An undetectable polymerase chain reaction signal in routine HIV plasma viral load monitoring is associated with better virological outcomes in patients receiving highly active antiretroviral therapy


  • This work was presented in part at the 19th Conference on Retroviruses and Opportunistic Infections, Seattle, WA, 5–8 March 2012 (poster L-134).

Correspondence: Dr Bruno Hoen, Service de Maladies Infectieuses et Tropicales, CHU de Besançon, Hôpital Saint-Jacques, 25030 Besançon Cedex, France. Tel: 33 381 218 533; fax: 33 381 218 772; e-mail:



The aim of the study was to assess whether patients with undetectable viraemia [a negative polymerase chain reaction result (PCRneg)] and those with plasma viral load (PVL) < 40 HIV-1 RNA copies/mL but a detectable (positive) PCR signal (PCRpos) had different outcomes in terms of the development of blips and virological failure (VF).


A multicentre observational database analysis was carried out. Data for patients whose highly active antiretroviral therapy (HAART) regime had been unchanged for ≥ 6 months by 1 January 2008, whose first two PVL measurements of 2008 were < 40 copies/mL and who had at least five PVL measurements between 1 January 2008 and 31 December 2010 were extracted from a multicentre observational database of 4928 patients receiving HAART. PVL assays used during this period had a detection threshold of 20 or 40 copies/mL. Undetectable PVL at baseline (BLPCRneg) was defined as PCRneg at the first two PVL determinations of 2008. Multivariable Cox regression analysis was performed to investigate factors associated with the occurrence of blips and VF, defined as two consecutive PVL measurements > 40 copies/mL.


Of the 1957 patients included in the study (mean age 47 years; median antiretroviral exposure 10.3 years), 1312 had BLPCRneg. Outcome events included 322 blips and 139 VFs, with incidence rates being significantly lower in patients with BLPCRneg than in those with BLPCRpos [13.0% vs. 23.4% (P < 0.0001) and 5.1% vs. 11.2% (P < 0.0001), respectively]. In multivariable analysis, BLPCRneg was associated with a reduced risk of blips [hazard ratio (HR) 0.58; 95% confidence interval (CI) 0.47–0.73; P < 0.0001] and VF (HR 0.44; 95% CI 0.31–0.62; P < 0.0001).


Patients with PCRneg had better virological outcomes than those with PVL < 40 copies/mL but detectable viraemia. This suggests that the ‘no-signal’ information provided by currently commercially available HIV RNA quantification assays should be used routinely.


Nowadays most HIV-infected patients receiving highly active antiretroviral therapy (HAART) have a plasma viral load (PVL) < 50 HIV-1 RNA copies/mL [1, 2], which is the recommended goal for HAART [3]. However, in a substantial proportion of these patients, plasma HIV RNA remains detectable below this cut-off of 50 copies/mL. Actually, when highly sensitive assays are used, HIV may remain detectable in the plasma of patients receiving HAART, especially during the third phase of viral decay after initiation of therapy [4]. While current commercially available plasma HIV-1 RNA real-time polymerase chain reaction (PCR) quantification assays have a lower limit of quantification of 20 or 40 copies/mL, they can report qualitative RNA detection below these thresholds [5]. In a recent study by Doyle et al. of 1247 patients receiving HAART with PVL < 50 copies/mL, 19% of patients had a PVL of 40–49 copies/mL, 41% had a detectable but nonquantifiable PVL (< 40 copies/mL), and 40% had an undetectable PVL (no PCR signal detected). The level of viraemia was associated with the rebound rate (> 50 copies/mL), which was 34.2, 11.3 and 4% in these three groups, respectively [6].

The aim of our study was to assess whether patients with undetectable viraemia [a negative PCR result (PCRneg)] and those with PVL < 40 copies/mL but detectable viraemia [a positive PCR signal (PCRpos)] had different outcomes regarding the occurrence of blips and virological failure (VF). We conducted this study using the multicentre Dat'AIDS observational database.


