Factors Associated with Mortality in Human Immunodeficiency Virus Type 1–Infected Adults Initiating Protease Inhibitor–Containing Therapy: Role of Education Level and of Early Transaminase Level Elevation (APROCO–ANRS EP11 Study)

Abstract

This study attempted to identify factors associated with mortality among human immunodeficiency virus (HIV)–infected adults starting a protease inhibitor (PI)–containing therapy. Among 1155 patients consecutively enrolled in the APROCO study between May 1997 and June 1998, clinical characteristics were as follows: median age, 36 years; median baseline CD4 cell count, 288 cells/mm³; and median baseline plasma HIV RNA load, 4.4 log₁₀ copies/mL. After a median follow-up of 27 months, 48 deaths had occurred, of which 44% were related to acquired immune deficiency syndrome. The mortality rate was 2.9% at 12 months. When both data at baseline and data at 4 months after the start of PI therapy were considered, factors independently associated with mortality were (Cox model) low baseline plasma creatinine level, low school education level, low CD4 cell count at 4 months, low hemoglobin level, and elevated hepatic transaminase levels. Thus, social context plus clinical and biologic data, including the 4-month response to treatment, must be considered in treatment of HIV-infected patients

Despite a dramatic decline in mortality among human immunodeficiency virus (HIV)–infected adults in industrialized countries since the advent of highly active antiretroviral therapy (HAART) [1], mortality remains higher among HIV-infected adults than in the general population [2]. Since the use of HAART began, the distribution of the causes of death has changed [3], and factors associated with mortality might also have changed. In this study, we attempted to identify factors associated with mortality among HIV-1–infected adults at initiation of protease inhibitor (PI)–containing HAART. We also assessed the impact of early effect of treatment up to 4 months after the initiation of treatment

Patients and Methods

PatientsThe Antiprotéases Cohorte (APROCO) Agence Nationale de Recherches sur le SIDA EP11 study is a French prospective multicentric cohort of HIV-1–infected adults (⩾18 years old) starting PI-containing therapy for the first time. Subjects were consecutively enrolled in 47 clinical centers in France between May 1997 and June 1998. Exclusion criteria were primary infection and postexposure prophylaxis

Characteristics studiedFollow-up was scheduled at months 1 and 4 after enrollment and every 4 months thereafter. Investigators were requested to report deaths to the study management center as soon as known. The last reminder to check the exhaustiveness of the notifications was sent in November 2000

The following baseline variables were included in this study: sex, age, birthplace (France, Africa, or other), HIV transmission category, HIV clinical stage, CD4 cell count, plasma HIV RNA level, hemoglobin level, body mass index (BMI), plasma creatinine level, hepatitis C virus (HCV) antibodies, hepatitis B virus surface antigenemia, prior antiretroviral therapy, type of PI prescribed, and the combination of nucleoside analogues prescribed. In addition, the following variables were assessed by self-administered questionnaires: highest diploma obtained (table 1), employment, type of residence, couple life, participation in a patient association, depression, smoking, alcohol consumption, and injection drug use. The following variables were measured at month 4: occurrence of an AIDS-defining event, CD4 cell count, plasma HIV RNA level, hemoglobin level, elevation of hepatic transaminase levels, and change of PI

Table 1

Characteristics of 1155 human immunodeficiency virus (HIV)–infected patients and probability of death 4, 12, and 24 months after initial start of protease inhibitor, by main baseline characteristics, in the APROCO study

HIV clinical stage was defined using the 1993 Centers for Disease Control and Prevention (CDC) clinical classification system [4]. We used a French translation of the Center for Epidemiologic Studies depression scale to assess the degree of depression (depression was defined by a score >16). Elevation of hepatic transaminase levels (aspartate aminotransferase or alanine aminotransferase) at month 4 of follow-up was defined as a change from a level of <40 IU/L at baseline to ⩾40 IU/L (upper limit of normal). Because of missing data for HCV antibodies in 10% of patients and missing data for education level in 11%, we introduced a category “unknown” in these 2 variables to maintain the same patient sample for analysis throughout

