| |
Association Between Efavirenz as Initial Therapy for HIV-1 Infection and Increased Risk for Suicidal Ideation or Attempted or Completed Suicide: An Analysis of Trial Data
|
| |
| |
Download the PDF here
Download the PDF here
Ann Intern Med. July 1 2014
"Suicidal Ideation and Attempted or Completed Suicide: In the primary analysis, suicidality was reported for 62 participants, and attempted or completed suicide accounted for 36% of cases (17 out of 47) in the efavirenz group and 33% (5 out of 15) in the efavirenz-free group. During ITT follow-up, suicidality was reported for 83 participants: 43% of cases (27 of 63) in the efavirenz group and 35% (7 of 20) in the efavirenz-free group involved attempted or completed suicide.
Incidence of first suicidality event was 8.08 per 1000 PYs (47 events) in the efavirenz group and 3.66 (15 events) in the efavirenz-free group (stratified IRΔ, 4.62 per 1000 PYs [95% CI, 1.62 to 7.62]). Time to suicidality was shorter in the efavirenz group, whereas time to nonsuicide death did not differ significantly between groups (Figure 1). The efavirenz association with time to suicidality did not differ significantly by study (Figure 2).
Participants in the efavirenz group had a greater hazard of suicidality than the efavirenz-free group (HR, 2.28 [CI, 1.27 to 4.10]; P = 0.006). The estimated efavirenz association seemed larger in the first 24 weeks, with an HR of 3.69 (CI, 1.41 to 9.63) versus 1.54 (CI, 0.71 to 3.34) beyond 24 weeks; this difference was not statistically significant and the proportional hazards assumption was not violated (P = 0.165; continuous log-transformed time, P = 0.24). Increased hazard of suicidality with efavirenz was seen across several sensitivity analyses (Table 2). In a secondary analysis of time to suicidal ideation; attempted or completed suicide; or death attributed to substance abuse, homicide, or accident, incidence was 9.28 (54 events) and 4.64 per 1000 PYs (19 events) in the efavirenz and efavirenz-free groups, respectively (stratified IRΔ, 4.84 per 1000 PYs [CI, 1.59 to 8.10]), with an estimated hazard that was 2-fold greater in the efavirenz group (HR, 2.06 [CI, 1.21 to 3.50]; P = 0.007)."
--------------------------
Editors' Notes
Context- Postmarketing reports have suggested that efavirenz increases risk for suicide.
Contribution
· In an analysis of data from 4 large, randomized trials in which patients with HIV were randomly assigned to either efavirenz-containing or efavirenz-free regimens for initial therapy, efavirenz was associated with a doubling of risk for suicidality (a composite of suicide, suicide attempt, and suicidal ideation).
Caution
· The clinical trials were not specifically designed to investigate suicidality.
Implication
· An increased risk for suicidality should be considered when choosing efavirenz as part of an initial antiretroviral regimen.
-The Editors
------------------------------
1 July 2014
Association Between Efavirenz as Initial Therapy for HIV-1 Infection and Increased Risk for Suicidal Ideation or Attempted or Completed Suicide: An Analysis of Trial Data,
Katie R. Mollan, MS; Marlene Smurzynski, PhD; Joseph J. Eron, MD; Eric S. Daar, MD; Thomas B. Campbell, MD; Paul E. Sax, MD; Roy M. Gulick, MD; Lumine Na, MS; Lauren O'Keefe, BS; Kevin R. Robertson, PhD; and Camlin Tierney, PhD
Ann Intern Med. 2014
Abstract
Background: The relationship between efavirenz use and suicidality is not well-defined.
Objective: To compare time to suicidality with efavirenz-containing versus efavirenz-free antiretroviral regimens for initial treatment of HIV.
Design: Participant-level data were analyzed from 4 AIDS Clinical Trials Group, antiretroviral-naive studies conducted from 2001 to 2010. Within each study, participants were randomly assigned to an efavirenz-containing (n = 3241) or efavirenz-free (n = 2091) regimen. (ClinicalTrials.gov: NCT00013520 [A5095], NCT00050895 [A5142], NCT00084136 [A5175], and NCT00118898 [A5202])
Setting: AIDS Clinical Trials Group sites; 74% of participants enrolled in the United States.
