Higher Specificity of Nucleic Acid Sequence-Based Amplification Isothermal Technology than of Real-Time PCR for Quantification of HIV-1 RNA on Dried Blood Spots

Abstract

Dried blood spots (DBS) are widely proposed as a plasma surrogate for monitoring antiretroviral treatment efficacy based on the HIV-1 RNA level (viral load [VL]) in resource-limited settings. Interfering coamplification of cell-associated HIV-1 DNA during reverse transcription (RT)-PCR can be avoided by using nucleic acid sequence-based amplification (NASBA) technology, which is based on an RNA template and isothermic conditions. We analyzed VL values obtained with DBS and plasma samples by comparing isothermic NASBA (NucliSENS EasyQ HIV-1 V2.0; bioMérieux) with real-time RT-PCR (Cobas TaqMan HIV-1 V2.0; Roche). Samples from 197 HIV-1-infected patients were tested (non-B subtypes in 51% of the cases). Nucleic acid extractions were performed by use of NucliSENS EasyMAG (bioMérieux) and Cobas AmpliPrep (Roche) before the NASBA and RT-PCR quantifications, respectively. Both quantification assays have lower limits of detection of 20 (1.3) and 800 (2.9) log₁₀ copies/ml (log) in plasma and DBS, respectively. The mean (DBS minus plasma) differences were −0.39 and −0.46 log, respectively, for RT-PCR and NASBA. RT-PCR on DBS identified virological failure in 122 of 126 patients (sensitivity, 97%) and viral suppression in 58 of 70 patients (specificity, 83%), yielding 12 false-positive results (median, 3.2 log). NASBA on DBS identified virological failure in 85 of 96 patients (sensitivity, 89%) and viral suppression in 95 of 97 patients (specificity, 98%) and yielded 2 false-positive results (3.0 log for both). Both technologies detected HIV-1 RNA in DBS at a threshold of 800 copies/ml. This higher specificity of NASBA technology could avoid overestimation of poor compliance or the emergence of resistance when monitoring antiretroviral efficacy with the DBS method.

INTRODUCTION

The increasing availability of antiretroviral drugs for HIV-infected populations in developing countries calls for more field-friendly methods for HIV-1 viral load monitoring during treatment.

The gold standard for HIV-1 viral load measurement is analysis of plasma samples by means of reverse transcription (RT)-PCR, isothermal nucleic acid-based amplification assay (NASBA), or branched DNA signal amplification assay (bDNA). As viral load assay methods currently require adequate facilities with complex equipment and specialized personnel, shipment of plasma samples to reference laboratories has been proposed as an alternative. However, this requires cool storage and compliance with strict regulations governing the transport of infectious substances. Dried fluid spot specimens have been shown to be a viable alternative to plasma specimens for PCR-based early diagnosis of HIV-1 infection in infants (1–3). Dried blood spot (DBS) specimens are particularly suited to resource-limited settings, because they require minimal training to prepare. Furthermore, DBS are stable at ambient temperature for several weeks (4), are deemed nonhazardous once dried, and can be shipped by regular mail to specialized reference laboratories for HIV-1 RNA quantification (5–9).

The sensitivity of viral load analysis on DBS is an important factor, especially for detecting viral rebound in antiretroviral treatment (ART)-experienced patients. Four main commercial assays for HIV-1 RNA quantification have been evaluated on DBS samples, NASBA (Organon Teknika, Durham, NC), NucliSENS (bioMérieux, Lyon, France), Amplicor HIV-1 monitor (Roche Diagnostics Molecular Systems, Alameda, CA), and Abbott Real-Time HIV-1 (Abbott Diagnostics Des Plaines, Chicago, IL), with detection thresholds between 3.5 and 4 log copies per ml (6, 7, 9). Sensitivity is lower with DBS than with plasma samples by around −0.3 to −0.8 log (10–12).

Recent evaluation of NucliSENS EasyQ HIV-1 V2.0 (based on NASBA technology) on DBS and plasma samples under different storage conditions (temperature, humidity, and duration) showed that DBS samples yielded 100% specificity, linear results over a wide range of viral loads, and a detection limit of 800 copies/ml (13). This complies with recent WHO guidelines defining virological failure as a plasma HIV-1 RNA level of >1,000 copies/ml (14).

