Several studies have been conducted to evaluate the predictors of initial virological response to highly active antiretroviral treatment (HAART) in previously antiretroviral-naive or experienced patients. In both naive and experienced patients, viral load at the time of starting HAART has been shown to be associated with the probability of reaching a viral load below 500 copies/ml; patients with higher pre-HAART viral loads are less likely to achieve this level [1–4]. However, little is known about the relationship between the viral response to HAART over the first 24 weeks of therapy and viral load values measured 1–2 months after the initiation of HAART [4,5]. Indeed, it would be beneficial to identify patients whose chance of achieving an undetectable viral load by 24 weeks of therapy is low, in order to investigate the possible reasons for a lack of virological response promptly and to prevent unnecessary periods of ongoing viral replication in the presence of drugs that can lead to the development of drug resistance [6,7]. This is particularly relevant for patients who initiate HAART for the first time. Powderly and colleagues  showed that a viral load reduction of 2.5 log10 copies/ml or greater in 4 weeks afforded the highest probability of viral suppression at week 48 (≤ 50 copies/ml) in patients starting a regimen containing nelfinavir, zidovudine and lamivudine. Huang and colleagues  reported that 3–25% of naive patients starting HAART were ‘off-track’ for viral suppression by weeks 2–4, but that being off-track was not predictive of the subsequent viral response. The conclusion of the authors was that it is problematical to identify future non-responders using the viral loads measured up to week 4.
We analysed data from a group of antiretroviral-naive patients for whom we evaluated the probability of achieving a viral load of 500 copies/ml or less by week 24 from the initiation of HAART (three or more antiretroviral agents) according to their baseline viral load, their week 4 viral load and their week 8 viral load concentrations.
Patients and methods
We used data from the J.W. Goethe Universität HIV clinic cohort, Frankfurt, Germany. This is an ideal study to investigate this issue as patients recruited in the cohort are frequently monitored (every 4 weeks) and are seen in a routine clinical setting, without selection. Plasma HIV RNA is measured using quantitative reverse transcriptase–polymerase chain reaction (Amplicor; Roche Molecular Systems, Inc., Pleasanton, CA, USA) with a lower limit of detection of 500 copies/ml. The ultrasensitive version of this assay with a lower limit of detection of 50 copies/ml was used in patients whose viral load was 500 copies/ml or less starting from 1 January 1998.
We considered 453 previously antiretroviral-naive patients who started at least three antiretroviral drugs and whose pre-HAART viral load (measured less than 4 weeks before starting therapy) was greater than 500 copies/ml, and for whom a viral load measured at week 4 (or at week 8) and at least one viral load in the window ranging from week 12 to week 24 of therapy was available. A window from day 15 to day 35 and from day 36 to day 63 was used, respectively, to define the weeks 4 and 8 viral loads. When more than one HIV-RNA level was measured in these windows, the measurement closest to weeks 4 and 8 was used. In the event of a tie, the later of the two measurements was used. Using these definitions, 361 patients (79.7%) had a measure of viral load 4 weeks after the initiation of therapy and 348 (76.8%) had a measure of viral load 8 weeks after the initiation of therapy. Two hundred and fifty-six patients (56.5%) had a measure of viral load both at week 4 and at week 8.
Virological suppression below 500 copies/ml by week 24
Patients were grouped according to their pre-HAART viral load using the cut-offs of 10 000, 100 000, and 1 000 000 copies/ml and according to the 4–8 weeks viral loads using the cut-offs of 500, 1000, and 10 000 copies/ml. The frequency distribution of patients starting at various levels of pre-HAART viral load according to the viral load levels achieved by 4–8 weeks was calculated. To study the prognostic value of the week 4 viral load in predicting the rate of achieving a viral load of 500 copies/ml or less by week 24, we only used patients whose week 4 viral load was above 500 copies/ml (n = 208). Similarly, to study the predictive value of the week 8 viral load we considered only patients whose viral load had not reached 500 copies/ml before week 8 (n = 103). We defined as ‘successes’ patients who, between weeks 12 and 24 (days 84–168), achieved a viral load of 500 copies/ml or less on at least one occasion.
