Certain three-drug regimens, including the combination of zidovudine, didanosine, and a non-nucleoside reverse transcriptase inhibitor such as nevirapine, inhibit wild-type HIV-1 in vitro more effectively than one or two of these drugs [12-15]. Nevirapine (400 mg/d), as monotherapy or in combination with zidovudine or zidovudine and didanosine, has resulted in sustained antiviral effect in vivo, although nevirapine-resistant virus was isolated from patients in open-label trials [16-18] and from in vitro selection experiments [19, 20]. No adverse drug-drug interactions were noted in patients receiving the triple combination of nevirapine, zidovudine, and didanosine over a short period [18]. Therefore, we tested whether this triple combination would improve immunologic and virologic effects in vivo by comparing it with a combination of zidovudine and didanosine in a 48-week phase II, randomized clinical trial in adults with HIV-1 disease who had previously received prolonged nucleoside therapy.

Methods

Top Methods Results Discussion Author & Article Info References

Study Design and Treatment Regimens

This multicenter, randomized trial was AIDS (acquired immunodeficiency syndrome) Clinical Trials Group Protocol 241. Patients received either nevirapine with open-label zidovudine and didanosine (the triple combination) or placebo with open-label zidovudine and didanosine (the double combination). The placebo tablets were identical in appearance to the active nevirapine tablets. Patients and investigators were blinded to treatment assignments; a permuted-blocks design within each institution was used for randomization. Patients were stratified according to their screening CD4 cell counts: 50 or fewer cells/mm³, 51 to 200 cells/mm³, and 201 to 350 cells/mm³.

Zidovudine (Retrovir, provided by Glaxo Wellcome, Research Triangle Park, North Carolina) was given orally as two 100-mg tablets three times a day. Didanosine (Videx, provided by Bristol-Myers Squibb, Wallingford, Connecticut) was given orally as two 100-mg chewable and dispersable tablets twice daily for patients weighing at least 60 kg and as one 100-mg tablet and one 25-mg tablet twice daily for patients weighing less than 60 kg. Nevirapine (Viramune, provided by Boehringer-Ingelheim Pharmaceuticals, Ridgefield, Connecticut) was given orally as a 200-mg tablet once daily for the first 2 weeks and thereafter as one 200-mg tablet twice daily. This schedule was used because a lead-in period with a lower daily dose of nevirapine had decreased the frequency of nevirapine-related rashes in earlier studies [17]. Data were gathered, analyzed, and interpreted independently of the pharmaceutical companies that provided the drugs, according to the standard operating procedures of the AIDS Clinical Trials Group.

Study Sample

Adults who had documented HIV-1 infection and CD4 cell counts of 350 cells/mm³ or less within 30 days of randomization and who had had previous nucleoside therapy with zidovudine, didanosine, or zalcitabine for at least 6 months were eligible. Eligibility criteria also included a Karnofsky performance status score of at least 70% within 30 days of randomization, a hemoglobin concentration of at least 91 g/L or more for men and 88 g/L for women, a neutrophil count of 1000 cells/mm³ or more, a platelet count of 75 000 x 10⁹/L or more, a serum creatinine concentration no more than 1.5 times the upper limit of normal, serum concentrations of alanine aminotransferase and aspartate aminotransferase no more than 3 times the upper limit of normal, and a serum amylase concentration no more than 1.5 times the upper limit of normal (unless the serum lipase concentration was 1.5 times the upper limit of normal or less). Patients were excluded if they were intolerant of zidovudine (at 500 or 600 mg/d) or didanosine (at 400 mg/d for tablets and 500 mg/d for sachets), had moderate peripheral neuropathy ( grade 2 according to the National Institute of Allergy and Infectious Diseases [NIAID] Division of AIDS Table for Grading Adult Adverse Experiences), had pancreatitis, or had previously used non-nucleoside reverse transcriptase inhibitors. The study was approved by the institutional review board at each institution, and all participants gave written informed consent.

Concomitant therapy with other antiretroviral agents, foscarnet, biological response modifiers, erythromycin, clavulanate-containing antibacterial agents, warfarin, phenytoin, or phenobarbital was not permitted. Prophylaxis for Pneumocystis carinii pneumonia was required for all patients with a CD4 cell count of 200 cells/mm³ or less or a history of P. carinii pneumonia. Maintenance therapies for other opportunistic infections were permitted. Patients could continue their prestudy antiretroviral nucleoside therapy until the day on which therapy with the study medications was started.

