The potential for CD4 cell increases in HIV-positive individuals who control viraemia with highly active antiretroviral therapy

Objectives: To study the long-term CD4 cell responses to highly active antiretroviral therapy (HAART) in treatment-naive patients whose viral loads remained below 500 copies/ml for prolonged periods.

Design: A total of 237 patients whose viral loads remained below 500 copies/ml for one year or more. Median follow-up was 1.9 years.

Methods: CD4 cell counts were analysed to investigate long-term immunological response using mixed-effects models with the slope allowed to change after 1, 12 and 24 months of HAART.

Results: The median baseline CD4 cell count was 175 cells/mm³. After an initial rapid increase in the first month after HAART (97.2 cells/mm³ a month), increases in CD4 cell counts continued less rapidly (11.6 cells/mm³ a month). This increase slowed by 2.4 cells/mm³ a month after one year. CD4 cell counts continued increasing after 2 years, but the rate of increase again slowed (estimated slope at 2 years 5.4 cells/mm³ a month; decrease in slope from year 2 compared with years 1-2 3.7 cell/mm³ a month). A total of 198 out of 211 patients (94%) with measurements at baseline and one year experienced an increase in CD4 cell counts in this interval; 81 and 67% had an increasing slope between 1 and 2 and 2 and 3 years, respectively. By the end of follow-up, CD4 cell counts had increased by 319 cells/mm³, and were more than 500 cells/mm³ in 40% of patients.

Conclusion: Although the rate of immune recovery slowed after 2 years, CD4 cell counts rose in most and began to return to levels seen in HIV-negative individuals.

Introduction

HIV-1 infection is associated with a loss of immune function, which is characterized by a decrease in the numbers of CD4 T lymphocyte cells [¹]. Since the introduction of highly active antiretroviral therapy (HAART) treated individuals have generally experienced large reductions in viral load and there have been dramatic reductions in both mortality and morbidity [^2-5]. There has also been evidence of short-term increases in CD4 cell counts in those receiving HAART [^6-8], preceded by an initial rapid increase in the CD4 cell count in the first 1-3 months after starting therapy, believed to be caused by a redistribution of cells trapped in the lymphoid tissue [⁹].

The potential for immune reconstitution up to 2 years after commencing HAART therapy is well known [^10-16]. There is some evidence that the rate of immune recovery, as measured by the CD4 cell count, at 2 years after starting treatment is slower than that at earlier times. However, because of the relatively short time since HAART was introduced the behaviour of the immune system over longer periods of time is less well documented [^17,18]. It is not known whether or when this recovery will eventually stop, and whether the immune system will ultimately recover to and remain at levels comparable to those seen in HIV-negative individuals. One study [¹⁷] has shown that the greatest increases in CD4 cell counts occurred in the first 2 years of therapy; however, this study had small numbers and not all participants were therapy naive.

We studied the potential recovery possible in the immune system as demonstrated by the CD4 cell count in a study of antiretroviral-naive patients treated with HAART with long periods of follow-up. As immune recovery is maximal when there is a suppression of plasma viral loads [^14,19,20], our study examined the behaviour of CD4 cell counts in patients who sustained virological suppression while on HAART for prolonged periods of time. Of particular interest was whether CD4 cell counts continued to increase after 2 years of treatment, and to what extent the immune system could potentially recover under optimal circumstances.

Methods

Patient population

All individuals included in the study were patients at the Ian Charleson Clinic at the Royal Free Hospital, London. Information on these patients is collected prospectively as patients attend for care. The information is audited every 6-9 months by a trained research assistant who updates information on deaths and all AIDS-defining illnesses from patient notes, and cross-checks the information in the notes with that on the database.

