ACTG protocol 116B/117 was a randomized, controlled, double-blind clinical trial that compared the efficacy of continued zidovudine (600 mg daily) with that of didanosine (either 750 or 500 mg daily of the sachet formulation) in patients who had tolerated zidovudine for a minimum of 16 weeks [16]. The working hypothesis for ACTG protocol 116B/117 was that the efficacy of zidovudine might diminish over time because of one or more of the following factors: HIV-1 resistance, intolerance to zidovudine, or zidovudine toxicity to lymphocytes. The clinical trial was designed to assess whether changing therapy from zidovudine to didanosine after at least 16 weeks of zidovudine monotherapy decreased the risk for disease progression. Progression to the main clinical outcome (a new AIDS-defining event or death) was delayed in patients who were randomly assigned to didanosine (500 mg daily) compared with those who continued to receive zidovudine [16]. However, the relative benefit of switching to didanosine therapy for the patients in ACTG protocol 116B/117 was not related to the duration of zidovudine therapy before the study [16].

We determined the prevalence of zidovudine-resistant HIV-1 at study entry and analyzed the relation of baseline HIV-1 isolate drug susceptibility, and HIV-1 isolate syncytium-inducing phenotypes, CD4⁺ T-lymphocyte count, HIV-1 disease stage, and treatment assignment to subsequent clinical progression of HIV-1 disease during ACTG protocol 116B/117.

Methods

Top Methods Results Discussion Author & Article Info References

Participants

A total of 913 patients who had received at least 16 weeks of previous zidovudine therapy were enrolled in ACTG protocol 116B/117. At entry, CD4⁺ T lymphocyte counts were 300 cells/mm³ or less for patients with a diagnosis of AIDS or AIDS-related complex and 200 cells/mm³ or less for patients with asymptomatic HIV-1 infection [16]. A previously undiagnosed AIDS-defining event [17] or death was the primary treatment efficacy end point in ACTG protocol 116B/117 [16]. One hundred eighty-seven patients had baseline HIV-1 isolates characterized in this virology study.

Virus Isolation

Human immunodeficiency virus type 1 was isolated from the peripheral blood mononuclear cells of the first five patients at each of the 36 clinical sites at the time of enrollment in ACTG protocol 116B/117. The HIV-1 isolates were obtained by cocultivation of patients' peripheral blood mononuclear cells with phytohemagglutinin-stimulated peripheral blood mononuclear cells from donors seronegative for HIV-1. Cell-free, supernatant fluids from HIV-1 core (p24) antigen-positive cultures were stored at –70°C [18]. The peripheral blood mononuclear cells of the second five patients at each clinical site were cryopreserved and not immediately cultured, according to the protocol. At some sites, these samples were also cultured in real time. After the trial had been completed, frozen samples from 255 patients were shipped on dry ice using overnight delivery to five laboratories participating in this virology study; 188 samples were culture supernatant fluids and 67 samples were cells. Each laboratory expanded virus stocks in a similar manner in cultures of phytohemagglutinin-stimulated peripheral blood mononuclear cells in the absence of drug; a maximum of two passages was allowed beyond initial cocultivation. Baseline virus stocks from 187 of these patients were characterized in the five laboratories. Recovery rates in the five laboratories ranged from 73% to 100% for culture supernatant fluid samples and from 46% to 56% for cell samples.

Drug-Susceptibility Assay

A standardized, previously validated method was used for the virus stock infectivity titration (expressed as 50% tissue culture infectious doses per milliliter [TCID₅₀/mL]) and for the drug-susceptibility assay in each of the five laboratories [19]. Human immunodeficiency virus type 1 stocks from 170 patients gave a Spearman-Karber infectivity titer of 2000 or more TCID₅₀/mL and were suitable for drug-susceptibility testing; stocks expanded from 17 patients had an insufficient viral titer for the drug-susceptibility assay. A standardized inoculum of 1000 TCID₅₀/10⁶ cells was then used in the second step that involved triplicate cultures either in the absence of drug (untreated controls) or at each of five different concentrations of zidovudine (0.001 to 5.0 µmolars) or didanosine (0.1 to 31.6 µmolars). Human immunodeficiency virus type 1 core (p24) antigen production was measured in culture supernatant fluids by enzyme-linked immunosorbent assay (ELISA) after 7 days in culture [19], and the 50% inhibitory concentrations of zidovudine and didanosine were calculated [20]. A reference strain of zidovudine-resistant virus (A018C) was tested in parallel in each assay [1]. A positive control for in vitro antiviral inhibition, the HIV-1 protease inhibitor saquinavir [21, 22], and drug toxicity controls were each included. Each of the five laboratories also evaluated a blinded panel of HIV-1 isolates, including one pair of reference isolates, A018A and A018C [1]. The 50% inhibitory zidovudine concentrations determined in the different laboratories ranged from 0.001 to 0.06 µmolars for A018A and from 2.2 to 5.0 µmolars for A018C.

