Didanosine (2',3'-dideoxyinosine) is a newer nucleoside analog that has been shown to be effective in vitro against HIV [9]. Didanosine has in vitro activity against viral isolates that have high-level resistance to zidovudine [10]. Early clinical trials showed that didanosine can have a persistent beneficial effect on surrogate markers of HIV infection, such as CD4 counts, p24 antigen levels, and constitutional symptoms [11-14]. The investigators who did these studies found that peripheral neuropathy and pancreatitis were the dose-limiting toxicities of didanosine. More recently, controlled studies have shown that a switch to didanosine can improve clinical outcome in persons with advanced HIV disease who have received zidovudine [15, 16]. More specifically, Kahn and colleagues [15] showed such a benefit in patients with AIDS or AIDS-related complex who were clinically stable while receiving zidovudine and who had CD4 counts of 300 cells/mm³ or less and in asymptomatic HIV-infected patients with CD4 counts of 200 cells/mm³ or less. Spruance and coworkers [16] showed a similar benefit in patients with CD4 counts of 300 cells/mm³ or less and signs of clinical deterioration while receiving zidovudine therapy.

No clinical data are available on the role of didanosine in stable patients in earlier stages of HIV disease who have received zidovudine. We therefore specifically compared the safety and efficacy of didanosine with that of continued zidovudine therapy in clinically stable HIV-infected persons who had CD4 counts between 200 and 500 cells/mm³ and had received zidovudine for at least 6 months. We hypothesized that a change to a second effective antiretroviral agent before the anticipated development of high-level resistance to zidovudine would prevent resistance and consequently delay the progression of HIV disease.

Methods

Top Methods Results Discussion Author & Article Info References

Study Design

Randomization was stratified by the study center and by the CD4 cell count at study enrollment (more than or less than 300 cells/mm³). Successfully screened patients were randomly assigned using computer-generated random numbers. Randomization was done at a central location to ensure that patients, research personnel, and pharmacists remained blinded to the treatment allocation. All study participants provided informed consent. The study protocol and informed consent were approved by the review boards of the participating institutions and by the Canadian HIV Trials Network (CTN), with which our study is registered as protocol CTN-002.

Patients

Eligible study participants were male and nonpregnant female patients 12 years of age or older. Other entry criteria were the following: 1) HIV infection documented by enzyme-linked immunosorbent assay; 2) two sequential prerandomization CD4 counts between 200 and 500 cells/mm³ obtained at least 72 hours apart within 30 days of randomization, with the most recent measurement done within 14 days of randomization; 3) zidovudine therapy received for at least 6 months before randomization at a dose of at least 500 mg/d for the month immediately preceding study entry; 4) zidovudine therapy at 500 mg/d or greater for at least 21 of the previous 26 weeks; 5) a Karnofsky performance status of greater than 60 at study entry; 6) a hemoglobin level greater than 85 g/L or a hematocrit greater than 0.25 (in the absence of blood transfusion in the preceding 2 weeks); 7) a neutrophil count greater than 0.75 x 10⁹/L; 8) a platelet count greater than 50 x 10⁹/L; 9) serum aminotransferase and alkaline phosphatase levels greater than five times the upper limit of normal; 10) a serum creatinine level greater than 1.5 times the upper limit of normal; 11) a serum uric acid level less than 530 µmol/L; and 12) a serum amylase level less than 2.1 times the upper limit of normal.

The following are the normal values for chemical variables: aspartate aminotransferase, as high as 0.67 µkat/L; alkaline phosphatase, 0.58 to 1.75 µkat/L; creatinine, 40 to 120 µmol/L; and amylase, 0.50 to 1.83 µkat/L. Study participants were required to take adequate birth control measures during the study.

