Correction of Underquantification of Human Immunodeficiency Virus Type 1 Load with the Second Version of the Roche Cobas AmpliPrep/Cobas TaqMan Assay
Initial evaluations of the Cobas AmpliPrep/Cobas TaqMan human immunodeficiency virus type 1 (HIV-1) test (CAP/CTM) demonstrated good performance but, afterwards, reports about underquantification were published. We investigated whether the problem was solved with a second version of this assay, the Cobas AmpliPrep/Cobas TaqMan HIV-1 test, version 2.0 (CAP/CTM v2.0). The remaining plasma of 375 consecutive HIV-1 positive samples with a viral load of >or=4,000 copies/ml was collected in three laboratories. The samples were diluted and retested with our routine method Cobas AmpliPrep/Cobas Amplicor HIV-1 monitor test v1.5 in ultrasensitive mode (CAP/CA PHS), as well as with the CAP/CTM and CAP/CTM v2.0 tests. An absolute difference between the results of two methods of >or=0.71 log(10) copies/ml was defined as moderately discrepant, and an absolute difference of >or=0.93 log(10) copies/ml was defined as severely discrepant. In addition, criteria for considering the new methods equivalent to the routine method were formulated. (i) For CAP/CTM compared to CAP/CA PHS, 36 (9.5%) and 20 (5.3%) samples were, respectively, considered moderately and severely underquantified by CAP/CTM. The mean difference between CAP/CTM and CAP/CA PHS was -0.32 log(10) copies/ml. Eight of nineteen of the severely underquantified samples were from patients infected with HIV-1 subtype B strain. (ii) For CAP/CTM v2.0 compared to CAP/CA PHS, no sample was moderately or severely underquantified by CAP/CTM v2.0. A mean difference of 0.08 log(10) copies/ml was found with CAP/CTM v2.0 compared to CAP/CA PHS. The underquantification problem of the CAP/CTM kit was clearly demonstrated. The criteria for the equivalence of CAP/CTM v2.0 to the routine test CAP/CA PHS were fulfilled.
JOURNAL OF CLINICAL MICROBIOLOGY, Apr. 2010, p. 1337–1342 Vol. 48, No. 4
Copyright © 2010, American Society for Microbiology. All Rights Reserved.
Correction of Underquantiﬁcation of Human Immunodeﬁciency Virus
Type 1 Load with the Second Version of the Roche Cobas
AmpliPrep/Cobas TaqMan Assay
A. De Bel,
* D. Marissens,
S. Van den Wijngaert,
and D. Pie´rard
AIDS Reference Laboratory of the Vrije Universiteit Brussel, Subunit Universitair Ziekenhuis Brussel, Laarbeeklaan 101, 1090 Brussels,
; AIDS Reference Laboratory of the Vrije Universiteit Brussel, Subunit Universitair Medisch Centrum Sint Pieter,
Hoogstraat 322, 1000 Brussels, Belgium
; and AIDS Reference Laboratory of the Universite´ Libre de Bruxelles,
Erasme University Hospital, Route de Lennik 808, 1070 Brussels, Belgium
Received 23 June 2009/Returned for modiﬁcation 18 September 2009/Accepted 4 February 2010
Initial evaluations of the Cobas AmpliPrep/Cobas TaqMan human immunodeﬁciency virus type 1 (HIV-1)
test (CAP/CTM) demonstrated good performance but, afterwards, reports about underquantiﬁcation were
published. We investigated whether the problem was solved with a second version of this assay, the Cobas
AmpliPrep/Cobas TaqMan HIV-1 test, version 2.0 (CAP/CTM v2.0). The remaining plasma of 375 consecutive
HIV-1 positive samples with a viral load of >4,000 copies/ml was collected in three laboratories. The samples
were diluted and retested with our routine method Cobas AmpliPrep/Cobas Amplicor HIV-1 monitor test v1.5
in ultrasensitive mode (CAP/CA PHS), as well as with the CAP/CTM and CAP/CTM v2.0 tests. An absolute
difference between the results of two methods of >0.71 log
copies/ml was deﬁned as moderately discrepant,
and an absolute difference of >0.93 log
copies/ml was deﬁned as severely discrepant. In addition, criteria for
considering the new methods equivalent to the routine method were formulated. (i) For CAP/CTM compared
to CAP/CA PHS, 36 (9.5%) and 20 (5.3%) samples were, respectively, considered moderately and severely
underquantiﬁed by CAP/CTM. The mean difference between CAP/CTM and CAP/CA PHS was ⴚ0.32 log
copies/ml. Eight of nineteen of the severely underquantiﬁed samples were from patients infected with HIV-1
subtype B strain. (ii) For CAP/CTM v2.0 compared to CAP/CA PHS, no sample was moderately or severely
underquantiﬁed by CAP/CTM v2.0. A mean difference of 0.08 log
copies/ml was found with CAP/CTM v2.0
compared to CAP/CA PHS. The underquantiﬁcation problem of the CAP/CTM kit was clearly demonstrated.
The criteria for the equivalence of CAP/CTM v2.0 to the routine test CAP/CA PHS were fulﬁlled.
Since the mid-1990, the human immunodeﬁciency virus type
1 (HIV-1) viral load (VL) assay is a major tool in the follow-up
of HIV-1-infected individuals to predict progression of HIV
disease and to monitor antiviral treatment response (5, 7, 14).