Syncytium-Induction Assay

A 96-well microtiter plate format was used to assess syncytium-inducing phenotype, defined as cytopathic effect in the MT-2 cell line, by cultivation of cell-free virus (200 TCID₅₀) with 5 x 10⁴ MT-2 cells [23, 24]. A control virus with the syncytium-inducing phenotype (A018C [1]) and a negative control (no virus) were run in parallel. The cultures were examined microscopically every 3 days for up to 14 days. Syncytium-inducing viruses had three or more multinucleated giant cells/well.

Statistical Analysis

We compared patients who did and did not reach a clinical end point during ACTG protocol 116B/117 [16] using baseline clinical, virologic, and immunologic markers. Progression was defined as development of a new AIDS-defining event or death, whichever occurred first [16]. Associations of baseline markers with death were also analyzed separately. Relative hazards for progression or death associated with baseline clinical, virologic, and immunologic factors were determined from Cox proportional-hazards regression models involving single (unadjusted relative hazard) or multiple (adjusted relative hazard) independent variables [25]. Factors evaluated included drug resistance and syncytium-inducing phenotype of a baseline isolate, CD4⁺ T-lymphocyte count (cells/mm³), disease stage (AIDS compared with AIDS-related complex or asymptomatic infection), and randomized treatment assignment (didanosine or continued zidovudine). Baseline HIV-1 isolates were categorized as highly resistant to zidovudine (50% inhibitory concentration of zidovudine 1.0 µmolars), as moderately resistant to zidovudine (50% inhibitory concentration of zidovudine 0.2 µmolars but < 1.0 µmolars), or as susceptible to zidovudine (50% inhibitory concentration of zidovudine < 0.2 µmolars). For variables treated categorically, a relative hazard indicates the rate of progression or death for patients with a specific level of that independent variable, relative to patients who did not have that level. Continuous variables, including CD4⁺ T-lymphocyte counts and HIV-1 p24 antigen values, were log_e-transformed before analysis; this transformation improved the fit of models that included these variables. The relative hazard for baseline CD4⁺ T-lymphocyte count was expressed per doubling of the marker level.

Multiple regression models controlled for additional factors, including duration of zidovudine therapy before patients were randomly assigned, the total cumulative previous zidovudine dose, baseline 50% inhibitory concentration of didanosine, quantitation of baseline serum HIV-1 p24 antigenemia (pg/mL), laboratory site of virology assays, and level of in vitro virus replication in the drug-susceptibility assay (measured by HIV-1 p24 antigen level in untreated control cultures). To determine whether treatment with didanosine had an effect regardless of the presence or absence of high-level zidovudine resistance at baseline, adjusted relative hazards for disease progression and death associated with assignment to didanosine were determined from multiple regression models that excluded patients with baseline isolates showing high-level resistance to zidovudine. To evaluate the associations among predictor variables, contingency table analyses were used for categorical variables [26], and Spearman correlation coefficients (r) were used for continuous variables. Differences in distributions of continuous variables were evaluated by Wilcoxon scores. All P values presented are two-sided.

Results

Top Methods Results Discussion Author & Article Info References

Patients in the Virology Study

Characteristics of the subgroup of patients whose HIV-1 isolates were analyzed were compared with those of all patients in ACTG protocol 116B/117. Pair-wise comparisons were made between the total number of patients in the trial (n = 913) and each of the three subgroups listed in Table 1 (255 patients from whom HIV-1 isolates were initially obtained, 187 patients whose HIV-1 isolates were tested for syncytium-inducing phenotype, and 170 patients whose HIV-1 isolates were tested for drug susceptibility). Rates of disease progression higher than those for the total number of patients in ACTG protocol 116B/117 were noted for patients from whom HIV-1 isolates were initially obtained (P = 0.0003), patients whose HIV-1 isolates were tested for syncytium-inducing phenotype (P = 0.001), and patients whose HIV-1 isolates were tested for drug susceptibility (P = 0.002). Patients for whom viral specimens were not available had a lower rate of progression than those from whom viral specimens were obtained (data not shown). Each subgroup listed in Table 1 had a lower median entry CD4⁺ T-lymphocyte count than did the overall sample, but these differences were not significant (P = 0.19 to 0.23 for the different comparisons). Other baseline characteristics were similarly distributed in each virologically studied subgroup, in the subgroup from whom viral specimens were not obtained, and in the entire protocol (Table 1 and data not shown). The distribution for treatment-arm assignment in the subset we studied was similar to that of all patients in the protocol [16]; 67% received didanosine and 33% received zidovudine in each group.