Patients: Antiretroviral-naive participants.
Intervention: Efavirenz versus efavirenz-free regimens.
Measurements: Suicidality was defined as suicidal ideation or attempted or completed suicide. Groups were compared with a hazard ratio and 95% CI estimated from a Cox model, stratified by study.
Results: Seventy-three percent of participants were men, the median age was 37 years, and 32% had documented psychiatric history or received psychoactive medication within 30 days before entering the study. Median follow-up was 96 weeks.
Suicidality incidence per 1000 person-years was 8.08 (47 events) in the efavirenz group and 3.66 (15 events) in the efavirenz-free group (hazard ratio, 2.28 [95% CI, 1.27 to 4.10]; P = 0.006). Incidence of attempted or completed suicide was 2.90 (17 events) and 1.22 (5 events) in the efavirenz and efavirenz-free groups, respectively (hazard ratio, 2.58 [CI, 0.94 to 7.06]; P = 0.065). Eight suicides in the efavirenz group and 1 in the efavirenz-free group were reported.
Limitation: There was not a standardized questionnaire about suicidal ideation or attempt. Efavirenz was open-label in 3 of 4 studies.
Conclusion: Initial treatment with an efavirenz-containing antiretroviral regimen was associated with a 2-fold increased hazard of suicidality compared with a regimen without efavirenz.
Primary Funding Source: National Institutes of Health.
Efavirenz is a preferred nonnucleoside reverse transcriptase inhibitor for treatment of HIV (1-4). Although efavirenz is generally safe and effective, it is associated with central nervous system side effects (5-8); prescribing information contains warnings of rare but serious psychiatric experiences, including suicide, but also notes that a causal relationship cannot be determined from postmarketing reports (5). Likewise, published cases and case series report suicidal thoughts or behavior with efavirenz (9-17). A literature review stated that clear evidence of association between efavirenz and suicide was not available and thus psychiatric history should not exclude patients from efavirenz treatment (18).
Given the widespread use of efavirenz and uncertainty about its relationship to suicide, suicide attempt, or suicidal ideation, we report an AIDS Clinical Trials Group cross-protocol analysis of 4 studies in which participants were randomly assigned to an initial efavirenz-containing or efavirenz-free antiretroviral regimen. Our primary goal was to compare the hazard of suicidality between participants assigned to an efavirenz-containing versus efavirenz-free antiretroviral regimen for initial treatment of HIV-1, a potential safety issue not reported in the original studies.
Discussion
Among treatment-naive patients with HIV-1 from 4 AIDS Clinical Trials Group studies, the hazard of suicidality (suicidal ideation or attempted or completed suicide) was significantly greater with a randomly assigned efavirenz-containing regimen, approximately twice that seen with an efavirenz-free regimen. This increased hazard was seen across sensitivity analyses. Moreover, 8 of 9 completed suicides were in the efavirenz group.
In multivariable analyses, factors associated with increased hazard of suicidality were random assignment to efavirenz, self-reported history of IDU, and documented psychiatric history or recent prestudy psychoactive medication. In some analyses, younger age and lower body weight were also associated with increased hazard of suicidality. The efavirenz association did not seem to differ substantially by any particular baseline characteristic (Supplement), but the subgroup analyses are limited by small event numbers within groups and have potential for confounding by concurrent cultural and patient characteristics (28).
To our knowledge, this is the first reported analysis to demonstrate a statistically significant association between randomly assigned efavirenz for initial treatment of HIV and suicidal thoughts or behavior. A French patient-recall questionnaire showed that 9% of patients reported emergent suicidal ideation more than 1 month after initiation of efavirenz (29). In contrast, a clinical trial of treatment-experienced patients did not detect an association between randomly assigned efavirenz and depressive disorder (HR, 1.32; P = 0.47), 4 suicide attempts were reported (1 of 178 in the efavirenz group and 3 of 177 in the protease inhibitor group) (30). Extended 4-year follow-up restricted to the efavirenz group of this same treatment-experienced study reported attempted suicide in 3 of 178 participants, including 1 suicide death after 3 years in the study (31). In HIV Prevention Trials Network 052, a randomized study of early versus delayed initiation of antiretroviral therapy for HIV-1 infection, 81% of participants used an initial efavirenz-containing regimen: As of February 2011, there were 10 deaths in the early group (all 10 participants were prescribed efavirenz), including 3 suicides and 3 deaths due to unknown causes; in the delayed group, there were 13 deaths, including 6 deaths due to unknown causes, 2 motor vehicle deaths, and 0 suicides (32).