The following were the exclusion criteria: 1) the presence of an uncontrolled AIDS-defining illness; 2) known or suspected pulmonary Kaposi sarcoma or Kaposi sarcoma requiring systemic cytotoxic chemotherapy; 3) grade II or greater dementia; 4) active substance abuse; 5) antiretroviral therapy other than zidovudine; 6) any use of biological-response modifiers or corticosteroids within 30 days of entry or therapy with ribavirin within 90 days of entry; 7) previous participation in studies involving didanosine or zalcitabine; 8) grade II or greater neurologic, allergic, or renal toxicities; 9) any history of pancreatitis, intractable diarrhea, or malabsorption; 10) unexplained seizures within the previous 6 months or need for anticonvulsant agents; 11] treatment with neurotoxic drugs within 30 days of entry; and 12) past or current heart disease or requirement for cardiac medication.

All study participants were encouraged to use prophylaxis for Pneumocystis carinii infection according to contemporary guidelines [17]. The use of megestrol acetate, foscarnet, aspirin, acetaminophen, nonsteroidal anti-inflammatory agents, oral acidifying agents, and oral acyclovir was discouraged. Treatment of opportunistic infections was permitted. In patients developing serious symptoms or laboratory abnormalities, study medications were withheld until the symptoms or laboratory abnormalities resolved; at this point, patients were encouraged to resume the study medication, according to a prespecified dose-reduction scheme.

Treatment Regimens

Zidovudine (Retrovir, Burroughs-Wellcome, Research Triangle Park, North Carolina) was provided in 100-mg capsules to be taken at a dosage of 600 mg/d divided into at least three daily doses. Didanosine (Videx, Bristol-Myers Squibb, Princeton, New Jersey) was provided in sachets containing 5.2 g of citrate-phosphate buffer and sucrose adjusted to yield a final net weight of 20 g. The contents of one sachet were to be dissolved in water and swallowed. Didanosine dosage was adjusted for weight: Patients weighing at least 60 kg received 500 mg/d in two divided doses; patients weighing less than 60 kg received 334 mg/d in two divided doses. The didanosine formulation was changed in October 1991 from 500- and 334-mg/d sachet formulations to 400- and 200-mg/d tablet formulations. Study participants were instructed to chew the didanosine tablets thoroughly either together or in rapid succession and then to rinse with approximately 120 mL of room-temperature drinking water, which was also to be swallowed. Alternatively, the two tablets were to be crushed and thoroughly dispersed in at least 120 mL of drinking water; this solution was to be drunk immediately, followed by approximately 120 mL of drinking water. Study participants were instructed to always take didanosine on an empty stomach, at least 2 hours after and 1 hour before meals. To maintain the double-blind nature of the protocol, patients assigned to receive didanosine were given identical zidovudine placebo, and those assigned to receive zidovudine were given identical didanosine placebo.

Follow-up

After completion of the baseline evaluation, patients were seen at biweekly intervals for the first 2 months and monthly thereafter. A safety profile, including a symptom-targeted questionnaire, hematologic assessment, and chemistry panel were done at each visit. The CD4 count and viral resistance studies were done at baseline; at weeks 2, 8, and 12; and every 3 months thereafter. Formal follow-up of this cohort, as per the study protocol, was completed on 12 October 1992. Long-term, off-protocol, follow-up information on survival, AIDS-defining illnesses, and CD4 lymphocyte counts was compiled on one occasion using standardized data collection forms in December 1993. This allowed us to collect additional follow-up information on all patients after study completion and information on the complete study period for patients who dropped out of the study.

Study End Points

The primary clinical end point was the occurrence of a new, previously undiagnosed AIDS-defining event (according to the revised 1987 criteria of the Centers for Disease Control and Prevention) or death [18]. Clinical end points were reviewed by study monitors at each clinical site and were confirmed in a blinded manner by the clinical end points committee.

Didanosine was licensed by the Food and Drug Administration in the United States and by the Health Protection Branch in North America in the fall of 1991. At that time, didanosine became the standard therapy in Canada for persons with zidovudine intolerance or disease progression despite zidovudine therapy. Thus, in late 1991, while the study remained blinded and before any data were analyzed, a 33% decline in CD4 counts from baseline was added as a primary study end point to maintain consistency with prevailing clinical practice.