Prevalence of Zidovudine Resistance at Entry

Baseline HIV-1 isolates with high-level resistance to zidovudine (defined as 50% inhibitory concentration of zidovudine 1.0 µmolars) were found in 26 of 170 patients (15%; 95% CI, 10% to 20%), and baseline HIV-1 isolates with any degree of zidovudine resistance (50% inhibitory concentration of zidovudine > 0.2 µmolars) were found in 69 of 170 (41%; CI, 37% to 45%) patients. For patients with drug-susceptibility and immunologic data (n = 169), the CD4⁺ T-lymphocyte count at study entry was associated with zidovudine resistance; high-level zidovudine resistance was noted in 28% (19 of 67) of those who had less than 50 CD4⁺ T lymphocytes/mm³ but in only 7% (7 of 102) of those with 50 or more CD4⁺ T lymphocytes/mm³ (P < 0.001).

Other factors were not significantly associated with high-level zidovudine resistance, including cumulative dose and duration of zidovudine therapy before study entry. Zidovudine resistance was not significantly associated with syncytium-inducing phenotype. The frequency of syncytium-inducing phenotype was 76% among highly zidovudine-resistant isolates, 64% among moderately zidovudine-resistant isolates, and 66% among zidovudine-susceptible isolates (P = 0.6). The quantity of HIV-1 p24 antigen in serum obtained from patients at study entry, as determined by enzyme-linked immunosorbent assay, did not correlate with 50% inhibitory concentrations of zidovudine of baseline HIV-1 isolates (r = –0.07;P = 0.34).

Risk Factors

Among patients from whom drug-susceptibility results were obtained (n = 170), 46 previously undiagnosed AIDS-defining events and 38 deaths occurred during the trial. Progression to either of these end points occurred in 77% (20 of 26) of the patients with highly zidovudine-resistant HIV-1 isolates at baseline compared with 53% (23 of 43) of those with moderately zidovudine-resistant isolates and 41% (41 of 101) of those with zidovudine-susceptible isolates (P = 0.003). Forty-two percent (11 of 26) of patients with isolates showing high-level zidovudine resistance were assigned to zidovudine and 58% (15 of 26) of these patients were assigned to one of two didanosine doses (750 or 500 mg daily). The unadjusted relative hazard of high-level zidovudine resistance for progression to a new AIDS-defining event or death was 1.93 (CI, 1.17 to 3.21) (Table 2). Baseline CD4⁺ T-lymphocyte count, baseline isolate syncytium-inducing phenotype, and diagnosis of AIDS at entry were also each associated with more rapid progression in analyses that did not control for other factors (Table 2).

High-level zidovudine resistance increased the risk for progression in multiple regression analyses that adjusted for each of the other factors studied (adjusted relative hazards) (Table 2). The adjusted relative hazard of progression for patients with HIV-1 isolates showing high-level zidovudine resistance was 1.74 (CI, 1.00 to 3.03). Baseline CD4⁺ T-lymphocyte count and diagnosis of AIDS also predicted progression in this multiple regression analysis (Table 2). The increased adjusted relative hazard for patients with highly zidovudine-resistant baseline HIV-1 isolates was not altered when the criteria for progression were expanded to include a recurrent AIDS-defining event and either a new AIDS-defining event or death.

High-level zidovudine resistance, baseline CD4⁺ T-lymphocyte count, syncytium-inducing phenotype, and AIDS diagnosis at study entry were each also predictive of death as single factors (Table 3). High-level zidovudine resistance remained associated with death when adjustments were made for each of the other factors (adjusted relative hazard, 2.78; CI, 1.21 to 6.39) (Table 3). High-level zidovudine resistance was also significantly associated with death in a different multiple regression analysis limited to didanosine-treated patients (adjusted relative hazard, 3.79; CI, 1.26 to 11.41; adjusted for baseline CD4⁺ T-lymphocyte count, syncytium-inducing virus phenotype, and AIDS diagnosis). The relative hazard for death among patients with HIV-1 isolates having a syncytium-inducing phenotype at study entry also remained increased when adjusted for the other predictors (adjusted relative hazard, 3.29; CI, 1.06 to 10.26) Table 3, although the increased risk for clinical progression for syncytium-inducing phenotype lost statistical significance when adjustments were made for those factors (Table 2).