Statistical analysisWe compared baseline characteristics by using χ² and Wilcoxon tests where appropriate. Survival time was calculated from the date of the first PI prescription to the date of death or to the latest follow-up date before 31 May 2000. Survival probabilities were estimated by the Kaplan-Meier product-limit method. A Cox proportional-hazards regression model was used to study the association between patient characteristics and mortality

Our strategy included 2 steps. First, we studied characteristics measured at baseline. Variables retained in the multivariate model were those associated with mortality in univariate analysis with P<.25. Because of the high number of variables retained, we performed an additional selection in 4 intragroup modeling strategies for variables describing HIV infection, nutritional status, lifestyle patterns, and sociodemographic characteristics. At each step, a reduced model was produced by backward elimination. In the second step, we updated the reduced model with baseline covariates by taking into account measures at month 4 of follow-up. This update involved a selection of patients for whom the period of follow-up was longer than 4 months (left truncature). The proportional-hazards assumption was checked using graphical methods by examining log [−log (survival probability)] versus log (time) plots for each covariate in the final model and by testing the interaction between time and these covariates. SAS software (version 8.0; SAS Institute) was used for statistical analysis

Results

Characteristics of the 1155 patients are described in table 1. The patients with unknown education level had more-advanced disease at baseline than other patients: 51% had <200 CD4 cells/mm³ (vs. 33%; P<.0001), 40% had HIV RNA ⩾5 log₁₀ copies/mL (vs. 27%; P<.001), and 33% had low hemoglobin levels (vs. 24%; P=.02)

The median duration of follow-up was 27 months (interquartile range, 22–32 months). During a total follow-up of 2402 person-years, 48 deaths occurred: 21 (44%) were AIDS related, 11 (23%) were possibly HIV related, and 16 (33%) were not directly related to HIV infection. The cumulative probability of death was 2.9% (95% confidence interval [CI], 1.9%–3.9%) at 12 months and 4.0% (95% CI, 2.8%–5.2%) at 24 months

In univariate analysis, baseline characteristics associated with a higher risk of death were as follows: AIDS stage, CD4 cell count <200 cells/mm³, plasma HIV RNA level ⩾5 log₁₀ copies/mL, BMI <20 kg/m², low hemoglobin level, low plasma creatinine level, injection drug use as HIV transmission category (vs. men who have sex with men), active injection drug use, low school education level, no employment, depression, no prior antiretroviral therapy, and treatment with stavudine-lamivudine (vs. with zidovudine-lamivudine). Probability of death did not differ by sex, age, birthplace, HCV serologic status, or type of PI prescribed. In the multivariate analysis, baseline characteristics associated with a higher risk of death were as follows: CD4 cell count <200 cells/mm³, low hemoglobin level, low plasma creatinine level, and low school education level (table 2)

Table 2

Risk of death according to baseline variables and update with variables measured at month 4 in 1155 human immunodeficiency virus (HIV)–infected patients after beginning therapy with protease inhibitors in the APROCO study

Among factors assessed at month 4 after start of treatment, those associated in univariate analysis with a higher risk of death were CD4 cell count <200 cells/mm³, plasma HIV RNA level ⩾4 log₁₀ copies/mL, low hemoglobin level, elevation of hepatic transaminase levels, and occurrence of an AIDS-defining event during the first 4 months of follow-up. In the multivariate analysis, characteristics associated with a higher risk of death were low plasma creatinine level at baseline, low school education level, and, at the 4-month time point, CD4 cell count <200 cells/mm³, low hemoglobin level, and elevation of hepatic transaminase levels (table 2)

Discussion

CD4 cell count and hemoglobin level have been associated with mortality in other studies since the use of HAART began [5, 6]. The persistence of this association when these factors were measured at month 4 of follow-up underlines the importance of early response to treatment in the midterm prognosis [7]