Sensitivity Testing

Samples for testing sensitivity to the study drugs were obtained from 102 of 120 patients (85%) enrolled at five clinical sites who were preselected on the basis of logistic issues. Forty-nine of these patients received didanosine, and 53 received zidovudine. Sensitivity testing was done at a single central laboratory (Lady Davis Institute, Jewish General Hospital, Montreal, Quebec, Canada). A modified version of a previously published consensus protocol was used to determine phenotypic resistance [19]. Briefly, HIV was recovered from peripheral blood mononuclear cells at study entry and at regular intervals thereafter by cocultivation of patients' cells with phytohemagglutinin-stimulated cord blood lymphocytes, which were obtained from the Department of Obstetrics, Jewish General Hospital [20]. Reverse transcriptase activity and p24 antigen levels in clarified culture supernatants were determined every 2 to 4 days [8]. Determinations of 50% inhibitory concentrations (IC₅₀) for viruses isolated at the time of study entry and thereafter were calculated on the basis of reverse transcriptase and p24 antigen assays [21]. Data reported here, based on reverse transcriptase activity, did not differ significantly from data based on p24 antigen determinations. The success of routine viral isolation in the study population exceeded 90%. High-level resistance to zidovudine was prospectively defined as an IC₅₀ greater than 0.8 µm. An IC₅₀ greater than 15 µm was prospectively defined as indicating high-level resistance to this agent on the basis of previous experience with wild-type isolates studied in our laboratories that routinely possessed sensitivities of 3 to 5 µm. To prevent unblinding of the study, the results of sensitivity testing were reported after the protocol was completed.

Study Termination

A sample size of 213 patients per arm was calculated to detect a relative risk of 1.8 in the rate of progression to death or new AIDS-defining illness, with 80% power and a significance level of 0.05. The median time to death or a new AIDS-defining illness was assumed to be 5 years for patients receiving continued zidovudine therapy. Patients were expected to be accrued within 12 months and were followed for 24 months after the last patient was enrolled [22]. An interim analysis was planned at 12 months using the boundary conditions of Fleming and colleagues [23].

The planned interim analysis was completed in June 1992 by the Safety and Efficacy Review Committee of the Canadian HIV Trials Network, which recommended that the trial be continued. However, the results of a controlled trial showing an improved clinical outcome of didanosine compared with continued zidovudine therapy in patients who had received zidovudine and had low CD4 counts, AIDS-related complex, or AIDS had been disclosed in April 1992 [15]. Although our study involved patients with higher CD4 counts at entry, a substantial number of study participants had shown decreased CD4 counts or clinical evidence of disease progression; these characteristics placed them within the inclusion limits of the recent trial [15]. For many patients, open-label didanosine had become a more desirable option than continued therapy with the study drug. Therefore, after discussions among study participants, investigators, and sponsors, it was jointly agreed to stop enrollment on 12 July 1992 and to continue follow-up as per the protocol until 12 October 1992. These decisions were made while the study remained blinded and before the investigators examined the data.

Statistical Analysis

Baseline characteristics were compared between treatment groups with two-tailed t-tests for normally distributed continuous variables, Wilcoxon rank-sum tests for skewed continuous variables, and Cochran-Mantel-Haenszel tests of general association for categorical variables [24, 25]. Intention-to-treat analyses were done on the primary and secondary outcomes, both including and excluding follow-up data collected after study completion. Probabilities of progressing to AIDS-defining illness or death, drug resistance, and prespecified changes in CD4 count were estimated using Kaplan-Meier curves [26]. Differences in rates of progression to various end points between treatment groups were tested with log-rank tests. Relative risks for progression to AIDS-defining illness or death and prespecified changes in CD4 counts were estimated with Cox proportional-hazards models [27]. A CD4 count response was defined as a decline of 33% of the baseline value. Changes from the baseline CD4 count for individual patients were compared between treatment groups at 2, 8, 12, 24, 36, and 48 weeks of follow-up with two-tailed t-tests.