For the purposes of this study, HAART is defined as simultaneously receiving three or more antiretroviral treatments. All patients included in the analyses were naive to antiretroviral therapy at the time of starting HAART, and were followed from this time forward. We chose to focus on the potential for immune reconstitution in those who were experiencing a virological response. All patients studied thus had to have achieved undetectable plasma HIV-1-RNA levels (defined as less than 500 copies/ml) within 6 months of starting therapy, and had to have remained below this level until at least one year after starting HAART. Individuals were required to have at least one year of follow-up to focus on whether complete immune reconstitution, as demonstrated in this study by changes in the CD4 cell count, would be possible after a long period of sustained virological response. However, 'blips' (transient episodes of detectable viraemia), defined as single viral load measurements that rose above 500 copies/ml but then returned to undetectable levels at the next reading, were allowed, as it has been shown that patients with single blips do not progress more quickly [²¹]. As we were interested in immunological outcomes regardless of the treatment regimen, providing viral load levels remained below 500 copies/ml, follow-up was not censored at treatment change.

All patients were required to have a baseline CD4 cell count measurement in the period between 6 months before and 2 days after starting treatment. In those with more than one measurement in this period the one closest to the date of starting HAART was chosen as the baseline measure. Furthermore, at least one CD4 cell count reading was required to have been taken within the first 6 months after starting treatment. From a possible 502 patients starting HAART at the Royal Free Hospital, 237 participants met these criteria and have been included in the analyses, 53 of whom participated in a clinical trial at some point during follow-up. Of the 265 subjects who did not meet the inclusion criteria, 81 did not attain a viral load measurement of less than 500 copies/ml within 6 months of starting treatment, 85 did not have a viral load measurement taken within 6 months of starting HAART (these were often patients starting HAART in the early periods when viral loads were measured less frequently), 19 did not maintain this level of virological suppression until one year after starting HAART, 37 did not have follow-up data for one year from starting HAART, 28 did not have a CD4 cell count measurement taken within 6 months of starting HAART, and 15 did not have a baseline CD4 cell count.

The characteristics of those included in the analysis at the time of starting HAART are shown in Table 1. Compared with the 265 treatment-naive patients starting HAART who did not meet the inclusion criteria mentioned above, patients included were more likely to have started therapy at a later year. However, baseline CD4 cell counts and viral load measurements were similar in the two groups. Of those who were excluded from the analysis, 54.7% (145/265) were on regimens of one protease inhibitor and two nucleoside reverse transcriptase inhibitors compared with 33.3% (79/237) of those included. A further 29.1% (77/265) of those excluded were on regimens containing one non-nucleoside reverse transcriptase inhibitor and two nucleoside reverse transcriptase inhibitors, compared with 45.1% (107/237) of those included.

Table 1

We were only interested in CD4 cell count changes occurring when a patient had suppressed viraemia. Therefore patients were followed until either their viral load rose above 500 copies/ml on two consecutive occasions, at which point data were censored on the date of the first measurement above 500 copies/ml, or, for those who did not experience virological rebound during follow-up, until 2 days after their final viral load measurement. If a patient's final measurement exceeded 500 copies/ml but the previous viral load was less than 500 copies/ml, then the date of the last viral load less than 500 copies/ml was taken as the censoring date for the analysis.

Laboratory methods

CD4 cell counts were performed using standard flow cytometry techniques and plasma HIV-1 RNA measured by a variety of commercially available methods. When first introduced in 1996 viral load monitoring was performed using the Amplicor polymerase chain reaction HIV-1 Monitor test 1.0 (Roche Diagnostics, Roche Products Ltd., Welwyn Garden City, Herts, UK) and then later upgraded to the 1.5 version with add-in non-B primers. More recently the Cobas assay was used, which is equivalent to the Roche Assay.

Statistical methods

The CD4 cell count at 12 months after starting HAART was selected by taking a window from 9 to 15 months. The CD4 cell count measurement during this time interval that was nearest to one year was chosen. Those without a measurement in this window were considered to have no measurement at one year. Readings at months 6 (using a window from 3 to 9 months), 18 (15-21 months), 24 (21-27 months), 30 (27-33 months) and 36 (33-39 months) were chosen in a similar way.

