Pulido, Federico MD*; Arribas, Jose R MD†; Miró, Jose M MD‡; Costa, María A MD*; González, Juan MD†; Rubio, Rafael MD*; Peña, Jose M MD†; Torralba, Miguel MD*; Lonca, Montserrat MD‡; Lorenzo, Alicia MD†; Cepeda, Concepcion MD*; Vázquez, Juan J MD†; Gatell, Jose M MD‡; the EfaVIP Cohort Study Group

Highly active antiretroviral therapy (HAART) consisting of 2 nucleosides and 1 protease inhibitor (PI) has been widely used for the treatment of advanced HIV-infected patients. Profoundly immunosuppressed AIDS patients treated with PI-based HAART (PI-HAART) experience substantial quantitative and qualitative recovery of CD4 cells. ¹ Numerous studies have shown that after CD4 cell reconstitution, AIDS patients can recover from previously untreatable opportunistic diseases. ^2,3 In addition, immune recovery after PI-HAART has made it possible to discontinue primary and secondary prophylaxis against opportunistic pathogens. ⁴

Efavirenz (EFV) is a nonnucleoside reverse transcriptase inhibitor that was licensed by regulatory agencies for the treatment of HIV infection in 1998. In clinical trials, EFV has shown superior antiviral efficacy compared with PIs in both naive (in comparison to indinavir and nelfinavir) ^5,6 and dual nucleoside–experienced patients (in comparison to nelfinavir). ⁷ In 1 clinical trial, there were not significant differences in CD4 cell recovery in patients treated with EFV- or indinavir-based HAART irrespective of baseline CD4 count. ⁸ These results clearly support the notion that it is possible to design PI-sparing HAART regimens using 2 nucleosides and EFV. ^9,10

EFV-based HAART (EFV-HAART) has not been widely used for the treatment of profoundly immunosuppressed HIV-infected patients. EFV was licensed approximately 2 years after the efficacy of PI-HAART had been demonstrated in advanced AIDS. Since EFV was licensed, the efficacy of EFV-HAART for the treatment of symptomatic AIDS has never been compared with that of PI-HAART in clinical trials. Consequently clinicians have been somewhat reluctant to use EFV-HAART in the group of patients with advanced AIDS and opportunistic diseases. ¹¹

In a previous report, we showed that EFV-HAART had high effectiveness in a cohort of 92 HIV-1–infected patients with profound immunosuppression. ¹² The high effectiveness of EFV-HAART was demonstrated in terms of virologic, immunologic, and clinical outcomes. To the best of our knowledge, however, there are not available data directly comparing different HAART strategies for the treatment of severely immunosuppressed HIV-infected patients. The lack of data regarding optimal treatment of advanced HIV-infected patients is quite worrisome, because patients with low CD4 counts represent a large proportion of all HIV-infected patients. In a national survey, 53% of all patients receiving medical care in the United States had CD4 counts of <200 cells/μL. ¹³ In recent cohort studies also performed in the United States ^14,15 and Spain, ¹⁶ 35% to 37% of all outpatients on initial HIV primary care presentation had CD4 counts of <200 cells/μL or developed AIDS shortly after presentation

The objective of our study was to compare the effectiveness of EFV-HAART versus PI-HAART in profoundly immunosuppressed HIV-infected patients suffering from opportunistic diseases. To achieve this objective, we have performed a retrospective comparative cohort study of all HIV-infected patients followed in 3 HIV clinics who had CD4 counts of <100 cells/μL and received EFV- or PI-based HAART as their first antiretroviral therapy.

PATIENTS AND METHODS

Patient Selection

This retrospective cohort study was carried out in 3 university-based public hospitals HIV units in Spain (2 in Madrid and 1 in Barcelona) that provide comprehensive primary care to HIV-infected adults. Taken together, these 3 HIV units are actively following 5900 HIV-infected patients. In these monographic HIV units, patients are reviewed following a similar protocol, with clinical and analytical visits at least every 3 months.

Patients were included in the EfaVIP study if they were naive for antiretroviral therapy; had a CD4 count of <100 cells/μL; received treatment with EFV- or PI-based HAART, including 2 nucleosides (patients with saquinavir hard-gel as a single PI were excluded); and had at least 1 follow-up visit after starting this antiretroviral regimen. Patients were identified through electronic databases that record outpatient visits, CD4 cell counts, and treatment received. Once a patient was identified, medical records were reviewed.

The following data were recorded on standardized case record forms at baseline: demographic data, HIV risk factors, HIV-related diseases, hepatitis C serum antibody, hepatitis B surface antigen, and baseline CD4 cell count and HIV-1 viral load (0–28 days before starting HAART). HIV-related events were diagnosed according to the Centers for Disease Control and Prevention (CDC) 1993 guidelines.