Most relevant studies have been performed on samples from patients living in Africa, where HIV-1 viral diversity is very high. However, the accuracy of HIV-1 quantification in DBS versus plasma samples has not been studied according to different HIV-1 subtypes.

In whole blood, HIV nucleic acids are present as RNA in the plasma matrix and as DNA in the cell matrix. Most platforms use RT-PCR to amplify conserved regions of the HIV-1 genome after nucleic acid extraction. RT-PCR may thus coamplify both RNA and DNA, if present, potentially leading to false-positive results for virological failure. Conversely, NASBA technology is based on the use of an RNA template to initiate amplification under isothermic conditions, thus avoiding amplification of any viral DNA. The frequency of false-positive results for virological failure has not been compared between RT-PCR and NASBA.

Here we compared the isothermic NASBA assay (NucliSENS EasyQ HIV-1 V2.0; bioMérieux) with a real-time RT-PCR-based assay (Cobas TaqMan HIV-1 V2.0; Roche) for HIV-1 RNA quantification in a large number of DBS and plasma specimens containing a broad range of viral loads and HIV-1 subtypes.

MATERIALS AND METHODS

Ethics statements.

This was a noninterventional study with no addition to usual procedures. Biological specimens were obtained for only standard viral diagnosis requested by physicians (no specific sampling, no modification of the sampling protocol, no supplementary questions in the national standardized questionnaire). The results were analyzed using an anonymized database. French Public Health Law CSP Art. L 1121-1.1 states that such studies without supplemental blood sampling do not require ethics approval and require only oral informed consent. The CPP Ile de France IV ethics committee waived the requirement for ethics approval (agreement no. IRB 00003835).

Preparation of DBS samples.

We prospectively collected 197 plasma and whole-blood samples from EDTA tubes sent to the laboratory for HIV-1 resistance genotyping, as recommended in antiretroviral-naive patients and in patients with suspected treatment failure (plasma viral load [pVL] above 50 copies/ml). This yielded a wide range of pVL values and allowed us to take HIV-1 strain diversity into account. Before centrifugation, fresh blood was thoroughly homogenized. Aliquots of 50 μl of whole blood from each patient were spotted on filter paper (five spots per card, no. 903 Protein Saver cards; Whatman) then air dried and stored in sealed bags at ambient temperature.

Plasma, collected after 20 min of centrifugation at 3,000 × g, was stored at −80°C in 1.5-ml aliquots for batch testing. One milliliter of each plasma sample was processed according to the instructions of the manufacturers of the NucliSENS EasyMAG and Cobas AmpliPrep assays.

Viral load determination by NASBA.

Each DBS on each card was cut out using sterile scissors. For each analysis, two spots from each patient (total 100 μl) were eluted in 2 ml of NucliSENS lysis buffer and incubated on a roller mixer for 2 h at room temperature. The filter paper was then removed from the tube and the lysate was further processed using NucliSENS EasyMAG according to the manufacturer's instructions; the total lysate volume of 2 ml was used for extraction (bioMérieux). Nucleic acids were isolated using Boom's silica-based technology.

Viral load was also quantified with NucliSENS EasyQ HIV-1 V2.0 (bioMérieux, Lyon, France) according to the manufacturer's instructions. The linear dynamic range of this assay spans 25 (1.4 log₁₀) to 3,000,000 (6.4 log₁₀) HIV-1 RNA copies/ml when 1 ml of plasma is used. The lower limit of quantification is 800 copies/ml when 2 spots of 50 μl of whole blood are used (13). The limit of detection is 100 copies/ml, and qualitative results are obtained below 800 copies/ml (negative, <100 copies/ml; positive, between 100 and 800 copies/ml).

Viral load determination by RT-PCR.

Two spots per patient were eluted as described above for the NASBA method. One milliliter of lysis buffer was extracted using Cobas AmpliPrep, leading to a doubling of real pVL values throughout the analysis.

