually defined as detectable viraemia of <1000 copies/ml. ³ Where LLVR is a single occurrence, with a succeeding (and preceding) undetectable level, usually measured within a period of 3 months, it is described as a VL ‘blip’. ⁴ There are multiple suggested reasons for the occurrence of LLVR, including sub-optimal ART adherence, ⁴ transient biological virological fluctuations associated with inter-current illness or vaccination,^5,6 drug interactions reducing ART drug levels ⁴ and episodic virus release from latent reservoirs. ⁷ A further significant cause, that has become more obvious of late, is laboratory-related or VL assay variability at the lower margins of detection. ⁸ The majority of LLVRs settle spontaneously without any need for a change of ART regimen,^4,9,10 but in a proportion of patients may herald future true virological failure.^11–13

At the Edinburgh Genitourinary Medicine (GUM) department HIV service, three separate assays were used consecutively over the relatively short period of November 2006 to June 2012, for measurement of patient VLs: the Roche HIV Cobas Taqman version-1 assay (used between 23 November 2006 and 10 May 2010), followed by Roche HIV Cobas Taqman version-2 (used between 11 May 2010 and 05 June 2011) and most recently the Abbott RealTime HIV-1 assay (used from 06 June 2011 onwards).

While the Roche Taqman version-2 assay had a manufacturer-reported LLOQ of 20 copies/ml, the Edinburgh laboratory reported all three assays reviewed as with a lower limit of detection of 40 copies/ml. Thus, patients who had a VL of between 20 and 39 on the Roche assay were reported as undetectable.

Anecdotally, clinicians at our service reported noticeable differences in the relative frequency of LLVR results identified by the three assays; with a rise in the number of non-significant detectable VL results identified with the change of the HIV Taqman version-1 (CTM v1) to the Roche HIV Taqman version-2 (CTM v2) assay, followed by an obvious fall when it was replaced with the Abbott RealTime (Abbott RT) assay. We therefore conducted a retrospective study to investigate whether these suspicions regarding assay result variability were justified. The aim of this study was to review episodes of detectable VLs using these three assays and determine whether certain assays are more sensitive with regard to producing LLVR or ‘blips’. As a secondary aim, we compared patients who had more frequent occurrences of LLVR with those who had none or only one over the study period, to identify any significant differences between these groups.

Methods

For the purposes of the study, we reviewed a roughly equal time period for each of the three VL assays, of 12–13 months each: between 10 April 2009 and 10 May 2010 for the Roche CTM v1 assay, between 11 May 2010 and 05 June 2011 for the Roche CTM v2 assay and between 06 June 2011 and 06 June 2012 for the Abbott RT assay. As noted earlier, all three assays were considered to have a LLOQ of 40 copies/ml.

We selected as our study population the first 250 patients from our HIV cohort who had at least one VL performed on each of the three assays over the relevant study period. Data were extracted from both clinical notes and patient electronic records. Information gathered included patient demographics (age, gender, ethnicity and sexual orientation), ART status, ART history and VL results over the study period. Of the VL rebound episodes in patients on ART, those <1000 copies/ml (LLVR) were scrutinised for any documented clinical explanation for the same (e.g. poor adherence, diarrhoeal illness, drug-drug interactions, etc). LLVR with no documented reason were classed as ‘NDR’ (no documented reason). Study data analysis was using SPSS 19.

Results

The original planned study cohort of 250 patients was reduced to 247 patients as notes for three patients could not be found.

Table 1 shows the comparison of episodes of LLVR with NDR across the three assays over the study period. Table 2 and Graph 1 demonstrate the frequency and value of detectable VLs with NDR <1000 for each separate assay.

OPEN IN VIEWER

Table 1.

Detectable viral loads (VLs) on the separate assays over the study period.

