The AIDS Clinical Trials Group (ACTG) study 116B/117 from the National Institutes of Allergy and Infectious Diseases showed that patients who had been previously treated with zidovudine reached study morbidity and mortality end points statistically less frequently if, irrespective of the course of the disease, they were switched to didanosine rather than if they continued zidovudine [10]. The present trial was designed to examine the benefit of the clinical practice of switching from zidovudine to alternate therapy when patients manifest signs of progressive disease, such as new clinical syndromes or a marked decrease in CD4 counts. These signs are often interpreted as zidovudine failure, but their clinical significance and the benefit of abandoning zidovudine use are unclear.

Methods

Top Methods Results Discussion Author & Article Info References

Study Design

Our study was a randomized, double-blind, multicenter, parallel, comparative trial of zidovudine (Retrovir; Burroughs Wellcome Co., Research Triangle Park, North Carolina) and didanosine (Videx; Bristol Laboratories, Princeton, New Jersey) among patients who had tolerated zidovudine for at least 6 months and who had signs of disease progression. Initial randomization was stratified by baseline CD4 cell count (<100 cells/mm³ or 100 to 300 cells/mm³). A provision existed for a blinded, compassionate crossover after 3 months of therapy. All participating patients were followed after the treatment medication was discontinued until death or the final study analysis. The study was designed to have 80% power to detect a 50% decrease in hazard rate for new or increased clinical morbidity for 12 months among patients who were treated with didanosine compared with those who continued to receive zidovudine, using a two-sided significance level of 5%.

Patients

Eligible patients for this study were men and women who were seropositive for HIV, were 12 years of age or older, had AIDS-related complex or AIDS, and had two sequential pretreatment CD4 cell counts that, when averaged, were 300/mm³. In addition, volunteers must have received and tolerated 600 mg/d or more of zidovudine therapy for 6 or more months and must have shown at least one of the following criteria for clinical deterioration within 12 weeks before randomization: 1) a new or recurrent AIDS-defining event [11]; 2) involuntary weight loss of more than 5% of the body weight; 3) a decrease of 20 or more in Karnofsky score from the onset of zidovudine therapy; 4) unexplained temperature of 38.0 °C or more for more than 7 days despite adequate investigation; 5) appearance of new or recurrent oral hairy leukoplakia or oral candidiasis; 6) appearance of new or recurrent herpes zoster; 7) appearance of new or recurrent dermatologic afflictions [for example, psoriasis, molluscum contagiosum, or seborrheic dermatitis]; 8) appearance of chronic herpes simplex ulcers not responsive to acyclovir therapy; or 9) a 50% decrease in CD4 cells from the time of initiation of zidovudine therapy.

Patients were excluded if they had clinical evidence of acute or chronic pancreatitis in the last 2 years, moderate-to-severe (grades II to IV Schaumberg score) peripheral neuropathy, a history of seizures requiring medication, or other serious complications. Women had a negative serum pregnancy test, and all patients signed an institutional review board-approved document of informed consent.

Study Medication

Patients were initially randomly assigned to receive zidovudine capsules, 600 mg/d (100 mg six times daily while awake) or didanosine sachets, 750 mg/d, adjusted for weight (375 mg twice daily for patients 60 kg or 250 mg twice daily for those < 60 kg) and placebo capsules or placebo sachets according to treatment group. Four months after the beginning of the study, the formulation of didanosine was changed to a chewable and dispersible buffered tablet given 600 mg/d with adjustment for weight (300 mg twice daily for patients 60 kg or 200 mg twice daily for those < 60 kg), a dose with a pharmacokinetic profile similar to that of the 750 mg/d sachet [12].