Management of Toxicities

Toxicity was graded according to the NIAID Division of AIDS Table for Grading Adult Adverse Experiences. Modification, interruption, or discontinuation of therapy with the study medications was done for many adverse effects one drug at a time, starting with the drug most likely to have caused the toxicity. Zidovudine was the first drug to be modified for severe anemia, myositis, neutropenia, fatigue, and headache; didanosine was the first to be modified for severe nausea, vomiting, diarrhea, constipation, hyperamylasemia, fasting hypertriglyceridemia, hyperuricemia, and peripheral neuropathy; and nevirapine or placebo was the first to be modified for rash, severe thrombocytopenia, and adverse neuropsychological effects. Therapy with all study medications was interrupted for other severe adverse experiences, including hepatotoxicity. Therapy with a medication or medications was interrupted until the toxicity resolved and was then resumed at half the original dose, except that didanosine therapy was permanently discontinued for pancreatitis and nevirapine or placebo use was permanently discontinued for severe rash. Recurrent or persistent adverse effects led to the permanent discontinuation of either zidovudine or didanosine therapy. All study treatment was stopped if therapy with nevirapine or placebo or any two study medications was permanently discontinued.

Patient Evaluation

The schedule of patient evaluations is shown in Table 1. Standardized assays were used for CD4 cell counts [21, 22] and serum p24 antigen levels (AIDS Clinical Trials Group Virus Quality Assurance Laboratory reference standards were used with enzyme-linked immunosorbent assay kits from either Coulter [Hialeah, Florida], Abbott Laboratories [North Chicago, Illinois], or DuPont [Wilmington, Delaware]). Plasma nevirapine concentrations were measured at Boehringer-Ingelheim Pharmaceuticals by a validated high-performance liquid chromatography assay [23]. Additional virologic and pharmacologic assessments were done on specimens from all patients enrolled at 8 of the 16 study sites (n = 198). The HIV-1 infectivity titers in peripheral blood mononuclear cells were calculated by using maximum likelihood estimation from a standardized HIV-1 quantitative microculture assay in quality-assured AIDS Clinical Trials Group virology laboratories [24]. Plasma HIV-1 RNA levels were measured by quantitative reverse transcriptase polymerase chain reaction (Roche Molecular Systems, Alameda, California, and Branchburg, New Jersey) done in batch at Roche Biomedical Laboratories (Research Triangle Park, North Carolina) [25]. Plasma zidovudine and didanosine concentrations were measured at the University of Alabama at Birmingham by validated radioimmunoassays [26, 27]. Viruses isolated at entry were classified as either syncytium-inducing or non-syncytium-inducing according to the results of an MT-2 cell assay [28].

Primary outcome measures were differences between treatment groups at the end of the study in mean absolute CD4 cell count, HIV-1 infectivity titer in peripheral blood mononuclear cells, serum p24 antigen level, and time to HIV-1 disease progression or death. Secondary outcome measures were differences between treatment groups at the end of the study in the mean percentage of T lymphocytes that were CD4 cells and in plasma HIV-1 RNA levels. These measurements at the end of the study were the geometric means of measurements obtained between weeks 40 and 48, expressed as the change from the geometric mean of the two measurements made before treatment. Standardized area-under-the-curve analyses of immunologic and virologic measures over time were additional secondary outcome measures used to assess differences between treatments that included short-term effects. Progression of HIV-1 disease was defined as the development of a new AIDS-defining clinical event according to the Centers for Disease Control and Prevention case definition (except that we excluded CD4 cell count as a disease progression criterion) [29], a newly diagnosed deep-seated bacterial infection or bacteremia unrelated to the use of injection drugs or an intravascular catheter, 1 month or more of symptomatic microsporidiosis, or the recurrence of either P. carinii pneumonia or central nervous system toxoplasmosis. All disease progression events, deaths, and potentially life-threatening rashes were reviewed in a blinded manner by the study chair.

Statistical Analysis and Interim Data Monitoring

Intention-to-treat analysis was used for all efficacy outcomes [30]. Following the intention-to-treat paradigm, four patients who had never received their assigned therapy because of study site pharmacy errors were analyzed as part of the group to which they had been randomly assigned. However, analyses done according to the treatment actually dispensed gave similar results. The differences between groups at the end of the study (on a log₁₀ scale) and standardized area-under-the-curve values for CD4 cell count, percentage of T lymphocytes that were CD4 cells, HIV-1 infectivity titer in peripheral blood mononuclear cells, and number of plasma HIV-1 RNA copies were compared between treatment groups by using linear regression and adjusting for study site [31]. For each patient, the standardized area-under-the-curve was formed by joining successive log₁₀-transformed measurements and assessing the area between this curve and the patient's baseline level, with standardization by the time between start of treatment and last measurement. Changes in the median log₁₀-transformed serum p24 antigen levels, as well as differences in the median plasma concentrations of zidovudine and didanosine, were compared between treatment groups by using the Wilcoxon test. In analyses of times to events, we used Kaplan-Meier estimators, log-rank tests, and proportional hazards models stratified by screening CD4 cell count with treatment assignment as a covariate [32]. Analyses of adverse events were censored 30 days after permanent discontinuation of treatment or at 366 days (48 weeks plus 30 days). Analyses of death were censored at 366 days or at time of loss to follow-up. All other analyses were censored at 336 days (48 weeks) or at the time of loss to follow-up. Subgroup analyses assessed the significance of the subgroup interaction with treatment using linear regression adjusted for study site and the subgroup main effect; these were exploratory because the study was not designed to assess a specified magnitude of difference between subgroups. The Fisher exact test was used to compare numbers of patients in each treatment arm who developed a particular severe adverse event.