Total nucleic acids were extracted from plasma and DBS samples by using the Cobas AmpliPrep (Roche) system, which uses a generic silica-based capture method. Viral load was then measured with Cobas TaqMan HIV-1 V2.0 (Roche) according to the manufacturer's instructions. The linear dynamic range of the assay spans 20 (1.3 log₁₀) to 10,000,000 (7 log₁₀) HIV RNA copies/ml when 1 ml of plasma is used. As no information was available for the use of this assay on DBS, we decided to use the same limit of quantification and detection cutoffs of 800 and 100 copies/ml for plasma and DBS, respectively, as for the NucliSENS EasyQ HIV-1 V2.0 assay.

Determination of HIV-1 subtypes.

The HIV-1 subtype was determined by sequencing the Pol region (protease and reverse transcriptase) on a 16-capillary sequencer (ABI PRISM Genetic Analyzer; Applied Biosystems, Les Ulis, France), with respect to the Los Alamos HIV sequence database (HIV BLAST [http://www.hiv.lanl.gov/]). When this sequencing method failed, serotyping was performed as previously described (15) and the results were expressed as subtype B or non-B.

Statistical analysis.

To evaluate correlations between the different technologies, all viral RNA values were recorded as log₁₀-transformed copy numbers of HIV-1 RNA per milliliter of DBS or plasma. Negative samples and samples below the detection threshold were attributed the lower threshold value, while those above 10,000,000 copies were attributed a value of 10,000,000 copies/ml. Results obtained for DBS specimens were not corrected for the hematocrit.

For each method, pVL was expressed as log₁₀ copies/ml, together with the median, interquartile ranges (25 to 75%), and mean ± standard deviation (SD), both for all the plasma and DBS samples and also according to the HIV-1 subtype (B versus non-B). The Mann-Whitney two-sample test was used to compare median VL values according to the viral subtype.

The Spearman rank correlation test and intraclass correlation coefficient were used to measure the correlation between RNA values obtained for DBS and plasma samples with each technology. The concordance between DBS and plasma samples was assessed by using the Bland-Altman approach, in which the differences in VL (equal and above 2.9 log) in the two sample types were plotted against the mean VL of each technology. In DBS and plasma samples, agreement of the results for the detection of virological failure obtained with each technology, using a threshold of 800 copies/ml (2.9 log₁₀), was assessed by using kappa statistics and the Landis-Koch interpretation scale (kappa values of <0.40 indicate poor agreement; >0.40 and <0.75, fair to good agreement; and >0.75, excellent agreement). P values were determined using Fisher's exact test.

All analyses were performed with STATA statistical software (version 12). Differences were considered significant when P values were below 0.05.

RESULTS

HIV-1 quantification according to the viral subtype.

Paired plasma-DBS specimens from 197 patients were collected and tested with both technologies. HIV-1 subtype results were available for 187 (94.4%) patients and were in keeping with the subtype diversity in France (16). Among these 187 patients, 91 (49%) were infected by the B subtype and 96 (51%) by non-B subtypes, as follows: 28 (15%) CRF02_AG, 17 (9%) complex recombinants, 4 (2%) G, 4 (2%) F, 3 (1.6%) A, 3 (1.6%) H, 2 (1%) CRF01_AE, 2 (1%) CRF22_01A1, 1 (0.5%) CRF18_cpx, 1 (0.5%) CRF36_cpx, and 1 (0.5%) CRF45_cpx. The remaining 30 samples (16%) could not be genotyped but were classified as non-B by serotyping. The viral subtypes were grouped together as B or non-B for subsequent analysis.

Table 1 shows the median (interquartile range [IQR; 25 to 75%]) VL values obtained on DBS and plasma samples with the two technologies, both for all samples and according to the B and non-B subtypes. Overall, median VL values in plasma and DBS were 3.2 and 3.4 log copies/ml, respectively, with the RT-PCR method and 2.8 and 2.9 log copies/ml, respectively, with the NASBA method. No differences were found between B and non-B subtypes with either method (P < 0.5 for all comparisons).

TABLE 1.

HIV-1 viral load determined by RT-PCR and NASBA on DBS and plasma for all samples and according to viral subtype

Assay and sample type	HIV-1 viral load (median [IQR]) (log10 copies/ml) of:
	All samples (n = 197)	B subtype (n = 91)	Non-B subtype (n = 96)
PCR
Plasma	3.2 (1.7–4.9)	3.5 (1.8–4.9)	3.3 (2.1–5)
DBS	3.4 (2.9–4.4)	3.4 (2.9–4.4)	3.6 (2.9–4.5)
NASBA
Plasma	2.8 (1.4–4.4)	2.8 (1.4–4.6)	3.4 (1.4–4.2)
DBS	2.9 (2.9–3.9)	2.9 (2.9–4.0)	2.9 (2.9–4.1)

Correlation between DBS and plasma values obtained by RT-PCR.