Generalized estimating equation models were used to assess the effect of treatment, week of follow-up, and baseline resistance to zidovudine on changes in CD4 count from the count at enrollment [28]. In secondary analyses, a confirmed 33% decline in CD4 count (that is, a decline in CD4 count confirmed by a second consecutive determination) and both single and confirmed 50% declines from baseline CD4 count were used. Statistical tests and counts of events for the study period incorporated data obtained up to and including week 48 (the end of the study period). Tests that encompassed the extended follow-up period included all available data.

Results

Top Methods Results Discussion Author & Article Info References

Two hundred forty-six patients were enrolled in the study from 10 Canadian centers between January 1991 and July 1992. Although 43 patients had minor eligibility violations regarding baseline CD4 count, all were included in the analysis. All patients had at least one CD4 count in the specified range within 30 days of randomization. After we excluded one patient in the didanosine arm who was found to be ineligible before any dosing because of a previous myocardial infarction, 245 randomly assigned patients were included in the analysis: 118 in the didanosine arm and 127 in the continued zidovudine therapy arm. All decisions on inclusions and exclusions were made before data were analyzed. Clinical data were available during the extended follow-up period for 227 patients (93%); data on CD4 counts were available for 215 patients (88%).

Baseline Characteristics

As shown in Table 1, the baseline characteristics of the two groups did not significantly differ. Study participants were primarily white (95%), male (94%), and homosexual or bisexual (84%) (mean age, 37 years). Sixty-six percent were asymptomatic, 30% had AIDS-related complex, and 4% had a diagnosis of AIDS at entry. The median CD4 count at entry was 320 cells/mm³; 43% of patients had CD4 counts less than 300 cells/mm³, and 57% had counts greater than 300 cells/mm³. The median duration of zidovudine therapy before randomization was 471 days, and the median daily average zidovudine dose was 530 mg/d.

Study Therapies

As shown in Table 2, significantly more patients in the didanosine arm received the assigned therapy for the entire study period (74% compared with 60%; P = 0.03). Patients assigned to receive continued zidovudine therapy were more likely to discontinue study therapy because of clinical deterioration (P = 0.007) or adverse drug reactions (P = 0.06). Three patients were lost to follow-up: two in the didanosine arm and one in the continued zidovudine arm. The median duration of therapy with the study drug was 322 days for the patients assigned to didanosine and 276 days for the patients assigned to continued zidovudine (P = 0.06). The median duration of follow-up (including extended follow-up) was 737 days for patients receiving didanosine therapy and 764 days for patients receiving continued zidovudine treatment (P = 0.44).

Concomitant therapies were frequently prescribed during the study. In particular, agents active against P. carinii, herpesviruses, and fungi were used by 60% of patients receiving didanosine and 66% of those receiving continued zidovudine therapy (P = 0.33). As shown in Table 2, the pattern of concomitant therapy did not significantly differ between treatment arms. This remained true even when specific medications were considered separately.

Antiretroviral therapy used during the extended follow-up period did not differ significantly between study groups. During the extended follow-up period, didanosine was used by 80 patients originally assigned to receive it and by 77 patients in the zidovudine group (P = 0.11). Zidovudine was used by 84 patients originally assigned to receive it and by 71 patients in the didanosine group (P = 0.66). Zalcitabine was used by 18 and 22 patients in the didanosine and zidovudine groups, respectively (P = 0.90). Stavudine and lamivudine were used by two patients each in each arm (P = 0.91).

