Deeks, Steven G.^a; Barbour, Jason D.^a,b; Grant, Robert M.^a,b; Martin, Jeffrey N.^a

Introduction

The recommended therapeutic approach to HIV infection is to use combination therapy to reduce plasma viremia to below the level of detection (< 50 copies HIV RNA/ml). Failure to achieve and maintain undetectable plasma viremia should result in a modification of the failing regimen [^1–3]. Although based on sound scientific principles [⁴], this aggressive therapeutic strategy may not be feasible for many patients. Current therapeutic regimens often lack sufficient potency to completely suppress viral replication, particularly in patients with drug-resistant virus. Even if effective options are available, their complexity and side effect profile make long-term adherence challenging. Finally, the expense of combination therapy and viral load monitoring may prevent the widespread application of standard therapeutic strategies in resource poor countries. Thus, clinicians are often faced with the challenge of continuing therapy despite evidence of ongoing viral replication.

Continued therapy in the face of incomplete viral suppression usually results in the emergence of a drug-resistant virus population. Theoretically, this should lead to progressively increasing levels of HIV replication and, as a consequence, progressive CD4 T-cell depletion [^1–3]. There are limited data, however, linking partial or incomplete viral suppression with combination antiretroviral therapy (virologic failure) with CD4+ T-cell depletion. Indeed, CD4 cell counts often remain elevated above pre-therapy levels during the initial phases of virologic failure [^5,6].

The durability of CD4 cell preservation during virologic failure is not known. Prior studies evaluating the relationship between the virologic and immunologic response to combination therapy had limited follow-up (often 1 to 2 years) [^5–7]. More importantly, these studies focused on the date therapy was initiated rather than on the date virologic failure occurred. Since virologic failure occurred at different times during treatment, the actual duration of virologic failure was not described.

To better characterize the immunologic consequences of continuing therapy despite virologic failure, we analyzed CD4 T-cell trends in patients experiencing a virologic failure of a protease inhibitor-based therapy (defined here as persistent plasma HIV RNA levels > 500 copies/ml). Our primary objective was to determine the durability of treatment-mediated CD4 T-cell gains during virologic failure. Our secondary objective was to quantify the relationship between viral replication and CD4 T-cell loss in the setting of partial viral suppression. We were particularly interested in evaluating the relative contributions of the absolute level of viremia achieved during treatment versus the treatment-mediated change in viremia from a pre-therapy set-point (`delta viral load').

Methods

Patients and measurements

This is an ongoing observational study of HIV-infected patients receiving protease inhibitor therapy at San Francisco General Hospital, an urban municipally-funded hospital. As previously described [⁵], we used an administrative database to identify patients who had been seen at least three times by a single clinician or who had enrolled in clinical trials. To be eligible for this analysis, these patients also (1) must have initiated a regimen containing at least one potent protease inhibitor prior to 1 January 1998 (indinavir, ritonavir, nelfinavir or saquinavir-soft gel capsule); (2) had an evaluable CD4 cell count prior to initiating therapy; and (3) remained on therapy for at least 16 continuous weeks. This study was approved by the Institutional Review Board of the University of California, San Francisco.

All measurements were obtained during the course of routine primary care, or as part of a clinical trial. Plasma HIV RNA levels measured during primary care visits were determined using the branched chain DNA (bDNA) technique (Quantiplex; Bayer, Emeryville, California, USA). HIV RNA levels measured in clinical trials were generally measured using a reverse transcriptase polymerase chain reaction assay (Amplicor; Roche Diagnostics, Branchburg, New Jersey, USA). For this analysis, all viral load determinations were converted to bDNA 2.0 equivalents, as previously described [⁵]. CD4 cell counts were determined according using flow cytometry. Only CD4 cell counts that had a simultaneous HIV RNA determination were used in this analysis. The clinical laboratory at San Francisco General Hospital follows Centers for Disease Control guidelines for performing CD4 T-cell determinations in HIV-infected patients [⁸] and follows National Committee for Clinical Laboratory Standards (NCCLS) for performing all clinical tests.