	Roche CTM v1 assay	Roche CTM v2 assay	Abbott RT assay
Number of patients tested during study period	(247)	(247)	(247)
Number of VL tests performed over study period	902	726	537
Study period during which assay was used	10/04/09–10/05/10	11/05/10–05/06/11	06/06/11–06/06/12
Minimum number of VLs per patient on each assay	1	1	1
Maximum number of VLs per patient on each assay	7	5	5
Average number of VLs per patient on each assay	3.65 (95% CI 2–6)	2.94 (95% CI 1.4–4)	2.17 (95% CI 1–4)
No. of detectable VLs (% of all VLs during study period)	374 (41.5%)	233 (32.1%)	76 (14.2%)
No. of detectable VLs in patients on ART	249	120	55
No. of detectable VLs of <1000 copies/ml in patients on ART with NDR (% of all detectable VLs in patients on ART)	161 (64.6%)	112 (93.3%)	19 (34.5%)
Mean detectable VL of <1000 copies/ml in patients on ART with NDR (range)	141 (41–959)	125.8 (41–917)	142.4 (41–487)

OPEN IN VIEWER

Table 2.

Frequency and value of detectable viral loads (VLs) NDR <1000 for each assay.

Value of detectable VL (copies/ml)	Roche 1 (Series 1)	Roche 2 (Series 2)	Abbott (Series 3)
40–100	91	71	10
100–200	37	18	7
200–300	18	14	0
300–400	7	6	0
400–500	2	1	2
500–600	2	0	0
600–700	2	1	0
700–800	0	0	0
800–900	0	0	0
900–1000	2	1	0

NDR = no documented reason.

OPEN IN VIEWER

Graph 1.

Frequency and value of detectable viral load (VLs) NDR <1000 for each assay.

It is noticeable that the average number of VLs per patient reduced over the time period reviewed, from 3.6 on the Roche CTM v1 assay, to 2.9 on the Roche CTM v2 assay, to only 2.2 on the Abbott RT assay. This likely reflects departmental policy change in relation to monitoring of patients stable on ART, from a previous routine repeating of blood tests at 3 months, to a now standard of care of 6 months. However, it is also possible, especially for the Abbott RT in comparison to the CTM v2 assay, that the drop in the relative frequency of LLVR results allowed a reduction in the number of related repeat tests requested.

Of the three assays, the Roche CTM v1 assay produced the highest overall number of detectable VL results in patients on ART. However, a significant proportion of these detectable results in all three assays is likely to be explained by poor treatment adherence or undisclosed stopping of therapy by patients, as well as individual assay variability. We therefore compared between the three assays the relative proportion of detectable results which were more likely to be a consequence of assay variability: i.e. results of <1000 copies/ml in patients stable on therapy with no documented clinical reason (NDR). Utilising the Chi square test, there was a significant difference between the three groups in the proportion of <1000 copies/ml results with NDR (p < 0.001). When the proportion <1000 copies/ml with NDR results were compared using pairwise analyses between each assay (i.e. assay 1 vs 2, 2 vs 3 and 1 vs 3), in each comparison there was a significant difference in each analysis, with the Roche CTM v2 assay highest and the Abbott RT lowest (p < 0.05).

It is interesting that the mean VL was relatively similar for all three assays when considering those results that were <1000 copies/ml with NDR. This is despite the VL result range being significantly more restricted on the Abbott assay, the upper value being only 487. As demonstrated by Table 2 and Graph 1, this is because relatively more of the results <1000 copies/ml and NDR picked up by the Roche CTM v1 and v2 assays were at a lower level, compared to the Abbott assay.

As a secondary aim, we also explored patient demographics of gender, race and sexual orientation, as well as likely mode of HIV acquisition (heterosexual sexual intercourse, homosexual/bisexual sexual intercourse or intravenous drug use), for any significant relationship to the study population developing none, single or multiple episodes of LLVR with NDR over the study period. For patients who had remained on the same regimen over the whole study period (n = 105, 67.7%), their ART regimen type (i.e. protease inhibitor-based ART, non-nucleoside reverse transcriptor-based ART or any other regimen type) was also considered. In our study, there was no significant association between any of the above reviewed factors and the frequency of patients developing LLVR with NDR.

