Impact of Late Presentation on the Risk of Death Among HIV-Infected People in France (2003–2009)

Abstract

Objective: 

A recent consensus defines “late presentation” (LP) during the course of HIV infection as presentation with AIDS whatever the CD4 cell count or with CD4 <350 cells per cubic millimeter. Here, using this new definition, we examined the frequency and predictors of LP and its impact on mortality.

Methods: 

In antiretroviral-naive patients enrolled in the French Hospital Database on HIV between 2003 and 2009, we studied risk factors for LP by multivariable logistic regression. The impact of LP on mortality was analyzed according to the level of immunodeficiency by using Cox multivariable models adjusted for potential confounders, with follow-up categorized into 0–6, 6–12, and 12–48 months.

Results: 

There were 11,038 (53.9%) late presenters among the 20,496 patients included in the study. Compared with patients presenting for care with CD4 ≥350 cells per cubic millimeter, patients presenting with AIDS had a very high risk of death with crude hazard ratio ranging from 48.3 during the first 6 months of follow-up to 4.8 during months 12–48; the corresponding values among AIDS-free patients with CD4 ≤200 cells per cubic millimeter were 8.1 and 2.3. Importantly, patients presenting with CD4 between 200 and 350 cells per cubic millimeter also had a significantly increased risk of death beyond 6 months of follow-up (hazard ratio: 3.0 and 1.8 for months 6–12 and 12–48, respectively). Results were similar after adjustment.

Conclusions: 

LP with HIV infection is still very frequent in France and is associated with higher mortality, even among patients with only moderate immunodeficiency. Encouraging early testing and access to care is still urgently needed.

INTRODUCTION

The prognosis of HIV infection has improved dramatically with the advent of combined antiretroviral therapy (cART).^1,2 Although antiretroviral drugs are increasingly effective and well tolerated, many individuals throughout Europe still present late in the course of the infection, with estimates ranging from 23% to 45% depending on the definition of late presentation (LP) and the country.³ Studies in this area are difficult to compare, owing to the use of different definitions of LP. Standardization of this definition would facilitate epidemiological studies and comparisons across different countries or periods.^3,4 Despite their heterogeneity, most studies suggest that mortality is higher among late presenters.⁵ One study showed that the risk of death associated with LP, defined by AIDS or CD4 <200 cells per cubic millimeter, was especially high during the first 6 months of follow-up and remained elevated long after care initiation.⁶ Therefore, LP results in a major loss of life expectancy for HIV-infected patients.

World Health Organization guidelines currently recommend cART for patients with a history of an AIDS-defining illness or a CD4 cell count below 350 cells per cubic millimeter.⁷ In line with these recommendations, LP was recently defined as presentation with AIDS whatever the CD4 cell count or with CD4 <350 cells per cubic millimeter, whereas advanced HIV disease at presentation was defined as AIDS or CD4 <200 cells per cubic millimeter.⁸

The aim of this study was to investigate risk factors associated with LP including calendar period and the impact of LP on mortality according to the immunodeficiency level using the new consensus definition.

METHODS

Description of the French Hospital Database

The French Hospital Database on HIV (FHDH ANRS CO4) is a nationwide hospital-based cohort created in 1989.⁹ Currently, 69 hospitals throughout France contribute data on HIV-infected patients. The only FHDH inclusion criteria are HIV-1 or HIV-2 infection and written informed consent. Trained research assistants prospectively collect clinical, biologic, and therapeutic data from the patients' medical records using specialized software. A standardized follow-up form is completed at least every 6 months or at each visit or hospital admission during which a new clinical manifestation is diagnosed, a new treatment is prescribed, or a change in biologic markers is noted. The database was approved by the French data security authority (Commission Nationale de l’Informatique et des Libertés).