Study endpoints

Virologic failure was defined as a plasma HIV RNA > 500 copies/ml on at least two consecutive measurements. Isolated plasma HIV RNA levels > 500 copies/ml (i.e., a measurement of > 500 copies/ml accompanied by levels of < 500 both immediately before and after) were not considered evidence of virologic failure. For patients whose HIV RNA levels fell below 500 copies/ml, we defined the date of virologic failure as the first confirmed HIV RNA above 500 copies/ml. For patients whose plasma HIV RNA levels never fell below 500 copies/ml we used the visit closest to week 16 as the virologic failure date.

Immunologic failure was defined as the return of the peripheral CD4 cell count to below the pre-protease inhibitor baseline level after at least 16 weeks of continuous therapy. Patients must have had at least two consecutive CD4 counts below pre-protease inhibitor baseline, or at least two CD4 cell counts below baseline in any 16-week period. A single determination was considered a valid endpoint only if it was the last recorded CD4 cell count. The date of immunologic failure was defined as the date the CD4 cell count first returned to pre-therapy levels.

Statistical analysis

The distribution of time to virologic failure after initiation of therapy was estimated using the Kaplan–Meier method. For this analysis, day 0 was defined as the date protease inhibitor therapy was initiated. Patients were censored at the time they were lost-to-follow-up (defined as not having a plasma HIV RNA determination for 6 months or longer).

For patients experiencing virologic failure, the time to immunologic failure was analyzed with the Kaplan–Meier method. Only patients who had a CD4 T-cell count greater than their pre-therapy level at the time of virologic failure were considered in this analysis. Day 0 was defined as the date virologic failure was confirmed. We censored observations either when patients became lost-to-follow-up or when they initiated a successful salvage regimen (defined as HIV RNA < 500 copies/ml through week 16 of a salvage regimen). In a secondary analysis, we also censored data at the time therapy was discontinued for 16 weeks or more.

Proportional hazards regression was used to determine the association between the following factors and immunologic failure: (1) baseline (pre-protease inhibitor) CD4 cell count; (2) age; (3) absolute plasma HIV RNA level; (4) change in HIV RNA levels from a pre-protease inhibitor therapy baseline (defined as the plasma HIV RNA level following the time therapy was initiated minus the plasma HIV RNA obtained prior to therapy); and (5) treatment status (either on or off protease inhibitor therapy). The latter three variables were considered as time-dependent variables, meaning that the patient's most recent measurement or status was used in the model when estimating associations. CD4 cell count, absolute HIV RNA level, and change in HIV RNA were classified in categories whose boundaries were set according quintiles present on the date virologic failure was confirmed.

Proportional hazards regression was performed with Stata (version 5; Stata Corporation, College Station, Texas, USA). All other statistical analyses were performed using SAS (version 6.12; SAS Institute, Cary, North Carolina, USA).

Results

Patient characteristics

Of the 483 patients treated with a protease inhibitor-based regimen, 303 eventually experienced virologic failure. The median time to virologic failure was 1.2 years (Fig. 1) and 65% experienced virologic failure by 3 years of observation. At the time virologic failure was confirmed, 291 of 303 (93%) patients had a CD4 T-cell count that was greater than their pre-therapy levels. The remainder of this analysis focuses on these 291 patients (see Table 1).

Fig. 1

Table 1

The median CD4 T-cell count at the time of virologic failure was 224 × 10⁶ cells/l [interquartile range (IQR), 128 to 360 × 10⁶ cells/l] and the median plasma HIV RNA level was 3.74 log₁₀ copies/ml (IQR, 3.11 to 4.38 log₁₀ copies/ml). The median change in CD4 T-cell count at the time of virologic failure was +90 × 10⁶ cells/l (IQR, 36 to 175 × 10⁶ cells/l) and the median change in plasma HIV RNA levels was −0.94 log₁₀ copies/ml (IQR, −0.29 to −1.53 log₁₀ copies/ml).