Several patented methodologies using a variety of techniques,
from reverse transcriptase PCR to the branched DNA assay,
are commercially available for quantitative HIV-1 RNA testing
in diagnostic laboratories.
VL assays were developed in industrialized countries, where
HIV-1 subtype B predominates. In contrast, subtype B is a
minor variant in developing countries and dissemination of
non-B subtypes has also started in Western countries, espe-
cially in Belgium, where many viruses of African origin are
circulating (17). Among HIV-1-infected patients attending
Universitair Ziekenhuis Brussel (UZB), as many as 59% of the
strains belong to non-B subtypes (6). This high genetic diver-
sity of HIV-1 is a challenge for the quantiﬁcation of plasma
HIV-1 RNA. Indeed, in the past, several studies have reported
the failure of commercial assays for viral load monitoring in
patients infected with non-B subtypes (1, 2, 4, 19). This ﬁnding
led Roche Diagnostics, Ltd. (Rotkreuz, Switzerland), the man-
ufacturer of the widely used Cobas Amplicor HIV-1 monitor
test version 1.0 assay, to modify this assay in 1997. By the
addition of new primers, the assay could cover a broader range
of viral diversity (Cobas Amplicor HIV-1 monitor test version
In the early years of 2000, automation in sample extraction
(for example, the Cobas AmpliPrep of Roche Diagnostics and
M1000 of Abbott Diagnostics, Abbott Park, IL) facilitated
high-volume testing and made viral load testing more robust.
With the ongoing introduction of new technologies, classical
endpoint ampliﬁcation techniques became more and more re-
placed by kinetic real-time ﬂuorescence based tests that com-
bine ampliﬁcation and detection in one step. A TaqMan-based
real-time technique (10) is simple, rapid, sensitive, speciﬁc, and
reproducible and makes further automation possible. More-
over, the risk of contamination is lower due to the closed tube
conﬁguration. The Roche Cobas AmpliPrep/Cobas TaqMan
(CAP/CTM) quantitative HIV-1 assay requires little manual
intervention between the initial addition of the sample to the
assay tube and the generation of the quantitative result. Initial
evaluations of this assay were good (11, 15, 16, 18), but with a
trend to lower viral load values on average than those obtained
with the Cobas Amplicor HIV-1 monitor version 1.5 PHS/
PHM assay (9). Afterward, reports about serious underquan-
* Corresponding author. Mailing address: AIDS Reference Labora-
tory of the Vrije Universiteit Brussel, Subunit UZ Brussel, Laarbeek-
laan 101, 1090 Brussels, Belgium. Phone: 32 2 477 50 00. Fax: 32 2 477
50 15. E-mail: Annelies.DeBel@uzbrussel.be.
Published ahead of print on 17 February 2010.
† The authors have paid a fee to allow immediate free access to
tiﬁcation became available (3, 9, 20). The incidence of under-
quantiﬁcation was low and seemed to be prominent only in
To overcome the issue of underquantiﬁcation, the CAP/
CTM assay was modiﬁed to Cobas AmpliPrep/Cobas TaqMan
HIV-1 test, version 2.0 (CAP/CTM v2.0). A dual-target strat-
egy was chosen: besides gag primers and a FAM-labeled gag
probe, additional ltr primers and a FAM-labeled ltr probe were
included in the assay. The two targets, gag and ltr, are ampliﬁed
with the same efﬁciency. The worst-case scenario is the com-
plete failure of one of the PCR amplicon targets to be ampli-
ﬁed. This would result in the loss of one-half of the ﬂuorescent
signal. The PCR of that specimen would then need one addi-
tional thermal cycle to achieve the threshold cycle (C
quantiﬁcation. Hence, with a C
⫹ 1, the specimen HIV-1
RNA titer would be quantiﬁed by half, equivalent to 0.3-log
difference, from the “true” value. Guidelines state that a dif-
ference of ⬍0.5 log copies/ml (cp/ml) is clinically not signiﬁ-
cant (5), and this variation has no clinical consequence.
The aim of the present study was to evaluate if the under-
quantiﬁcation issue of the CAP/CTM HIV-1 test is solved with
the introduction of the CAP/CTM v2.0 test. A comparison was
performed with the routine Cobas Amplicor HIV-1 monitor
version 1.5, ultrasensitive mode (CAP/CA PHS), the ﬁrst ver-
sion of CAP/CTM and CAP/CTM v2.0.
(Part of this research was presented at the 7th European
HIV Drug Resistance Workshop, Stockholm, Sweden, 25 to 27
March 2009, poster 86.)
MATERIALS AND METHODS
Samples. From May to September 2008, 375 consecutive EDTA anticoagu-
lated plasma samples (collected in Sarstedt EDTA “lavender” tubes) stored at
below ⫺70°C after acquisition with HIV-1 viral load ⱖ4,000 cp/ml in the
CAP/CA PHS (or with manual RNA extraction at Universite´ Libre de Bruxelles
[ULB]) routine test were selected. Apart from this condition and the availability
of a sufﬁcient sample volume, no selection for the inclusion of samples was done.
At UZB, 5 of 83 (6.0%) samples were not included due to insufﬁcient sample
volume. Only one sample per patient was included.