Conduct of the Study

Baseline pretreatment studies included medical history; physical examination; CD4 counts; serum HIV p24 antigen levels; laboratory measures of hematopoietic, renal, and hepatic function; and electrocardiograms and chest roentgenograms. Patients were monitored on a biweekly basis for 12 weeks, then monthly, for clinical and laboratory variables of efficacy and potential drug toxicity. Patients were treated with study medication until they voluntarily withdrew, had severe, uncontrollable drug toxicity, were noncompliant with the protocol, or the study was terminated, whichever occurred first. If a study end point was reached or a severe, uncontrollable drug toxicity occurred 12 weeks or more after study entry, crossover to the alternate treatment arm was mandated. If patients discontinued study medication, they were reevaluated 4 weeks after discontinuation of dosing and were then followed periodically by clinic visit or telephone contact to determine their course.

Study End Points

Primary end points for evaluation of drug efficacy were evidence of increased clinical morbidity that was defined in the protocol as the occurrence of one of the following: 1) death; 2) development of a new, appropriately documented AIDS-defining event [11], excluding Kaposi sarcoma; or 3) development of two new or recurrent HIV-related opportunistic infections in concert with a decrease in the CD4 cell count below 50% of the entry level. Human immunodeficiency virus-related diagnoses included recurrent AIDS-defining events, oral candidiasis, multidermatomal herpes zoster, and oral hairy leukoplakia. Secondary end points defined in the protocol were measurements of HIV p24 antigenemia, CD4 cell counts, and drug intolerance. In addition, we analyzed the data by several nonspecified end points to facilitate comparison with other anti-retroviral studies.

Statistical Analysis

The method of Kaplan and Meier [13] was used to estimate the distribution of times to study end points and to estimate adverse events. Treatment group comparisons were based on the estimated hazard ratios of zidovudine to didanosine (relative risk [RR]) using a Cox proportional-hazards model stratified for baseline CD4 cell count (<100 or 100 cells/mm³) [14]. Analysis of primary and secondary efficacy end points was on the basis of intent to treat: Once randomly assigned to primary study therapy, patients were included in the analysis of efficacy by original drug assignment for as long as data could be collected on the occurrence of study end points and irrespective of changes in HIV-specific treatment. The association of study end points with potentially prognostic baseline characteristics (CD4 count, diagnosis of AIDS or AIDS-related complex) was examined using a Cox proportional-hazards model that included as a covariate the characteristic of interest. To look for uniform treatment effects across important subgroups, relative risk was calculated for subsets defined by baseline CD4 count (<100 or 100 cells/mm³, a stratification variable), diagnosis (AIDS or AIDS-related complex), duration of previous zidovudine exposure ( 12 months or >12 months), and the criteria for clinical deterioration required for entry while controlling, as appropriate, for baseline CD4 count. These subset analyses were not prespecified in the protocol and should be interpreted with caution [15].

For analysis of clinical and laboratory adverse events, the data were censored at crossover or 30 days after discontinuation of study therapy. Treatment group comparisons by longitudinal measurement of CD4 cells were done on values obtained up to 24 weeks on study using the generalized Wilcoxon test [16]. A "CD4 response" was defined as an increase of 25% and 25 cells or greater over the baseline CD4 cell count that was maintained for at least 4 weeks. A "p24 antigen response" was defined as a decrease of 50% from the baseline value or to below the limit of detection (30 pg/mL) that was maintained for at least 4 weeks. Treatment group comparisons by CD4 cell and p24 antigen responses were done by logistic regression analysis adjusted for baseline CD4 cell count [17].

Interim analyses of the study were done in October 1991 and April 1992 by an external data safety monitoring board. On the basis of the April 1992 analysis and the results of the AIDS Clinical Trials Group 116B/117 study [10], the Board terminated the study. Reported P values are not adjusted for the interim analyses.

Results

Top Methods Results Discussion Author & Article Info References

Study Sample

A total of 312 patients from 19 study sites at tertiary care medical centers were entered into the trial, of whom 160 were randomly assigned to receive didanosine and 152, to receive zidovudine. Patients entered the trial between June 1990 and February 1992, and the overall median duration of follow-up was 47 weeks. Analysis of follow-up showed that 84% of the patients who did not reach a primary study end point had their last visit within 4 months of the study termination.