An interim analysis of HIV-1 disease progression and adverse events was presented to the AIDS Clinical Trials Group Data and Safety Monitoring Board in February 1994. The Board recommended that the study be continued without modification; patients and investigators remained blinded to treatment assignments until the completion of the study.

Results

Top Methods Results Discussion Author & Article Info References

Study Sample

Four hundred patients were randomly assigned to treatment between May and July 1993; 25 patients were enrolled at each of the 16 study sites. Two patients did not receive study treatment, were not followed, and were excluded from analyses: One had refused treatment, and the other did not meet eligibility criteria because of an abnormal chest radiograph. Thirteen patients had minor eligibility violations (7 in the triple-combination group and 6 in the double-combination group). Throughout the study, 3 patients assigned to receive nevirapine were given placebo, and 1 patient assigned to receive placebo was given nevirapine because of study site pharmacy errors.

Selected characteristics of the patient sample before treatment are shown in Table 2. The median length of previous nucleoside therapy was 25 months, and 131 patients (33%) had received therapy for more than 36 months. Three hundred eighty-six patients (97%) had previously used zidovudine, and 187 (47%) had previously received didanosine as well as zidovudine, including 135 (34%) who had had previous simultaneous combination therapy with zidovudine and didanosine. In addition to receiving zidovudine, 263 patients (66%) had previously received didanosine or zalcitabine as either a component of a combination therapy or as sequential monotherapy.

Forty-seven patients (12%) were lost to follow-up; 21 (11%) were in the triple-combination group and 26 (13%) were in the double-combination group (P > 0.2). The rates of premature permanent discontinuation of treatment were similar in the two groups; 66 patients (33%) in the triple-combination group and 64 patients (32%) in the double-combination group discontinued treatment prematurely. No difference in the time to discontinuation was noted between the two treatment groups (P > 0.2). Rates of discontinuation were higher among patients with lower CD4 cell counts: Fifty-three percent of patients with 50 or fewer cells/mm³ compared with 28% of patients with 51 to 200 cells/mm³ and 24% of patients with 201 to 350 cells/mm³ discontinued treatment. Most discontinuations (86%) were patient initiated, and the percentage of discontinuations that were patient initiated did not differ between treatment groups. In 327 patients (82%), the CD4 cell count was measured at least once at or after week 40 for the evaluation of long-term change. Sixty-six of 89 patients (74%) had their serum p24 antigen levels measured at or after week 40. Of the 198 patients who participated in the virology substudy, 153 (77%) had their HIV-1 infectivity titers in peripheral blood mononuclear cells measured at least once, and 151 (76%) had their plasma HIV-1 RNA quantitated at least once at or after week 40.

Patient compliance with nevirapine therapy was evaluated by assaying nevirapine levels in 738 plasma specimens taken at random from 301 patients; this was feasible because nevirapine has a long half-life in plasma [17, 33]. Detectable levels of nevirapine (200 ng/mL) were found in 2 of 362 plasma samples (0.5%) from patients randomly assigned to the double-combination group and 334 of 376 plasma samples (91.5%) from patients assigned to the triple-combination group. One of the 2 patients who was randomly assigned to placebo and had nevirapine detectable in plasma had erroneously been given nevirapine throughout the study.

Concomitant nevirapine administration did not increase plasma concentrations of zidovudine or didanosine (Figure 1). Indeed, median zidovudine concentrations were significantly lower in the triple-combination group than in the double-combination group at three of the five measured time points after the simultaneous administration of all study medications (Figure 1). Median plasma didanosine concentrations did not differ between groups.