Data on CD4 cell count and HIV-1 viral load were recorded after starting HAART. If HAART was changed or discontinued, the date and reason for the change or discontinuation were recorded. A careful review of medical records was performed to identify AIDS-related diseases (new or relapsing) after starting HAART. The lower limit of detection of the HIV-1 RNA assays used was <400 copies/mL, because that was the limit for the assay available during the first years of the study.

Study End Points

For all the analyses, observation started when HAART was first initiated. The primary end point was time to treatment failure. The analysis used an intention-to treat approach that considered any of the following as failure: (1) patients never achieving an HIV-1 RNA viral load of <400 copies/mL or having an increase above the limit of quantification in 2 consecutive determinations (only 1 determination was required if the clinician decided to change therapy) during follow-up after an initial virologic treatment response, (2) patients lost to follow-up, (3) therapy discontinuation or change not due to simplification, and (4) incidence of opportunistic infections or death. Time was recorded as 0 days if the HIV-1 RNA viral load never reached <400 copies/mL. For this analysis, data from patients who changed their therapy due to simplification were censored at the date of simplification.

Secondary end points included (1) percentage of patients with an HIV-1 viral load <400 copies/mL at months 3, 6, 12, 18, and 24; (2) time to virologic failure; (3) time to reach a CD4 lymphocyte count >200 cells/μL; and (4) incidence of HIV-related diseases (new or relapsing).

The percentage of patients with an HIV-1 viral load <400 copies/mL was analyzed using 3 different methods: (1) intention-to-treat analysis with noncompleters or changes not due to simplification considered as failures, (2) intention-to-treat analysis with noncompleters considered as failures but ignoring changes of therapy, and (3) according to treatment received (“on treatment”), excluding patients lost to follow-up or with a change in their therapy or with missing viral loads at specified time points. For both intention-to-treat analyses, patients who continued receiving HAART but had not reached the specified time point were censored.

Virologic failure was defined as never achieving an HIV-1 RNA viral load of <400 copies/mL or having an increase above the limit of quantification in 2 consecutive determinations (only 1 determination was required if the clinician decided to change therapy) during follow-up after an initial virologic treatment response. Time was recorded as 0 if the HIV-1 RNA viral load never reached <400 copies/mL. Data for patients lost to follow-up while their viral load was <400 copies/mL were censored at the time of the last viral load available. Patients who had a single viral load above 400 copies/mL at the last recorded measurement in the database were not considered virologic failures at that point unless the clinician decided to change antiretroviral treatment. For this analysis, any change in therapy not caused by virologic failure was ignored.

Time to CD4 cell counts >200/μL was recorded as the elapsed time between starting HAART and the first of 2 consecutive CD4 counts >200 cells/μL or the time to the only CD4 count >200 cells/μL if that was the last count available for the patient. Changes of therapy or simplifications were ignored when analyzing this end point.

Statistics

Baseline characteristics were compared using the Fisher exact test and χ² test for categoric variables and the Wilcoxon 2-sample test for continuous variables. Time-to-event analyses were performed using Kaplan-Meier survival curves, the log-rank test, and Cox proportional hazards regression. Assumptions of the Cox proportional hazard analyses were ascertained using graphic methods and Schoenfeld residuals. ¹⁷ Variables considered in all Cox multivariate analyses included group of therapy, age, sex, centers of the study, calendar period (year of starting HAART), hepatitis virus coinfections (hepatitis B or C virus), previous diagnosis of AIDS, nucleosides used in the combination (zidovudine- or stavudine-containing regimens and use of lamivudine), and baseline CD4 cell count and HIV-1 RNA level (log₁₀ transformed).

RESULTS

Description of the Cohort

All 310 antiretroviral-naive patients with a CD4 lymphocyte count <100 cells/μL who started EFV- or PI-based HAART in any of the 3 hospitals and had at least 1 return visit to the clinic were included. Ninety-two were treated with EFV-based HAART, and 218 were treated with PI-based HAART (101 with nelfinavir, 74 with indinavir, 31 with ritonavir, 6 with indinavir plus ritonavir, and 6 with saquinavir plus ritonavir).

Basal characteristics of the cohort for both groups of therapy are shown in Table 1. Virologic, immunologic, and clinical characteristics showed that the cohort was extremely immunosuppressed. Of all the patients included in the cohort, 193 (62%) had <50 cells/mL before they started HAART. Baseline HIV-1 RNA was higher than 100,000 copies/mL in 239 patients (77%), with a median value of 301,500 copies/mL. Before starting antiretroviral treatment, 250 patients (81%) suffered symptomatic HIV infection and 173 of them (56%) had been diagnosed with an AIDS-defining disease.