After routine testing of the original specimen the remaining EDTA anticoag-
ulated plasma was diluted 1:5 or 1:10 in HIV-1 RNA negative EDTA plasma
pool (Roche Diagnostics GmbH, Germany) and divided into aliquots. Aliquots
were stored frozen below ⫺70°C until tested. All aliquots were subjected to equal
numbers of freeze-thaw cycles.
Samples were selected and divided into aliquots at 3 laboratories in Brussels:
the AIDS Reference Laboratory of the VUB, subunit UZ Brussel (UZB; n ⫽ 78)
and subunit University Medical Center Saint-Pierre (STP, n ⫽ 149) and the
AIDS Reference laboratory of the ULB, Erasme University Hospital (EHB, n ⫽
Viral load assays. All samples were analyzed in the AIDS reference laboratory
of the VUB. The subunit UZ Brussel analyzed the samples collected at EHB and
UZB, while the subunit STP tested their own samples.
Each specimen was tested in singlet in the three study tests. The assays were
performed in accordance to the package inserts of the manufacturer. For quan-
tiﬁcation of the target nucleic acid and control of PCR inhibition an internal
quantiﬁcation standard was used. The results were expressed in copy number of
HIV-1 RNA copies per milliliter. The prevention of back-contamination was
ensured by the use of uracil-N-glycosylase, an enzyme that inactivates dUTP-
(i) Cobas AmpliPrep/Cobas Amplicor HIV-1 monitor test, version 1.5, ultra-
sensitive mode (CAP/CA PHS; Roche Diagnostics, Ltd., Rotkreuz, Switzerland).
After probe-speciﬁc extraction with the ultrasensitive protocol on the Cobas
AmpliPrep (CAP), ampliﬁcation was done with the Cobas Amplicor HIV-1
Monitor Test, v1.5, and endpoint detection was performed using the Cobas
Amplicor (CA) instrument. The primers target the highly conserved gag region.
If the result was described above the upper quantiﬁcation limit (⬎100,000 cp/ml),
samples were diluted 1:10 or 1:100 with HIV-1 RNA negative human EDTA
plasma and retested to obtain a measurable concentration of HIV-1 RNA.
Several lots of reagents were used to perform the assays.
(ii) Cobas AmpliPrep/Cobas TaqMan HIV-1 test (CAP/CTM; Roche Diagnos-
tics Ltd.). After generic extraction with the CAP, ampliﬁcation was performed
using the Cobas TaqMan HIV-1 Test on the Cobas TaqMan 48 (CTM) instru-
ment. A dually labeled hybridization probe targeting the gag region was used in
the assay. Multiple kit numbers were used to perform the assays.
(iii) Cobas AmpliPrep/Cobas TaqMan HIV-1 test, version 2.0 (CAP/CTM
v2.0; Roche Diagnostics Ltd.). This assay simultaneously targets the gag and the
ltr region with two dually labeled hybridization probes. For the present study, a
prerelease version, CAP/CTM version 1.5 provided by Roche Diagnostics was
used. Afterward, based on extended validation, a software modiﬁcation was
performed to extend the lower quantiﬁcation limit from 40 to 20 cp/ml. There-
fore, this test version was later renamed as version 2.0, but the reagents of both
assay versions are the same. This change did not inﬂuence our results since no
samples with low viral load were included in this evaluation. CAP/CTM v2.0 is
not yet available in the United States.
After generic extraction with the CAP, ampliﬁcation was performed using the
Cobas TaqMan HIV-1 Test, v2.0 on the CTM instrument. All assays were
performed using one lot of reagents.
Controls. Each run consisted of 20 samples and one high positive, one low
positive and one negative control included in the kit. In addition, in order to have
an independent control, one in-house produced internal run control was ana-
lyzed in each run.
Criteria for equivalency of two VL methods and discrepancy deﬁnitions. The
minimal change in viral load considered as clinically signiﬁcant is a 0.5 log
cp/ml (5). Thus, the maximum total error (TE
), deﬁned as systematic error
(SE) ⫹ 1.96 ⫻ the random error (RE), must be lower than 0.5 log
Because in Belgium HIV patients are monitored at one AIDS reference center
and we know from our validation ﬁle that the bias is very small, we can assume
that the SE is negligible. Thus, the maximum allowable RE
is 0.26 log
cp/ml. Because we allow a TE of 0.5 log
cp/ml for both methods in a com
parison the total TE
) or 0.71 log
cp/ml. Thus, the
absolute difference between CAP/CA PHS and CAP/CTM or CAP/CTM v2.0
should be ⬍0.71 log
cp/ml for at least 95% of all samples. Samples that deviate
by more than 0.71 log
cp/ml are referred to as moderately discrepant.
According to the analogous principles described above we can state that for at
least 99% of all samples the absolute difference between CAP/CA PHS and
CAP/CTM or CAP/CTM v2.0 should be less than 2.58
) or 0.93
cp/ml. Samples that deviate by more than 0.93 log
cp/ml are referred to
as severely discrepant.
Graphs and statistics. Absolute bias plots were used to represent the degree
of agreement between the assays. The x axis of the absolute bias plot shows the
mean of the results, and the y axis represents the absolute difference between the
values obtained by the two platforms. The dotted lines represent the above-
deﬁned criteria for considering the new assay equivalent to the routine assay.
To obtain a normal distribution of the values, results were transferred to log
values. Normality was checked by using a modiﬁed Shapiro-Wilk test. Mean
values were compared to the t test for paired data. Epidemiological data were
compared to the chi-square test. Statistics were performed with Analyze-It for
Microsoft Excel (version 2.12; Analyze-It Software, Ltd., United Kingdom).