Contrary to results of some studies, in our study, neither CDC clinical stage, early occurrence of a new AIDS-defining event, nor plasma HIV RNA level at baseline or at 4 months of follow-up remained associated with the risk of death after taking into account the CD4 cell count. If the occurrence of an AIDS-defining event can be predictive of death when considered as a time-dependent variable [8], the early occurrence of an AIDS-defining event did not seem to impair the vital prognosis in our study. These early events may reflect either delayed immune reconstitution or immune reconstitution syndrome, which unmask a latent opportunistic event. In the same way, increased plasma HIV RNA level can independently be related to the risk of death when considered as a time-dependent variable [5], but, in our study, the early virologic response was not independently predictive of death when early immunologic response was taken into account

Plasma creatinine level was first introduced in the analysis to take into account renal function. The higher risk of death was actually for the subnormal category level, even after adjustment for BMI. In the absence of specific measurements in our study, level of plasma creatinine could be considered as a reflection of muscular mass, and low levels could reflect some malnutrition. Lean body mass, which is mostly skeletal muscle, was an independent predictor of survival among HIV-infected patients before HAART [9], and low BMI is associated with mortality [10]. In the era of HAART, we can assume that the adjustment of prognostic analysis on measures, taking into account lean body mass, may be more precise than BMI. This result also suggests that nutritional interventions may remain necessary in the era of HAART

We did not find any association between HCV seropositivity and mortality, in agreement with others [11]. Some have reported an association between HCV seropositivity and disease progression [12]. The presence of the category “HCV antibody status unknown” did not seem to raise questions about this result. We analyzed sensitivity by attributing HCV seropositivity to patients with unknown HCV status who died and HCV seronegativity to patients with unknown HCV status who did not die. This hypothesis and its opposite did not change our result. HCV seropositivity was not associated with higher mortality in any case. Hepatic cytolysis was more frequent in HCV-coinfected patients in this study and in a previous analysis [13]. We can assume that the association between elevation of hepatic transaminase levels at 4 months and mortality may correspond, in part, to the most severe HCV coinfections or to a flare of hepatitis after HAART. Difficulty in continuing a potentially toxic treatment may be another explanation. An association between elevation of hepatic transaminase levels in patients receiving antiretroviral therapy and mortality was found in another prospective cohort [14]

School education level is probably a proxy for psychological, behavioral, or social characteristics. In the general population, the association between mortality and both income and education level has been established [15]. Since the start of use of HAART, an ecologic study has found an association between neighborhood socioeconomic status and survival after AIDS [16]. In our study, the association between school education level and mortality remained after taking into account early response to treatment. Nevertheless, we cannot exclude problems of adherence among persons with low education levels. Although we did not measure income in our study, the type of residence category produced similar information. We did not find any association between this variable and the risk of death

The fact that the risk of death in the group of patients with unknown education level was significantly higher than in patients who had at least a high school education may be related in part to advanced disease status. Moreover, patients with a low level of education may have declined to fill in the questionnaire more frequently. On that account, the category “school education level unknown” does not seem to raise questions about the association between school education and mortality

Although it was not significant, we found a trend for a higher risk of death for women than for men (hazard ratio, 1.6; 95% CI, 0.7–3.4), whereas, in the general population, mortality rates are lower for women than for men in these age groups. This trend might be explained by a difference in school education level: 20% of women had no school diploma, versus 12% of men, which might reflect socioeconomic difficulties that might have influenced the risk of death

The few factors associated with mortality in this cohort are easy to measure during routine patient treatment in industrialized countries. CD4 cell count, plasma creatinine level, and hemoglobin level are linked to disease progression. Toxicity of treatments, possibly reflected in this study by elevation of hepatic transaminase levels and hemoglobin levels, seemed to play a role in midterm prognosis through a mechanism that remains to be clarified. The higher risk of death in patients with a low level of education needs to be further evaluated to identify potential reasons (e.g., difficulty in treatment adherence, social problems, or high-risk behavior)