Table 1

There were no differences between groups at baseline regarding sex, age, CD4 cell count, viral load, and clinical stage. Three major imbalances were noticeable in both groups of patients (see Table 1). First, consistent with the time of introduction of these drugs in clinical practice in Spain, more patients received EFV in more recent years (see Table 1). Second, there was a trend to more heterosexual transmission and fewer intravenous drug users among EFV-treated patients, reflecting epidemiologic changes in the HIV epidemic in Spain during the last years. ¹⁸ Third, lamivudine was more frequently used with EFV, whereas didanosine was more frequently used with PIs (see Table 1). Again, this could be related to patient treatment practices in more recent years. ^19,20

One hundred seventy-three patients had been diagnosed with an AIDS-defining condition before starting HAART (52 with Pneumocystis jiroveci pneumonia, 43 with tuberculosis, 38 with esophageal candidiasis, 15 with cytomegalovirus end-organ diseases, 13 with cerebral toxoplasmosis, 12 with Kaposi sarcomas, 8 with cryptococcosis, 7with intestinal cryptosporidiosis, 7 with progressive multifocal leukoencephalopathy, and 6 with disseminated Mycobacterium avium disease). There were no differences in the opportunistic infection distribution between the 2 therapy groups, although there was a trend to more patients with tuberculosis in the group receiving EFV (20% vs. 11%; P = 0.08). This could be due to the possibility of simultaneously using EFV and rifampin. Seventy-seven patients did not suffer an AIDS-defining disease before starting HAART but had been diagnosed with an HIV-related disease. Median time from the onset of an AIDS-defining or HIV-related disease and the beginning of HAART was 32 days.

During follow-up, 143 patients (46%) discontinued the prescribed HAART. Reasons for discontinuation are shown in Table 2.

Table 2

Virologic Response

After 6 months of therapy, a significantly higher proportion of patients receiving EFV-based HAART reached a viral load below the detectability level (400 copies/mL). These differences persisted during the complete follow-up period and were observed when analyzed by 3 different methods (intent-to-treat analysis considering a noncompleter or change not due to simplification as failure, intent-to-treat analysis considering a noncompleter as failure and ignoring changes of therapy, or on treatment) (Fig. 1).

Figure 1

Time to Failure

Kaplan-Meier plots (Fig. 2) showed that time to treatment failure (discontinuations and changes of therapy not due to simplification were considered as failure) was shorter with PI-based HAART than with EFV-based HAART (P = 0.0009 by log-rank test).

Figure 2

The relative hazard (RH) of therapy failure for patients receiving PIs (vs. those receiving EFV) was 2 (95% confidence interval [CI]: 1.3–3.08; P = 0.002). After adjusting for year of starting HAART, age, gender, hepatitis C serology, hepatitis B serology, center, previous diagnosis of AIDS, lamivudine use, baseline viral load, and baseline CD4 cell count, time to treatment failure remained longer with EFV-based HAART (RH = 2.19, 95% CI: 1.23–3.89; P = 0.008).

A sensitivity analysis including only virologic failure was also performed. For this analysis, treatment changes not due to virologic failure were ignored. The RH of virologic failure for patients receiving PIs (vs. those receiving EFV) was 4.91 (95% CI: 1.77–13.63; P = 0.002). Again, a significantly lower risk of virologic failure was observed in patients treated with EFV after adjusting by year of starting HAART, age, gender, hepatitis C serology, hepatitis B serology, center, previous diagnosis of AIDS, lamivudine use, baseline viral load, and baseline CD4 cell count (RH = 2.52; 95% CI: 1.14–5.61; P = 0.024).

Progression to CD4 Cell Count >200 Cells/μL

The probability of reaching a CD4 cell count >200 cells/μL was not significantly different between groups treated with PI-based HAART versus EFV-based HAART (RH = 0.8, 95% CI: 0.57–1.12; P = 0.19;Fig. 3). In multivariate Cox regression analysis, time to CD4 cell count >200 cells/μL was independently associated with baseline CD4 cell count (RH = 1.51 per 25 cells/μL, 95% CI: 1.31–1.75; P < 0.0001), log baseline HIV RNA level (RH = 1.78, 95% CI: 1.23–2.58; P = 0.002), and history of intravenous drug use (RH = 0.55, 95% CI: 0.39–0.78; P = 0.001) but not PI use (RH = 0.87, 95% CI: 0.53–1.44; P = 0.60). When those patients with only the last available CD4 count reaching >200 cells/μL (16.5% [n = 51]) were considered as not reaching the end point, results remain unchanged (data not shown).