Sequencing and subtyping. Samples with severe discrepancies (n ⫽ 20) were
sequenced by Roche Molecular Diagnostics (Pleasanton, CA) with unpublished
primers to identify possible mismatches in all primer and probe regions.
Nucleic acids were extracted from 200 l of the original plasma (n ⫽ 12) or,
when the original plasma was no longer available, from 200 l of the 1:5-diluted
(n ⫽ 8) plasma using sample preparation reagents from the Cobas HIV-1
Monitor test. A 600-l portion of lysis reagent was mixed with a 200-l sample,
followed by incubation for 10 min at room temperature. Nucleic acids were
precipitated by the addition of 800 l of isopropanol and then pelleted by
centrifugation for 15 min in a microcentrifuge. The pellets were washed with 1 ml
of 70% ethanol (EtOH) and centrifuged for 5 min. The alcohol was removed, the
tubes were pulse spun for 30 s, and the residual EtOH was removed. The pellets
were resuspended in 100 l of sample diluent from the Cobas HIV-1 Monitor
test and stored at ⫺80°C until used.
HIV-1 sequences ﬂanking the CAP/CTM HIV-1 version 2 target regions were
ampliﬁed by reverse transcription-PCR. Murine leukemia virus reverse tran-
scriptase (Applied Biosystems, Foster City, CA) was used for reverse transcrip-
tion, and PCR was carried out with AmpliTaq DNA polymerase (Applied Bio-
systems). Heminested ampliﬁcations were then performed to increase the yield
of amplicon. Ampliﬁcation products were analyzed by agarose gel electrophore-
sis. The amplicons from reactions with one predominant product band were
puriﬁed for sequencing by using either QiaQuick PCR puriﬁcation (Qiagen,
1338 DE BEL ET AL. J. CLIN.MICROBIOL.
North Rhine-Westphalia, Germany) or ExoSAP-IT kit reagents (USB, Cleve-
land, OH). If multiple major amplicon populations were present, the products of
appropriate size, separated by agarose gel electrophoresis, were excised from the
gel and puriﬁed using reagents from a QiaQuick gel extraction kit (Qiagen).
Sequencing was either performed on the ABI 3730xl analyzer using BigDye
chemistry and Collection Software v3.0 (Applied Biosystems) or by Sequetech
Corp. (Mountain View, CA).
To determine the subtype of the virus in each sample, the gag sequence from
each sample was aligned to sequences from HIV-1 isolates of known genotypes
from the GenBank sequence database (http://www.ncbi.nlm.nih.gov/GenBank
/index.html) using the CLUSTAL V sequence alignment program, which also
generates a phylogenetic tree, via the Lasergene software package (MegAlign
version 5.07; DNASTAR, Inc., Madison, WI).
Demographic information. For each new HIV diagnosis in Belgium the at-
tending physician is asked to ﬁll in a questionnaire and send it to the AIDS
Reference Laboratory that made the diagnosis. All coded data are processed
once a year by the Institute of Public Health to map and characterize the Belgian
HIV epidemic. These data, although not complete, were used to compare the
populations consulting the three different centers and to track the probable
origin of the HIV-1 strains yielding severely discrepant results.
Study ethics. The present study was approved by the biomedical ethical com-
mission of the Universitair Ziekenhuis Brussel and the Medical Faculty of the
Vrije Universiteit Brussel (reference number BUN B14320084122).
Because the initial result was outside the dynamic range of
the CAP/CA PHS assay, 22 samples had to be diluted 1:10 for
the CAP/CA PHS analysis and 1 sample was diluted 1:100 for
additional analysis. No dilutions had to be made for CAP/CTM
and CAP/CTM v2.0 due to the broader dynamic range of these
One sample (EHB_141) demonstrated an extremely high
quantitative result with both CAP/CTM tests (⫹1.29 and
cp/ml) compared to CAP/CA PHS. This result was
considered as an outlier and was not taken into account in our
calculations. The sequence of this sample contains six mis-
matches to the upstream CA PHS primer. The subtype of this
sample was determined as CRF18_cpx.
For one sample (EHB_084) the result for CAP/CTM was
⬍40 cp/ml, although HIV-1 RNA was detected with CAP/CA
PHS (3.70 log
cp/ml) and CAP/CTM v2.0 (3.75 log
For the calculations we made an approximation and used 40
cp/ml, the best-case scenario for EHB_084.
CAP/CTM compared to CAP/CA PHS. The results are rep-
resented in an absolute bias plot (Fig. 1).
The overall mean difference between CAP/CTM and
CAP/CA PHS was ⫺0.32 log
cp/ml (P ⬍ 0.05). The median
difference was ⫺0.29 log
cp/ml with a range between ⫺2.64
and 0.78 log
A total of 36 samples (9.6%; EHB [n ⫽ 16], STP [n ⫽ 15],
and UZB [n ⫽ 5]) were moderately and 20 samples (5.3%;
EHB [n ⫽ 10], STP [n ⫽ 7], and UZB [n ⫽ 3]) were severely
underquantiﬁed with CAP/CTM compared to CAP/CA PHS.
The observed differences between the different sites are not
statistically signiﬁcant (P ⫽ 0.55 and 0.59, respectively).