Clinical Events

No patient died during the 48-week study. As shown in Table 3, nine new AIDS-defining illnesses developed during the study period; all but one were in patients receiving continued zidovudine therapy (relative risk, 7.9 [95% CI, 1.0 to 63.3; P = 0.02]). Kaplan-Meier plots of the probabilities of progressing to a new AIDS-defining illness or death for each treatment group are shown in Figure 1.

View larger version (15K): [in this window] [in a new window]   Figure 1. Kaplan-Meier plots of the probabilities of not progressing to a new AIDS-defining illness (ADI) or death for each treatment group during the study period. Tick marks represent censored patients. AZT equals zidovudine; ddI equals didanosine.

Seventeen patients died during the extended follow-up period. When new AIDS-defining illnesses and deaths were considered during the extended follow-up, 14 events (including 9 deaths) occurred in the didanosine group and 28 events (including 8 deaths) occurred in the continued zidovudine therapy group (relative risk, 1.9 [CI, 1.0 to 3.6; P = 0.05]). Kaplan-Meier plots of the probabilities of progressing to a new AIDS-defining illness or death, including the extended follow-up period, are shown in Figure 2.

View larger version (15K): [in this window] [in a new window]   Figure 2. Kaplan-Meier plots of the probabilities of not progressing to a new AIDS-defining illness (ADI) or death during the study period and the extended follow-up period. Tick marks represent censored patients. AZT equals zidovudine; ddI equals didanosine.

CD4 Counts

As shown in Figure 3, a change to didanosine led to a statistically significant increase in CD4 counts by week 2. Statistically significant differences in CD4 counts favoring didanosine were seen at all time points until the study was completed at week 48 (P < 0.01). Kaplan-Meier plots of the probability of having a confirmed decline in CD4 cell counts of 33% from baseline are shown in Figure 4. The number of events in each treatment group and the relative risk estimates are shown in Table 3. According to the log-rank test, patients in the continued zidovudine therapy group were significantly more likely to have decreased CD4 counts than patients in the didanosine group (P < 0.001). As shown in Table 3, similar conclusions were obtained for single 33% and 50% declines from baseline, both with and without extended follow-up data, and for confirmed 50% declines when extended follow-up data were included in the analysis. No statistically significant difference was seen between treatment groups in the frequency of confirmed 50% CD4 count declines during the study period, probably because of the small number of events.

View larger version (16K): [in this window] [in a new window]   Figure 3. Mean change from baseline CD4 cell count by treatment group during the study period. Bars represent 95% CIs. AZT equals zidovudine; ddI equals didanosine.

View larger version (16K): [in this window] [in a new window]   Figure 4. Kaplan-Meier plots of the probability of avoiding a confirmed 33% decrease in CD4 cell count from the enrollment value. Tick marks represent censored patients. AZT equals zidovudine; ddI equals didanosine.

Viral Sensitivity Studies

Viral sensitivity studies were done in 102 unselected patients: 53 in the continued zidovudine treatment group and 49 in the didanosine group. This subgroup did not significantly differ at baseline from the entire study population (data not shown). Twenty-eight percent of patients in the continued zidovudine therapy group and 21% of patients in the didanosine group had high-level (IC₅₀ more than 0.8 µm) resistance to zidovudine (P = 0.49) at baseline.

As shown in Figure 5, only one patient in the didanosine group developed new high-level resistance to zidovudine during the study. This patient had an IC₅₀ for zidovudine of 0.74 µm at baseline, which indicates that the patient had already nearly achieved the prospectively established threshold for high-level resistance (0.8 µm). In addition, it is highly unlikely that this patient continued receiving zidovudine therapy after randomization: Mean corpuscular volume was 106 fL at baseline, immediately decreased to within the normal range (82 to 100 fL) when the study therapy was initiated, and remained stable throughout the study period.

View larger version (14K): [in this window] [in a new window]   Figure 5. Kaplan-Meier plots of the probability of remaining free of resistance to zidovudine (AZT). Tick marks represent censored patients. ddI equals didanosine.