All analyses were carried out on change from baseline. Initially, the change from baseline in CD4 cell count for every measurement taken on each individual was calculated. These measurements were then plotted against the time from starting treatment, disregarding the individual to whom the observation belonged. A locally weighted smoothed spline estimate of the overall trend (Lowess) [²²] was added to this graph to describe the overall behaviour of CD4 cell counts over time. Second, in order to provide an alternative and simpler presentation of the data, the median and interquartile ranges of the change in CD4 cell count at different timepoints defined by the windows described above were plotted. Although this plot illustrates the overall trends in CD4 cell count change, the presence of missing values means that the individuals contributing data to each timepoint may differ.

A mixed-effects regression model [²³] was used to examine the general trend in the changes in CD4 cell count over time. A mixed-effects model fits an overall mean pattern of change in CD4 cell counts over time, but it allows each individual's CD4 cell count to follow a path that randomly differs from this mean pattern. The model assumes that there is a normal error variance structure, but the covariance between observations is non-zero. This enables more than one observation from each individual at different timepoints to be included in the model, and so every CD4 cell count measured during follow-up was studied. No intercept was included, as the analysis is carried out on change from baseline. The change from baseline on the day of starting treatments is, thus, by definition, zero. Fixed terms, that is those assumed to be the same for each individual, were included that allowed the slope to change in gradient at 4 weeks (to allow for any initial redistribution effect), at one year and at 2 years after starting HAART.

As a measure of immune reconstitution, the percentage of patients who had attained CD4 cell counts of 500 cells/mm³ or greater was calculated at certain timepoints. Statistical tests were performed using SAS version 8.0 (SAS Institute Inc., Cary, NC, USA). Mixed level models were implemented using proc mixed. The Lowess plot was generated using an existing macro (http:|http://www.math.yorku.ca/SCS/sssg/lowess.html) All P values are two-sided.

Results

The median follow-up information on CD4 cell counts from the time of starting HAART in these 237 patients was 1.9 years [interquartile range (IQR) 1.27-2.72 years, range 0.82-4.49 years]. A total of 2559 CD4 cell counts were measured on the 243 individuals, with a median of 11 (range 1-36) measurements per patient.

A Lowess plot showing the changing CD4 cell count after starting HAART, is shown in Fig. 1. All 2559 CD4 cell counts have been used in Fig. 1. It demonstrates an overall increasing trend in CD4 cell counts. A plot of median change in baseline at different timepoints is shown in Fig. 2.

Fig. 1

Fig. 2

A summary of the change in CD4 cell count between years is shown in Table 2. Of 211 subjects with measurements at both baseline and one year, 198 (93.8%) experienced an increase in CD4 cell count from baseline. Overall, CD4 cell counts increased by a median of 197 cells/mm³ (IQR 111-296) during the first year of follow up. One hundred and nine out of 134 participants with values at years 1 and 2 (81.3%) experienced an increase from their 1-year count at their 2-year reading. The CD4 cell counts of a further 51/76 participants (67.1%) increased at 3 years compared with their 2-year reading. Therefore, 32.9% experienced a decrease in their CD4 cell count. Comparing the lower limits of the range and IQR shows that 25% of those with a CD4 cell count at both 2 and 3 years experienced a decrease of between 311 and 19 cells/mm³ in this time interval.

Table 2

The predicted mean curve obtained from the mixed-level model is shown in Fig. 3. All 2559 observed CD4 cell counts were included in this model. After an initial rapid increase in CD4 cell count in the first month after starting therapy, increases continued to occur, but at a less rapid rate. At each timepoint when the slope was allowed to change, a decrease occurred, although overall the CD4 cell count continued to increase. The initial slope is an increase of 97.2 cells/mm³ per month after starting HAART therapy [95% confidence interval (CI) 87.5-107.0] in the first month. There is a significant slowing down in the rate of increase in the CD4 cell count at 4 weeks (change in slope -85.7, -96.5 to -74.8; P ≤ 0.0001, Wald test) and at 2 years (change in slope from year 1 -3.7, -5.8 to -1.6; P = 0.0006). The change at one year was marginally non-significant (change -2.4, -4.9-0.1; P = 0.06) but was still included in the final model. After 2 years of therapy there was still an increase in CD4 cell counts of 5.4 cells/mm³ per month (95% CI 3.8-7.0).