The efavirenz prescribing information describes psychiatric events from 2 early randomized studies (5-6, 33), with open-label indinavir and double-blinded, nelfinavir-containing comparator regimens, respectively. The frequency of suicidal ideation was 0.7% of 1008 patients in the efavirenz group and 0.3% of 635 patients in the efavirenz-free group, with nonfatal suicide attempts among 0.5% and 0% of patients, respectively (5). Publicly available medical and statistical review documentation for traditional U.S. Food and Drug Administration approval of efavirenz states that efavirenz-containing treatment (relative risk, 2.1 [CI not reported]) and history of psychiatric disorder (relative risk, 4.2) were associated with serious nervous system or psychiatric experiences. Events included aggravated or severe depression, hallucination, suicidal ideation or attempt, seizure, aggressive reaction, paranoid reaction, and manic reaction (34); the prescribing information contains similar qualitative statements (5). Similar estimates were identified in the current analysis with respect to relative hazard of suicidality. The data presented in the current analysis contribute important information on suicidality, with a total of 83 patients with reported suicidality (34 attempted or completed suicide).
The overall incidence of suicidality observed here was similar to the 5.4-per-1000 PYs incidence of depression or attempted or completed suicide reported in a French cohort with HIV (35). Nonetheless, it remains possible that suicidal ideation was underreported here. Secondary analysis of suicidality or fatal injury (substance abuse, homicide, or accident) was conducted to help capture deaths possibly related to suicidal behavior; this analysis also demonstrated a 2-fold increased hazard with efavirenz, consistent with the primary result. Death attributed to suicide, injury, or unknown cause was reported for 28 participants (3.2 per 1000 PYs) in the efavirenz group and 7 (1.2 per 1000 PYs) in the efavirenz-free group. Suicide is sometimes misclassified as accidental death or unknown cause and could have been underreported (36-37).
The 4 randomized studies presented here (and combined through a stratified analysis) provide a large and diverse patient population that enhances the strength of our findings. Each study had a concurrent efavirenz-free comparison group, and patient characteristics were balanced through randomization. Patients were followed closely every 2 to 3 months for a median of nearly 3 years, regardless of treatment modification or adverse events. Our primary analyses were ITT, an approach supported by the better reporting of harms CONSORT (Consolidated Standards of Reporting Trials) extension (38), and results from as-treated and multivariable analyses supported the primary results. Although several published case reports describe suicidality onset within 1 month after efavirenz initiation (9-13, 15), late-onset suicidality has also been documented (14, 34). In our analysis, suicidality onset occurred throughout follow-up, both early and late (Figure 1), emphasizing the importance of long-term assessment.
This study has limitations. Three of 4 trials were open-label and thus suicidality, particularly suicidal ideation, may have been susceptible to reporting bias; suicide was listed under postmarketing experiences in the efavirenz prescribing information before the time frame of these studies (5). There was no standardized questionnaire about suicidality or depression, such as the Hamilton scale (39), and psychiatric or suicidal history may have been underreported or undisclosed to care providers. Further, patients with psychiatric history could have been selectively not referred or recruited to these clinical trials because of concerns for potential adverse events. The efavirenz-free comparison groups here included 2 or 3 nucleosides with or without a protease inhibitor, respectively; efavirenz was not compared with another nonnucleoside reverse transcriptase inhibitor or integrase inhibitor in these studies. Occurrence, disclosure, and method of suicidality may vary by sex, race, and across regions of the world due to social, cultural, and economic differences (40-44). This large study included 5332 patients enrolled from 9 countries, but small suicidality event numbers limited our ability to assess potential differences among countries (Supplement).