Among the 105 samples with a plasma VL of ≥2.9 log by RT-PCR, the mean ± SD values for plasma and DBS samples, respectively, were 4.7 ± 0.86 log copies/ml and 4.3 ± 0.80 log copies/ml, and the median (IQR) values were 4.8 log (4.2 log, 5.2 log) and 4.3 log (3.7 log, 4.9 log).

The correlation coefficient for plasma and DBS VL values was r² = 0.76 (Fig. 1a) (Spearman rank correlation test) and the intraclass correlation coefficient (confidence interval [CI]) was 0.77 (0.69 to 0.85). The Bland-Altman plot in Fig. 1b shows that the mean of the difference between DBS and plasma was −0.39 log (limits of agreement, −1.26 and 0.47), regardless of the VL level and viral subtype. A significant difference was observed for 6 samples (5.7%), with 5 (3 non-B and 2 B subtypes) above the upper limit (0.47 log) in DBS compared to those in plasma and 1 non-B subtype above the lower limit (−1.26 log) in plasma compared to those in DBS (Fig. 1b).

FIG 1.

(a) Correlation curve for VL on plasma and DBS using RT-PCR. (b) Bland-Altman representation of VL in plasma and DBS using RT-PCR.

Correlation between DBS and plasma values obtained by NASBA.

Using NASBA technology, among the 93 samples with a pVL of ≥2.9 log, the mean ± SD values for plasma and DBS, respectively, were 4.5 ± 0.81 log₁₀ copies/ml and 4.0 ± 0.79 log copies/ml, and the median (IQR) values were 4.5 log (3.9 log, 5.2 log) and 4.0 log (3.4 log, 4.6 log).

The correlation coefficient for viral loads in plasma versus DBS was 0.65 (Fig. 2a) (Spearman rank correlation test) and the intraclass correlation coefficient (CI) was 0.67 (0.56 to 0.78). The Bland-Altman plot in Fig. 2b shows that the mean of the difference between DBS and plasma was −0.46 log (limits of agreement from −1.46 to 0.55), regardless of the VL level and viral subtype. A significant difference was observed for 7 (7.5%) samples, with two B subtypes above 0.55 log on DBS compared to those in plasma and 5 (3 B and 2 non-B subtypes) above −1.46 log in plasma compared to those in DBS.

FIG 2.

(a) Correlation curve for VL on plasma and DBS using NASBA. (b) Bland-Altman representation of VL in plasma and DBS using NASBA.

Concordance of HIV-1 RNA viral loads measured in DBS and plasma with RT-PCR and NASBA.

The reliability of DBS specimens for detecting virological failure was evaluated by using plasma specimens as the reference for each method, and the recommended threshold (800 copies/ml). As shown in Table 2, there was excellent overall agreement between DBS and plasma specimens for detecting VL above 800 copies/ml, with P values of <0.001 for both methods and kappa statistics of 0.82 (RT-PCR) and 0.86 (NASBA). RT-PCR correctly identified virological failure on DBS specimens from 122 of the 126 patients concerned (sensitivity, 97%; positive predictive value, 91%) and viral suppression on DBS specimens from 58 of 70 patients (specificity, 83%; negative predictive value, 94%). PCR on DBS specimens yielded 12 false-positive results (median 3.2 [range, 2.9 to 3.6] log, with 8 B, 1 non-B, and 3 undetermined subtypes).

TABLE 2.

Concordance between RT-PCR and NASBA on plasma and DBS for identifying virological failure

Assay type and VL (log copies/ml) for DBS specimens	No. of plasma samples with VL of:		Total no.	K value (mean ± SE [95% confidence interval])	P	Performance values (%) of PCR and NASBA assaysa
	<2.9 log copies/ml	>2.9 log copies/ml
RT-PCR				0.82 ± 0.04 (0.74–0.90)	<0.001	Sensitivity, 97; specificity, 83; PPV, 91; NPV, 94
VL < 2.9	58	4	62
VL > 2.9	12	122	134
Total	70	126	196
NASBA				0.86 ± 0.04 (0.78–0.94)	<0.001	Sensitivity, 89; specificity, 98; PPV, 98; NPV, 90
VL < 2.9	95	11	106
VL > 2.9	2	85	87
Total	97	96	193

^aPPV, positive predictive value; NPV, negative predictive value.