Fig. 3

The sensitivity of the results to the timepoints chosen at which the slope was allowed to change was studied by fitting a further model in which the slope was allowed to change at 3 months (and not 4 weeks as before), one year and 2 years. As 4 weeks was chosen arbitrarily as the timepoint at which the initial rapid increase of CD4 cell counts ended, changing this timepoint to 3 months allowed the investigation of the effect varying this cut-off would have on the results. This resulted in a slower initial increase in the CD4 cell count, but after 3 months the rates of increase are similar to those seen in the main analysis described above.

In order to examine the extent to which the reduction in the rate of increase in CD4 cell counts after 2 years of HAART is caused by some patients reaching CD4 cell counts above 500 copies/ml, and thus in the range of CD4 cell counts seen in HIV-negative individuals [^24,25], the mixed-level model was repeated but with each patient's follow-up censored at the first occasion that their CD4 cell count reached or exceeded 500 cells/mm³. In this model, there was no significant change in the rate of increase in CD4 cell counts at either one or 2 years. From 2 years after commencing HAART onwards, the mean predicted rate of increase was 10.2 cells/mm³ per month (95% CI 5.2-15.2). Furthermore, of the 76 individuals with a CD4 cell count taken at both 2 and 3 years, 36 out of 43 (83.7) of those with CD4 cell counts below 500 cells/mm³ at 2 years compared with 15 out of 33 (45.5%) of those with CD4 cell counts above 500 cells/mm³ at 2 years experienced an increase in their CD4 cell count in this interval.

By the end of follow-up, CD4 cell counts had risen by a median of 319 (range -224-1546) cells/mm³ and CD4 cell counts had risen above 500 cells/mm³ in 116 out of 237 (40.0%) of those in the study.

Discussion

The results from this study imply that in those starting HAART who manage to maintain virological suppression, the number of CD4 T lymphocytes will continue to increase for at least 3 years after starting HAART. After an initial biphasic effect, the CD4 cell count continues to rise at a steady rate but the rate of this increase slowed after 2 years. This is partly to do with the fact that some patients have started to reach 'normal' levels.

Among HIV-negative individuals it is estimated that the mean (SD) [²⁵] of CD4 cell counts observed is 830 (290) cells/mm³. Another study [²⁴] found that the mean (95% CI) CD4 cell count for HIV-negative men and women was 710 (350-1440) cells/mm³ and 820 (450-1450) cells/mm³, respectively. As the CD4 cell counts of most individuals in the cohort being studied before the time of HIV infection are unknown, it is not possible to establish whether their CD4 cell counts have returned to 'normal' levels when on HAART. However, as the risk of clinical disease is very low in individuals with a CD4 cell count above 500 cells/mm³ [²⁶], and as this falls at the lower end of the range of CD4 cell counts expected in HIV-negative individuals, this value was taken as a measure of immune reconstitution in this study. As CD4 cell counts begin to return to levels above 500 cells/mm³, the rate of increase in CD4 cell levels would be expected to slow as CD4 cell counts return to those seen in HIV-negative individuals and remain stable [²⁷]. There is some evidence from the above analysis that this is the case.

Of those who had a CD4 cell count taken at 2 and 3 years, approximately 35% experienced a decrease in their CD4 cell count. As this does not take into account the magnitude of the fall in CD4 cell count, these changes could be caused by random fluctuations. However, as 25% of those with a CD4 cell count at both 2 and 3 years experienced a decrease of between 311 and 19 cells/mm³ (Table 2), some of those who were maintaining virological suppression and so remaining adherent to therapy were no longer sustaining immunological reconstitution.