Efavirenz is associated with central nervous system side effects and reports of suicidality (5, 9-17), and there is limited mechanistic evidence linking efavirenz to neurotoxicity (45-48). In our analysis, suicidal ideation, suicide attempt, suicide death, and death due to unknown causes each followed a pattern of greater frequency with efavirenz. The random assignment to efavirenz or nonefavirenz therapy increases the likelihood that these results represent a causal relationship between efavirenz and suicidality. Given the widespread use of efavirenz and severity of these adverse events, the observed increased risk is clinically relevant. Suicidality occurred uncommonly but with increased frequency in patients treated with efavirenz or those with a history of psychiatric comorbid conditions or injection drug use. In general, suicide risk factors include, but are not limited to, previous suicide attempt, substance or alcohol abuse, fixed hopelessness, agitation, severe depression, anxiety or panic, impulsiveness, complete loss of pleasure, uncontrolled pain, and poor social support (40, 49-50). Care should be taken to avoid stigmatization of persons living with HIV and psychiatric comorbid conditions (51-53). When efavirenz is used as a component of antiretroviral therapy, patients should be monitored carefully for exacerbation of depression or evidence of suicidal thoughts or behavior.
Results
Participant Characteristics and Study Follow-Up
In total, 5332 antiretroviral-naive participants were randomly assigned to an initial efavirenz-containing (n = 3241) or efavirenz-free (n = 2091) regimen. Seventy-four percent of participants were enrolled in the United States. At entry, median age was 37 years, 73% of participants were men, 8% reported a history of IDU, 13% had received recent prestudy psychoactive medication, 10% had received antidepressant medication, and 32% had documented psychiatric history or recent prestudy psychoactive medication use. Baseline characteristics were balanced between groups through randomization (Appendix Table 1). Median follow-up was 96 weeks (25th, 75th percentiles, 62, 132 weeks) for the primary analysis; the efavirenz and efavirenz-free groups had a similar length of follow-up in each study (Table 1).
Suicidal Ideation and Attempted or Completed Suicide
In the primary analysis, suicidality was reported for 62 participants, and attempted or completed suicide accounted for 36% of cases (17 out of 47) in the efavirenz group and 33% (5 out of 15) in the efavirenz-free group. During ITT follow-up, suicidality was reported for 83 participants: 43% of cases (27 of 63) in the efavirenz group and 35% (7 of 20) in the efavirenz-free group involved attempted or completed suicide.
Incidence of first suicidality event was 8.08 per 1000 PYs (47 events) in the efavirenz group and 3.66 (15 events) in the efavirenz-free group (stratified IRΔ, 4.62 per 1000 PYs [95% CI, 1.62 to 7.62]). Time to suicidality was shorter in the efavirenz group, whereas time to nonsuicide death did not differ significantly between groups (Figure 1). The efavirenz association with time to suicidality did not differ significantly by study (Figure 2).
Participants in the efavirenz group had a greater hazard of suicidality than the efavirenz-free group (HR, 2.28 [CI, 1.27 to 4.10]; P = 0.006). The estimated efavirenz association seemed larger in the first 24 weeks, with an HR of 3.69 (CI, 1.41 to 9.63) versus 1.54 (CI, 0.71 to 3.34) beyond 24 weeks; this difference was not statistically significant and the proportional hazards assumption was not violated (P = 0.165; continuous log-transformed time, P = 0.24). Increased hazard of suicidality with efavirenz was seen across several sensitivity analyses (Table 2). In a secondary analysis of time to suicidal ideation; attempted or completed suicide; or death attributed to substance abuse, homicide, or accident, incidence was 9.28 (54 events) and 4.64 per 1000 PYs (19 events) in the efavirenz and efavirenz-free groups, respectively (stratified IRΔ, 4.84 per 1000 PYs [CI, 1.59 to 8.10]), with an estimated hazard that was 2-fold greater in the efavirenz group (HR, 2.06 [CI, 1.21 to 3.50]; P = 0.007).