It is reasonable to assume that treatment modifications made before week 16 of therapy are more likely to be the result of intolerance or toxicity rather than poor viral response. However, any treatment modification occurring after week 16 with a viral load greater than 500 copies/ml could be caused by perceived poor viral response, and failing to take this into account could have biased our estimates. Viral loads obtained after the date of first therapy modification after week 16 have thus been ignored. Patients included in the analysis were those who had at least one viral load measurement in the window ranging between weeks 12 and 24, obtained while they were still on the same treatment that they were receiving at week 16.
The observed probabilities of achieving a viral load of 500 copies/ml or less by week 24 according to the viral load pre-HAART, the week 4 viral load and the week 8 viral load concentrations were calculated. In addition, we performed a logistic regression of the binary response ‘virological success by week 24’ with the pre-HAART, weeks 4 and 8 viral loads, as continuous covariates. Separate models including the change in viral load from pre-HAART levels to week 4 and the change from pre-HAART levels to week 8 were also fitted.
We chose to report the observed probabilities of virological success according to the absolute value of pre-HAART viral load and weeks 4 and 8 viral load rather than the weeks 4 and 8 changes from pre-HAART values, because we felt that it would be easier to utilize the results directly. Indeed, the probability of success for a given change in viral load from pre-HAART values can be approximately calculated from these data. However, the probabilities of achieving a viral load of 500 copies/ml or less per 1 log10 decrease in viral load from pre-HAART levels, as estimated from the logistic regression, were also given.
Virological suppression below 50 copies/ml by week 24
The logistic regression analysis was repeated to evaluate the prognostic value of weeks 4 and 8 viral loads in predicting patients who achieved a viral load below 50 copies/ml between weeks 12 and 24. This analysis was only performed on 165 patients who started HAART after 1 January 1998.
Sustained virological suppression up to week 48
In patients who had reached a viral load below 500 copies/ml before week 24, we evaluated whether weeks 4 and 8 viral loads were useful in predicting which patients were more likely to sustain the achieved viral suppression up to week 48. Patients whose viral loads remained below 500 copies/ml for the time period ranging between weeks 24 and 48 were defined as ‘successes', all other patients were defined as ‘failures'. A logistic regression with this binary outcome was performed.
The characteristics of the patients included in the analysis are shown in Table 1. The patients were predominantly men (80.3%), 37 years old on average, and acquired HIV infection via homosexual contacts (49.2%). The patients were all previously antiretroviral naive, having started HAART with relatively advanced disease, with a median CD4 count pre-HAART of approximately 210 cells/μl and a median plasma viral load of 5.4 log10 copies/ml. The majority of patients (52.1%) started two nucleoside reverse transcriptase inhibitors (NRTI) and one protease inhibitor (PI), or two NRTI and one non-nucleoside reverse transcriptase inhibitor (NNRTI) (25.4%). The NRTI most frequently used were: lamivudine (97.6%), zidovudine (62.0%), didanosine (19.2%), and stavudine (23.6%). Some patients started the NNRTI nevirapine (19.6%) or efavirenz (10.6%). Among those starting a PI, the majority of patients started indinavir (41.7%), but some also started nelfinavir (18.1%) and ritonavir (7.7%). The characteristics of patients with available measures of viral load at weeks 4 or 8 were very similar to those observed overall (Table 1). Of the 361 patients whose week 4 viral load was available, 208 (57.6%) had not yet achieved a viral load of 500 copies/ml or less, and were followed on average for 17.1 weeks from week 4 to the last available viral load measured within 24 weeks of the initiation of HAART. Of the 348 patients with available week 8 viral loads, 103 (29.6%) had not yet achieved a viral load of 500 copies/ml or less by week 8 and were followed on average for 13.6 weeks from week 8 to the last available viral load before week 24. Of all patients, 70 (15.5%) had only one available measure of viral load between weeks 12 and 24 from starting HAART, 173 (38.2%) had two measurements, and 200 (46.3%) had three or more determinations.