The Dat'AIDS observational database

The Dat'AIDS observational database originates from a consortium of nine French HIV medical centres that share the same electronic medical record system (EMRS), Nadis® (Fedialis Medica, Marly Le Roi, France), for the medical follow-up of HIV-infected patients [7]. All the patients gave their written consent to have their medical chart recorded in this EMRS, which collects the following data along with the dates of events: demographic details, medical history, HIV disease history, data on clinical events, antiretroviral history, PVL measurements, CD4 cell count, and laboratory data throughout follow-up.

Standardized quality control assessments of databases are performed twice yearly in each centre. A merged database is maintained and regularly updated with the Nadis® medical records of all patients from all the Dat'AIDS consortium centres. This database forms the Dat'AIDS observational database.

Participating centres

From 1 January 2008 onwards, in five centres contributing to the Dat'AIDS observational database, for each PVL result below the quantification threshold (20 or 40 copies/mL depending on the assay used), a specific box was ticked in the patient's EMRS file whenever the PCR signal was not detected (PCRneg). These five centres (in Besançon, Marseille, Nice, Strasbourg and Toulouse) formed the participating centres in the current study.

Patient selection

The date of 1 January 2008 was chosen as an arbitrary baseline to allow 3 years of follow-up of patients, until 31 December 2010. Among the observational database patients, we selected those who were on stable HAART (unchanged for at least 6 months by 1 January 2008), whose first two PVL determinations in 2008 were < 40 copies/mL, and who had at least five PVL determinations between 1 January 2008 and 31 December 2010.

PVL techniques and definitions

PVL assays used throughout the study period were commercially available PCR assays which had a detection threshold of 40 [Abbott Diagnostics RealTime HIV-1 (Abbott Molecular, Rungis, France) and Cobas AmpliPrep/Cobas TaqMan HIV-1 V1.0 assays (Roche Diagnostics, Meylan, France) until mid-2009] or 20 (Cobas AmpliPrep/Cobas TaqMan HIV-1 V2.0 assays after mid-2009) copies/mL. Undetectable PVL (PCRneg) was defined as an undetected PCR signal whatever the assay used. A blip was defined as a single PVL measurement > 40 copies/mL, immediately followed by a PVL measurement < 40 copies/mL. VF was defined as two consecutive PVL measurements > 40 copies/mL. Patients were sorted into two groups according to baseline PVL: BLPCRneg, when the first two PVL determinations of 2008 were undetectable, or BLPCRpos, when a PCR signal was detected in at least one of the first two PVL determinations of 2008.

Statistical analysis

The outcome endpoints were the occurrence of the first blip and the occurrence of VF. Data were censored at the date of the last PVL determination of 2010 or at the occurrence of a virological endpoint, whichever came first.

Patients with BLPCRneg and BLPCRpos were compared at baseline for age, sex, HIV risk group, whether the last plasma hepatitis C virus (HCV) PCR determination was positive, duration of known HIV infection, CD4 lymphocyte count nadir, CD4 count at baseline, history of AIDS, time since initiation of antiretroviral therapy (ART), highest pre-ART PVL, number of prior ART regimens, number of prior VFs, time since last PVL > 40 copies/mL, and type of HAART by 1 January 2008 [two nucleoside reverse transcriptase inhibitors (NRTIs) + one protease inhibitor (PI), two NRTIs + one nonnucleoside reverse transcriptase inhibitor (NNRTI), or other].

Bivariable analysis was performed to identify predictors of the occurrence of blips and VF, using the log-rank test. Multivariable Cox regression analysis with backward selection was performed based on all variables associated with each outcome (blip and VF) at P < 0.20 in the bivariable analysis. Multivariable models were adjusted on the centre.


Study population

Of the 6202 patients who participated in the Dat'AIDS observational database on 1 January 2008, 4928 were on HAART and 1957 met the inclusion criteria. There were 1364 men and 593 women, with a mean age of 47 years, and a median ART duration of 10.3 years. Among these, 1312 patients (67.0%) belonged to the BLPCRneg subgroup. Table 1 displays their main baseline characteristics as well as a comparison of these characteristics between BLPCRneg and BLPCRpos patients. The number of PVL measurements after baseline were not different between the two groups.