Study Population

Antiretroviral-naive HIV-1 patients enrolled in the FHDH between January 1, 2003, and June 30, 2009, and aged 15 years or more were selected for this study. To ensure that they were not previously under care, patients with a previous history of AIDS or a previous viral load result were excluded. Patients with no available CD4⁺ cell count within 90 days after enrollment except if they had an AIDS-defining event, those with a diagnosis of HIV infection more than 90 days after enrollment, and those with an AIDS-defining event of an unknown type were excluded from this analysis. AIDS was defined using the 1993 definition.¹⁰

LP to Care

In line with both the new consensus definition of LP and the therapeutic guidelines, LP was defined as presentation with AIDS whatever the CD4 cell count or with CD4 <350 cells per cubic millimeter at the time of enrollment in the database. For the analysis of the risk of death, “Disease stage at first hospital contact” was categorized as AIDS whatever the CD4, CD4 ≤200 cells per cubic millimeter (these 2 categories correspond to presentation with advanced HIV disease), 200 cells per cubic millimeter < CD4 < 350 cells per cubic millimeter, and CD4 ≥350 cells per cubic millimeter (no LP).

Statistical Analysis

Associations between sociodemographic variables and LP were studied using multivariable logistic regression. The following covariates were considered as potential confounders: age, sex/transmission group/migrant status, geographic area of the hospital, and time since diagnosis of HIV infection. Time since diagnosis of HIV infection was categorized as less than 3 months to characterize patients who have recently been diagnosed or 3 months or more to characterize patients who knew their diagnosis but were not under care as defined in.⁸ Differences in the risk of death according to disease stage at first hospital contact could be confounded by cART initiation. Therefore, we decided to study time between first hospital contact and cART initiation (in months) according to disease stage at first hospital contact and calendar period. The probability of cART initiation according to the calendar period and disease stage at first hospital contact was evaluated using the Kaplan–Meier estimate. For each category of disease stage at first hospital contact, we evaluated the hazard ratio (HR) for cART initiation by using a univariable Cox proportional hazard model. Time (in months) between the date of enrollment and the date of death or the last visit to an FHDH center, or 4 years, whichever occurred first, was calculated. The 4-year mortality rate was estimated with the Kaplan–Meier method. The impact of LP on mortality was analyzed with univariable and multivariable Cox proportional hazard models. HRs were estimated for patients with AIDS, CD4 ≤200 cells per cubic millimeter, or 200 < CD4 < 350 cells per cubic millimeter, compared with those with CD4 ≥350 cells per cubic millimeter. Multivariable Cox proportional hazard models were adjusted for the same variables as in the logistic regression. In addition, models were adjusted either for the calendar period of the first hospital contact (model 1) or for the cART initiation, modeled as a time-dependent variable (model 2). We checked the proportional hazards assumption in the Cox models. As the hazards were not proportional over the 48-month period, follow-up was divided into periods of 0–6, 6–12, and 12–48 months. We also tested interactions between “disease stage at first hospital contact,” on the one hand, and “calendar period of first hospital contact,” “geographic area,” or “time since HIV diagnosis,” on the other hand. The P values below 0.05 were considered to denote statistical significance. All analyses were done with SAS v9.3 software (SAS Institute, Inc, Cary, NC).

RESULTS

Among the 22,553 antiretroviral-naive patients aged 15 years or more who were enrolled in the FHDH cohort between January 1, 2003, and June 30, 2009, 2057 patients were excluded from the analysis for the following reasons: 1655 had no available CD4⁺ T-cell count within 90 days after enrollment and no AIDS diagnosis, 63 had a diagnosis of HIV infection more than 90 days after enrollment, 128 had an AIDS-defining illness of an unknown type at enrollment, and 211 had an unknown date of the first AIDS-defining event. Excluded patients were more likely to be followed in Paris area and more likely to be nonmigrant men and women. Among the 2057 excluded patients, 82 deaths were observed within 4 years. Median follow-up among the remaining 20,496 patients was 46.3 months (interquartile range: 23.0–70.8).

