Correlation between Human Immunodeficiency Virus Type 1 (HIV-1) RNA Measurements Obtained with Dried Blood Spots and Those Obtained with Plasma by Use of Nuclisens EasyQ HIV-1 and Abbott RealTime HIV Load Tests

Abstract

The plasma human immunodeficiency virus (HIV) RNA load is used in the clinical routine for the monitoring of HIV infection and the patient's response to antiretroviral therapy. Other body fluids or dried blood spots (DBS) can be used, however, to assess the level of viremia. The use of DBS may be especially helpful for the monitoring of HIV-infected patients in resource-poor settings, where access to adequate laboratory facilities is often difficult. However, the correlation between the HIV RNA levels in plasma and those in DBSs has not been well established. Paired plasma and DBS samples obtained from HIV type 1 (HIV-1)-infected patients were tested for HIV RNA copy numbers by using two different commercial assays, the Nuclisens EasyQ HIV-1 (version 1.1) test (the Nuclisens test; Biomerieux) and the m2000rt RealTime HIV test (the m2000rt test; Abbott). Nucleic acid extraction was performed manually by using either the Nuclisens isolation kit (which uses the Boom methodology) or the m2000rt sample preparation kit (an iron particle-based method). A total of 103 paired plasma and DBS samples were tested. Viral load results were obtained for 97 (94.2%) samples with the Nuclisens isolation kit and 81 (78.6%) samples with the m2000rt kit. The overall correlation between the RNA loads in plasma and DBS was good, although better results were obtained by the Nuclisens test (R² = 0.87, P < 0.001) than by the m2000rt test (R² = 0.70, P < 0.001). While the specificities were excellent and similar for both the Nuclisens and the m2000rt tests (97.1% and 100%, respectively), the sensitivity was greater by the Nuclisens test than by the m2000rt test (75.8% and 56.6%, respectively). Overall, the viral loads in DBS tended to be lower than those in plasma, with mean differences of 0.3 log unit (standard deviation, 0.5 log unit) and 0.76 log unit (standard deviation, 0.8 log unit) for the Nuclisens and the m2000rt tests, respectively. The levels of agreement between the measurements in plasma and DBS were assessed by using the Bland-Altman plot for each assay. The Nuclisens test gave results within its defined limits (−0.65 to 1.26) for 95.9% of the samples, while the m2000rt test gave results within its limits (−0.83 to 2.33) for 100% of the samples. In summary, the HIV-1 load can accurately be quantified by testing DBS by either the Nuclisens or the m2000rt test, although the Nuclisens test may outperform the m2000rt test when nucleic acids are extracted manually.

Viral load measurement has turned out to be the major indicator of antiretroviral treatment success or failure in human immunodeficiency virus (HIV)-infected patients (16). Accordingly, periodic viral load monitoring is recommended for HIV-infected patients receiving antiretroviral therapy. Several commercial tests have been approved for the quantitation of HIV RNA in plasma, including (i) the Cobas TaqMan HIV type 1 (HIV-1) test (Roche Molecular Systems, Basel, Switzerland) (24), (ii) the Nuclisens EasyQ HIV-1 (version 1.1) test (the Nuclisens test; Biomerieux, Durham, NC) (12), (iii) the branched DNA Versant HIV-1 RNA (version 3.0) assay (Siemens, Berkeley, CA) (14), and (iv) the m2000rt RealTime HIV-1 test (the m2000rt test; Abbott, Chicago, IL) (13, 30).

The implementation of antiretroviral therapy in developing countries through the Toward Universal Access program, carried out by the World Health Organization (35) in collaboration with UNAIDS and UNICEF (33), is making the availability of viral load testing of HIV-infected patients in resource-limited settings urgently necessary. Since viral load testing is relatively expensive and requires adequate facilities with complex equipment and specialized personnel, the shipment of plasma samples to reference laboratories has been proposed as an alternative to the development of new facilities with the necessary equipment and personnel. However, that option is complicated because of the requirement of cool temperatures for storage and transportation. Therefore, the use of dried blood spots (DBS) is increasingly being considered as an alternative that can be used to overcome the troubles inherent to the testing of plasma. Patient's blood placed on DBS can be shipped at room temperature to specialized reference laboratories, where the specimens may be subject to HIV RNA quantitation (1, 2, 3, 8, 11, 22), HIV subtyping, and/or drug resistance testing (5, 15, 19, 28). DBS have also been successfully used to assist with the confirmation of a diagnosis of HIV infection (10, 18, 27), measurement of CD4 counts (23), and even analysis of antiretroviral drug levels (17). Nevertheless, the sensitivity and the degree of correlation between the viral loads obtained with DBS and those obtained with plasma have been a matter of concern. We report here on the correlation between the viral load in plasma and that in DBS, as well as the lower limit of detection in DBS, obtained by two commercial assays, the Nuclisens assay (Biomerieux, Durham, NC) and the m2000rt assay (Abbott, Chicago, IL).

MATERIALS AND METHODS

Specimens.