The pretreatment characteristics of the patients are shown in Table 1. Most of the patients were white, gay men in their mid-30s. The median duration of previous zidovudine therapy was 18 months. Prophylactic medication for Pneumocystis carinii infection was taken by 85% of the patients, approximately half of whom used aerosolized pentamidine and half of whom used trimethoprim-sulfamethoxazole (Bactrim and Septra) or dapsone. The most common signs of clinical deterioration at entry were, in descending order, oral candidiasis or oral hairy leukoplakia, CD4 cell decrease, skin diseases, and weight loss. No statistically significant differences were noted in these variables between the two treatment groups.

As shown in Table 2, 40% of the didanosine recipients and 30% of the zidovudine recipients continued to receive primary therapy (randomly assigned treatment given at entry) until the decision was made to terminate the study. The reasons for discontinuing primary therapy are listed in Table 2. Approximately half of the patients stopped receiving all study therapy, whereas 10% of the didanosine-treated patients and 27% of zidovudine-treated patients crossed over to the alternate study arm. Overall, the median time patients received primary therapy was 29 weeks, a large proportion of the total duration of patient follow-up (median, 47 weeks). Sixty-seven percent of clinical end points occurred while patients were receiving primary therapy.

The rate of progression to primary study end points by treatment group is shown in Figure 1 and Table 3. Patients who continued to receive zidovudine had a 50% higher risk for developing a primary study end point compared with those who were switched to didanosine (RR = 1.5; CI, 1.1 to 2.0; P = 0.02). Using only development of a new AIDS-defining diagnosis or death as end points, the magnitude of the difference between treatment groups was similar but of borderline statistical significance (RR = 1.3; CI, 1.0 to 1.9; P = 0.09). One fifth of the patients in the study died. When death was used as the end point, no difference was noted between the two treatments.

View larger version (19K): [in this window] [in a new window]   Figure 1. Kaplan-Meier distributions of time to primary study end points by treatment group. Top. All patients (RR for zidovudine: didanosine =1.5; P = 0.02). Bottom. Patients with pretreatment CD4 cell counts 100 cells/mm3 or more (RR for zidovudine:didanosine = 2.2; P = 0.03). The numbers within the figure are the patients remaining in the study at different points in time. DDI = didanosine; ZDV = zidovudine.

When the results of treatment were examined by clinical diagnosis at entry, the findings for patients with AIDS (RR = 1.6; CI, 1.0 to 2.6; P = 0.06) and for those with AIDS-related complex (RR = 1.5; CI, 0.9 to 2.3; P = 0.09) were both consistent with the overall results (see Table 3). The CD4 cell counts at entry were highly predictive of the response to treatment. Patients with CD4 counts of 100 cells/mm³ or more had a marked benefit from switching to didanosine (RR = 2.2; CI, 1.1 to 4.4; P = 0.03), whereas those with counts of 100 cells/mm³ or less showed a much smaller effect (RR = 1.3; CI, 0.9 to 1.8; P = 0.22). Patients who received zidovudine for more than 12 months seemed more likely to benefit from didanosine (RR = 1.5; CI, 1.1 to 2.3; P = 0.02) than patients who received zidovudine for 12 months or less (RR = 1.2; CI, 0.6 to 2.3; P > 0.2). However, relatively few patients entered the study who had received zidovudine for 12 months or less (see Table 1). No apparent association was noted between relative risk and the type of prophylaxis used for P. carinii infection (data not shown).

Comparison of the two treatment groups by the different categories of recent clinical deterioration required for study entry showed a consistent benefit from didanosine therapy in most subgroups. The numbers of patients were too small within groups defined by each individual sign or opportunistic disease to draw firm conclusions. It appeared that, while controlling for baseline CD4 count, a greater risk for clinical end points (zidovudine:didanosine) was associated with relatively mild manifestations of disease progression at entry (oral infections, dermatologic disease, herpes zoster, or a decrease in Karnofsky score) rather than with more severe manifestations (AIDS-defining opportunistic infection, acyclovir-unresponsive herpes, unexplained fever, or a 50% decrease in CD4 count) (data not shown).