The sequence results and the probable origin of the severely
underquantiﬁed strains are represented in Table 1. For one
sample (EHB_039) the ampliﬁcation of the gag region repeat-
edly failed. The other 19 samples were from patients infected
with 9 different subtypes. Eight of these nineteen (42.1%)
patients were infected with a subtype B strain. In 18 of the
severely underquantiﬁed samples, there are mismatches to
CAP/CTM gag primers or probe that may affect assay perfor-
mance. No mismatches were found in one sample (UZB_001),
and thus the observed difference of ⫺1.06 log
cp/ml is prob
ably due to an anomalous error.
The ampliﬁcation of the ltr region that was done for further
sequence comparisons succeeded in 16 of these 20 samples.
There were no mismatches found to the CAP/CTM ltr region
that would be likely to affect CAP/CTM v2.0 assay perfor-
mance. Two (0.5%; EHB [n ⫽ 1], STP [n ⫽ 1], and UZB [n ⫽
0]) samples were moderately higher quantiﬁed with CAP/
CTM. These two samples were also higher quantiﬁed with
CAP/CTM v2.0. As for sample EHB_141, the sample deﬁned
as an outlier, two (EHB_090) or three (SPB_083) mismatches
to the upstream CA PHS primer were detected.
CAP/CTM v2.0 compared to CAP/CA PHS. The results are
represented in an absolute bias plot (Fig. 2). The titer values
FIG. 1. Modiﬁed Bland and Altman plot CAP/CTM versus CAP/CA PHS v1.5. The dotted lines represent the deﬁned acceptance criteria.
OL. 48, 2010 CAP/CTM AND CAP/CTM v2.0 ASSAYS FOR HIV-1 VL 1339
obtained with CAP/CTM v2.0 were on average ⫹0.08 log
cp/ml higher compared to CAP/CA PHS (P ⬍ 0.05). The
median difference was ⫹0.05 log
cp/ml, with a range from
⫺0.68 to 1.02 log
cp/ml. No sample was moderately or se
verely underquantiﬁed. Seven (1.9%; EHB [n ⫽ 3], STP [n ⫽
3], and UZB [n ⫽ 1]) samples were moderately and three
(0.8%; EHB [n ⫽ 1], STP [n ⫽ 1], and UZB [n ⫽ 1]) were
severely higher quantiﬁed by CAP/CTM v2.0. Five out of seven
moderately higher quantiﬁed samples were sequenced. Three
of them had ﬁve mismatches to the CA PHS upstream primer.
For two of them no mismatches were found.
Demographic comparison of the patient population attend-
ing the centers. Three parameters (nationality, risk category,
and probable country of infection) were compared by using a
chi-square test. A statistically signiﬁcant difference (P ⬍ 0.05)
was observed for all three parameters.
The patient population consulting at UZB contains a higher
fraction of people with the Belgian nationality (44.9% versus
24.2% at STP and 25.2% at EHB). The associated risk cate-
gory is more often men having sex with men (MSM; 44.9%
versus 22.8% at STP and 20.9% at EHB), and the probable
country of infection is more often Belgium (39.4% versus
16.8% at STP and 10.8% at EHB).
The patients seeking medical advice at STP are more often
European (not Belgian), North African, American, and Asian
(11.4% versus 5.1% at UZB and 4.0% at EHB). Moreover, the
risk categories mother-to-child transmission (MTCT) and in-
travenous drug use (IVD) are more prevalent at the STP
patient population (7.4% versus 0.0% at UZB and 2.0% at
Both American (5) and European (7) guidelines for the
treatment of HIV-1-infected adults make use of two markers—
the viral load and the CD4
T-cell count—to assess the level of
HIV-1 viremia and the immune function of infected patients.
These tests are used as predictors for deciding when to begin
antiviral therapy and to assess virologic and immunologic ef-
ﬁcacy of treatment. Therefore, accurate measurement of
HIV-1 viral load is essential to provide clinicians with valuable
information to determine treatment decisions. Although the
new quantitative HIV-1 assays are designed to cope with in-
creasing molecular diversity of the virus, there were several
reports about serious underquantiﬁcation issues with the ﬁrst
version of the CAP/CTM test (3, 9, 20). To overcome this
problem, additional primers and a probe, located in the highly
TABLE 1. Subtype and number of mismatches to gag primers and
probe of the severely underquantiﬁed samples
(n ⫽ 20) in CAP/CTM
Sample Origin Subtype
gag region (no. of mismatches)
EHB_003 No data F1 or CRF12_BF 2 5
EHB_027 No data B 4
EHB_028 Belgium B 1
AF AF AF
EHB_057 Belgium B 1 3
EHB_084 Belgium B 1
EHB_149 No data B 3
SPB_027 Netherlands CRF02_AG 1 1
SPB_046 Belgium B 1
A1 1 3
SPB_133 Belgium B 2
F1 or CRF12_BF 1
UZB_039 Belgium B 4
AF, ampliﬁcation failed.
FIG. 2. Modiﬁed Bland and Altman plot CAP/CTM v2.0 versus CAP/CA PHS v1.5. The dotted lines represent the deﬁned acceptance cri-
1340 DE BEL ET AL. J. CLIN.MICROBIOL.
conserved ltr region of HIV-1, were included in the second
version of the kit in addition to the gag primers and probe. We
report here the results of a three-site multicenter evaluation
study of the CAP/CTM and CAP/CTM v2.0 tests in compari-
son to the routine CAP/CA PHS test.