Among patients receiving continued zidovudine therapy, the cumulative probability of developing high-level resistance to zidovudine was 59% at 1 year (Figure 5). The rates of developing resistance significantly differed between the treatment groups (P = 0.01). Average IC₅₀ values for zidovudine progressively increased among patients assigned to receive continued zidovudine therapy. In contrast, IC₅₀ values for zidovudine decreased among patients assigned to receive didanosine therapy. Cross-resistance between study drugs was seen in less than 10% of patients in each group (P = 0.51). Didanosine resistance (IC₅₀ more than 15 µm) was rarely seen during the study, whether patients were receiving zidovudine or didanosine therapy. Nine patients had decreased sensitivity to didanosine (IC₅₀ more than 5 µm) at some point during the study. In seven of these patients, this decreased sensitivity was present at baseline. The other two patients had decreased sensitivity at 169 and 192 days of didanosine therapy, respectively (IC₅₀ in each case, 12.7 µm). The only case of high-level resistance to didanosine was documented at baseline in a patient who denied ever receiving didanosine therapy.

Longitudinal Analysis

We used generalized estimating equation models to assess the effects of treatment, the week of follow-up, and resistance to zidovudine at study enrollment on changes in CD4 counts from enrollment (Table 4). This model for longitudinal data accounts for the correlation among multiple CD4 counts from a single patient. We included in this analysis the 102 patients for whom viral sensitivity data were available. Similar to the results shown in Figure 3, this model showed that CD4 counts increased by approximately 50 cells/mm³ soon after study enrollment in patients receiving didanosine and began to decline after 24 weeks of follow-up. The CD4 counts of patients in the continued zidovudine arm were an average of 68.1 cells/mm³ lower at weeks 2 to 48 than those of patients receiving didanosine.

Any resistance to zidovudine at enrollment was associated with a decrease in CD4 counts during the study (change in CD4 count, –20.3 cells/mm³; P = 0.04). When the degree of resistance to zidovudine was included in the model, there was a trend toward greater decreases in CD4 counts in patients with high-level resistance to zidovudine at enrollment (there was a change in CD4 count of –26.8 cells/mm³, P = 0.10 for IC₅₀ more than equals 0.8 µm; and a change in CD4 count of –15.1, P = 0.16 for IC₅₀ of 0 to 8 µm). These differences, however, did not reach conventional significance levels. The duration of zidovudine therapy before enrollment, disease stage at enrollment, and CD4 count at enrollment were not found to be significantly associated with changes in CD4 count after we used generalized estimating equation models to control for treatment, resistance status, and time since enrollment.

Adverse Events

As shown in Table 5, abdominal pain was reported significantly more frequently in patients receiving continued zidovudine treatment. In addition, musculoskeletal pain, dizziness, and fatigue tended to be reported more frequently among patients receiving continued zidovudine therapy. Two cases of pancreatitis were diagnosed during the study period, 1 in each treatment arm. Twenty-three cases of neuropathy were seen among patients receiving didanosine, and 22 cases were seen among patients receiving continued zidovudine therapy. Leukopenia was more frequent among zidovudine recipients, hyperuricemia was more frequent among didanosine recipients, and there were trends toward fewer cases of anemia and more cases of hyperamylasemia among didanosine recipients. When only AIDS Clinical Trials Group grade III hyperamylasemia (2.1 to 5 times the upper limit of normal) was considered, 9 and 3 cases were encountered among patients in the didanosine and zidovudine groups, respectively (P = 0.11). No AIDS Clinical Trials Group grade IV hyperamylasemia was diagnosed during the study.

Discussion

Top Methods Results Discussion Author & Article Info References

Our results show that a change to didanosine among clinically stable patients with moderately advanced HIV disease who have tolerated zidovudine for at least 6 months leads to a decrease in the rate of progression of HIV disease and a sustained increase in the CD4 count. In addition to showing an improved clinical outcome, our data indicate that a change to didanosine decreases the chances of developing in vitro high-level resistance to zidovudine. Our findings also show that the benefit obtained during the controlled phase of the study continued for 1 year after the completion of the trial.