Table 1. Patients' characteristics at baseline for the whole population and by baseline plasma viral load (PVL) [negative (BLPCRneg) vs. positive (BLPCRpos) for polymerase chain reaction (PCR) signal]
 Whole population (n = 1957)BLPCRneg (n = 1312)BLPCRpos (n = 645)P
  1. ARV, antiretroviral; HAART, highly active antiretroviral therapy; HCV, hepatitis C virus; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; PI, protease inhibitor; SD, standard deviation.
Male sex [n (%)]1364 (69.7)892 (68.0)472 (73.2)0.02
Age (years) [mean (SD)]47.1 (9.8)47.0 (9.8)47.4 (9.6)0.42
Duration of follow-up (months) [mean (SD)]32.1 (1.9)32.1 (1.8)32.0 (2.0)0.30
Number of PVL measurements post baseline [mean (SD)]10.8 (2.9)10.8 (2.8)10.6 (3.2)0.24
Duration of known HIV infection (years) [mean (SD)]13.1 (6.4)13.2 (6.3)13.0 (6.4)0.63
Risk group [n (%)]    
Homosexual701 (35.8)447 (34.1)254 (39.4)0.17
Heterosexual792 (40.5)544 (41.5)248 (38.5)
Drug user282 (14.4)193 (14.7)89 (13.8)
Other84 (4.3)56 (4.3)28 (4.3)
Unknown98 (5.0)72 (5.5)26 (4.0)
Last plasma HCV PCR positive [n (%)]211 (10.8)136 (10.4)75 (11.6)0.40
History of AIDS [n (%)]432 (22.1)285 (21.7)147 (22.8)0.59
CD4 count nadir [n (%)]    
< 200 cells/μL1035 (52.9)685 (52.2)350 (54.3)0.51
200 < 350 cells/μL745 (38.1)511 (39.0)234 (36.3)
≥ 350 cells/μL177 (9.0)116 (8.8)61 (9.5)
CD4 count at baseline (cells/μL) [mean (SD)]609 [265]602 [262]622 [271]0.13
Highest PVL ≥ 5 log copies/ml [n (%)]957 (48.9)598 (45.6)359 (55.7)< 0.0001
HAART duration at baseline [n (%)]    
1st quartile (≤ 5 years)459 (23.5)293 (22.3)166 (25.7)0.21
2nd quartile (> 5–10 years)491 (25.1)345 (26.3)146 (22.6) 
3rd quartile (> 10–12 years)463 (23.7)308 (23.5)155 (24.0) 
4th quartile (> 12 years)544 (27.8)366 (27.9)178 (27.6) 
Number of prior ARV regimens [n (%)]    
1st quartile (≤ 3)672 (34.3)454 (34.6)218 (33.8)0.01
2nd quartile [> 3–5]401 (20.5)279 (21.3)122 (18.9)
3rd quartile [> 5–8]464 (23.7)325 (24.8)139 (21.6)
4th quartile (> 8)420 (21.5)254 (19.4)166 (25.7)
Type of HAART at baseline [n (%)]    
2 NRTIs +1 PI863 (44.1)535 (40.8)328 (50.9)< 0.0001
2 NRTIs +1 NNRTI639 (32.6)460 (35.1)179 (27.8)
Other455 (23.3)317 (24.2)138 (21.4)
≥ 2 prior virological failures [n (%)]499 (25.5)314 (23.9)185 (28.7)0.07
Time since last PVL > 40 copies/mL    
1st quartile (≤ 9 months)514 (26.3)285 (21.7)229 (35.5)< 0.0001
2nd quartile (> 9–24 months)478 (24.4)300 (22.9)178 (27.6)
3rd quartile (> 24–40 months)485 (24.8)361 (27.5)124 (19.2)
4th quartile (> 40 months)480 (24.5)366 (27.9)114 (17.7)

Outcome events

A blip occurred in 322 patients (16.5%), less frequently in BLPCRneg than in BLPCRpos patients (13.0% vs. 23.4%, respectively; P < 0.0001). VF occurred in 139 patients (7.1%), also less frequently in BLPCRneg than in BLPCRpos patients (5.1% vs. 11.2%, respectively; P < 0.0001).