A total of 103 whole-blood samples were collected from HIV-infected patients receiving care at the Hospital Carlos III in Madrid, Spain, and placed in tubes containing EDTA. DBS were prepared by dropping 50 μl of blood in each circle (five spots per card) of filter paper (S&S903; Schleicher & Schull, BioScience GmbH, Barcelona, Spain) and were left to dry overnight at room temperature. They were then stored at 4°C until they were processed. Plasma aliquots were obtained by centrifugation of the whole blood, and the supernatant was stored at −20°C until use.

Before nucleic acid isolation, blood was eluted from the filter papers. For elution, scissors were used to cut one spot from the filter paper containing DBS for each patient (50 μl), and the filter paper spot was left in 2 ml of Nuclisens lysis buffer (Biomerieux, Madrid, Spain) for 2 h with gentle rotation. The filter paper was then removed from the tube, and the solution containing the material eluted from the DBS was further processed according to the manufacturer's instructions.

Viral load determination by Nuclisens assay.

Nucleic acid isolation was carried out by the method of Boom et al. (7), which is a silica-based technology. For DBS samples, extraction was performed manually with the Nuclisens isolation kit, whereas the Nuclisens EasyMag automatic extractor was used for the plasma specimens. Viral load quantitation was performed according to the manufacturer's instructions. The linear dynamic range of this assay ranges from 50 to 3,000,000 HIV RNA copies/ml when 1 ml of plasma is used (12).

Viral load determination by Abbott m2000rt assay.

RNA was extracted manually from both plasma and DBS samples by using the Abbott mSample preparation system, which uses an iron particle-based method. The viral load was then measured according to the manufacturer's instructions. The linear dynamic range of the assay ranges from 40 to 10,000,000 HIV RNA copies/ml when 1 ml of plasma is used (30).

Statistical analysis.

All values were log₁₀ transformed before they were analyzed. Paired sample t test statistics were used for comparisons of the loads obtained with plasma and those obtained with DBS, as well as for comparison of the results obtained by the two distinct assays. For analysis of the concordance between the results obtained with plasma and those obtained with DBS, the method of Bland and Altman (6) was used. All analyses were performed by using SPSS statistical software (version 15.0; SPSS Inc., Chicago, IL). Differences were considered significant when P values were less than 0.05.

RESULTS

Correlation between viral load values obtained with DBS and those obtained with plasma by Nuclisens assay.

Viral load measurements were originally intended to be made for 103 samples, but paired viral load results for DBS and plasma could be obtained for only 97 samples (94.2%). Invalid results for DBS were repeatedly obtained with four pairs of samples and for both DBS and plasma for another two pairs of specimens.

Figure 1a shows the correlation between the viral load determinations obtained with plasma and those obtained with DBS. The overall R² value was 0.87 (P < 0.001). Agreement between the undetectable and the detectable viral load results for plasma and DBS was seen for 83.5% of the samples (the true-positive rate), and the kappa coefficient was 0.67 (P < 0.001). This means that 15.8% of the DBS for which undetectable viremia was reported had measurable HIV RNA levels in plasma.

FIG. 1.

Linear regression comparing viral loads obtained with DBS and those obtained with plasma.

A deeper examination of these false-negative results obtained with DBS showed a relationship of the results to a lower detection limit threshold for DBS than for plasma. Overall, 97.7% of the samples (42/43) with more than 3.48 log HIV RNA copies/ml in plasma yielded detectable viremia by the use of DBS, while levels over a range of 3.08 to 3.48 log HIV RNA copies/ml measurable in plasma were measurable in only 60% (3/5) of the DBS samples.

The Bland-Altman plot (Fig. 2a) shows that the mean difference between the values obtained with DBS and those obtained with plasma by the Nuclisens assay was 0.31, with two times the standard deviation ranging from 0.65 to 1.27. All but four samples provided results within these limits. Overall, differences in viral load values of greater than 0.5 log unit obtained by testing DBS and plasma were seen for 33% of the specimens by the Nuclisens assay.

FIG. 2.

Bland-Altman plots of average viral loads in DBS compared to those in plasma.

Correlation between viral load values obtained with DBS and those obtained with plasma by m2000rt assay.

Pairwise determinations of the viral loads in plasma and DBS could be obtained for 81 of 103 samples (78.6%) that had been intended to be analyzed. Invalid results were repeatedly obtained for 13 DBS specimens and 8 plasma samples. Moreover, one sample pair yielded invalid results with both the DBS and plasma. The overall R² value between the measures obtained with DBS and those obtained with plasma was 0.70 (P < 0.001) (Fig. 1b).

Agreement between undetectable and detectable viral load results in plasma and DBS was recognized for 75.3% of the samples. Up to 23.8% of the specimens with measurable viremia in plasma yielded undetectable values when DBS were used. The kappa agreement of coefficient was 0.53 (P < 0.001).