View larger version (33K): [in this window] [in a new window]   Figure 1. Median zidovudine and didanosine plasma concentrations after administration of the combination of zidovudine and didanosine with either nevirapine (black bars) or placebo (white bars). Top. Plasma concentrations of zidovudine. Bottom. Plasma concentrations of didanosine. Plasma samples were collected at visits during week 8 (between 0.25 to 0.5 hours and 1.0 to 1.5 hours after simultaneous administration of all drugs), week 24 (between 2.0 and 2.5 hours after simultaneous administration of all drugs), week 32 (between 3.5 and 4.0 hours after simultaneous administration of all drugs), and week 44 (between 7.0 and 8.0 hours after simultaneous administration of all drugs). The number of patients sampled is indicated within each bar, and the P value (determined using the Wilcoxon test) for the difference between treatment groups at each time interval is indicated above each set of bars.

CD4 Cell Counts

Mean absolute CD4 cell counts decreased to less than baseline by the end of the study in both treatment groups, although the decrease was smaller in the triple-combination group (Figure 2, top). The mean long-term absolute CD4 cell count was 15% lower than the baseline value for the triple-combination group and 33% lower than the baseline value for the double-combination group. Thus, an 18% difference (95% CI, 7% to 29%; P = 0.001) was seen between treatment groups in mean absolute CD4 cell count at the end of the study, favoring the triple combination. At the end of the study, 46% of the triple-combination group and 33% of the double-combination group had absolute CD4 cell counts that were higher than values before treatment. The difference between groups was established within 4 weeks of the start of study treatment and was sustained for 48 weeks (Figure 2, top). The secondary standardized area-under-the-curve analysis of absolute CD4 cell count also significantly favored the triple combination (P < 0.001). At the end of the study, the mean percentage of T lymphocytes that were CD4 cells was 5% less than the baseline value for the triple-combination group and 20% less than the baseline value for the double-combination group, for a difference between treatments of 15% (CI, 8% to 24%; P < 0.001) favoring the triple combination.

View larger version (15K): [in this window] [in a new window]   Figure 2. Mean changes from baseline, by treatment group. Top. Mean changes in CD4 cell counts as percentage of baseline count. Middle. Mean changes in peripheral blood mononuclear cell human immunodeficiency virus type 1 (HIV-1) infectivity titers in log10 infectious units per million cells (IUPM). Bottom. Mean changes in plasma HIV-1 RNA levels in log10 plasma HIV-1 RNA copies/mL. Dashed lines indicate the triple-combination (nevirapine, zidovudine, and didanosine) treatment group; solid lines indicate the double-combination (zidovudine and didanosine) treatment group. Bars represent the SE. The numbers of patients tested at each time point are indicated above the x-axis.

Quantitative Virologic Measures

The mean long-term titer of infectious HIV-1 in peripheral blood mononuclear cells was 0.34 log₁₀ infectious units per million cells less than the baseline value (a 54% reduction from baseline) for the triple-combination group and 0.02 log (₁₀) infectious units per million cells less than the baseline value (a 5% reduction from baseline) for the double-combination group (Figure 2, middle). Therefore, at the end of the study, a difference of 0.32 log₁₀ infectious units per million cells (CI, 0.05 to 0.59 log₁₀ infectious units per million cells; P = 0.023) favoring the triple combination was seen. Similarly, the mean long-term plasma HIV-1 RNA level was 0.10 log₁₀ copies/mL less than the baseline value (a 20% reduction from baseline) for the triple-combination group and 0.16 log₁₀ copies/mL higher than the baseline value (a 44% increase from baseline) for the double-combination group (Figure 2, bottom). The mean plasma RNA level was thus 0.25 log₁₀RNA copies/mL lower (CI, 0.03 to 0.48 log₁₀RNA copies/mL; P = 0.028) for the triple-combination group at the end of the study. In comparing the patterns of change over time in peripheral blood mononuclear cell HIV-1 infectivity titer (Figure 2, middle) and plasma HIV-1 RNA level (Figure 2, bottom), it is important to note that quantitative peripheral blood mononuclear cell microcultures were not done at weeks 4 or 32. The difference between groups in these quantitative virologic measures was established at the first measurement. For both peripheral blood mononuclear cell viral infectivity titers and plasma HIV-1 RNA levels, a similar rate of increase was seen in each treatment group after 16 weeks (Figure 2, top and middle). The standardized area-under-the-curve analyses also showed significantly less circulating virus in the triple-combination group (P = 0.034 for peripheral blood mononuclear cell HIV-1 infectivity titer and P = 0.004 for plasma HIV-1 RNA level). The long-term reduction in serum p24 antigen level was also larger for the triple-combination group (P = 0.020). The median long-term serum p24 antigen level was lower for the triple-combination group (28 pg/mL) than for the double-combination group (38 pg/mL).