Several factors were not associated with progression to a new AIDS-defining event or death. Duration and cumulative dose of previous zidovudine therapy, baseline 50% inhibitory concentration of didanosine, baseline serum HIV-1 p24 antigen, and amount of p24 antigen in untreated control wells of the drug-susceptibility assay did not predict progression. Controlling for which laboratory did the virologic assays did not alter the risk for zidovudine resistance. The analysis suggested minimal potential for bias in zidovudine-resistance risk estimates because of differential loss to follow-up; no difference was noted between patients with baseline isolates showing high-level zidovudine resistance and others for the length of time patients received treatment (P = 0.84) or the length of time patients were in the study (P = 0.66).

In our subset of patients with baseline HIV-1 drug-susceptibility results (n = 170), patients randomly assigned to either dose of didanosine progressed more slowly than those assigned to zidovudine (unadjusted relative hazard, 0.59; CI, 0.41 to 0.85) (Table 2). Similar results were reported for the total sample in ACTG protocol 116B/117 (n = 913); only the 500-mg dose of didanosine showed clinical superiority to zidovudine in the earlier analysis [16] of all the patients. Although the benefit of didanosine did not reach statistical significance when adjusted for other factors in our study (Tables 2 and 3), the estimates of decreased risk associated with random assignment to didanosine were similar when the 26 patients with highly zidovudine-resistant baseline HIV-1 isolates were excluded from analyses (adjusted relative hazard of assignment to didanosine for progression, 0.76 [CI, 0.44 to 1.30]; adjusted relative hazard for death, 0.49 [CI, 0.22 to 1.09]).

Discussion

Top Methods Results Discussion Author & Article Info References

Our study is the first to show that high-level zidovudine resistance predicts accelerated disease progression or death in patients with advanced HIV-1 disease, after controlling for other predictors of progression. Patients with highly zidovudine-resistant HIV-1 isolates (50% inhibitory concentration of zidovudine 1 µmolars) at entry into ACTG protocol 116B/117 showed more rapid progression than patients having more susceptible isolates. Moderate levels of zidovudine resistance were not associated with disease progression or death (Tables 2 and 3). Patients with baseline high-level zidovudine resistance did not differ from the others in regard to duration of follow-up, indicating that differential duration of observation did not explain the associations between high-level resistance and disease progression or death in the proportional hazards regression models.

Previous reports [7-11] have described an association between high-level zidovudine resistance and clinical disease progression. However, one study [7, 8] did not control for baseline CD4⁺ T-lymphocyte count or other predictors, and a second study [9] adjusted only for baseline CD4⁺ T-lymphocyte count. One previous case–control study [14] with a limited sample size found a stronger association between syncytium-inducing phenotype and progression than was found between zidovudine resistance and progression. In contrast to the earlier investigators, we studied mortality alone as an end point and controlled for additional predictors of progression, including the syncytium-inducing phenotype.

For patients with high-level zidovudine resistance in our study, the increased risk for progression to either a new AIDS-defining event or death Table 2 or for death alone Table 3 was not significantly decreased when adjusted for other factors that were also associated with progression, including baseline CD4⁺ T-lymphocyte count, diagnosis of AIDS, syncytium-inducing phenotype, and treatment assignment. High-level zidovudine resistance is clinically important because it is predictive of death and the more broadly defined end point of either a new AIDS-defining event or death. Clinicians should note that the predictive ability of high-level zidovudine resistance in this group of patients was not explained by baseline host clinical and immunologic status or by virus syncytium-inducing phenotype. The low prevalence (15%) of high-level zidovudine resistance among HIV-1 isolates recovered from patients who had received a median of 14 months of zidovudine therapy suggests that duration of therapy is not an adequate marker for high-level zidovudine resistance.

Among HIV-1 isolates from our study patients with advanced disease, high-level zidovudine resistance was not associated with the syncytium-inducing phenotype, as had been found previously in a study [12] of a smaller sample of patients with less advanced disease. The previous studies [7-11, 13-15] evaluating the association of one of these factors (syncytium-inducing phenotype and drug susceptibility) with disease progression did not control for the possible contribution of the other factor to the association. In our study, an increased risk for death was associated with both syncytium-inducing phenotype and high-level zidovudine resistance after adjustment for the alternate virologic factor and for other variables.