Virological suppression below 500 copies/ml by week 24
Table 2 (a,b) shows the distribution of patients according to their pre-HAART viral load and the 4–8 weeks viral loads. This table can be read as the ‘expected proportion of patients’ falling in each of the 4–8 weeks viral load level categories for a given pre-HAART viral load level. Therefore, for example, of the 79 patients whose pre-HAART viral load was above 1 million copies/ml, 14 (17.7%) achieved a viral load of 500 copies/ml or less by week 4 but the majority (n = 43, 54.4%) still had a viral load between 1000 and 10 000 copies/ml at that time (Table 2 a). However, most patients had already achieved a viral load of 500 copies/ml or less by 4 weeks when starting with 100 000 copies/ml or less. Many more patients achieved a viral load of 500 copies/ml or less by week 8 irrespective of their pre-HAART level (n = 245 out of 348, 70.4%) and, again, not surprisingly, the rate of achievement was higher in patients with low levels of pre-HAART viral load (Table 2 b).
Overall, 175 out of 208 patients (84.1%) included in the prospective week 4 viral load analysis achieved a viral load of 500 copies/ml or less by week 24. The observed probabilities of success for different levels of the pre-HAART viral load and of the week 4 viral load are shown in Table 3. The pre-HAART viral load did not show any association with the observed chance of reaching a viral load of 500 copies/ml or less (chi-square 1.00;P = 0.80;Table 3). In contrast, the week 4 viral load appeared to be a strong predictor of the subsequent virological response. The probability of achieving a viral load of 500 copies/ml or less by week 24 decreased dramatically for values of the week 4 viral load above 10 000 copies/ml (chi-square 17.1;P = 0.001;Table 3). However, even in patients whose week 4 viral load was above 10 000 copies/ml, the overall chance of a value of 500 copies/ml or less by week 24 was still 61%. Patients starting HAART with a viral load higher than 1 million copies/ml, and whose week 4 viral load was still above 10 000 copies/ml, showed a lower chance of achieving a viral load of 500 copies/ml or less by week 24 (four out of nine patients, 44.4%, Table 3).
The observed probabilities of achieving a viral load of 500 copies/ml or less by week 24 according to the week 8 viral load are shown in Table 4. As expected, in this group of 103 patients whose viral loads had not reached 500 copies/ml before week 8, the overall chance of reaching a viral load of 500 copies/ml or less by week 24 was lower than in the previous analysis (79 successes, 76.7%). Again, the crude prognostic value of week 8 viral load was higher (chi-square 23.2;P = 0.001) than that of pre-HAART viral load (chi square 2.19;P = 0.53). In this analysis, in patients whose week 8 viral load was still above 10 000 copies/ml, the overall chance of achieving a viral load of 500 copies/ml or less over the subsequent 8 weeks was only 42.3% (Table 4).
In order to make a better comparison of the relative prognostic value of the pre-HAART viral load, the week 4 viral load, and the week 8 viral load we performed a logistic regression analysis. Table 3 and Table 4 suggested that the chance of achieving a viral load of 500 copies/ml or less associated with the pre-HAART viral load could be different according to the level of the weeks 4–8 viral load. This method of analysis allowed us to test formally for a possible interaction between the two covariates expressing the viral load levels pre-HAART and at weeks 4–8 (included in the model as continuous covariates). The test for interaction was, however, not significant (Wald test P = 0.19 and P = 0.06, for weeks 4 and 8, respectively). From fitting a logistic model including the pre-HAART viral load, the week 4 viral load, and the week 8 viral load as continuous covariates, we found that the week 8 viral load was the strongest predictor (Wald tests 0.02, 6.11, and 12.63, respectively, for pre-HAART, week 4 and week 8 viral loads). The estimated adjusted relative odds of achieving a viral load of 500 copies/ml or less by week 24, irrespective of the previous viral load levels, was 0.42 [95% confidence interval (CI) 0.26–0.68;P = 0.0004] per 1 log10 higher week 8 viral load (Table 5). The change in viral load was also more predictive after 8 weeks than after 4 weeks of HAART. The univariate relative odds of viral success per 1 log10 greater week 8 decrease in viral load from pre-HAART levels was 1.99 (95% CI 1.45–2.73;P = 0.0001;Table 5). Several other co-factors including sex, Centers for Disease Control and Prevention stage, age, and CD4 cell count at the time of starting HAART failed to be significantly associated with the risk of virological success.