Kaplan–Meier estimates of the probability of a blip after 12, 24 and 32 months of follow-up were 10.9% [95% confidence interval (CI) 9.0–13.6%] vs. 4.8% (95% CI 3.7–6.1%), 18.3% (95% CI 15.4–21.6%) vs. 10.1% (95% CI 8.4–11.8%), and 24.1% (95% CI 20.8–27.9%) vs. 13.0% (95% CI 11.3–15.0%) for BLPCRpos and BLPCRneg, respectively (P < 0.0001 for all comparisons).

Kaplan–Meier estimates of the probability of VF after 12, 24 and 32 months of follow-up were 5.4% (95% CI 3.8–7.7%) vs. 2.1% (95% CI 1.3–3.1%), 8.6% (95% CI 6.5–11.3%) vs. 3.8% (95% CI 2.8–5.1%), and 10.3% (95% CI 8.0–13.3%) vs. 4.2% (95% CI 3.1–5.5%), respectively, for BLPCRpos and BLPCRneg (P < 0.0001 for all comparisons).

At the end of follow-up, mean CD4 lymphocytes counts had increased in both groups, without a significant difference between groups (667 cells/μL in BLPCRneg vs. 673 cells/μL in BLPCRpos; P = 0.70).

Predictors of blip

In multivariable analysis, predictors of the first blip included BLPVLneg, the highest pre-ART PVL value, the duration of controlled viraemia under HAART, and replicative HCV infection (Table 2). Patients with BLPCRneg had a 42% reduced risk of a blip, as compared to those with BLPCRpos. Also, the longer the duration of controlled viraemia, the lower the risk of a blip, with the lowest HR (0.31; 95% CI 0.21–0.44) for the longest duration (4th quartile; > 40 months) of controlled viraemia.

Table 2. Cox proportional hazards regression analyses: predictors of first blip and virological failure
 Predictors of first blipPredictors of virological failure
Univariable analysisMultivariable analysis*Univariable analysisMultivariable analysis*
HR (95% CI)PHR (95% CI)PHR (95% CI)PHR (95% CI)P
  1. ARV, antiretroviral; CI, confidence interval; HAART, highly active antiretroviral therapy; HR, hazard ratio; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; PCR, polymerase chain reaction; BLPCRneg, no PCR signal at baseline; PI, protease inhibitor; PVL, plasma viral load.
  2. *Backward selection, centre-adjusted model.
Male sex1.16 (0.91–1.48)0.25  0.80 (0.57–1.14)0.22  
≤ 40 years1   1   
> 40–45 years1.08 (0.78–1.50)0.63  1.28 (0.81–2.04)0.04  
> 45–50 years1.16 (0.84–1.61)  1.07 (0.66–1.75)  
> 50 years0.96 (0.70–1.32)  0.65 (0.39–1.08)  
Risk group        
Homosexual1   1   
Heterosexual1.03 (0.80–1.33)0.12  1.08 (0.72–1.62)0.01  
Drug user1.33 (0.97–1.84)  2.06 (1.31–3.24)  
Other1.34 (0.81–2.21)  1.37 (0.62–3.05)  
Unknown0.62 (0.32–1.18)  0.64 (0.23–1.78)  
Last plasma HCV PCR positive1.72 (1.27–2.32)< 0.0011.52 (1.11–2.06)0.012.42 (1.61–3.65)< 0.00011.69 (1.10–2.58)0.02
History of AIDS1.17 (0.91–1.51)0.23  1.26 (0.86–1.84)0.23  
CD4 count nadir        
< 200 cells/μL10.97  10.39  
200 < 350 cells/μL0.97 (0.77–1.23)  0.90 (0.62–1.29)  
≥ 350 cells/μL0.97 (0.66–1.44)  1.32 (0.78–2.23)  
HAART duration at baseline        
1st quartile (≤ 5 years)10.23  10.00210.001
2nd quartile (> 5–10 years)0.72 (0.53–0.99)  1.04 (0.59–1.85)1.27 (0.71–2.27)
3rd quartile (> 10–12 years)0.88 (0.65–1.20)  1.71 (1.01–2.90)1.73 (1.01–2.98)
4th quartile (> 12 years)0.93 (0.69–1.24)  2.21 (1.35–3.63)2.50 (1.50–4.16)
Number of prior ARV regimens        
1st quartile (≤ 3)10.13  1< 0.001  
2nd quartile [> 3–5]0.84 (0.61–1.16)  0.84 (0.48–1.47)  
3rd quartile [> 5–8]0.99 (0.74–1.32)  1.69 (1.08–2.64)  
4th quartile (> 8)1.25 (0.94–1.67)  2.11 (1.36–3.27)  
Type of HAART at baseline        
2 NRTIs + 1 PI10.004  10.001  
2 NRTIs + 1 NNRTI0.65 (0.50–0.85)  0.48 (0.31–0.75)  
Other0.99 (0.76–1.29)  1.12 (0.76–1.64)  
Number of prior virological failures        
010.21  10.003  
11.04 (0.78–1.40)  1.38 (0.89–2.16)  
≥ 21.25 (0.97–1.61)  1.91 (1.32–2.79)  
Highest PVL > 5 log copies/mL1.66 (1.32–2.07)< 0.00011.45 (1.16–1.82)0.0011.14 (0.81–1.58)0.46  
BLPCRneg0.50 (0.40–0.62)< 0.00010.58 (0.47–0.73)< 0.00010.41 (0.30–0.57)< 0.00010.44 (0.31–0.62)< 0.0001
Time since last PVL > 40 copies/mL        
1st quartile (≤ 9 months)1< 0.00011< 0.00011< 0.00011< 0.0001
2nd quartile (> 9–24 months)0.45 (0.34–0.60)0.49 (0.36–0.65)0.42 (0.27–0.65)0.46 (0.29–0.71)
3rd quartile (> 24–40 months)0.46 (0.35–0.61)0.55 (0.41–0.74)0.35 (0.22–0.55)0.36 (0.23–0.59)
4rd quartile (> 40 months)0.25 (0.18–0.36)0.31 (0.21–0.44)0.23 (0.13–0.38)0.29 (0.17–0.52)