In Belgium, 58% of newly diagnosed HIV patients were of
foreign origin in 2007 (17). Among HIV-1-infected patients
attending UZB, subtype B virus is only present in 41% of the
patients (6). Therefore, the sample collection used was ex-
pected to contain a wide variety of subtypes, including a lot of
non-B subtypes, and consequently could represent a great chal-
lenge for the new generation of viral load tests.
The patient population attending the three centers is signif-
icantly different. The most prominent differences were that the
patient population consulting at UZB contains a higher frac-
tion of MSM with the Belgian nationality and the probable
country of infection is more often Belgium. The patients seek-
ing medical care at STP are more often from European (non-
Belgian), North African, American, and Asian nationalities.
Moreover, the risk categories MTCT and IVD are more prev-
alent at the STP patient population.
The results of the comparison of CAP/CTM and CAP/CA
PHS are represented in an absolute bias plot (Fig. 1). A total
of 36 of 375 (9.6%) samples were moderately and 20 of 375
(5.3%) samples were severely underquantiﬁed. Although not
statistically signiﬁcant (P ⫽ 0.59), the prevalence of severe
underquantiﬁcation was highest at EHB: 6.7% of samples
compared to 4.6% at STP and 3.8% at UZB. No particular
subtype is affected by the underquantiﬁcation problem. The 19
subtyped, severely discrepant samples were from patients in-
fected with nine different subtypes. Eight of these nineteen
(42.1%) patients were infected with a subtype B strain. Thus,
in contrast to earlier-generation HIV-1 VL assay problems (1,
2, 4, 19), even subtype B strains could be affected by the
underquantiﬁcation issue of the ﬁrst version of the CAP/CTM
test, indicating that the mismatches were not associated with
the HIV-1 subtype.
As illustrated in Table 1, both the primer and the probe
binding region polymorphisms were identiﬁed as the root
cause in the rare cases of signiﬁcant underquantiﬁcation. Due
to patented primers and probes, Roche did not wish to disclose
the exact location of the mismatches. Therefore, we cannot
support or refute the theory of Korn et al. (12) that mutations
at nucleotide position 1488 of the HXB2 reference sequence,
the ⫺3 position of the suggested downstream primer, are an
important cause of the underestimation of HIV-1 RNA levels.
Nevertheless, there should be other critical mutations because
8 of the 19 sequenced underquantiﬁed samples had no mis-
matches to the downstream primer.
We found that 2 of 375 (0.5%) samples were moderately
higher quantiﬁed by CAP/CTM. These two samples are also
moderately higher quantiﬁed (⫹0.79 and ⫹0.91 log
respectively) with CAP/CTM v2.0. Although not well docu-
mented since we only had access to the number of mismatches
and not to the sequences as such, this ﬁnding is probably due
to a better primer-probe match than with the CAP/CA PHS
assay, the method used as the “reference” or comparison
method. This is also the case with sample EHB_141, which we
considered an outlier.
The overall mean difference between CAP/CTM and
CAP/CA PHS was ⫺0.32 log
cp/ml (P ⬍ 0.05). This differ
ence is statistically signiﬁcant. Although this difference is ⬍0.5
cp/ml, which is generally accepted as clinically relevant
(5), it clearly illustrates the underquantiﬁcation issue of the
ﬁrst version of CAP/CTM. Moreover, the high number of
moderately and severely underquantiﬁed samples is clinically
problematic. The criteria used above to consider the new
method equivalent to the routine method are not fulﬁlled for
The results of the comparison of CAP/CTM v2.0 and
CAP/CA PHS are represented in an absolute bias plot (Fig. 2).
No sample was moderately or severely underquantiﬁed. Eight
(2.1%) samples were moderately and three (0.8%) of 375 sam-
ples were severely higher quantiﬁed with CAP/CTM v2.0. In
all, four samples (three ⫹ EHB_141) had several mismatches
to the upstream CA PHS primer and were probably under-
quantiﬁed with the older CAP/CA method. Thus, it appears
that the high genetic diversity of HIV-1 also affects the per-
formance of the “reference” or comparison method.
A mean difference of ⫹0.08 log
cp/ml was found with the
CAP/CTM v2.0 compared to CAP/CA PHS (P ⬍ 0.05). This
difference is statistically signiﬁcant but not clinically relevant
cp/ml). The above-described criteria to consider
the new method equivalent to the routine method are fulﬁlled
for CAP/CTM v2.0.
In addition, one sample (EHB_141) quantiﬁed extremely
high with both Cobas TaqMan tests (⫹1.29 and ⫹1.84 log
cp/ml; CAP/CTM and CAP/CTM v2.0, respectively) compared
to the routine CAP/CA PHS test and was considered to be an
outlier. The patient’s CD4 count and clinical presentation cor-
related better with the higher VL (this patient stopped taking
all antiretroviral therapy at sampling time). The sequence of
the HIV-1 strain from this sample revealed six mismatches to
the upstream CA PHS primer. Therefore, we consider that it
was underquantiﬁed by the CAP/CA PHS test.
The present study has three limitations. (i) Lot-to-lot vari-
ability for the CAP/CTM v2.0 kit was not evaluated during our
study. Only one lot of reagents was used during the evaluation.
