The quantification of HIV-1 RNA in plasma has become the standard for assessing a patient’s plasma viral load (VL) at the time of diagnosis and throughout the course of the disease. Real-time PCR-based nucleic acid tests (NAT) have a high sensitivity and broad dynamic range [1] and the use of fine-tuned primer/probe sets has made it possible to detect isolates of HIV-1 groups M, N, and O as well as all known subtypes and circulating recombinant forms. However, due to the highly variable nature of HIV, primer/probe binding to the target region occasionally can be sub-optimal, resulting in underestimation of VL [2, 3]. It is important to establish such underestimation at the time of diagnosis, as VL monitoring will be used for monitoring disease progression and the success of antiretroviral therapy (ART). Falsely low or even undetectable viral load due to underestimation may impact the interpretation of the success of ART and increase the risk of inadvertent virus transmission. In Switzerland, regulations on HIV confirmation testing issued by the Swiss Federal Office of Public Health (SFOPH) in 2006 therefore request that a VL ≤1’000 cp/ml be confirmed by the product-enhanced reverse transcriptase (PERT) assay, a sequence-independent test that quantifies retrovirus particles based on their content of enzymatically active reverse transcriptase (RT) [4]. This test has been available in our country since 1994, and its sensitivity is comparable to NAT [5, 6]. VL based on the PERT assay, or on other RT-based tests, have been shown to correlate well with RNA-based VL over a wide dynamic range (10²-10⁶ cp/ml) [6–8].

Here, we analyzed the frequency of NAT-based VL ≤1’000 cp/ml among newly diagnosed, untreated HIV-1 patients in Switzerland and investigated whether such low VL were true or due to underestimation. To this end, we compared viral loads derived from NAT and PERT at the time of diagnosis and complemented these data with repeat and/or alternative NAT measurements. Alternative NAT involved testing in a different primer/probe target region, addressing the possibility of sequence polymorphism-associated failure of viral RNA detection.

In this study we investigated whether commercial viral load assays in use during the past few years provided reliable concentrations of HIV-1 RNA in plasma. We focused our analysis on patients with ≤1’000 cp/ml, because the possible risks for untoward consequences of VL underestimation, namely a delay of ART or further virus transmission in a false belief of being non-infectious [17, 18], are highest in this group. Confirmatory testing of the VL was possible in only 32 of the 71 patients with an initial NAT VL ≤1’000 cp/ml (45%). Possible reasons for this low rate may include the following: Usually, a new blood sample was needed for a PERT assay at the SNCR, which required the patient returning to the physician who received the information that a PERT assay was considered necessary. If the initial physician had meanwhile transferred the patient to an HIV specialist, this information may not have been forwarded to the specialist. Moreover, the decision of whether a test is performed or not ultimately remains in the physician’s responsibility; conceivable reasons for not performing a PERT assay include transient stay in the country or costs. Finally, it may not always have been possible to link a PERT result obtained for a patient sample received under full personal identity, with the coded personal information in the HIV notification.

Aside from these limitations, our analysis showed that the PERT assay confirmed the low diagnostic NAT VL in 78% of patients who had both tests. Seven patients had a clinically significant higher VL by PERT than NAT, with four patients showing a >10-fold difference. In three of these four patients we confirmed technical underestimation as the reason for the diagnostic low RNA VL, as demonstrated by significantly higher VL measurements when using alternative NAT platforms. These four most discrepant cases had all been tested with earlier NAT versions now outdated. Indeed, underestimation has been observed repeatedly for both CAM1.5 [13, 15] and CTM1.0 [2, 14, 16] in the past. One patient showed a ≥3-fold NAT-VL underestimation on the CTM2.0 compared to PERT, which disappeared in later samples. In this case we cannot exclude some technical variance of the PERT assay.

The majority of patients were classified as having an older infection, thus minimizing concerns about large changes in VL between the first and second sample collected. Four patients classified as having a recent infection (patient ID 7, 16, 26, 27), and all had a NAT and PERT VL of ≤1’000 cp/ml. As such low VL are rarely seen in patients with recent HIV-1 infection, we cannot exclude the possibility that the diagnostic NAT VL was falsely low, but nevertheless “confirmed” by PERT because the second sample was no longer in the acute, but in the chronic phase of infection in which the VL was already at the lower set-point. We can also not exclude the possibility that the classification as recent infection was false, because the diagnostic specificity of the algorithm used in these four cases was only 93% [9]. Therefore, it is impossible to draw any safe conclusion regarding the true VL at the time of diagnosis in these four cases.

Both B and non-B subtypes were affected by NAT-based VL underestimation in our study, in agreement with others who found that underestimation was associated with a rare polymorphism in the viral genome [19]. With the introduction of CTM2.0, which uses dual target detection in gag and LTR, VL underestimation decreased dramatically [16, 20, 21]. The ARm2000 platform uses the IN region as target, and VL measurements performed by CTM2.0 and ARm2000 were found to correlate well [15]. Despite all technological improvement, VL underestimation may still occasionally occur even with the contemporary NAT generation [22]. This could also be the reason that one of the patients shown in Figure 1 has a PERT VL well outside the lower limit of the 95% confidence interval.

In our study population, severe NAT VL underestimation was restricted to patients for whom the initial diagnostic VL measurement was performed on NAT platforms outdated today. Although we did not find similar severe VL underestimation for any of the 23 patients assessed on either CTM2.0 or m2000, the upper limit of the 95% confidence interval of this zero frequency is at 12.2%, suggesting a certain possibility for underestimation even with contemporary viral load tests. Continuation of the rule to confirm low VL with an alternative test is thus justified from a statistical viewpoint.

In light of the rapidly evolving global HIV-1 quasispecies it is virtually unavoidable that variants will arise which will be detected sub-optimally by a given sequence-based VL test. Confirmation of low VL by alternative methods like the sequence-independent PERT assay or a NAT targeting a different region of the viral genome will help to minimise such cases.