Disease Progression or Death

Sixty-one patients (15%) had disease progression or died within 48 weeks of starting treatment; 34 (17%) were in the triple-combination group and 27 (14%) were in the double-combination group, for a relative hazard rate for disease progression or death of 1.24 (CI, 0.75 to 2.06; P > 0.2). This includes 7 patients who died without first developing disease progression. Seventy-four percent of events in the triple-combination group and 78% of events in the double-combination group occurred while patients were receiving study medications or within 30 days of permanent discontinuation. Some patients had more than one confirmed disease progression event: Thirty-nine events were recorded among 30 patients in the triple-combination group and 28 events were recorded among 24 patients in the double-combination group. The most common HIV-1 disease progression events were P. carinii pneumonia (in 10 patients in the triple-combination group and 5 in the double-combination group), invasive cytomegalovirus disease (in 10 patients in the triple-combination group and 6 in the double-combination group), and Kaposi sarcoma (in 3 patients in each treatment group). Nineteen patients died within 366 days of starting study treatment; 11 (6%) were in the triple-combination group and 8 (4%) were in the double-combination group (P > 0.2). Only one death was not attributed to HIV-1 disease progression.

In the stratum of patients who had 50 or fewer CD4 cells/mm³, 21 patients (45%) in the triple-combination group and 17 patients (35%) in the double-combination group had at least one disease progression event or died, giving a relative hazard for disease progression or death for this stratum of the triple-combination group of 1.26 (CI, 0.67 to 2.39; P > 0.2). Among patients with 51 to 200 cells/mm³ and 201 to 350 cells/mm³, the numbers of patients in the triple-combination group and the double-combination group who had disease progression or died were also similar: 11 compared with 9 and 2 compared with 1, respectively. The numbers of deaths were also similar for the two groups in each screening CD4 count stratum.

Subgroup Analyses

The triple combination had an advantage over the double combination in immunologic and virologic measures in comparisons among subgroups categorized according to age, sex, race, risk factors for HIV-1 acquisition, AIDS classification, months of previous nucleoside use, type of previous nucleoside therapy at any point or within 14 days of study entry, baseline isolate syncytium-inducing phenotype (syncytium-inducing or non-syncytium-inducing), baseline serum p24 antigenemia (detectable or not detectable), baseline infectious virus titer in peripheral blood mononuclear cells ( 10; 11 to 100; or more than 100 infectious units per million cells), or baseline number of plasma HIV-1 RNA copies/mL ( 10 000; 10 001 to 100 000; or more than 100 000). A similar advantage to triple-combination therapy was seen in both short-term (baseline to first measurement) and long-term comparisons within each of these subgroups. The magnitude of the long-term relative increase in the percentage of baseline CD4 cell count in the triple-combination group was greater in the middle screening CD4 cell stratum (51 to 200 cells/mm³) than in the other strata. However, a differential effect of nevirapine across CD4 cell strata ( 50 cells/mm³; 51 to 200 cells/mm³; and 201 to 350 cells/mm³) was not seen in standardized area-under-the-curve analyses of CD4 cell counts or changes from baseline plasma HIV-1 RNA level or peripheral blood mononuclear cell infectious virus titer.

Some pretreatment characteristics, including any previous use of didanosine or zalcitabine, predicted treatment effects in both groups even though they were not associated with a differential advantage for the triple combination. In the double-combination group, patients who had previously been treated only with zidovudine had a long-term decrease in plasma HIV-1 RNA level that was 0.57 log₁₀RNA copies/mL more of a decrease than that in patients who had also previously received either didanosine or zalcitabine (P < 0.001). Similarly, in the triple-combination group, patients who had not previously used didanosine or zalcitabine had a long-term decrease in plasma HIV-1 RNA level that was 0.87 log₁₀RNA copies/mL more of a decrease than that in patients who had previously used these drugs (P < 0.001). At the end of either study treatment, mean HIV-1 RNA levels were less than baseline values in patients who had not previously used didanosine or zalcitabine but were higher than baseline values in the patients who had.

Adverse Events

Severe rash was reported in 17 patients (9%) assigned to receive the triple combination and in 3 patients (2%) assigned to receive the double combination (P = 0.002) (Table 3). Rashes graded as severe or potentially life-threatening occurred early in the course of treatment (median, 4 weeks). Seven rashes were considered potentially life-threatening; six occurred in the triple-combination group and one occurred in the double-combination group. In each case, the rash resolved after the discontinuation of treatment. In all six cases in the triple-combination group, diffuse maculopapular rash and fever were reported, and four cases included concomitant oral ulcerations. The Stevens-Johnson syndrome was diagnosed in one patient who was receiving the triple combination and was considered probable in a second patient; a third patient was treated presumptively to prevent the development of this syndrome. Short courses of systemic cortico-steroid therapy were given without reported complications to three patients who had severe rash.