Frequency and Risk Factors of LP

A total of 11,038 patients (53.9%) presented late and 6396 patients (31.2%) had advanced HIV disease (including 2253 patients with AIDS). Sociodemographic characteristics and their association with LP are shown in Table 1. Men who have sex with men were less likely than all other transmission groups to present late, with respective LP rates being 40.7%–69.2%. The risk of LP increased with age as illustrated with the odds ratio among patients aged 60 years or more (2.85) and the odds ratio among those aged between 50 and 59 years (2.24) relative to patients aged less than 30 years. However, 44.3% of patients aged less than 30 years were late presenters. The frequency of LP was higher in French Departments of the Americas (1.11) and in Reunion Island (1.13) than in the rest of France. The frequency of LP fell between 2003 and 2009, from 57.1% among patients enrolled in 2003–2004 to 49.7% in 2007–2009.

TABLE 1: Characteristics of the Patients and Their Association With LP

LP and cART Initiation

For this analysis, our working sample included 19,763 patients, which correspond to all patients with at least one follow-up. During 4 years of follow-up, 15,429 patients (78.1%) had initiated cART. Whatever the period during which the patients presented for care, those with advanced HIV disease quickly initiated cART whether they had AIDS or CD4 ≤200 cells per cubic millimeter (Kaplan–Meier estimates were >90% at 6 months and >94% at 18 months). Patients presenting for care with either 200 cells per cubic millimeter < CD4 < 350 cells per cubic millimeter or CD4 ≥350 cells per cubic millimeter initiated cART more rapidly during the more recent period (Table 2).

TABLE 2: Percentage of Individuals on cART According to the Time Since the First Hospital Contact, Disease Stage at First Hospital Contact, and Calendar Period and Crude Hazards Ratios for cART Initiation

LP and Mortality

During the 4 years of follow-up, 509 patients [4-year death rate, 3.0%; 95% confidence interval (95% CI): 2.7% to 3.3%] died of whom 435 patients (85.5%) were late presenters. The 4-year mortality rate in late presenters was 4.8%, whereas it was 1.1% in nonlate presenters. Among the 509 deaths, 223 (43.8%) were deaths from AIDS causes, 50 (9.8%) from non–AIDS-defining malignancies, 28 (5.5%) from infection, 26 (5.1%) from vascular and heart diseases, 16 (3.1%) from liver diseases, and 52 (10.2%) from other causes. Patients presenting with AIDS were more likely to die from AIDS causes than other patients 146 (59.8%) versus 77 (29.1%, P < 10⁻³). Figure 1 shows Kaplan–Meier survival curves according to the disease stage at first hospital contact. Mortality among patients presenting for care with AIDS was very high, especially during the first 6 months of follow-up where 156 deaths were observed, corresponding to a 6-month mortality rate of 7.4% (95% CI: 6.3–8.3); the 4-year mortality rate was estimated as 12.4% (95% CI: 10.9–13.9), and the risk of death fell with higher CD4 cell counts at presentation. Figure 2A shows the crude and adjusted HRs for death and their associated 95% CIs according to the disease stage at first hospital contact during each follow-up period (0–6, 6–12, and 12–48 months). Compared with patients presenting for care with CD4 ≥350 cells per cubic millimeter, the risk of death was very high among patients presenting with AIDS, with HRs ranging from 48.3 (95% CI: 28.0–83.5) during the first 6 months of follow-up to 4.8 (3.3–7.0) during months 12–48. Among patients with CD4 ≤200 cells per cubic millimeter without AIDS-defining event, the corresponding figures were 8.1 (4.5–14.6) and 2.3 (1.6–3.4). The pejorative prognostic value of LP was also observed among patients with only moderate immunodeficiency (CD4 between 200 and 350 cells per cubic millimeter) during the periods 6–12 months (HR: 3.0, 95% CI: 1.2–7.4) and 12–48 months (HR: 1.8, 95% CI: 1.2–2.7). Adjustment for sociodemographic variables and for either the calendar period of enrollment or the cART initiation did not modify the results. As shown in Figure 2B, patients included in the last period were less likely to die (the 4-year mortality rates were estimated as 3.6% in 2003–2004 vs 2.0% in 2007–2009). Adjustment for sociodemographic variables and for disease stage at first hospital contact did not modify the results. None of the tested interactions was significant.

FIGURE 1: Four-year survival curves and 4-year mortality rates according to disease stage at first hospital contact.