Figure 3

Almost a half of these severely immunosuppressed patients receiving the initial regimen had a CD4 cell count >200 cells/μL at month 12, without differences between treatment groups (48.6% [35/72] with EFV-based HAART vs. 47.9 [90/188] with PI-based HAART; P = 0.92). There was no difference in mean increase of CD4 cell count at 1 year (173 vs. 185; P = 0.9 by Mann-Whitney U test), although a significant difference favoring EFV was observed at month 24 (336 vs. 253; P = 0.024 by Mann-Whitney U test;Fig. 4). If only those patients with an undetectable viral load at month 24 are included in the analysis, the increase in CD4 cell count remains higher with EFV (348 vs. 264; P = 0.018 by Mann-Whitney U test).

Figure 4

Disease Progression and Death

There were no differences between groups in the incidence of HIV-related events during the follow-up period (7.7 per 100 patient-years with EFV vs. 7.3 per 100 patient-years with PIs; P = 0.99 by χ² test).

There were 42 new HIV-related events during the study period. Of these 42 patients, 7 were treated with EFV and 35 with PIs, and 12 relapses of previously diagnosed diseases (3 with EFV and 9 with PIs) were diagnosed during the follow-up without differences between groups in their distribution. More than a half of these diagnoses (53.7%) were made during the first 16 weeks of therapy.

Nine patients died during the follow-up period (4 treated with EFV and 5 treated with PIs; P = 0.46), with a rate of death of 3.06 and 0.82 per 100 patient-years of follow-up for the PI and EFV groups, respectively. Only 4 patients died because of progression of their AIDS disease (1 treated with EFV [progressive wasting] and 3 treated with PIs [cryptococcal meningitis, progression of pulmonary Kaposi sarcoma, and pneumonia]).

DISCUSSION

Pooled cohorts ²¹ have demonstrated that the risk of death or AIDS progression is unquestionably increased in patients who start HAART with CD4 cell counts of <200 cells/μL. Paradoxically, 6 years after the advent of HAART, little is known about the optimal antiretroviral regimen for patients with low CD4 cell counts. The most important reason for this uncertainty is that patients with advanced HIV infection and/or recently diagnosed opportunistic diseases are underrepresented in clinical trials during the HAART era. In an overview of 23 trials of HAART in naive patients, the mean CD4 cell count and median viral load at baseline were 375 cells/μL and 4.69 log₁₀ copies/mL, respectively. ²² Of note, only 1 trial in this overview included patients with a median baseline CD4 cell count of <200 cells/μL. ²³

Here, we report a retrospective comparative cohort study showing that EFV-HAART had superior effectiveness compared with PI-based HAART (mostly nelfinavir and indinavir) for the treatment of HIV-infected patients with <100 CD4 cells/μL, many of whom suffered from recently diagnosed AIDS-defining conditions. To the best of our knowledge, this is the first study that has analyzed the comparative effectiveness of 2 different HAART strategies for the treatment of patients suffering advanced HIV infection.

In the unadjusted analysis, the probability of reaching a combined treatment failure end point, which included virologic failure, treatment discontinuation (due to adverse events or losses to follow-up but censoring simplifications), and treatment change, was almost 2 times higher for PI-HAART than for EFV-HAART. Although factors known to be strongly associated with virologic response were well balanced at baseline, it was important to adjust for other relevant factors. ²⁴ The PI-HAART cohort was, on average, 2.5 years older than the EFV-HAART cohort, and it is possible that patients treated with EFV might have benefited from the incremental learning curve of their clinicians. Patients in the EFV group were treated more frequently with lamivudine than patients in the PI group. In addition, there was a trend toward more patients in the PI-HAART group with a history of injecting drug use, a factor that can be associated with a poorer response to therapy. Finally, even in patients with low CD4 cell counts, small differences in the numbers of CD4 cells might be relevant. ²⁵ When we performed an analysis adjusting for calendar time, risk group, HIV infection, lamivudine use, and CD4 cell count among other variables, the risk of treatment failure for the PI-HAART group was even higher (more than 2 times) than in the unadjusted analysis. It should be recognized that despite this adjustment, residual bias may persist, especially with regard to the timing of initiation of therapy.