Of the 398 patients, 107 (27%) had a severe or potentially life-threatening sign or symptom; 60 of the patients (30%) were in the triple-combination group and 47 (24%) were in the double-combination group (Table 3). With the exception of rash, no significant difference was seen in the frequency of these events or the distribution of times to first sign or symptom (P = 0.16). Six patients in the triple-combination group and 1 in the double-combination group developed pancreatitis (P = 0.12). One hundred eleven patients (28%) had at least one severe or potentially life-threatening abnormal chemical or hematologic result. No differences in these laboratory abnormalities were noted between the two treatment groups; 59 patients (30%) in the triple-combination group and 52 (26%) in the double-combination group had such abnormalities. No significant differences were seen between treatment groups in the frequency of any specific laboratory abnormalities Table 3 or distributions of times to first abnormality (P > 0.2).

Discussion

Top Methods Results Discussion Author & Article Info References

The triple combination of nevirapine, zidovudine, and didanosine had better immunologic and virologic activity than the double combination of zidovudine and didanosine among HIV-1-infected adults who had CD4 counts of 350 cells/mm³ or less and had had extensive previous nucleoside therapy. After 48 weeks, the triple-combination group had a higher mean absolute CD4 cell count and a greater percentage of T lymphocytes that were CD4 cells than did the double-combination group. For a hypothetical patient with a baseline CD4 cell count of 138 cells/mm³ (the median for this sample), this corresponds to a loss by the end of the study period of 21 CD4 cells/mm³ with triple-combination therapy compared with a loss of 46 cells/mm³ with double-combination therapy. Maximal antiviral response was seen soon after either study treatment was started. The triple-combination group maintained a lower mean peripheral blood mononuclear cell HIV-1 infectivity titer, a lower plasma HIV-1 RNA level, and a lower serum p24 antigen level than did the double-combination group over 48 weeks, although there was a parallel increase in circulating virus from the early nadir in each group (Figure 2, middle and bottom).

Our results lead to two conclusions. First, the superiority of nevirapine to placebo was maintained over a long period, which supports the efficacy of this non-nucleoside reverse transcriptase inhibitor. Second, the three-drug combination strategy appeared to be significantly better than the two-drug combination strategy with regard to CD4 cell counts and virologic benefit. Although this comparison of combination strategies is complicated by the as-yet undetermined proportion of patients who had zidovudine- and didanosine-resistant virus at baseline, the long-term relative advantage of the triple combination was also seen in the subgroup of patients who entered the study having previously been treated only with zidovudine. Most patients in that subgroup were likely to have had virus at baseline that was neither resistant to didanosine nor highly resistant to zidovudine; this assumption was made on the basis of the results of earlier studies of other patients who had received zidovudine [34]. Our conclusion about the three-drug combination strategy is also supported by the results of other trials that have evaluated combinations of three different antiretroviral agents in previously treated patients [35, 36] and is consistent with the enhanced antiviral and clinical effects seen with two-drug regimens compared with zidovudine monotherapy [6-1016-18, 37-39]. Our results thus encourage the further investigation of triple antiretroviral combinations and the inclusion of patients with less previous treatment in future studies.

Our study had some limitations. One concern involves the percentages of patients who were lost to follow-up or who prematurely discontinued study treatment, especially among patients with lower baseline CD4 counts. However, these percentages were similar in the two treatment groups, and their associations with disease status did not differ between groups. Thus, the overall rates of loss to follow-up (12%) and premature discontinuation of study treatment (21% by 24 weeks and 33% by 48 weeks) may have led to either under- or overestimation of the relative activity of the triple combination. Moreover, these rates are similar to those reported in many other clinical trials of HIV-1- infected patients, which include an 11% loss to follow-up and a 25% premature treatment discontinuation by 24 weeks in a group of previously untreated patients who had higher median CD4 counts at study entry than the patients in our study [38].

Another important caution that should be considered when interpreting our results is that the relative hazard for disease progression or death over 48 weeks did not differ significantly between treatment groups, either overall or in the stratum of patients with 50 or fewer CD4 cells/mm³ at screening. The trial was designed to have moderate power to detect a 2.5-fold increase in the median time to clinical progression favoring one treatment group among the patients with fewer than 50 CD4 cells/mm³. This was based on the assumption that the 1-year progression rate would be 60% to 80% among patients with 50 or fewer CD4 cells/mm³ who were assigned to the double combination strategy; such rates had been seen among similarly advanced patients in earlier studies [40]. However, 35% of the patients in the double-combination group and 45% of the patients in the triple-combination group who had 50 or fewer CD4 cells/mm³ progressed during our study. This may be attributable to improved prophylaxis and treatment of complications of HIV-1 infection in 1993-1994 compared with earlier periods, the use of combination therapy rather than zidovudine monotherapy, or both. Although our comparison of the triple-combination regimen to the state-of-the-art two-drug control regimen, based on clinical end points, had limited statistical power, several studies now indicate that differences in plasma HIV-1 RNA levels predict the future course of HIV-1 disease [41] and the relative rate of clinical progression during antiretroviral therapy [42]. It cannot be determined from these data whether the observed difference in antiviral effect between treatments was large enough to be clinically relevant.