Attempted or Completed Suicide
In a secondary ITT DSMB analysis, the outcome was restricted to attempted or completed suicide (Appendix Figure); incidence was 2.90 per 1000 PYs (17 events) in the efavirenz group and 1.22 (5 events) in the efavirenz-free group (stratified IRΔ, 1.86 per 1000 PYs [CI, 0.03 to 3.69]). The ITT DSMB analysis of association between efavirenz and time to attempted or completed suicide resulted in an estimated HR of 2.58 (CI, 0.94 to 7.06; P = 0.065). Results from the ITT sensitivity analysis reached statistical significance (HR, 2.56 [CI, 1.10 to 5.92]), whereas the as-treated result was not significant (HR, 2.24 [CI, 0.88 to 5.69]) (Table 2).
Causes of Death
Incidence of suicide was 0.90 per 1000 PYs (8 deaths) in the efavirenz group and 0.18 (1 death) in the efavirenz-free group. The 9 reported suicides were men; 7 were white, non-Hispanic and from the United States, and 2 were Asian and from India. Incidence of death due to suicide, injury, or unknown cause in the efavirenz and efavirenz-free groups, respectively, was 2.93 and 1.73 per 1000 PYs (17 and 7 deaths) among U.S. participants and 3.58 and 0 per 1000 PYs (11 and 0 deaths) among multinational participants. Other causes of death (such as infection, cancer, or organ failure) occurred with similar incidence in the efavirenz and efavirenz-free groups among U.S. participants (5.17 and 5.18 per 1000 PYs, respectively); among multinational participants, incidence of other causes of death was 8.46 and 9.66 per 1000 PYs in the efavirenz and efavirenz-free groups, respectively (Supplement).
Methods
Study Design and Participants
Individual-level data from antiretroviral-naive participants with HIV-1 in AIDS Clinical Trials Group studies conducted from 2001 to 2010 that involved random assignment to an efavirenz-containing or efavirenz-free regimen were included in this prespecified retrospective cross-study analysis. Four studies met these criteria: A5095 (ClinicalTrials.gov: NCT00013520) (19-20), A5142 (NCT00050895) (21), A5175 (NCT00084136) (22), and A5202 (NCT00118898) (23). Components of the antiretroviral regimen were randomly assigned, except for the nucleoside analogue choice in A5142.
The studies varied by antiretroviral regimen and slightly by duration and eligibility criteria (Table 1); each study excluded participants with substantially abnormal baseline laboratory values. Histories of suicidal ideation or attempt were not exclusion criteria. Studies A5095 and A5202 enrolled participants in the United States and Puerto Rico; A5142 enrolled participants in the United States and South Africa; and A5175 enrolled participants from 9 countries in North and South America, Africa, and Asia.
Study protocols required reporting of signs, symptoms, or diagnoses at each visit, which were recorded with both open-text and data-entry codes. In A5175, diagnosis reporting instructions included specific codes for suicidality; the other studies used general psychiatric event codes (for example, "psychiatric disorder, specify") plus open-text description. Each study required reporting of severe and life-threatening graded signs or symptoms per the Division of AIDS grading table (24), as well as any sign or symptom, regardless of grade, that led to change in study treatment; diagnoses were not graded. Further, study A5142 required report of all moderate signs or symptoms, and A5095 and A5202 required report of all moderate central nervous system symptoms. Site institutional review boards approved each study; participants provided written informed consent.
Randomization
Each study used permuted-block randomization; stratification factors and treatment groups are listed in Table 1. Efavirenz was formulated as one 600-mg pill given once daily, with three 200-mg pills given initially in A5095. Efavirenz assignment was open-label in A5142, A5175, and A5202 and was blinded and placebo-controlled in A5095 before a data safety monitoring board (DSMB) recommendation to unblind efavirenz. The DSMB released recommendations mid-study about inferior efficacy in the efavirenz-free group of A5095 (20 February 2003) and A5175 (23 May 2008), after which participants in the efavirenz-free groups were given the option to switch treatment (19, 22).
Outcomes
The primary outcome of this cross-study analysis was suicidality, defined as suicidal ideation or attempted or completed suicide and identified from signs, symptoms, diagnoses, adverse events, and death data via Medical Dictionary for Regulatory Activities, version 15.0. Prespecified Medical Dictionary for Regulatory Activities preferred terms were "suicide," "completed suicide," "suicide attempt," "intentional overdose," "multiple drug overdose intentional," "poisoning deliberate," "suicidal ideation," "suicidal behavior," and "depression suicidal." Attempted or completed suicide was a secondary outcome. Clinical investigators, blinded to treatment and previous adverse events, independently reviewed death data categorized as suicide, substance abuse, homicide, accident, unknown cause, or other cause (for example, infection, cancer, or organ failure); a secondary outcome included suicidality or fatal injury attributed to substance abuse, homicide, or accident.