Virological suppression below 50 copies/ml by week 24
One hundred and twenty-four patients (75.2%) of the 165 who started HAART after 1 January 1998 had a viral load greater than 50 copies/ml at week 4 of therapy. Of these, 90 (72.6%) achieved a viral load below 50 copies/ml by week 24; the probabilities of viral suppression were 88.7% (47 out of 53) in patients with week 4 viral loads of 50–500 copies/ml, 68.4% (13 out of 19) in those with week 4 viral loads of 501–1000 copies/ml, and 57.7% (30 out of 52) in those with a viral load greater than 1000 copies/ml (chi-square 16.1, P = 0.001). One hundred and ten patients had a viral load greater than 50 copies/ml by week 8, and 72 (65.5%) achieved suppression below 50 copies/ml by week 24; the probabilities of viral suppression according to the week 8 viral load were 75.3% (55 out of 73), 76.9% (10 out of 13), and 29.1% (seven out of 24), respectively, in the same categories above (chi-square 19.4, P = 0.001). In this group of patients the viral load at week 4 was the strongest predictor of viral success in a multivariate logistic regression analysis [adjusted odds ratio (OR) of achieving a viral load of 50 copies/ml or less by week 24; OR 0.39 per 1 log10 higher week 4 viral load; 95% CI 0.16–0.92;P = 0.03].
Sustained virological suppression up to week 48
Two hundred and ninety-two patients achieved a viral load of 500 copies/ml or less before week 24 and had at least one viral load measurement in the window ranging between weeks 24 and 48. Of these, 251 (86.0%) remained virologically suppressed below 500 copies/ml up to week 48. A week 8 viral load of 0–500 copies/ml was associated with a 85.2% chance (173 out of 203 patients) of maintaining viral suppression up to week 48, a week 8 viral load of 501–1000 copies/ml with a 87.5% chance (21 out of 24 patients), a week 8 viral load 1001–10 000 with a 78.8% chance (26 out of 33 patients), and a week 8 viral load greater than 10 000 with a 33.3% chance (three out of nine; chi-square 17.1;P = 0.001). Viral load at week 8 was the only significant independent predictor in the logistic regression analysis (adjusted OR of sustained viral suppression; OR 0.54 per 1 log10 higher week 8 viral load; 95% CI 0.32–0.93;P = 0.03).
We described the initial response to HAART in a group of previously antiretroviral-naive patients attending the J.W. Goethe Universität clinic in Frankfurt. These patients are very frequently monitored and continuously motivated to be adherent to their antiretroviral treatment. This may explain the high rate of patients achieving a viral load of 500 copies/ml or less by weeks 4 and 8 in this population compared with other HIV clinics in Europe . However, we believe that the estimated proportions of patients achieving a certain level of viral load by weeks 4 and 8 when starting from some defined pre-HAART levels given in Table 2 (a,b) can be used as a rough description of the expected viral response of naive patients in other clinical settings. These data also clearly show that the weeks 4 and 8 viral loads are strong predictors of the viral response over the first 24 weeks of HAART. The chance of achieving a viral load of 500 copies/ml or less associated with the weeks 4 and 8 viral loads seemed unlikely to differ according to various levels of pre-HAART viral load (the test for interaction was not significant). This result is in agreement with the findings of another previous study . In contrast, the viral load measured at the time of HAART was not predictive of virological success by week 24 (Table 3;Table 4;Table 5). This result appears to contradict those of several previous reports [1–4]. However, this may be explained by the fact that, in this analysis, the predictive value of pre-HAART viral load was assessed in ‘selected’ populations of patients with a viral load above 500 copies/ml by weeks 4–8.