Patients were observed in a state of continuous virologic failure for a median of 27.9 months. During this period, 128 patients (42%) stopped therapy for 16 weeks or more, and 89 patients (29%) switched to a salvage regimen and had a successful response (plasma HIV RNA < 500 through week 16). Forty-seven (15.5%) patients died after the onset of virologic failure. Thirty-one of these deaths were AIDS-related (wasting n = 8; non-Hodgkin's lymphoma, n = 7; progressive Kaposi's sarcoma, n = 5; HIV encephalopathy/progressive multifocal leukoencephalopathy, n = 3; invasive fungal disease, n = 2;Pneumocystis carinii pneumonia, n = 2; other, n = 4). The remaining 16 deaths were not associated with an AIDS-defining event (cirrhosis, n = 5; cancer, n = 3; suicide, n = 3; cardiac, n = 2; other, n = 3).

Immunologic failure after the onset of virologic failure

The median time to immunologic failure (return of CD4 to pre-therapy baseline) after the onset of virologic failure was determined. Patient data were censored at the time a successful salvage regimen was initiated. The median time to immunologic failure was 3.1 years (see Fig. 2a). This includes the observation of patients who discontinued therapy sometime after the onset of virologic failure. To determine the incidence of immunologic failure in patients who remained on antiretroviral therapy despite detectable viremia, we performed an analysis where patient data were also censored after therapy was discontinued therapy for 16 weeks or more (Fig. 2b). Only 36.8% of this group experienced a return of CD4 cell counts to pre-therapy levels by 3 years.

Fig. 2

Factors associated with immunologic progression after onset of virologic failure

We next examined the determinants of immunologic failure among those patients experiencing virologic failure. For the initial analysis, we censored patient data at the time a successful salvage regimen was initiated or at the time therapy was discontinued for 16 weeks or more. In an unadjusted analyses using measurements obtained at the date of virologic failure, both the absolute HIV RNA level and the change in HIV RNA levels from pre-therapy baseline (`delta viral load') were associated with subsequent immunologic failure (Fig. 3). When we examined time-dependent measurements in an unadjusted analysis, both virologic measurements – absolute plasma HIV RNA levels and change from pre-therapy baseline – were again associated with the risk of immunologic failure (Table 2). For example, compared to the reference group (> 1.68 log₁₀ decline in plasma HIV RNA levels), patients who had less than a 1.13 log₁₀ decline in plasma HIV RNA levels were at increased risk of immunologic failure. This risk became progressively greater as the ‘delta viral load’ became progressive smaller. We saw a similar dose–response relationship between absolute level of viremia and immunologic failure (Table 2).

Fig. 3

Table 2

In multivariable proportional hazards regression analysis where absolute HIV RNA levels and change in HIV RNA levels were again considered as time-dependent variables, there continued to be a strong independent association between the change in plasma HIV RNA from a pre-therapy baseline and the risk of immunologic failure (Table 2). Compared to the reference group (> 1.68 log₁₀ copies/ml decline in HIV RNA levels), patients who had less profound viral suppression had a higher risk of CD4 cell decline. This risk was most apparent if the degree of viral suppression was less than 0.69 log₁₀ copies/ml below baseline (relative hazard = 5.47, P = 0.04). The absolute level of viremia achieved during virologic failure had a less important effect on the risk of immunologic progression. Compared to patients with a plasma HIV RNA level < 2.98 log₁₀ copies/ml, only those with the very highest levels of viremia (> 4.5 log₁₀ copies/ml) were at increased risk of immunologic failure after adjustment for the change in change in HIV RNA levels from pre-therapy levels.