Surrogate Markers

The association between treatment groups and the time course of the CD4 cell count is shown in Figure 2. The median CD4 count decreased progressively from baseline among those who continued to receive zidovudine. For those switched to didanosine, CD4 counts remained above baseline for the first 12 weeks and then decreased. The longitudinal differences between the two treatment groups were significant for weeks 2 through 24 (P < 0.0001). The difference was even more striking among the subgroup of patients with CD4 counts of 100 cells/mm³ or more. Among didanosine recipients, 29% had a 25% and 25-cell increase in CD4 count over baseline compared with 10% of zidovudine recipients (P = 0.0001). No significant differences were noted between treatment groups in changes in the concentration of plasma p24 antigen (data not shown).

View larger version (20K): [in this window] [in a new window]   Figure 2. Median change in the CD4 cell count from pretreatment values by treatment group and time in study. Top. All patients. Bottom. Patients with pretreatment CD4 cell counts of 100 cells/mm3 or more. The numbers within the figure are the patients remaining in the study at different points in time. DDI = didanosine; ZDV = zidovudine.

The frequency of selected adverse events is shown in Table 4. A higher frequency of granulocytopenia occurred among zidovudine recipients (P = 0.005). Increases of serum amylase levels were twice as frequent among didanosine recipients, but this difference was not statistically significant (P = 0.15). Clinical pancreatitis was infrequent in both treatment arms. Neuropathy occurred to an equal extent in both groups.

Discussion

Top Methods Results Discussion Author & Article Info References

Our findings are consistent with those of the AIDS Clinical Trials Group 116B/117 study [10], but we extend their findings in several important ways. The 116B/117 trial studied patients who had received zidovudine for a median duration of 14 months and found a 40% decrease in the risk for progression only among the group that took 500 mg of didanosine per day ("low dose" didanosine). The 116B/117 trial found no significant benefit among those who took 750 mg of didanosine per day ("high dose" didanosine), the dosing level used in the present study, and found no apparent benefit among the subgroup of patients with AIDS. Despite the smaller size of the current study, we showed a benefit among patients with AIDS-related complex and with AIDS. Using comparable study end points (new AIDS diagnosis or death) and drug regimens, the overall magnitude of benefit appeared greater in our study (RR = 1.3; CI, 1.0 to 1.9) than in the high-dose arm of the AIDS Clinical Trials Group 116B/117 study (RR = 1.1; CI, 0.9 to 1.4). The current study included patients with a longer median duration of zidovudine therapy (18 compared with 14 months) and signs of progressive disease. Thus, our new findings might be explained by the selection of patients who were more likely to have developed zidovudine-resistant strains of HIV. The benefit of switching to didanosine, however, may be independent of zidovudine resistance [7].

The data from the AIDS Clinical Trials Group study 116A, a comparison of switching to didanosine with continuing to receive zidovudine among patients who had received zidovudine for less than 4 months, have recently been analyzed [18]. Overall, no significant difference was noted between the treatment groups. Among the subgroup of patients who had not previously received zidovudine, the zidovudine group did better, but among the subgroup who had previously received zidovudine for 8 to 16 weeks, switching to didanosine was associated with a more favorable outcome. Although an increasing duration of zidovudine therapy may be associated with a greater likelihood for emergence of zidovudine-resistant HIV, no fixed amount of time receiving zidovudine is likely to be optimal for all patients, and additional risk factors are needed to guide physician decisions. We have shown that patients with clinical signs of disease progression clearly benefit from a switch to didanosine. In our study and the AIDS Clinical Trials Group 116B/117 study, patients with greater CD4 cell counts were most likely to benefit, possibly because of a decreased viral load and less viral heterogeneity associated with greater CD4 counts.