In the clinical management of previously antiretroviral-naive patients starting HAART, it is crucial to detect early signs that there will be incomplete suppression of viral load. Inadequate viral suppression under therapy may lead to the early development of drug-resistant virus and eventually to virological failure [6,7]. The virological response to a second-line HAART regimen in patients experiencing virological failure after receiving their first HAART regimen is usually inferior compared with the response in antiretroviral-naive patients . There is thus a need in these patients to assess the reasons for the observed incomplete suppression so that the patient's management can be adjusted accordingly. There is evidence that, in patients who experience virological failure (i.e. failure to suppress viral load below 500 copies/ml or viral rebound), switches in therapy that occur in patients with lower levels of viral load are associated with a better subsequent viral response than those in patients with higher viral loads [10,11]. However, to date, the long-term clinical outcome of these patients is unknown, and a change in therapy alone, or a treatment intensification, may not improve patients’ future outcomes (e.g. if a patient is not adherent). Currently, suggestions on when to change or intensify the initial therapy in patients who show an incomplete response to their first HAART regimen within the first 24 weeks of starting HAART are based on theoretical considerations rather than ‘evidence-based’ medicine (e.g. a change in therapy is often recommended if the viral load is still detectable by week 24). Similarly, a minimum 2 log10 copies/ml decline has recently been indicated as the average ‘expected’ 4 week decline in previously antiretroviral-naive patients starting HAART, so that patients whose viral load does not drop by at least 2 log10 copies/ml by week 4 of HAART should be considered to be possible early non-responders . Unfortunately, there are only limited data to support this recommendation .
A previous report  indicated that it would be problematical to identify early non-responders using the viral loads measured up to 2–4 weeks among patients starting HAART for the first time. Our analysis shows that approximately 85% of patients whose viral load was still above 500 copies/ml at week 4 achieved a viral load of 500 copies/ml or less before week 24 from the date of starting HAART. Even patients whose week 4 viral load was still above 10 000 copies/ml, regardless of their pre-HAART levels, had a 61% chance of reaching a viral load of 500 copies/ml or less by 24 weeks of therapy (Table 3). The week 4 viral load was a strong predictor of viral suppression below 50 copies/ml. However, again, patients with a viral load greater than 1000 copies/ml at week 4 retained a 58% chance (30 out of 52 patients) of achieving a viral load of 50 copies/ml or less by week 24. Overall, we can conclude that the week 4 viral load does not have enough power to discriminate a group of patients whose chance of subsequent viral suppression is so low that an investigation of possible factors associated with virological failure would be recommended. In contrast, the chance of reaching a viral load of 500 copies/ml or less by week 24 in patients whose week 8 viral load was still above 10 000 copies/ml was as small as 42% (Table 4). A week 8 viral load greater than 1000 copies/ml was associated with a 29% chance of reaching a viral load of less than 50 copies/ml by week 24. Furthermore, in patients who achieved a viral load of 500 copies/ml or less by week 24, those who had a week 8 viral load greater than 10 000 copies/ml had only a 33% chance of maintaining viral suppression below 500 copies/ml. From fitting a logistic regression model, we estimated that the probability of achieving a viral load of 500 copies/ml or less by week 24 was 60% lower for each log10 higher level of the week 8 viral load, regardless of previous viral load values (Table 5). Once viral suppression below 500 copies/ml was achieved, the chance of maintaining it was 45% lower for each log10 higher level of the week 8 viral load (data not shown). So our data indicate that the week 8 viral load can indeed be useful in the early management of naive patients starting HAART to predict the patients who have little chance of suppressing and maintaining a viral load of 500 copies/ml or less.