Predictors of VF

Four predictors of VF were identified in multivariable analysis. Three of them had also been identified as predictors of a blip: BLPVLneg, the duration of controlled viraemia under HAART and replicative HCV infection (Table 2). The fourth predictor of VF was the total duration of ART, with the longest duration of ART (4th quartile; > 12 years) associated with the highest risk of failure (HR 2.50; 95% CI 1.50–4.16).


In this study, 67% of patients with PVL < 40 copies/mL had no detectable PCR signal at ‘baseline’, an arbitrary cross-sectional time-point chosen both to reflect the recent situation and to allow a 3-year follow-up period. This percentage is somewhat higher than that observed in the only other cohort study that addressed the same issue, in which such patients represented only 40.7% of the cohort [6]. This difference may be explained by a longer duration of HIV infection and HAART exposure in our study population [4]. It may also result from the fact that two PCR methods were used in this study, as it was shown that the analytical performances of commercial virus load assays may differ [8].

We found that patients with BLPCRneg, by comparison with BLPCRpos patients, were more often women, had a lower number of prior HAART regimens, had a lower pre-ART PVL, had a longer duration of PVL < 50 copies/mL, and were more often receiving an NNRTI-containing HAART regimen, the latter three having also been found in the study by Doyle et al. [6]. Although the relationship between exposure to a HAART regimen containing an NNRTI (especially nevirapine) and permanently undetectable PVL has also been reported by others [9, 10], this relationship may not be causal and may rather reflect different ART history, tolerability and/or adherence profiles. However, we feel that this finding deserves further evaluation.

We found that patients with BLPCRneg had better outcomes than patients with BLPCRpos, both in terms of the occurrence of blips and in terms of the development of VF, as shown by an independent and significant risk reduction of 42% and 56% for these two outcomes, respectively. These findings corroborate Doyle et al.'s, with a longer duration of follow-up and slightly different definitions for rebound and VF in our study.