The only difference between the treatment groups with regard to adverse events was that severe rash occurred among 9% of patients who received the triple combination and 2% of those who received the double combination (P = 0.002). The risk for nevirapine-associated rash appears to be greatest within the first weeks after starting nevirapine therapy. Clinicians must recognize that nevirapine-associated rash can be life-threatening and must intervene as soon as possible to limit the severity of this toxicity. Analyses are under way to evaluate whether predisposing factors for nevirapine-associated rash identified in our study are associated with an increased risk among patients in other ongoing, blinded trials of nevirapine.

Our controlled comparison, done over 48 weeks, confirmed observations from earlier, open-label trials that antiviral response to nevirapine can be prolonged [16-18], and our pharmacologic results extended understanding of this antiviral effect. Plasma drug levels were consistent with the hypothesis that nevirapine itself directly increases the inhibition of HIV-1 replication rather than indirectly improving virologic response through a drug-drug interaction that increases plasma concentrations of another antiretroviral agent (Figure 1). Indeed, our data and other analyses (data not shown) confirm that the concomitant administration of nevirapine and didanosine decreased zidovudine plasma levels by about 25%, probably because of an interaction affecting zidovudine bioavailability. This seems unlikely to be clinically important, given that larger changes are generally seen in clinically important bioavailability interactions [43] and that half the standard daily zidovudine dose used in our present study appeared equivalent to the standard dose in an earlier study [44].

The mechanism for the sustained difference in viral load between treatment groups seen in our study (Figure 2, middle and bottom) remains under investigation. In earlier trials [16-18], antiviral responses to nevirapine persisted despite the development of replication-competent, nevirapine-resistant HIV-1 isolates within the first few weeks of therapy. Virologic and pharmacologic studies are now in progress to evaluate whether the better long-term virologic response associated with nevirapine in our study is correlated with plasma levels of nevirapine exceeding the 50% inhibitory concentration of nevirapine-resistant virus, as suggested previously [17]. Other possible explanations include relatively decreased replicative potential or increased nucleoside susceptibility of viruses selected during therapy [39, 45]. Another hypothesis—that didanosine resistance was associated with the loss of antiviral response in both treatment groups—was suggested by the lower mean plasma RNA levels in patients who had received no didanosine or zalcitabine before the study than in those who had received these drugs.

In this randomized trial, nevirapine added long-term immunologic and virologic benefits to the combination of zidovudine and didanosine. Given that HIV-1 resistance, other pathogenetic factors, or both may limit the duration of the effectiveness of even more potent combinations, including other new agents such as HIV-1 protease inhibitors [35, 36, 46], it is important to maximize the number of antiretroviral options available for changes in combination regimens over an increasing duration of treatment. Nevirapine is likely to be clinically useful in combination regimens in the future, but the data do not yet suggest specific clinical guidelines.

Appendix

The AIDS Clinical Trials Group Protocol 241 Investigators who worked on this study included the members of the Protocol 241 Team, investigators at NIAID AIDS Clinical Trials Units, and investigators at the NIAID Division of AIDS.

Additional members of the Protocol 241 Team were Lyn Costanzo, RN, MSN, and Sharon Ruben (AIDS Clinical Trials Group Operations Office, Rockville, Maryland); Baiba Berzins, MPH (Northwestern University, Chicago, Illinois); Ana Martinez, RPh, and Irene Fishman, MA (NIAID Pharmaceutical and Regulatory Affairs Branch, Bethesda, Maryland); Karen Kazial, RN (Statistical and Data Management Center, Frontier Science and Technology Research Foundation, Amherst, New York); Susannah Cort, MD, Patrick Robinson, MD, David Hall, PhD, and Heather Macy (Boehringer-Ingelheim Pharmaceuticals, Inc., Ridgefield, Connecticut); Colin McLaren, PhD (Bristol-Myers Squibb Co., Wallingford, Connecticut); James Rooney, MD, and John Warwick, PharmD (Glaxo Wellcome Co., Research Triangle Park, North Carolina); Marc Cavaille-Coll, MD, PhD (Food and Drug Administration, Bethesda, Maryland); Fred Valentine, MD (New York University Medical Center, New York, New York); and David Booth (Clinical Site Monitoring Group, Durham, North Carolina).