Finally, we performed a separate proportional hazards model assessing the determinants of immunologic failure in which patient data after discontinuation of therapy were included. Change in plasma HIV RNA levels, absolute HIV RNA levels and treatment status (on or off therapy) were considered in a time-dependent manner. Change in plasma HIV RNA levels remained a strong independent predictor of time to immunologic failure (data not shown). Discontinuing antiretroviral therapy was associated with immunologic progression even after adjustment for viral load (hazard ratio, 2.0; 95% confidence interval, 1.23 to 3.14;P = 0.004).

Discussion

We have previously reported that CD4 cell counts initially remain elevated in HIV-infected patients who fail to achieve and maintain an undetectable plasma viral load while receiving a protease inhibitor-based antiretroviral regimen [⁵]. Here, with a longer period of observation, we report that immunologic progression does occur during virologic failure, but only after prolonged periods. Among patients who failed to achieve durable viral suppression (HIV RNA < 500 copies/ml) there is a median delay of 3 years between the onset of virologic failure and the return of the absolute CD4-cell count to pre-therapy levels. The delay is longer if we excluded observations after patients discontinued therapy. Change in viral load from pre-therapy levels, and not the absolute level of viremia achieved is the most important determinant of CD4 cell changes among patients with virologic failure.

The optimal therapeutic approach to virologic failure remains unclear. One of the central unanswered questions pertains to the virologic goal of therapy in pre-treated patients with limited therapeutic options. Based on the concern that incomplete viral suppression will lead to the emergence of drug resistance, current guidelines recommend that therapy be switched as soon as virologic failure is confirmed, and that complete viral suppression remain the immediate goal of therapy [^1–3]. Many patients, however, do not have enough sufficiently potent agents remaining to achieve durable viral suppression. Even if salvage regimens exist, their cost, toxicity and/or complex dosing requirements may preclude a successful response. Finally, considering the high virologic failure rates commonly observed during salvage therapy [⁹], aggressively switching therapy may rapidly deplete a patient's future therapeutic options. Our data provide support for a more conservative strategy, particularly for those patients with limited therapeutic options. For patients achieving some degree of viral suppression below pre-therapy levels, continuing a well-tolerated regimen despite ongoing viral replication may be beneficial. Such an approach, however, must be weighed against the risk of continued viral evolution and the emergence of high-level drug resistance (which we did not examine in this study).

Patients in our study who remained on therapy had a more prolonged CD4 T-cell response than patients who discontinued therapy, and this effect was independent of the level of viral replication. This observation suggests that therapy selects for a virus with reduced ability to deplete CD4 T cells (i.e., reduced virulence or reduced pathogenicity) and that this altered phenotype occurs independent of the level of viral replication. These observations are consistent with recent data indicating that protease inhibitor-resistant virus has reduced ability to replicate in thymic tissue [¹⁰]. Theoretically, the reduced ability of resistant virus to inhibit T-cell regeneration in thymic tissue will result in durable CD4 T-cell benefit even as virus replicates at high titer.

There are limitations in our data. Foremost, this is an uncontrolled patient population, and we did not have the ability to purely observe the natural history of virologic failure. Patients who entered a state of virologic failure had several strategic options available, including remaining on a stable regimen, switching therapy or stopping therapy. It is likely that patients doing poorly were more likely to stop therapy, and hence in our analyses where we censored observations at the time patients discontinued therapy we likely underestimated the true incidence of immunologic failure. On the other hand, including patients who discontinue therapy likely overestimated the risk of immunologic progression during virologic failure, because CD4 T-cell counts usually decline rapidly after therapy is discontinued [^11,12]. Achieving an unbiased estimate of the true incidence of immunologic failure will require an experimental setting in which patients are randomized to immediate versus deferred switching.

In conclusion, immunologic failure of protease inhibitor therapy occurs among patients experiencing long-term virologic failure, but is delayed by the continued use of a partially effective antiretroviral regimen. Change in viral load from pre-therapy ‘set point’ is the single most important predictor of maintaining a CD4 cell increase. For patients with limited therapeutic options, strategies based on maintaining some degree of partial viral suppression may be warranted.

References