(ii) A bias was possibly generated by selecting only samples
withaVLofⱖ4,000 copies/ml, which were diluted in order to
obtain enough sample volume to run the three tests in parallel.
Indeed, most of the included samples were from untreated
patients, harboring viral strains without mutation pressure
from antiretroviral drugs. However, gag and ltr are not current
drug targets and, although gag cleavage site mutations have
been described to be involved in protease inhibitor resistance
(21), all known gag mutations associated with drug resistance
lie outside the gag PCR amplicon target region used by the
CAP/CTM assays (S. Rose, Roche Molecular Diagnostics, per-
sonal communication). Hence, mutations associated with drug
resistance should not affect the test performance. Moreover,
dilution with HIV-1 RNA negative EDTA plasma might have
inﬂuenced the efﬁciency of the extraction and ampliﬁcation.
Nevertheless, even with a possibly biased sample pool, we were
able to clearly demonstrate the underquantiﬁcation problem
with CAP/CTM. (iii) The present study did not evaluate the
performance of the tests in the low quantiﬁcation range. In
order to ﬁnd signiﬁcant differences in quantiﬁcation, speci-
mens with higher nominal viral loads were used. However, as
seen with the ﬁrst version of this kit (16, 18), the second version
VOL. 48, 2010 CAP/CTM AND CAP/CTM v2.0 ASSAYS FOR HIV-1 VL 1341
will probably also show an increased sensitivity at low viral
loads in comparison with the CAP/CA PHS assay, as this fact
is linked to the real-time technology. The important practical
impact of this issue was recently discussed (8, 13).
In conclusion, the criteria needed to consider the new
method equivalent to the routine method are fulﬁlled only for
CAP/CTM v2.0. We clearly demonstrated the underquantiﬁ-
cation issue with the ﬁrst version of the CAP/CTM test, not
only in 13 samples of patients infected with a non-B subtype
but also in 7 samples from patients infected with subtype B.
Although the samples tested are all of Belgian origin (or at
least collected in Belgium), the CAP/CTM test is not adequate
for HIV-1 viral load testing on a general, worldwide basis.
With the CAP/CTM v2.0 test no underquantiﬁed sample was
identiﬁed compared to the routine test CAP/CA PHS. Thus,
only the second version of this kit can be used for routine
HIV-1 viral load testing in a routine clinical laboratory.
In spite of the good correlation between CAP/CA PHS and
CAP/CTM v2.0, clinical biologists and clinicians should remain
aware of the high degree of genetic variability of HIV-1 and
the difﬁculties that this entails in the design of appropriate
primer and probe sets for a real-time PCR that should be able
to quantify all known HIV-1 variants. When viral load status
does not develop as expected from clinical picture and/or
T-cell counts, other plasma HIV-1 RNA assays should
be used in order to detect possible variation of the virus,
whatever its subtype or origin.
We thank Roche Diagnostics, Ltd., Rotkreuz, Switzerland for sup-
plying the reagents used in this study, for sequencing the severely
discrepant samples and for lending us the Cobas TaqMan 48 analyzer.
We gratefully acknowledge A. Wyns, C. Vanneste, M.-H. Jurion, K.
Miller, and N. Gijbels for their excellent technical assistance. We thank
A. Sasse from the Belgian Institute of Public Health for his support.
1. Alaeus, A., K. Lidman, A. Sonnerborg, and J. Albert. 1997. Subtype-speciﬁc
problems with quantiﬁcation of plasma HIV-1 RNA. AIDS 11:859–865.
2. Damond, F., C. Apetrei, D. Descamps, F. Brun-Vezinet, and F. Simon. 1999.
HIV-1 subtypes and plasma RNA quantiﬁcation. AIDS 13:286–288.
3. Damond, F., B. Roquebert, A. Benard, G. Collin, M. Miceli, P. Yeni, F.
Brun-Vezinet, and D. Descamps. 2007. Human immunodeﬁciency virus type
1 (HIV-1) plasma load discrepancies between the Roche Cobas Amplicor
HIV-1 Monitor Version 1.5 and the Roche Cobas AmpliPrep/Cobas
TaqMan HIV-1 assays. J. Clin. Microbiol. 45:3436–3438.
4. Debyser, Z., E. Van Wijngaerden, K. Van Laethem, K. Beuselinck, M. Rey-
nders, E. De Clercq, J. Desmyter, and A. M. Vandamme. 1998. Failure to
quantify viral load with two of the three commercial methods in a pregnant
woman harboring an HIV type 1 subtype G strain. AIDS Res. Hum. Ret-
5. Department of Health and Human Services. Guidelines for the use of anti-
retroviral agents in HIV-1-infected adults and adolescents. U.S. Department
of Health and Human Services, Washington, DC. http://www.aidsinfo.nih
6. Echahidi, F., A. De Bel, O. Soetens, S. Lauwers, and D. Pie´rard. 2007.
Subtype distribution among HIV-1-infected patients attending UZ-Brussel.
Poster, 13th International Bioinformatics Workshop on Virus Evolution and
Molecular Epidemiology, Lisbon, Portugal, 9 to 14 September 2007.
7. European AIDS Clinical Society. 2009. Guidelines for the clinical manage-
ment and treatment of HIV-infected adults in Europe. European AIDS
Clinical Society, Paris, France. http://www.europeanaidsclinicalsociety.org
8. Gatanaga, H., K. Tsukada, H. Honda, J. Tanuma, H. Yazaki, T. Watanabe,
M. Honda, K. Teruya, Y. Kikuchi, and S. Oka. 2009. Detection of HIV type
1 load by the Roche Cobas TaqMan assay in patients with viral loads
previously undetectable by the Roche Cobas Amplicor Monitor. Clin. Infect.