FIGURE 2: Risk of death according to (A) disease stage at first hospital contact: with adjustment for age, sex/transmission group/sub-Saharan African migrant status, geographic area, time since HIV diagnosis, +period of first hospital contact (model 1) or cART (model 2) and (B) calendar period of first hospital contact: with adjustment for age, sex/transmission group/sub-Saharan African migrant status, geographic area, time since HIV diagnosis, disease stage at first hospital contact. The reported HR is the crude HR.

DISCUSSION

Among patients enrolled in the FHDH database from 2003 to 2009, 54% presented late, including 31% with advanced HIV disease. Even though they were more likely than other patients to initiate cART, late presenters had a higher risk of death than those presenting with CD4 ≥350 cells per cubic millimeter. This increased risk of death was observed up to 4 years of follow-up. Adjustment on potential confounders did not modify the results.

The main strengths of the study are the large size of the FHDH cohort and its representativeness of HIV-infected patients receiving care in France.¹¹ The main weakness is possible underascertainment of mortality because of losses to follow-up. Indeed, the percentage of patients with a last contact more than 1 year before the end of the study was 15.7%, and this might have influenced the estimates of survival rates and the relative risk of death according to the disease stage at first hospital contact. However, patients presenting late were less likely to be lost to follow-up compared with those presenting with CD4 ≥350 cells per cubic millimeter (P = 0.02) and it seems therefore unlikely that our results can be explained by loss to follow-up. In a previous study, we investigated the impact of losses to follow-up on survival estimates and found that relative risks were robust, even though the risk of death was slightly underestimated.¹² Among the 339 excluded patients with AIDS of unknown date or type, 37 died within 4 years. Their noninclusion may therefore have lead to a slight underestimation of the risk of death of patients presenting for care with AIDS.

In our previous study,⁶ covering the period 1997–2002, 35.7% of patients presented with advanced HIV disease, compared with 31.2% in the present study, a result in keeping with a gradual decline in presentation with advanced HIV disease. The high overall rate of LP in our study (54%) is similar to that found in other European countries such as Germany, Spain, and Italy (49.5%–63%), in studies also using the new definition of LP (CD4 <350 cells/mm³ or clinical AIDS).^13–16 Interestingly, the risk factors for LP found in the present study are the same as those found to be associated with presentation with advanced HIV disease in our previous study.⁶ Our results are also in line with those of other studies conducted in Europe, which showed that older age, migration status, and heterosexual transmission were risk factors for LP.^13–17 Risk factors for LP are well known and do not seem to change over time, across countries of care or across definitions of LP, even in countries with free HIV testing and care. Deblonde et al¹⁸ investigated barriers to HIV testing at the individual, medical, and public health policy levels. They found that HIV testing is a complex process that involves several stakeholders, which may explain the difficulty of improving HIV screening and reducing LP. In addition, some late presenters are unaware of their infection, whereas others are aware but had not previously sought care. Most previous studies and efforts to improve the situation have focused on late diagnosis and only few studies underlined the question of linkage to care and suggested the importance of handling linkage to care along with HIV testing.^19,20