Our results extend those of Kahn and colleagues [15], who found that a change in treatment from zidovudine to didanosine can slow the progression of HIV disease among patients who have AIDS or AIDS-related complex and a CD4 count of 300 cells/mm³ or less or among asymptomatic patients who have a CD4 count of 200 cells/mm³ or less. Recently published results of the intensive virologic monitoring done in this same study showed that high-level resistance to zidovudine at entry was independently associated with more rapid clinical progression and death, even with adjustment for other important factors, such as CD4 count or viral phenotype [29]. This effect persisted whether patients were randomly assigned to receive continued zidovudine or didanosine therapy despite cessation of the selection pressure naturally exerted by the drug [8, 30]. Our results further support this idea—we showed that resistance to zidovudine at baseline negatively affected CD4 count over time after adjustment for treatment effect. In this context, it has been proposed that HIV strains highly resistant to zidovudine may have increased virulence. Alternatively, the level of resistance at baseline could merely act as a surrogate marker for increased viral load. Whatever the underlying mechanism, the development of therapeutic strategies aimed at decreasing the rate of development of resistance to zidovudine seems desirable. Our results provide an important new insight into this issue and show that a change from zidovudine to didanosine at an earlier stage of HIV disease than that suggested by Kahn and colleagues (that is, when the prevalence of high-level resistance to zidovudine was relatively low) led to a reduction in the rate of development of high-level resistance to zidovudine, a decrease in the IC₅₀ to zidovudine, and an improved clinical outcome.

Recently, Spruance and colleagues [16] reported that patients with CD4 cell counts of less than 300 cells/mm³ and evidence of clinical deterioration while receiving zidovudine had fewer HIV-related clinical events when treatment was changed to didanosine. This effect was enhanced among patients with CD4 counts greater than 100 cells/mm³. We extend these results by showing that a change to didanosine leads to improved clinical and laboratory outcomes in patients with higher CD4 counts who are clinically stable while receiving zidovudine. When taken together with these recently published reports, our data also indicate that the magnitude of the didanosine effect on surrogate markers is greater when this strategy is applied at earlier stages of the disease [15, 16, 31].

Dolin and colleagues [32] recently reported the results of a clinical trial comparing didanosine with zidovudine among patients who had received zidovudine therapy for at least 4 months before study entry and patients who had received no previous zidovudine therapy. Although no statistically significant differences were found between treatment groups in the overall analysis, zidovudine was superior to didanosine therapy in patients who had not previously received zidovudine. Among patients who had received zidovudine for 8 to 16 weeks, however, a change to didanosine led to a more favorable outcome. These data suggest that although zidovudine is more effective among persons who have not previously received nucleosides, changing to didanosine after 8 to 16 weeks of zidovudine therapy leads to an improved outcome. The mechanism responsible for these differences is still unclear. Our data provide no further insight into the best time to switch from zidovudine to didanosine because we found that the extent of previous zidovudine therapy did not influence our results. We only included patients who had previously received zidovudine therapy for at least 6 months. This decision could have precluded us from identifying a threshold, if one exists, within the first 6 months of zidovudine therapy.

Recent evidence suggests that the simultaneous combination of zidovudine and didanosine can have a greater effect on surrogate markers of progression to HIV disease. Specifically, Collier and coworkers [33] recently reported that combination therapy with varying doses of zidovudine and didanosine produced larger and more sustained increases in CD4 cell counts, more frequent decreases in plasma HIV RNA titers, and more stable hematologic measurements than zidovudine therapy alone. Similarly, Yarchoan and associates [34] reported results of a pilot study that compared alternating full doses of zidovudine and didanosine at three weekly intervals with the simultaneous administration of a half dose of zidovudine plus a half dose of didanosine among symptomatic persons with HIV infection. Results of this study suggest that although adverse effects were generally similar between regimens, simultaneous combination therapy provided a greater and more sustained increase in CD4 cell count and body weight. The same investigators have also reported preliminary findings indicating that the simultaneous regimen was associated with a greater and more prolonged decrease in viral load. Ongoing studies assessing the effect of different combination therapy regimens on clinical outcomes are expected to provide definitive answers on the relative merits of monotherapy, sequential monotherapy, alternating therapy, or simultaneous combination therapy regimens.