9. Gueudin, M., J. C. Plantier, V. Lemee, M. P. Schmitt, L. Chartier, T.
Bourlet, A. Ruffault, F. Damond, M. Vray, and F. Simon. 2007. Evaluation of
the Roche Cobas TaqMan and Abbott RealTime extraction-quantiﬁcation
systems for HIV-1 subtypes. J. Acquir. Immune Deﬁc. Syndr. 44:500–505.
10. Holland, P. M., R. D. Abramson, R. Watson, and D. H. Gelfand. 1991.
Detection of speciﬁc polymerase chain reaction product by utilizing the 5⬘-3⬘
exonuclease activity of Thermus aquaticus DNA polymerase. Proc. Natl.
Acad. Sci. U. S. A. 88:7276–7280.
11. Katsoulidou, A., M. Petrodaskalaki, V. Sypsa, E. Papachristou, C. G. Anas-
tassopoulou, P. Gargalianos, A. Karafoulidou, M. Lazanas, T. Kordossis, A.
Andoniadou, and A. Hatzakis. 2006. Evaluation of the clinical sensitivity for
the quantiﬁcation of human immunodeﬁciency virus type 1 RNA in plasma:
comparison of the new COBAS TaqMan HIV-1 with three current HIV-
RNA assays–LCx HIV RNA quantitative, Versant HIV-1 RNA 3.0 (bDNA)
and COBAS Amplicor HIV-1 Monitor v1.5. J. Virol. Methods 131:168–174.
12. Korn, K., B. Weissbrich, C. Henke-Gendo, A. Heim, C. M. Jauer, N. Taylor,
and J. Eberle. 2009. Single-point mutations causing more than 100-fold
underestimation of human immunodeﬁciency virus type 1 (HIV-1) load with
the Cobas TaqMan HIV-1 real-time PCR assay. J. Clin. Microbiol. 47:1238–
13. Manavi, K. 2008. The signiﬁcance of low-level plasma HIV viral load on
COBAS TaqMan HIV-1 assays for patients with undetectable plasma viral
load on COBAS Amplicor monitor version 1.5. HIV Clin. Trials 9:283–286.
14. Mellors, J. W., A. Munoz, J. V. Giorgi, J. B. Margolick, C. J. Tassoni, P.
Gupta, L. A. Kingsley, J. A. Todd, A. J. Saah, R. Detels, J. P. Phair, and C. R.
Rinaldo, Jr. 1997. Plasma viral load and CD4
lymphocytes as prognostic
markers of HIV-1 infection. Ann. Intern. Med. 126:946–954.
15. Oliver, A. R., S. F. Pereira, and D. A. Clark. 2007. Comparative evaluation
of the automated Roche TaqMan real-time quantitative human immunode-
ﬁciency virus type 1 RNA PCR assay and the Roche Amplicor Version 1.5
conventional PCR assay. J. Clin. Microbiol. 45:3616–3619.
16. Perrin, L., J. M. Pawlotsky, M. Bouvier-Alias, C. Sarrazin, S. Zeuzem, and
G. Colucci. 2006. Multicenter performance evaluation of a new TaqMan
PCR assay for monitoring human immunodeﬁciency virus RNA load. J. Clin.
17. Sasse, A., A. Defraye, J. Buziarsist, D. Van Beckhoven, and S. Wanyama.
2008. Epidemiologie van AIDS en HIV-infectie in België, Toestand op 31
December 2007.Wetenschappelijk Instituut voor Volksgezondheid, Afdeling
Epidemiologie. http://www.iph.fgov.be/pdf/jaarrapport_aids_2007.pdf. Ac-
cessed 20 May 2009.
18. Schumacher, W., E. Frick, M. Kauselmann, V. Maier-Hoyle, and R. Babiel.
2007. Fully automated quantiﬁcation of human immunodeﬁciency virus
(HIV) type 1 RNA in human plasma by the COBAS AmpliPrep/COBAS
TaqMan system. J. Clin. Virol. 38:304–312.
19. Simons, P., G. Muyldermans, P. Lacor, G. Zissis, and S. Lauwers. 1997.
False-negative HIV viral load in AIDS patients. AIDS 11:1783–1784.
20. Taylor, N., M. Jauer, K. Bijuklic, A. Egle, R. Greil, W. Patsch, M. Ober-
meier, and J. Eberle. 2008. Elite controllers or misquantiﬁcation of virus load
by Cobas TaqMan? Poster, 9th International Congress on Drug Therapy in
HIV Infection, Glasgow, United Kingdom, 9 to 13 November 2008.
21. Verheyen, J., E. Litau, T. Sing, M. Daumer, M. Balduin, M. Oette, G.
Fatkenheuer, J. K. Rockstroh, U. Schuldenzucker, D. Hoffmann, H. Pﬁster,
and R. Kaiser. 2006. Compensatory mutations at the HIV cleavage sites
p7/p1 and p1/p6-gag in therapy-naive and therapy-experienced patients. An-
tivir. Ther. 11:879–887.
1342 DE BEL ET AL. J. CLIN.MICROBIOL.