Discussion

The increased incidence of LLVR of uncertain clinical significance following a change of the VL assay utilised is well described, most notably with the replacement of the Roche Cobas Amplicor with the Roche CTM v1.^8,14–16 Our study was undertaken to investigate this phenomenon, specifically in relation to the Roche CTM v1, Roche CTM v2 and Abbott assays, and within the clinical context.

Our results confirmed the initial clinical suspicions raised about a discrepancy between three VL assays in the relative proportion of LLVR results with NDR in patients on therapy; with Roche CTM v2 producing a significantly higher proportion of LLVR with NDR than the CTM v1, which was again considerably more than the relative proportion generated by the Abbott RT assay.

Previous research comparing the Roche CTM v1, CTM v2 and Abbott assays in terms of propensity to produce LLVR results have produced conflicting results. While some studies suggested these assays to be comparable in this regard, ¹⁷ others have shown a lack of concordance of assay results especially at the lower level of detectability. ¹⁸ Thus, similar to the results of this study, a number of other research studies have revealed the Roche CTM v2 to yield more VL blips and/or higher values on the same detectable sample than the Abbott RT assay.^19–21 Equally, again agreeing with the results of this study, some other studies comparing the CTM v1 and v2 assays have suggested that the CTM v2 identified more LLVR or higher blip values than the CTM v1 assay. ²²

We failed to identify any association in terms of gender, race, sexual orientation or type of regimen with frequency of LLVR. In comparison, in Kuyper et al.’s study, women were more likely to experience virological rebound. ²³ However, the authors concluded that this was secondary to worse ART adherence, usually related to social stressors, in the female subjects in their study. Their results therefore may not be directly extrapolated to other patient populations. Equally, unlike our study, in research by Philips et al and Smith et al, some ART regimens were more commonly associated with VL rebound.^24,25 Most such regimens however are no longer commonly used in clinical practice, due to a relatively high level of side effects. Certainly there are no patients in our department remaining on regimens including, for example, indinavir and nelfinavir. In addition, due to the relatively small patient numbers in our study, we were unable to review specific antiviral regimens, instead considering PI-based and NNRTI-based regimens as groups, possibly explaining the discrepancy between our and these two studies’ results.

A majority of above-mentioned studies reviewing the new VL assays assessed their performance by comparing their results when utilised to test the same clinical samples. Objectively, retrospective testing utilising all three assays of all the study VL samples would be the most exact and direct method of comparison between them. This approach would have the benefit of excluding any effect due to, for example, confounding factors such as variations in patient adherence or lab operator techniques over the study time period. This is therefore a significant limitation of our study. However, it seems doubtful that there would have been a significant change in adherence in the large study patient cohort overall, over the relevant time periods, accounting for demonstrated changes in the detectable proportion. Similarly, results were all generated at the same reference laboratory, making lab technique variability unlikely as the grounds for change in results. On the other hand, arguably, our study has the advantage of reviewing results obtained when utilising the assays in ‘real-life’ clinical practice. In addition, a further strength to our research is that it evaluates three assays relative to each other, rather than the more commonly reviewed two in the above detailed studies.

Our study however raises some interesting questions; most importantly, whether the increased relative frequency of LLVR with NDR with some assays is arbitrary and clinically irrelevant or whether, as has sometimes been suggested, these more sensitive assays are identifying patients whose virological control is relatively borderline or just under the assay limit of detection.^26,27 Our results appear to indicate the latter: thus, at least for the Roche 1 and 2 assays, most of the LLVR with NDR were results near the border of detectability.

Yet, even if this is the case, opinion and data are conflicting whether there is any long-term virological outcome difference between, for example, those whose VL is between 20 and 50 copies/ml and those with levels consistently <20 copies/ml.^28–33 Our study however did not gather data regarding later progression to failure of patients with LLVR and NDR, and we are therefore unable to shed any further light on this issue. This is an important question that should continue to be investigated in future research studies.