In our study, both medications were generally well tolerated. Patients who would have shown early intolerance to zidovudine were excluded from our study because of the eligibility criteria. However, the rate of discontinuation of therapy with the study drug was greater among patients randomly assigned to receive zidovudine therapy; even discontinuation of therapy attributed to adverse drug reactions was twice as frequent in this arm of the study. We identified no unexpected adverse effects associated with either study medication. In keeping with previous reports, bone marrow dysfunction, as shown by the development of anemia and leukopenia, was more frequently seen among patients receiving zidovudine, whereas hyperuricemia and hyperamylasemia were more commonly associated with didanosine therapy [15, 16, 35].

As is often the case in long-term therapeutic HIV trials, a substantial number of patients withdrew from the study. We attempted to account for this effect by obtaining follow-up data outside of the protocol-required visits and extending the follow-up beyond the termination of the study. This allowed us to confirm the clinical benefit shown by our data and further document its persistence beyond the second year of follow-up. However, the actual estimates of the magnitude of this effect may be conservative because systematic biases are probably present in the context of any study such as ours. For example, patients whose condition deteriorates are less likely to be available for long-term follow-up and therefore will not contribute to the estimate of the difference between treatment groups. In addition, dropouts would tend to increase the mean CD4 count of the group, thereby indicating that the treatment could sustain the CD4 count to a greater extent than would be the case if failures could have been accounted for [36].

In summary, our results show that a change to didanosine leads to a lasting benefit even in asymptomatic patients with stable disease who have previously received zidovudine therapy for at least 6 months. The extent to which this benefit could be improved with the use of simultaneous combination regimens must be established.

Appendix

The following are additional members of the Canadian HIV Trials Network 002 Study Group: Clinical site investigators: Mary Fanning, Richard Lalonde, Michel Bergeron, Walter Schlech, Irving Salit, and Peter Phillips; Laboratory investigators: Bonnie Spira, Brian Conway, Sharon Cassol, and Michael O'Shaughnessy; Data analysis: Anona Thorne and Joel Singer; Bristol-Myers Squibb, Canada: Claude Auclair.

Dr. Rachlis: Sunnybrook Medical Centre, Room 226, 2075 Bayview Avenue, Toronto M4N 3M5, Ontario, Canada.

Dr. Gill: Southern Alberta (HIV) Clinic, Foothills Hospital, Calgary T2N 2T9, Alberta, Canada.

Dr. Beaulieu: Hotel Dieu de Montreal, 3840 St. Urbain, Montreal H2W 1T8, Quebec, Canada.

Dr. Tsoukas: Montreal General Hospital, Room 7112, 1650 Cedar Avenue, Montreal H3G 1A4, Quebec, Canada.

Dr. Raboud: Canadian HIV Trials Network, St. Paul's Hospital, University of British Columbia, 200-1033 Davie Street, Vancouver V6E 1M7, British Columbia, Canada.

Dr. Cameron: Department of Infectious Disease, Ottawa General Hospital, 501 Smyth Road, Ottawa K1H 8L6, Ontario, Canada.

Drs. Salomon and Wainberg: Lady Davis Institute for Medical Research, Jewish General Hospital, 3755 Chemin Cote Ste. Catherine, Montreal H3T 1E1, Quebec, Canada.

Drs. Dunkle and Smaldone: Bristol-Myers Squibb, Research and Development, 102-5 Research Parkway, Wallingford, CT 06492-7660.