We also found that the risk of blips and VF was inversely correlated with the duration of viral suppression below the limit of quantification, as previously demonstrated [11-13]. This contrasted with an increasing risk of VF (but not of blips) with the duration of exposure to antiretrovirals. The latter finding may be explained by the fact that patients with longer exposure to antiretrovirals may also have had a longer period of suboptimal control of viraemia before receiving HAART.

Additionally, we found HCV infection to adversely affect the risk of both blips and VF, a finding that has not been previously reported, although PVL < 50 copies/mL was observed in 54% of HCV-positive vs. 64% of HCV-negative patients (p = 0.04) in one study that compared the two subgroups of patients in a population of 902 HIV-infected people attending a single HIV clinic [14]. A lower adherence to HAART has also been reported in HCV-infected patients [15], which may, at least partly, explain our findings.

One strength of our study, in comparison to Doyle et al.'s, is that we evaluated a larger number of patients recruited in different centres over a longer follow-up period, which consolidates their findings that patients with unquantifiable but detectable viraemia have a higher risk of blips and VF. However, a limitation of this study is that we could not assess the impact of adherence to HAART on outcomes. However, adherence is difficult to assess accurately and was not identified as a predictor of viral rebound in the study by Doyle et al. either.

We believe that two practical messages can be taken from our results for the clinical care of HIV-infected patients. First, attention should be paid to patients who have unquantifiable but detectable viraemia, especially when this is observed repeatedly. Whether an intervention in terms of the HAART regimen −a change or intensification, for example− should be made can obviously not be inferred from our findings. However, this question does deserve further consideration, ideally within a randomized clinical trial, as suggested by Gandhi and Deeks [16]. Until this question has been addressed, we suggest that such patients with unquantifiable but detectable viraemia should receive interventions such as adherence reinforcement, improvement of drug intake conditions, especially for drugs whose pharmacokinetics is influenced by food intake, and therapeutic drug monitoring. Secondly, we feel that the ‘no signal’ information provided by current commercially available PVL assays should be taken into account when PVL is below the limit of quantification and we would recommend that it be now recorded in routine clinical practice. This would be beneficial to the patients, at no additional cost, especially as compared with the recently developed single-copy assays [17].


The Nadis® EMRS was developed and is maintained by Fedialis Medica GlaxoSmithKline Group, with the support of ViiV Healthcare, France.

Conflicts of interest: The authors have no conflicts of interest to declare.

Financial disclosure: No funding was received for this study.

Appendix: Appendix: the Dat'AIDS Study Group (

P. Enel, V. Obry-Roguet, O. Faucher, S. Bregigeon and I. Poizot-Martin, (Marseille); B. Marchou, P. Massip, E. Bonnet, M. Obadia, M. Alvarez, L. Porte, L. Cuzin, M. Chauveau and I. Lepain (Toulouse); P. Pugliese, L. Bentz, C. Ceppi, E. Cua, J. Cottalorda, J. Durant, S. Ferrando, J. G. Fuzibet, R. Garraffo, A. Naqvi, V. Mondain, I. Perbost, S. Pillet, B. Prouvost-Keller, C. Pradier, S. Pugliese, P.-M. Roger, F. Sanderson, V. Rahelinirina, E. Rosenthal, M. Vassallo and P. Dellamonica (Nice); C. Allavena, E. Billaud, C. Biron, B. Bonnet, S. Bouchez, D. Boutoille, C. Brunet-François, N. Feuillebois, T. Jovelin, O. Mounoury, P. Morineau, F. Raffi, V. Reliquet, H. Hue and D. Brosseau (Nantes); Y. Yazdanpanah and P. Choisy (Tourcoing); C. Duvivier, M. A. Valantin, R. Agher, C. Katlama, M. Shoai-Tehrani, O. Lortholary, P. H. Consigny, G. Cessot, F. Touam and K. Benhadj (Paris); A. Cabié, S. Abel, S. Pierre-François and B. Liautaud (Fort de France); D. Rey, E. Ebel, P. Fischer and M. Partisani (Strasbourg); C. Chirouze, C. Drobacheff-Thiébaut, J. P. Faller, J. F. Faucher, A. Foltzer, H. Gil, B. Hoen, L. Hustache-Mathieu and C. Bourdeaux (Besançon).