TEXT

Top Abstract Text References

The advent of techniques that reliably quantify levels of human immunodeficiency virus type 1 (HIV-1) RNA in patient plasma has been pivotal in the development of new insights into in vivo viral dynamics and HIV disease pathogenesis (2). Various commercial RNA quantification techniques differing in sensitivity, dynamic range, and variability have now been introduced into clinical practice (4, 5). In this study we compare performance characteristics of an improved nucleic acid sequence-based analysis (NASBA) assay (NucliSens; Organon Teknika, Durham, N.C.) with the reverse transcription-PCR Amplicor Monitor assay (Roche Diagnostic Systems, Branchburg, N.J.).

(The results of this study were presented in part at the 12th World AIDS Conference, Geneva, held from June 28 to July 3, 1998 [1a].)

Plasma samples were obtained from adult and pediatric patients recruited into clinical trials of antiretroviral therapies. Informed consent was obtained from all patients or their legal guardians prior to enrollment in the studies. Plasma was separated from anticoagulated blood within 6 h of collection in EDTA or acid-citrate-dextrose and stored at 70°C until analyzed. All samples were tested on either the first or second thaw. Spiked plasma sample standards, provided by the Virology Quality Assurance Laboratory (VQAL; Rush-Presbyterian Hospital, Chicago, Ill.; sponsored by the Division of AIDS, NIH), containing a known number of HIV RNA copies (9) were used in direct comparisons of the assays and as external standards in all experiments with patient samples.

Both the Monitor and the NucliSens assays were performed following the manufacturers' instructions. Because of the nonnormal distribution of RNA concentrations, continuous variables were analyzed following log transformation, usually by nonparametric statistical methods. HIV RNA concentrations that were undetectable or invalid in one or both of the assays were omitted from statistical analysis. All analyses were performed with Statview 4.5 software (Abacus Concepts, Berkeley, Calif.).

Twofold serially diluted plasma samples containing between 250 and 2,000 HIV RNA copies/ml were tested in four or five replicates of both assays. Eight aliquots with nominal RNA concentrations greater than 500 copies/ml gave detectable results with both assays. Of the 10 plasma samples containing fewer than 500 copies/ml, 7 had detectable RNA results with Monitor and 6 had detectable results with NucliSens.

Serial threefold dilutions were made of 4 HIV-1-seropositive patient plasma specimens containing high concentrations of HIV RNA (4.2 × 10⁶ to 6.7 × 10⁶ copies/ml) and tested in both assays. A total of 36 samples were tested. Three samples containing more than 10⁶ HIV-1 RNA copies/ml gave invalid results with Monitor, one sample with a predicted value of 640 copies/ml was undetectable in both assays, and two samples with predicted concentrations of 1,920 and 320 RNA copies/ml were undetectable by NucliSens but gave values of 1,011 and 212 copies/ml with Monitor. The mean overall difference between the results of serially diluted specimens was closer to the true dilution factor and showed less variability when Monitor was used (Table 1). The linearity of Monitor was better than that of NucliSens in the lower half of the dynamic range (<10^4.4 to 10^4.75 RNA copies/ml), while the linearities of both assays were comparable in the upper range (Table 1). Log-transformed RNA values from the four dilution series were combined to perform linear regression analysis of the NucliSens- and Monitor-derived results against predicted RNA levels. The curve generated with Monitor provided a very close one-to-one linear fit (y = 1.002x  0.058, r = 0.997), whereas NucliSens had a linear relationship (y = 0.921x + 0.262, r = 0.964).

Spiked plasma samples containing known quantities of HIV-1 RNA (0, 1.5 × 10⁴, 1.5 × 10⁵, 7.5 × 10⁵, or 1.5 × 10⁶ copies/ml), provided by the VQAL, were analyzed in replicate runs with Monitor (88 runs with each standard, with 7.5 × 10⁵ copies/ml as the high standard) and NucliSens (69 runs, with 1.5 × 10⁶ copies/ml as the high standard). Both assays showed a high degree of reproducibility. The interassay variability, given by the standard deviation of the log values of replicate assays, was narrower for NucliSens than for Monitor. The reproducibility of NucliSens improved progressively with higher input copy number, while the interassay variability of Monitor remained constant at around 0.20 log₁₀ from 1.5 × 10⁴ to 7.5 × 10⁵ RNA copies/ml (Table 2). HIV-1 RNA concentrations obtained with Monitor were, on average, 0.27 log₁₀ (close to twofold) greater than those generated by NucliSens P < 0.0001) (Table 2). All samples containing zero copies of HIV RNA were negative, giving a specificity of 100% for both assays.