The estimated 4-year mortality rate among patients who had advanced HIV disease at presentation was 13.9% in our previous work⁶ compared with 6.7% in the present study. This improvement between 1997–2002 and 2003–2009 and the recent improvement evidenced in this study in 2007–2009 compared with 2003–2004 may be explained by the availability of new more effective and better tolerated antiretroviral drugs and by cART simplification, which improves adherence and reduces the risk of drug resistance.²¹ However, compared with patients presenting for care with CD4 ≥200 cells per cubic millimeter, the pejorative impact of advanced HIV disease at first hospital contact on mortality have not varied between our previous study and our current study: during the first 6 months after FHDH enrollment, the HR for death was 13.2 (95% CI: 9.0–19.0) in 1997–2002 and 13.4 (8.8–20.4) in 2003–2009. The corresponding figures for months 6–12 were 5.3 and 4.7. Even if patients with advanced HIV disease rapidly initiated treatment, as recommended, they had a higher risk of death regardless of the study period. Thus, available antiretroviral regimens are not sufficiently effective in late presenters, who remain at an elevated risk of death for as long as 4 years after presentation. In addition, a previous history of an AIDS-defining event in a patient's HIV infection course could be responsible of irreversible damages. The higher risk of death among patients with CD4 ≤200 cells per cubic millimeter or AIDS is well documented.^22–25 Here, we show that presentation with AIDS, regardless of the CD4 cell count, carries an especially poor prognosis and that patients presenting with moderate immunosuppression also have a higher risk of death. The higher risk of death among patients presenting for care with CD4 between 200 and 350 cells per cubic millimeter relative to patients presenting earlier became statistically significant only beyond 6 months, possibly because of a lack of power (only 11 deaths occurred among these patients at 6 months). Alternatively, these patients may be at a higher risk of clinical events that would result in death only after a certain amount of time. Indeed, these patients had a higher risk of AIDS-defining illnesses and hospitalization for non-AIDS events during the first 6 months of follow-up (data not shown). These results are in line with those of a study conducted in Spain, which showed that health care costs were significantly higher for patients with LP (AIDS or CD4 <350 cells/mm³) and that much of the difference relative to patients who did not present late was due to HIV-related hospital care costs.¹⁷ Also in line with our results, a recent study conducted in Italy, which used the European consensus definition of late diagnosis, showed that LP for care was associated with a high risk of new AIDS event or death.¹⁶ It is also well known that immunodepressed HIV-infected patients, including those with moderate immunodepression, have an increased risk of non-AIDS illnesses such as cardiovascular events and non–AIDS-defining malignancies.^26,27

The high rate of LP observed here suggests that the HIV testing policy in France in 2003–2009 was inadequate. Our results suggest that HIV testing must be reinforced in population subgroups who are at an increased risk of LP. However, it should also be noted that LP was still very frequent in other subgroups, such as men who have sex with men and patients less than 30 years old. Moreover, although most recommendations focus on late diagnosis, public health policy should not ignore patients who are aware of their seropositivity but who are not receiving care because they represented 26% of the new enrolled patients, despite free access to care in France. This is why linkage to care just after HIV diagnosis must also be reinforced. The stakes of the first consultations are crucial regarding the quality of the future medical patient's follow-up.²⁰ The first consultations have to meet the specific needs of patients, which are different from those of patients followed for several years. Developing specialized consultations for the 2 or 3 initial outpatient visits is thus essential. Moreover, point of care HIV tests, which deliver results quickly (less than 1 hour) and allow immediate knowledge of HIV status (in a single visit), could favor immediate entry into care.²⁸ French guidelines issued in 2009 recommend that HIV testing be expanded to cover the general population and underline the importance of linkage to care.²⁹

Even if HIV screening improves in future, some patients will continue to present late and will need more effective management than that currently available. Patients presenting late require close management, especially during the first 6 months, and treatment must be tailored to their specific situation. The recent literature recommends cART initiation within 15 days of onset of an AIDS-defining illness,³⁰ but the choice of cART regimen for these patients is particularly delicate because it must take into account drug interactions, higher risk of adverse effects, and potential resistance.³¹ Current cART regimens clearly fail to reduce the relative risk of death among patients with LP compared with other patients. The OPTIMAL ANRS 146 trial is designed to evaluate the impact of maraviroc adjunction to cART on clinical outcomes in patients with advanced HIV disease in France, Italy, and Spain (NCT01348308). Last, it must be remembered that LP has also community consequences because people may not act the same way toward transmission risk if they know they are infected or not.³² In addition to individual benefits of reducing LP in terms of mortality, decreasing the rate of LP may also take part in the decrease of HIV transmission and therefore in the decrease of the incidence of HIV infection.

LP with HIV infection is still very frequent in France and fell only slightly from 2003 to 2009. Late presenters, including those with only moderate immunosuppression, are at an increased risk of death, especially during the first 6 months after their first hospital contact but also for up to 4 years. Patients presenting with AIDS have an especially poor prognosis. Measures to reduce the frequency of LP would improve both the prognosis of HIV-infected individuals and the overall control of the epidemic.

ACKNOWLEDGMENT

The authors are grateful to all the participants in the cohort and especially the local research assistants.