Isolates of HIV-1 were not recovered from and characterized for every available specimen. Nevertheless, virus recoverability and stock expansion did not appear to bias analysis because each characteristic evaluated, including rate of progression, was similar between patients who had viral specimens obtained during ACTG protocol 116B/117 and those who had assays (syncytium-inducing phenotype and drug-susceptibility) done on their HIV-1 isolates in our study (Table 1). However, the subset of patients whose viral specimens were obtained during the trial had a higher progression rate than did the overall sample in ACTG protocol 116B/117 Table 1 and the subset of patients in whom specimens were not obtained (data not shown). A trend toward lower median entry CD4⁺ T-lymphocyte count was also noted among patients whose specimens were collected to isolate HIV-1 relative to the overall sample in ACTG protocol 116B/117. These findings suggest that the subgroup of patients whose specimens were analyzed for HIV had more advanced disease when compared with the entire sample in ACTG protocol 116B/117. Further investigation is necessary to determine whether the results of our study are applicable only to patients with such advanced disease or are generalizable to a broader group of patients infected with HIV-1, including patients with less advanced disease.

A decreased risk for disease progression was associated with switching patients who previously received zidovudine to didanosine treatment in unadjusted analyses of the entire group of patients in ACTG 116B/117 [16], in our subgroup of patients whose HIV-1 isolates were studied, and in a separate group with different characteristics [27]. Before our study, the leading hypothesis to explain this decreased risk was that didanosine benefit would be limited to patients failing zidovudine treatment who had HIV-1 isolates showing high-level zidovudine resistance. However, three results from our study were not consistent with this hypothesis. First, the number of patients with isolates showing high-level zidovudine resistance in our study was too small to explain the observed benefit of didanosine. Second, high-level zidovudine resistance was associated with an increased risk for death among patients assigned to didanosine. Third, the adjusted relative hazards of didanosine assignment were essentially unchanged when the few patients with highly zidovudine-resistant HIV-1 were excluded; the uncertainty in this latter analysis is evident from the relatively wide 95% CIs for the adjusted relative hazards. Taken together, these results suggest that patients with advanced HIV-1 disease may benefit from a switch from zidovudine to didanosine monotherapy, regardless of whether high-level resistance of HIV-1 to zidovudine is present and that laboratory assessment of zidovudine resistance is not necessary to decide when to switch patients from zidovudine to didanosine.

Other hypotheses to explain the mechanism underlying the relative benefit of didanosine therapy in patients previously treated with zidovudine require investigation. Plasma HIV-1 RNA copy number has been noted to decrease in patients who have received zidovudine and have switched therapy from zidovudine to didanosine [28, 29]. Further, such a decrease in plasma HIV-1 RNA copy number after didanosine initiation has been associated with delayed clinical progression in a preliminary analysis of patients in the ACTG 116B/117 protocol [29]. Didanosine may have optimal antiretroviral activity in cell types that are different from those in which zidovudine is most active [30]. A decrease in lymphocyte proliferation in the presence of zidovudine relative to didanosine has been noted in vitro [31] and may suggest another explanation for the relative benefit of didanosine therapy.

We do not understand why high-level zidovudine resistance was associated with an increased risk for death among didanosine-treated patients. Potential mechanisms may include greater immunologic damage, not fully reflected by CD4⁺ T-lymphocyte count in patients who had highly zidovudine-resistant HIV-1 for some time before entry to ACTG protocol 116B/117. Augmentation of didanosine resistance by zidovudine-resistance mutations may also be involved [32, 33].

Isolation of highly zidovudine-resistant HIV-1 from peripheral blood mononuclear cells did predict more rapid progression or death during treatment after adjustment for baseline virus syncytium-inducing phenotype and for host clinical and immunologic status. Identification of high-level zidovudine resistance did not appear practically useful as an adjunct to deciding whether to switch from zidovudine to didanosine monotherapy. However, the association with progression and death provides insight into mechanisms of pathogenesis and suggests that antiretroviral-resistance assays may have clinical utility in other situations. The potential for clinical utility requires further investigation in patients who have received zidovudine but are different from the patients whom we studied, for example, patients with less advanced HIV-1 disease or those given other secondary therapeutic options, such as combination therapies. Selective polymerase chain reaction and other genetic analyses of HIV-1 reverse transcriptase mutations may serve as more rapid markers for resistance to zidovudine and didanosine than did the phenotypic assay using peripheral blood mononuclear cells used in our study [10, 34-38]. Further clinical validation of virologic assays, continuing improvement in laboratory methods to identify HIV-1 drug resistance, and further understanding of mechanisms other than drug resistance that contribute to loss of antiretroviral efficacy may facilitate clinical application of HIV-1 monitoring in the future.

Presented in part at the IXth International Conference on AIDS, Berlin, Germany; 6 to 11 June 1993.

Abbreviation

TCID₅₀ : 50% tissue culture infectious doses