We have considered the potential for immune reconstitution in those who maintain virological suppression. Therefore, we have, by the nature of our analysis, chosen a very select and highly adherent population. Clearly, these results are unlikely to be generalizable to those who do not maintain an undetectable viral load, but they emphasize the importance of good adherence, and reinforce the need to provide support to enable patients to be able to tolerate their medications.

Our mixed-level model of change from baseline CD4 cell counts was relatively simple and did not adjust for any potential confounders such as age, sex and treatment group. However, our aim was to describe the overall behaviour of CD4 cell count, and not to study potential predictive factors of immune response. Adjusting for such confounders did not lead to any substantial changes in the qualitative results, and as the study group were very selective and our interest was the overall pattern and potential for immune recovery, we felt that the choice of this model was appropriate.

Furthermore, our cohort started at a relatively high baseline CD4 cell count. The median baseline CD4 cell count was 175 cells/mm³ (IQR 70, 284 cells/mm³). As with other potential confounders, we wanted to describe the overall response to therapy among the whole cohort. However, when we investigated whether patterns of CD4 cell count increase varied in those who started therapy at different CD4 cell count levels, we found no evidence for a difference in response. In particular, those starting HAART at low CD4 cell count levels (< 100 cells/mm³) appeared to respond in a similar manner to those starting at higher CD4 cell count levels.

Our analysis only considered one marker of immune function, and did not examine other markers such as the CD8 cell count or activate markers. Furthermore, we studied the absolute levels of CD4 cell counts, but did not take into account whether these CD4 cells were naive or memory cells. Also, we did not examine changes in CD4 T cell reactivity in these patients over time [^28-30]. However, this information is not readily available, as it is not routinely collected in our database.

Our study defined virological suppression as maintaining a viral load of less than 500 copies/ml. However, viral replication may still be occurring in those with viral load measurements that would be detectable on assays with a lower limit of detection of 50 copies/ml. In order to gain a long follow-up period we included patients starting therapy before such assays entered into routine use, and so were not able to investigate this further.

Our findings suggesting the potential for immune reconstitution over such long time periods are in agreement with other studies. Gulick et al. [¹⁶] conducted a double-blinded trial in which 97 protease inhibitor-naive participants received zidovudine, lamivudine and indinavir with either simultaneous or sequential initiation. The study reported an increase of 209 cells/mm³ from baseline at 100 weeks in CD4 cell counts in these patients. Kaufmann et al. [¹⁵] studied 73 therapy-naive patients from an observational database who had received HAART for 2 years. They reported a median rise of 20 cells/mm³ a month during the first 3 months of treatment, followed by an increase of 7 cells/mm³ a month thereafter. At 2 years, 54% of patients had attained CD4 cell counts of more than 500 cells/mm³. These results are similar to those obtained in our study. Staszewski et al. [¹⁴] obtained similar results with an increase in CD4 cell counts from baseline in 154 previously therapy-naive patients treated at a hospital in Germany at 72 weeks of 143 cells/mm³. Research by Tarwater et al. [¹⁸], in a multicentre study of homosexual men, found that CD4 cell counts reached a plateau between 2 and 3.5 years after commencing potent antiretroviral therapy. Kaufmann et al. [¹⁷] also reported a steady slowing down in the increase in CD4 cell counts studied for 4 years, although CD4 cell counts continued to increase. However, their study studied 95 individuals, of whom only 51% were therapy naive. More recently [³¹], there is evidence of continued CD4 cell count gains after 4 years of therapy, although that study included previously naive patients starting HAART, some of whom did not maintain low viral load measurements.

These results suggest that immune reconstitution can continue for prolonged periods of time in those maintaining virological suppression. There is evidence that CD4 cell counts could return to those seen in HIV-negative individuals. Further follow-up of these patients will show whether this will occur in the majority of patients who maintain virological suppression.