The conclusions in this report must be considered with the understanding that comparisons across trials are potentially misleading. The subgroup analyses within this trial are important to address clinical questions but should be interpreted with caution [15]. Lastly, the present trial and other similar studies of antiretroviral agent chemotherapy [1-3, 10] have been notable for a high rate of discontinuation of primary study therapy, whether by design (mandated or optional crossover for disease progression or drug toxicity) or for unscheduled reasons (see Table 2). Decreasing the duration of treatment potentially decreases the ability of study protocols to identify important long-term drug benefits, particularly survival.

In this study, a strong association was noted between the CD4 count response and the clinical response. The positive effect of didanosine therapy on CD4 cell counts paralleled the clinical benefit of didanosine; thus, an increase in the median cell count occurred in the first 3 months of the trial, a difference in cell counts between treatment groups persisted throughout the trial, and a greater positive effect occurred among patients whose pretreatment CD4 counts were 100/mm³ or more. These findings were qualitatively similar to those seen in the AIDS Clinical Trials Group study 116B/117, but the magnitude of the early increase in CD4 counts among our patients who switched to didanosine (+10 cells) appeared greater than that seen in either their 500 mg/d or 750 mg/d didanosine treatment group (+2 and +5 cells, respectively) [10]. We determined that serial CD4 measurements could explain 59% of the didanosine treatment effect [19], a larger fraction than in the AIDS Clinical Trials Group study 116B/117, possibly because of the larger magnitude of the CD4 response in this study and the earlier development of differences in the rate of clinical end points (see Figure 1 and Figure 2).

Didanosine therapy was well tolerated, with only a 4% rate of clinical pancreatitis and a 13% rate of hyperamylasemia, neither of which were significantly different from the rates in the patients receiving zidovudine (see Table 4). Compared with patients in the AIDS Clinical Trials Group study 116B/117 receiving 750 mg of didanosine per day, [10], our patients receiving didanosine had apparent lower rates of pancreatitis and hyperamylasemia (annual rate of 13 compared with 4, and 30 compared with 13, respectively). Comparison of the two protocols failed to detect a simple explanation for this difference, such as entry criteria for alcoholism and previous pancreatitis. Because our trial was the first large clinical trial to use didanosine tablets, it is possible that didanosine tablets provoke pancreatitis less often than the sachet formulation used in the AIDS Clinical Trials Group study 116B/117. Twice as many patients in the 116B/117 trial used aerosolized pentamidine as in our study (81% compared with 45%). Among our patients with pancreatitis, 6 cases occurred among 144 (4%) patients using aerosolized pentamidine and 1 case occurred among 168 (0.6%) patients using trimethoprim-sulfamethoxazole or dapsone. These observations suggest that concurrent aerosolized pentamidine may enhance the risk for pancreatitis. Differences in the use of aerosolized pentamidine could partially explain the different rates of pancreatitis between the two studies.

One could speculate from our results that greater benefits would have been obtained if we had used a lower dose of didanosine [10]. However, the optimal dose of didanosine is unknown at present. This study, the AIDS Clinical Trials Group studies, and the European/Australian Alpha trials indicate that didanosine sachets in doses of 750, 500, and 200 mg/d all may be effective [10, 18, 20]. Similar to monotherapy with zidovudine, monotherapy with didanosine promotes didanosine resistance [9, 21-24]. Continued exploration of new therapeutic approaches is necessary to find a more effective and lasting treatment for HIV infection.

Appendix

The Bristol-Myers Squibb Study Group includes Tammy Becker, Frank J. Bia, Carol Bujwit, Haig Donabedian, Thomas G. Evans, Martha Fiellin, Suzanne Kaempfer, J. Paul O'Keefe, Lynn Lipton, Rebecca J. Mark, Gregg Ott, Larry G. Reimer, Kristen Ries, Karen Waterman, and Margaret M. West.