Likewise, NASBA correctly identified virological failure on DBS specimens from 85 of the 96 patients concerned and viral suppression in 95 of 97 patients (positive predictive value, 98%; negative predictive value, 90%). NASBA on DBS yielded 2 false-positive results (2 B subtypes, 3.0 log for both).

DISCUSSION

The NASBA and RT-PCR methods each gave similar VL values for DBS and plasma specimens, with high correlation coefficients (>0.80) and few values outside the agreement limits of the Bland-Altman plot (<7.5%). In keeping with previous data obtained in countries with strong HIV diversity (17), we found no significant difference in VL values between B and non-B subtypes (8, 9, 12, 18), although it should be noted that we grouped together all non-B subtypes for analysis, while these subtypes encompass broad genetic diversity.

Overall, VL values obtained with DBS specimens tended to be lower than those obtained with the corresponding plasma samples, with a mean difference of around 0.4 log₁₀ copies/ml. This difference has previously been noted with the NucliSENS, Abbott, and Roche methods (10, 11) and is likely due to differences in specimen volume and in the quality of nucleic acid extraction. VL values obtained with DBS specimens should thus not be compared with VL values previously obtained for plasma samples when monitoring therapeutic efficacy in a given patient.

A lower limit of HIV-1 RNA detection above 5,000 copies/ml has been reported for DBS specimens (10, 11). This was in keeping with 2012 World Health Organization (WHO) recommendations in which persistent VL above 5,000 copies/ml was considered to confirm treatment failure. However, the new WHO guideline presented in June 2013 defines virological failure as a plasma VL above 1,000 copies/ml (14). We therefore analyzed the performance of DBS and plasma samples for detecting virological failure at a threshold of 800 copies/ml, a threshold validated for the NucliSENS EasyQ HIV-1 V2.0 test. Excellent overall agreement was obtained between DBS and plasma specimens for detecting VL above 800 copies/ml. Both methods detected virological failure on DBS specimens with high sensitivity (from 89% to 97%). However, specificity was lower with RT-PCR (83%) than with NASBA (98%), leading to more false-positive results for virological failure. This is in keeping with a previous study of whole-blood samples (18), in which 64 (23%) of 278 “undetectable” plasma specimens (≤500 copies/ml) analyzed with Cobas TaqMan real-time RT PCR (Roche) had positive corresponding DBS samples. PCR coamplifies cell-associated HIV DNA with HIV RNA, whereas the NASBA method with its isothermal amplification design minimizes the contribution of proviral DNA to VL measurement. Our study confirms this higher rate of false positives with RT-PCR (11%) than with NASBA (2%), at a threshold of 800 copies/ml. A modified extraction procedure specific for RNA has recently been proposed to avoid this DNA coamplification. Using the Abbott RealTime HIV-1 assay based on RT-PCR technology, Mbida et al. found a lower rate of false-positive amplification, as all plasma samples that were undetectable were also undetectable on the corresponding DBS (9). As RT-PCR is widely available in developing countries, DNA coamplification could be avoided by specific RNA extraction, the use of DNase, or raising the VL quantification threshold above 3 log. In contrast, coamplification of HIV DNA and RNA by PCR may be useful for diagnosing HIV infection in newborns receiving preventive antiretroviral treatment.

In conclusion, dried blood specimens are a suitable alternative to plasma samples for HIV-1 viral load measurement. Indeed, both PCR and NASBA gave satisfactory results on DBS in comparison with plasma, for both B and non-B subtypes, the latter being most prevalent in resource-limited settings (17). DBS correctly identified virological failure at the new WHO threshold of 1,000 copies/ml, but the NucliSENS EasyQ HIV-1 v2.0 assay showed better specificity than RT-PCR. This NASBA method can thus accurately detect treatment failure at an early stage, allowing rescue intervention to avoid accumulation of drug resistance mutations.