The reasons for incomplete suppression after starting HAART for the first time include drug intolerance, drug toxicity, drug resistance, non-adherence, insufficient drug absorption or some other, so far unrecognized, factor. On the basis of our findings, we believe that for patients whose viral load after 8 weeks of therapy is still above 10 000 copies/ml, there is an urgent need to assess adherence to therapy, drug levels and resistance, and if appropriate, change the therapy. We also noted that the assessment of non-adherence and drug absorption in a clinical setting is a very difficult task [13–15], and that in an average European clinical setting patients not sufficiently motivated to take the drug combination prescribed are often only ‘discovered’ after at least 12 weeks when their viral load, at best slightly reduced from pre-therapy levels, is finally measured. Even when viral loads are monitored every 4 weeks after the initiation of HAART, to our knowledge, there are no published data containing information that can help clinicians in associating a probability of reaching a viral load of 500 copies/ml or less by week 24 with viral load determinations measured before 12 weeks of therapy. A viral load above 10 000 copies/ml after 8 weeks of HAART (which in our analysis was found in 7.5% of our patients and was associated with a 42% chance of achieving a viral load of 500 copies/ml or less by 24 weeks) in patients who harbour no resistant virus may be indirect evidence of non-adherence or poor drug absorption.
1. Staszewski S, Miller V, Sabin CA. et al. Virological response to protease inhibitor therapy in a HIV clinic cohort.
AIDS 1999, 13: 367 –373.
2. Mocroft A, Gill MJ, Davidson W, Phillips AN. Predictors of viral response and subsequent virological treatment failure in patients with HIV starting a protease inhibitor.
AIDS 1998, 12: 2161 –2167.
3. Casado JL, Perez-Elias MJ, Antela A. et al. Predictors of long-term response to protease inhibitor therapy in a cohort of HIV-infected patients.
AIDS 1998, 12: F131 –F135.
4. Powderly WG, Saag MS, Chapman S, Yu G, Quart B, Clendeninn NJ. Predictors of optimal virological response to potent antiretroviral therapy.
AIDS 1999, 13: 1873 –1880.
5. d'Arminio Monforte A, Testori V, Adorni F. et al. CD4 cell count at the third month of HAART may predict clinical failure.
AIDS 1999, 13: 1669 –1676.
6. Molla A, Korneveva M, Gao Q. et al. Ordered accumulation of mutations in HIV protease confers resistance to ritonavir.
Nat Med 1996, 2: 760 –766.
7. Condra JH, Schleif WA, Blahv OM. et al. In vivo emergence of HIV-1 variants resistant to multiple protease inhibitors.
Nature 1995, 374: 569 –571.
8. Huang W, De Gruttola V, Fischl M, et al
. Patterns of plasma HIV RNA responses in antiretroviral treatment success and failure. 7th Conference on Retroviruses and Opportunistic Infections.
San Francisco, 30 January–2 February 2000 [Abstract 451].
9. Ledergerber B, Egger M, Opravil M. et al. Clinical progression and virological failure on highly active antiretroviral therapy in HIV-1 patients: a prospective cohort study.
Lancet 1999, 353: 863 –868.
10. Deeks SG, Hecht FM, Swanson M. et al. HIV RNA and CD4 cell count response to protease inhibitor therapy in an urban AIDS clinic: response to both initial and salvage therapy.
AIDS 1999, 13: F35 –F43.
11. Mocroft A, Phillips AN, Miller V, et al
. Virological response to a second-line protease inhibitor (PI) regimen: laboratory markers and choice of treatment. 7th Conference on Retroviruses and Opportunistic Infections
. San Francisco, 30 January–2 February 2000 [Abstract 539].
12. Carpenter CJ, Cooper DA, Fischl MA. et al. Antiretroviral therapy in adults.
:Updated recommendations of the International AIDS Society – USA Panel.
JAMA 2000, 283: 381 –390.
13. Kastrissios H, Suarez J-R, Katzenstein D, Girard P, Sheiner LB, Blaschke TF. Characterizing patterns of drug-taking behaviour with a multiple drug regimen in an AIDS clinical trial.
AIDS 1998, 12: 2295 –2303.
14. Eldred LJ, Wu AW, Chaisson RE, Moore RD. Adherence to antiretroviral and Pneumocystis prophylaxis in AIDS disease.
J Acquir Immune Defic Syndr 1998, 18: 117 –125.
15. Hoetelmans RM, Reijers MH, Weverling GJ. et al. Effect of plasma drug concentrations on HIV-1 clearance rate in antiretroviral naive patients during quadruple drug combination therapy.
AIDS 1998, 12: F111 –F115.