The superior antiviral activity of EFV-HAART over PI-HAART was consistently demonstrated in all sensitivity analyses performed. Simplification of PI-based HAART to more convenient regimens and switching among PIs because of intolerance are common clinical practices. In our cohort, 24% of patients treated with PI-HAART suffered an adverse event that prompted a PI switch and 7% switched their PI for reasons of convenience. It could be argued that simplification and switching among PIs should not be considered treatment failures in the same way as virologic failure and consequently should not be included as a combined treatment failure end point. When we analyzed only virologic failure, however, the relative risk was 4 times higher for the PI-HAART group than for the EFV-HAART group. Taking together the results of all these sensitivity analyses suggests that EFV-HAART was not only less likely to be discontinued than PI-HAART but had superior antiviral activity compared with PI-HAART in this cohort of severely immunosuppressed patients.

The most important limitation of our study was obviously the lack of randomization, which can induce important selection bias. In particular, physicians might have been more inclined to prescribe EFV to patients believed to be more adherent and PIs to patients more likely to be nonadherent (eg, intravenous drug users). In such a case, the observed differences in efficacy might be related not to antiviral activity but to patient level of adherence. Because we do not have data on adherence, we can only hypothesize about the impact of this variable in our overall results.

Despite the limitations of an observational cohort, a number of findings support the validity of our results. Importantly, the PI and EFV cohorts had no significant imbalances at baseline with respect to factors known to be strongly associated with virologic response. The percentage of patients with an AIDS-defining diagnosis, CD4 cell count, and HIV viral load was well balanced at baseline. One of the advantages of our study is that by focusing on a restricted group of patients (only antiretroviral naive, with <100 CD4 cells/μL) we have greatly limited in advance the risk of imbalances between groups at baseline. In our opinion, this advantage compensates for the relatively small sample of patients studied. The absence of significant imbalances at baseline is possibly the most important factor for reducing the probability of bias in cohort studies comparing the activity of antiretrovirals. ²⁴

Three previous cohort studies ^26–28 have shown that therapeutic responses in antiretroviral-naive patients with moderately advanced disease were better with EFV-HAART than with PI-HAART. In the study by Friedl et al, ²⁶ the probability of reaching HIV-1 RNA levels of <400 copies/mL was 1.75 times higher for EFV-HAART than for PI-HAART. Interestingly, this result was maintained in the subgroup of patients with <200 CD4 cells/μL at baseline. Lucas et al ²⁷ found that the odds ratio for achieving <400 copies/mL within 8 months of starting HAART was 2.7 times higher for patients receiving EFV than for patients receiving a single PI. Finally, Matthews et al ²⁸ reported that the probability or reaching <500 copies/mL at 6 months was 1.3 times higher in patients receiving EFV than in patients receiving PIs. It is remarkable that across 3 cohort studies and 1 clinical trial comparing EFV-HAART versus PI-HAART, the difference in favor of EFV in the percentage of patients reaching a viral load <400 copies/mL (8–12 months after starting HAART) has been large and almost identical: 22% in the Friedl et al ²⁶ study, 23% in the Lucas et al ²⁷ study, 24% in our study, and 22% in the 006 trial. ⁵

Our cohort offered a suitable opportunity for describing quantitative CD4 cell recovery in patients who start HAART with extremely low CD4 cell counts. Patients included in our study experienced a substantial CD4 cell recovery after starting HAART, irrespective of the antiretroviral regimen received. At 12 months of follow-up, the mean CD4 cell increases were 173 and 185 cells/μL for the EFV-HAART and PI-HAART groups, respectively. This result is highly concordant with data reported in the literature for patients with advanced and nonadvanced HIV infection. ^29,30 In clinical trials of HAART for the treatment of nonadvanced HIV infection, ²² the mean increases of CD4 cell counts at week 48 were 153 and 168 cell/μL for trials using nonnucleoside reverse transcriptase inhibitors and PIs, respectively. These data suggest that despite being extremely immunosuppressed, patients included in the EfaVIP-2 study achieved absolute CD4 cell gains at 1 year similar to those of nonadvanced HIV-infected patients treated with HAART.

In summary, our study suggests that EFV-HAART has superior effectiveness in comparison to nonboosted PI-based HAART for the treatment of severely immunosuppressed HIV-infected patients. The results of our retrospective cohort study are strengthened by the large difference found and the consistent results in all adjusted sensitivity analyses. Because our study is not randomized, however, residual bias cannot be completely excluded. Our results support the use of EFV-based HAART for the treatment of extremely immunosuppressed HIV-infected patients.

REFERENCES

APPENDIX

The EfaVIP Cohort Study members include the following individuals:

* Hospital La Paz: Susana Hernández and Maria Luisa Montes

* Hospital 12 de Octubre: Rafael Hervás, José R. Costa, and Angel del Palacio

* Hospital Clinic-Idibaps: Esteban Martínez, Felipe García, Elisa de Lazzari, Jose Mallolas, and José L. Blanco