Stratification of the samples into different plasma viral load intervals showed that viral loads greater than 3.72 log HIV RNA copies/ml were detectable by the use of DBS for 100% (22/22) of the specimens. In contrast, only 42.9% (three of seven) of the specimens with plasma viral loads between 3.23 and 3.72 log HIV RNA copies/ml had measurable viremia in DBS.

The Bland-Altman plot (Fig. 2b) showed that two times the standard deviation ranged from −0.80 to 2.34 for the Abbott m2000rt assay. All samples gave results within the limits defined by the standard deviation. The mean difference between measures obtained with DBS and those obtained with plasma was 0.77. Overall, 48.1% of the samples had values with differences greater than 0.5 log HIV RNA copies/ml when the levels obtained with DBS and those obtained with plasma were compared.

Correlation between Nuclisens and Abbott m2000rt assay results.

A comparison of the viral load results obtained by the Nuclisens and m2000rt assays showed a greater correlation when plasma was tested than when DBS were tested. The Pearson value for plasma determinations was 0.960 (P < 0.001) and the R² value was 0.921, while for the determinations with DBS, the Pearson value was 0.845 (P < 0.001) and the R² value was 0.714.

DISCUSSION

More than two-thirds of the estimated 34 million people living with HIV infection worldwide reside in low-income countries. About 3 million individuals in these countries currently have access to antiretroviral treatment (35). In this context, the availability of viral load testing is crucial to ensure the early recognition of treatment failure, which in turn favors the success of rescue interventions and avoids the accumulation of drug resistance mutations (4, 9, 20, 26). Since viral load testing requires specialized facilities and experienced personnel, alternative strategies have been explored. Recent reports have suggested that DBS may be suitable for use for the testing of HIV RNA levels, although most studies have so far examined only small sets of samples (1, 2, 25, 32, 34). Both the reliability and the sensitivity of tests that use DBS compared to those that use plasma have been matters of concern.

In the present study, we analyzed the reliability of two commercial real-time amplification assays used for the determination of viral loads, the Nuclisens and m2000rt assays, with more than 100 paired DBS and plasma specimens. In all instances, the corresponding manual methods for nucleic acid extraction were used. Both the Nuclisens and the m2000rt assays provided results with DBSs comparable to those provided with plasma, although the Nuclisens assay tended to be more accurate. In a similar study conducted by Zazzi and colleagues (18a), the preliminary results obtained with 175 paired samples tested by the m2000rt assay were better than the results obtained in our study. It should be noted, however, that nucleic acid isolation was performed automatically in their study, while we used a manual protocol. Since almost all commercial viral load tests have automated extraction methods, it seems reasonable to conduct studies like ours by using those updated procedures. The Nuclisens automated nucleic acid system is called EasyMag and is based on the same chemical principle as the manual procedure, the method of Boom et al. (7). For the m2000rt viral load test (7), the use of an automated method for the extraction of nucleic acids may improve the performance over that obtained by manual isolation, which is particularly tedious.

The correlation between the results obtained with plasma and those obtained with DBS by either the Nuclisens or the m2000rt assay was satisfactory. Moreover, the R² values between the two assays were 0.96 (P < 0.001) when plasma specimens were tested and 0.92 when DBS were tested. These results reinforce the suggestion that the measurement of viral loads in DBS by the use of either of these two commercial assays is reliable and may facilitate the more adequate monitoring of patients receiving antiretroviral treatment in resource-limited settings. A suspicion of antiretroviral treatment failure on the basis of only clinical manifestations or CD4 counts is clearly insufficient. A note of caution must also be kept in mind for the testing of HIV-1 non-B subtypes, which are the most prevalent in resource-limited settings (31), since the reliability of HIV RNA measurements in non-B subtypes for different HIV viral load assays is not uniform (12, 29).

The lower limit of detection for HIV RNA in DBS by either of the two commercial assays was greater than 3.5 log HIV RNA copies/ml. This threshold could be considered unsatisfactory, as the occurrence of a level of plasma viremia of greater than 3 log units while a patient is receiving antiretroviral therapy is well established to be associated with the selection of drug resistance and even clinical progression (21). Moreover, the World Health Organization has advised that drug resistance testing be performed for all patients with plasma HIV RNA loads greater than 1,000 copies/ml (36). The maintenance of a good correlation between the sensitivity obtained by the use of DBS and that obtained by the use of plasma in the lower viral load range seems to be important to support the use of DBS for viral load testing. Our study was carried out with just one blood spot for each determination, which means that the amount of blood used in the assay was only 50 μl. It would be interesting to examine whether the use of two blood spots for nucleic acid extraction would increase the amount of RNA and enhance the sensitivity of viral load testing with DBS. Moreover, the evaluation of viral load assays by the use of automated nucleic acid extraction protocols is warranted.

In summary, HIV RNA can be reliably quantified in DBS for samples with more than 3 log units by using either the Nuclisens or the m2000rt assay, although the first methodology seems to be the most precise, at least when manual nucleic acid extraction protocols are used. Further improvements in sensitivity are warranted to avoid the misrecognition of virological failures in patients who are receiving antiretroviral therapy and who have low-level viremia.