Additional investigators at NIAID AIDS Clinical Trials Group Units were Ruy Soeiro, MD, David Stein, MD, Barry Zingman, MD, and Jenny Schliosberg, MD (Albert Einstein College of Medicine, New York, New York); Bruce Polsky, MD, Kent Sepkowitz, MD, Victoria Sharpe, MD, and Michael Giordano, MD (Cornell University, New York, New York); Christine Wanke, MD, Roy Gulick, MD, Donald Craven, MD, and Carress Grodman, RN (Harvard University and Boston City Hospital, Boston, Massachusetts); Kenneth Fife, MD, PhD, John Black, MD, Kristin Todd, RN, CS, MSN, and Heather Nixon, RN, CS, MSN (Indiana University, Indianapolis, Indiana); Kirk Sperber, MD, Pieter Gerits, RN, Donna Mildvan, MD, and Peter Nicholas, MD (Mt. Sinai Medical Center, New York, New York); Robert L. Murphy, MD, Harold Kessler, MD, and Joseph Pulvirenti, MD (Northwestern University, Chicago, Illinois); Kathleen Squires, MD, Michael Saag, MD, Jill Weingarten, RN, and John Gnann, MD (University of Alabama, Birmingham, Birmingham, Alabama); Diane Havlir, MD, Chris Fegan, RN, Stephen Spector, MD, and Douglas Richman, MD (University of California, San Diego, San Diego, California); Mark Jacobson, MD, Kathy Dybeck, RN, Patrick Joseph, MD, and Kathleen Clanon, MD (University of California, San Francisco, San Francisco, California); Stacey McKenzie, MD, Pam Daniel, RN, Dale Dayton, RN, and Jill Leonard, BSN (University of Cincinnati, Cincinnati, Ohio); Robert Schooley, MD, Daniel Kuritzkes, MD, Graham Ray, RN, MSN, and Beverly Putnam, RN, ANP (University of Colorado Health Sciences Center, Denver, Colorado); Dushyantha Jayaweera, MD, Janie Patrone-Reese, RN, BSN, Thomas Tanner, RN, and Jo Moebus, RN (University of Miami, Miami, Florida); Nancy Reed, RN, MS, CS, Renee St. Jacque, RN, Keith Henry, MD, and Susan Swindells, MD (University of Minnesota, Minneapolis, Minnesota); Joe Eron, MD, David Ragan, RN, BSN, James Horton, MD, and Timothy Lane, MD (University of North Carolina at Chapel Hill, Chapel Hill, North Carolina); Ian Frank, MD, Anne Norris, MD, Roger Pomerantz, MD, and Stephen Hauptman, DO (University of Pennsylvania, Philadelphia, Pennsylvania); and Jan Geiseler, MD, John Leedom, MD, Frances Canchola, RN, and Connie Olson, RN (University of Southern California, Los Angeles, California).

Additional investigators at the NIAID Division of AIDS (Bethesda, Maryland) were Lawrence Deyton, MD, and Carla Pettinelli, MD.

Dr. Hughes: AIDS Clinical Trials Group Statistical Data and Analysis Center, Department of Biostatistics, Harvard School of Public Health, 677 Huntington Ave. Boston, MA 02115.

Dr. Johnson: Department of Medicine, Division of Infectious Diseases, University of Alabama at Birmingham School of Medicine and Birmingham Veterans Affairs Medical Center, 229 Tinsley Harrison Tower, 1900 University Boulevard, Birmingham, AL 35294.

Dr. Fischl: University of Miami School of Medicine, Department of Medicine, R-60A, PO Box 016960, Miami, FL 33101.

Dr. Sommadossi: Department of Pharmacology, Division of Clinical Pharmacology, University of Alabama at Birmingham School of Medicine, 1670 University Boulevard, Birmingham, AL 35294.

Mr. Liou: Biostatistics Department, MS #66, Genentech, Inc., 460 Point San Bruno Boulevard, South San Francisco, CA 94080.

Dr. Timpone: Department of Medicine, Division of Infectious Diseases, 110 Kober Cogan, Georgetown University Medical Center, 3800 Reservoir Road, NW, Washington, DC 20007.

Dr. Myers: Boehringer-Ingelheim Pharmaceuticals, Inc., 900 Ridgebury Road, Ridgefield, CT 06877.

Drs. Basgoz and Hirsch: Infectious Disease Unit, Massachusetts General Hospital and Harvard Medical School, 55 Fruit Street, Boston, MA 02114.

Dr. Niu: Division of Biostatistics and Epidemiology, Food and Drug Administration, 1401 Rockville Pike, HFM 220, Rockville, MD 20852.