Covariates
Each study protocol required report of prescription medication ongoing within 30 days before entry (denoted "recent prestudy"); prestudy psychoactive and antidepressant medications were identified from a medication list on the National Institute of Mental Health Web site (25). Psychiatric history was defined as any event in the Medical Dictionary for Regulatory Activities system organ class "psychiatric disorders," and depression-related events were classified according to review of psychiatric events data by a psychologist. Prestudy psychiatric measures included psychiatric event history, recent psychoactive medication, depression-related event history, and recent antidepressant medication; presence of event history or prestudy medication was combined into 1 covariate. Additional a priori baseline covariates included geographic region, sex, race or ethnic group, age, pretreatment CD4 count, history of AIDS-defining event, and history of injection drug use (IDU); pretreatment HIV-1 RNA levels, body weight, and body mass index (BMI) at study entry were evaluated post hoc (Appendix Table 1). Analysis of race or ethnic group was limited to white, black, and Hispanic from the United States because of potential social and ethnic differences among countries and low frequencies in other groups and was self-reported and classified according to National Institutes of Health categories. Covariate misclassification was possible; for example, history of psychiatric events or IDU could have been undisclosed or underreported.
Data Synthesis and Statistical Analysis
The primary analysis approach was intention-to-treat (ITT). Participant-level data were analyzed according to randomized treatment allocation, with follow-up from randomization to last on-study contact or death; all follow-up in A5095 and A5175 was censored after a DSMB recommendation related to the efavirenz comparison (denoted "ITT DSMB"). In a sensitivity analysis, follow-up included time from randomization to last on-study contact or death, regardless of DSMB recommendations (denoted "ITT"); deaths were summarized using the ITT approach. As-treated analyses excluded participants who never started treatment and included follow-up from treatment initiation through the earliest of the following: discontinuation of the assigned efavirenz-containing or efavirenz-free strategy plus 28 days for washout, discontinuation of all antiretroviral therapy plus 28 days, or last on-study contact (denoted "as-treated"). A sensitivity approach further censored as-treated follow-up at the time of DSMB recommendations (denoted "as-treated DSMB"). Antiretroviral modifications were allowed for reasons such as toxicity, virologic failure, or DSMB recommendations. Missing baseline data were rare (<1%); thus, covariate-adjusted analyses used a complete-case approach.
Crude incidence rates were calculated as the number of cases per total person-years (PYs) at risk, presented as events per 1000 PYs. Incidence rate difference (IRΔ) between treatment groups was quantified by a Mantel-Haenszel estimate, stratified by study, with a 95% CI computed using a rare events variance estimator (26). The primary end point, time to suicidality, is presented with cumulative incidence curves and compared between groups with a Gray test (27), stratified by study, with nonsuicide death considered a competing risk. Estimated efavirenz and baseline covariate associations were quantified by a hazard ratio (HR) from a Cox proportional hazards model, stratified by study. Modification of efavirenz association by covariates was evaluated with interaction terms. The Cox model proportional hazards assumption was evaluated with a piecewise constant hazard with time (≤24 weeks vs. >24 weeks) and with a log-transformed time variable. An incidence rate ratio for the efavirenz association was estimated from an exact Poisson model, stratified by study, to evaluate sensitivity of the Cox model to low event frequencies. Analyses were conducted 2-sided with a significance level of 0.05, without adjustment for multiplicity, in SAS, versions 9.2 or 9.3 (phreg, genmod) (SAS Institute, Cary, North Carolina), and in R, version 2.15.1, competing risks package (cmprsk) (www.r-project.org).
Role of the Funding Source
The National Institute of Allergy and Infectious Diseases funded all 4 studies and this combined data analysis. Industry sponsors provided most antiretroviral medications, including efavirenz, and served as members of individual study teams but were not involved in this analysis. The funding source had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
|
|
| |
| |
|
|
|