APROCO Study Group

Scientific committeeSteering committee: principal investigators: Catherine Leport and François Raffi; methodology: Geneviève Chêne and Roger Salamon; social sciences: Jean-Paul Moatti and Janine Pierret; virology: Françoise Brun-Vézinet and Hervé Fleury; pharmacology: Gilles Peytavin and Rodolphe Garraffo; and other members: Dominique Costagliola, Pierre Dellamonica, Christine Katlama, Laurence Meyer, Michel Morin, Didier Sicard, Alain Sobel, and Françoise Vincent-Ballereau

Events validation committeeLise Cuzin, Michel Dupon, Xavier Duval, Vincent Le Moing, Bruno Marchou, Thierry May, Philippe Morlat, Christian Rabaud, and Anne Waldner-Combernoux

ObserversMarie-Anne Bach, François Bourdillon, Patrick Choutet, Jean-François Delfraissy, Jean Dormont, Yves Souteyrand, and Jean-Louis Vildé

Data monitoring and statistical analysisChristine Alfaro, Catherine Barennes, Soraya Boucherit, Valérie Cailleton, Marie-Patrizia Carrieri, Cécile Charlois-Ou, Cécile Droz, Ségolène Duran, Jean-Luc Ecobichon, Caroline Egouy, Valérie Journot, Régis Lassalle, Loïc Latour, Vincent Le Moing, Charlotte Lewden, Samira Marrakchi, Bernard Masquelier, Walid Nouioua, Gilles Palmer, Edwige Pereira, Marie Préau, Marianne Savès, Marc Souville, Bruno Spire, Jeanina Surzyn, Rodolphe Thiébaut, and Raphaëlle Winum

PromotionAgence Nationale de Recherches sur le Sida (Action Coordonnée no. 7)

Clinical centers and coordinatorsAmiens (Jean-Luc Schmit), Angers (Jean-Marie Chennebault), Belfort (Jean-Pierre Faller), Besançon (Jean-Marie Estavoyer, René Laurent, and Dominique Vuitton), Bordeaux (Jacques Beylot, Michel Dupon, Jean-Yves Lacut, Michel Le Bras, and Jean-Marie Ragnaud), Bourg-en-Bresse (Philippe Granier), Brest (Michel Garré), Caen (Claude Bazin), Compiègne (Pierre Veyssier), Corbeil Essonnes (Alain Devidas), Créteil (Alain Sobel), Dijon (Henri Portier), Garches (Christian Perronne), Lagny (Pierre Lagarde), Libourne (Joël Ceccaldi), Lyon (Dominique Peyramond), Meaux (Christian Allard), Montpellier (Jacques Reynes), Nancy (Thierry May and Philippe Canton), Nantes (François Raffi), Nice (Jill Patrice Cassuto and Pierre Dellamonica), Orléans (Philippe Arsac), Paris (Elisabeth Bouvet, François Bricaire, Charles Caulin, Jacques Frottier, Pierre-Marie Girard, Serge Herson, Jean-Claude Imbert, Willy Rozenbaum, Didier Sicard, François Vachon, and Jean-Louis Vildé), Poitiers (Bertrand Becq-Giraudon), Reims (Gérard Rémy), Rennes (Christian Michelet), Saint-Etienne (Frédéric Lucht), Saint-Mandé (René Roué), Strasbourg (Jean-Marie Lang), Toulon (Dominique Jaubert and Jean-Pierre de Jaureguiberry), Toulouse (Patrice Massip), and Tours (Patrick Choutet)

Presented in part: 40th Interscience Conference on Antimicrobial Agents and Chemotherapy, Toronto, 17–20 September 2000 (abstract 1909)

All patients provided written informed consent. The Comité Consultatif de Protection des Personnes dans la Recherche Biomédicale, Paris-Cochin, approved the study

Financial support: Agence Nationale de Recherches sur le Sida (Action Coordonnée no. 7); Association des Professeurs de Pathologie Infectieuse et Tropicale; Abbott Laboratories; Boehringer-Ingelheim; Bristol-Myers Squibb Pharma; Glaxo Wellcome; Merck, Sharp, and Dohme; Roche

Study group members are listed after the text