Quantitative HIV-1 RNA analysis was performed with both the Monitor and the NucliSens assays on plasma samples from 51 HIV-seropositive patients being treated with a variety of antiretroviral drugs or receiving no treatment. Log-transformed HIV-1 RNA values that were above the level of detection limit of both assays (n = 41) were compared before and after adjustment with a regression equation, making use of the nominal log₁₀ RNA concentration in external VQAL standards run concurrently with patient samples (1, 9). The concordance between the assays for RNA detectability was 92%, with three samples which were negative by Monitor measuring 630, 1,100, and 1,200 copies/ml (1,181, 643, and 382 copies/ml, respectively, after adjustment) by NucliSens and one sample which was negative by NucliSens measuring 1,192 copies/ml (384 copies/ml after adjustment) with Monitor. In seven samples where unadjusted HIV RNA concentrates were detectable at less than 1,000 copies/ml with at least one assay, all but one had detectable levels by both methods.

In the 41 samples testing positive by both methods, the median unadjusted HIV RNA value with NucliSens was 2.51 × 10⁴ compared with 3.98 × 10⁴ copies/ml with Monitor (P = 0.016 by Wilcoxon signed rank test). This difference persisted, although to a lesser degree, after adjustment using the results of concurrently run standards (median NucliSens HIV RNA concentration of 3.09 × 10⁴ compared with 3.32 × 10⁴ copies/ml by Monitor; P = 0.081). Simple regression analysis (Fig. 1a and b) showed a close linear relationship between the results obtained with the two assays, with little change following adjustment in the correlation coefficient (r = 0.880 and P < 0.001 after, compared with r = 0.872 and P < 0.001 before adjustment), but much closer approximation to equivalence (y = 0.931x + 0.167, compared with y = 0.815x + 0.685, where y is log NucliSens and x is log Monitor HIV RNA concentration). The disparity between NucliSens and Monitor in these patient samples was greater at lower RNA levels before and after adjustment (Fig. 1c and d).

View larger version (20K): [in this window] [in a new window]   FIG. 1.   HIV-1 RNA concentrations in patient samples tested by NucliSens plotted against data obtained with Monitor, before (a) and after (b) adjustment by using VQAL standards. The solid line is the line of unity. (c and d) Difference between log10 Monitor and NucliSens data plotted against the mean of the log10 HIV-1 RNA concentrations measured with the two assays, before (c) and after (d) adjustment.

In this study we compared the performance characteristics of NucliSens isothermal HIV-1 RNA amplification with those of Roche Monitor reverse transcription-PCR, which is already in widespread clinical use. The linearities of Monitor and NucliSens assay were comparable from approximately 500 to 1,000,000 RNA copies/ml, although Monitor had superior linearity and sensitivity in the lower range and performed variably at RNA levels greater than 1,000,000 copies/ml. These results resemble those found in comparisons of Monitor with the previous generation NASBA assay, which has a lower detection limit of 1,000 RNA copies/ml (3, 8). Both assays gave correlation coefficients and linear slopes close to unity when compared with the nominal RNA concentrations present in serially diluted samples. Both assays were close to 100% sensitive in detecting HIV-1 RNA in samples containing 1,000 copies/ml and had comparable sensitivities between 60 and 70% for RNA concentrations between 250 and 1,000 copies/ml.

Interassay variability assessments made with replicate measurements of standard HIV RNA concentrations were similar for both assays. NucliSens had lower interassay variability, especially at higher RNA concentrations (Table 2). Within both assays, linearity and reproducibility fell away at levels close to the lower detection limit. Direct comparison of the two assays made with seropositive patient samples showed that they correlated closely in their ability to detect viral RNA and in actual quantitative RNA measurement. Following adjustment by regression on the measured concentration of concurrently run VQAL standards, there was some improvement in the agreement between the assays with respect to the absolute HIV-1 RNA concentration, though this was not as marked a correction as has been reported in previous comparisons of Monitor and NASBA (1, 9). On the other hand, following adjustment, the linear regression equation relating log₁₀ NucliSens values to log₁₀ Monitor values much more closely resembled the line of equivalence, with an increase in the slope from 0.82 to 0.93 and a reduction in the intercept from 0.69 to 0.17.

In conclusion, the two assays appear to have similar performance characteristics. The linear dynamic range of NucliSens extends above 10⁶ RNA copies/ml and thus may be more suitable for assessing viral load in infants and children, who usually have higher HIV-1 RNA concentrations than adults (6, 7). In contrast, the linearity, and probably the sensitivity, of the Monitor assay is superior in the lower range.