Impact of HAART advent on admission patterns and survival in HIV-infected patients admitted to an intensive care unit

Background: Several studies found increased survival times and decreased hospitalization rates since the introduction of highly active antiretroviral therapy (HAART).

Objective: To examine the impact of HAART on admission patterns and survival of HIV-infected patients admitted to an intensive care unit (ICU).

Design: Prospective observational cohort study.

Setting and subjects: All HIV-infected patients admitted from 1 January 1995 to 30 June 1999, to an infectious diseases ICU located in Paris.

Main outcome measures: ICU utilization and admission patterns, and survival.

Results: A total of 426 HIV-related admissions were included. Sepsis increased from 16.3% to 22.6% from the pre- to the post-HAART era, whereas AIDS-related admissions decreased from 57.7% to 37% (P < 0.05). No significant difference in ICU utilization was found. In both periods, half of the patients were not on antiretroviral treatment at ICU admission. In-ICU mortality was 23%, without significant difference between the study periods. By multivariable analysis, in-ICU mortality was significantly associated with SAPS II > 40, Omega score > 75 and mechanical ventilation; and long-term survival with admission in the HAART era and AIDS at ICU admission. Cumulative survival rates after ICU discharge were 85.3% and 70.8% after 12 and 24 months, respectively.

Conclusions: HAART had little impact on ICU utilization by HIV-infected patients. After the introduction of HAART AIDS-related conditions decreased and sepsis increased as reasons for ICU admission. Whereas ICU survival was dependent on usual prognostic markers, long-term survival was clearly dependent on HIV disease stage and HAART availability. In both study periods, at least a half of the HIV infected patients were not on anti-retroviral treatment at the time of ICU admission.

Introduction

The use of highly active antiretroviral therapy (HAART) is associated with a dramatic reduction in HIV RNA levels and with a concomitant increase in CD4 lymphocyte counts. After the introduction of HAART in 1997, the rates of HIV-related complications and hospital admissions dropped sharply [^1-3]. It has been suggested that HAART has changed the natural history of HIV infection [⁴].

The reasons for admission to an intensive care unit (ICU), prognostic markers, and short- and long-term survival of HIV-infected patients admitted to the ICU have been extensively reported [^5-15]. However, there are few data on HIV-infected patients admitted to the ICU in the HAART era [^5-7]. Moreover, to our knowledge, no studies have evaluated potential changes in ICU utilization, admission patterns, and short- and long-term survival associated with the advent of HAART.

To determine whether the introduction of HAART has affected ICU admission patterns and outcomes, we investigated ICU utilization, reasons for admission, and survival in HIV-infected patients admitted to an ICU.

Methods

A prospective, observational cohort study was conducted at the ICU of the Bichat-Claude Bernard hospital, a 1100-bed university hospital in Paris, France where more than 2500 HIV-infected patients are currently receiving follow-up, a number not significantly modified during the study period.

Data were recorded for all HIV-infected patients admitted to our ICU during the 42-month period from 1 January 1995 to 30 June 1999. The patients were divided into two groups based on whether they were admitted before or after the advent of HAART. The cut-off date was 31 December 1996; although protease inhibitors were introduced in France during the second quarter of 1996, their use became widespread only in the last quarter of the same year.

After ICU admission, each patient was assigned to one of five groups based on the main organ failure requiring ICU admission, as previously defined [⁸]. Admission was considered AIDS-related if the main admission diagnosis was an AIDS-defining condition (1987 Centers for Disease Control and Prevention definition [¹⁶]). Severity of the acute illness was evaluated using the simplified acute physiologic score II (SAPS II) [¹⁷], and the Omega score which reflects patient-care activity during the ICU stay [¹⁸].

Survival curves were constructed for patients discharged alive after a first ICU stay during the study period. All patients were followed up until 30 June 2000. Between June 2000 and December 2000, information on the condition of the patients was obtained from their physicians. In addition, death certificate registers were checked.

The Statistica 4.5 software package (StatSoft^TM 1994, Tulsa, Oklahoma USA) was used for data collection and statistical analysis. Differences in frequency and value of baseline characteristics across groups of patients and across periods of ICU admission (by month or quarter) were tested using the chi-square test and analysis of variance (ANOVA) or Student's t test, as appropriate. All tests were two-sided. Trends in ICU utilization were evaluated using interrupted time series Autoregressive Integrated Moving Average (ARIMA). Monthly admissions of HIV-infected patients from other hospitals and from our hospital were analysed. Reasons for admission were evaluated by 3-month periods, admission groups, and main definitive diagnosis. Median survival time was estimated using the product-limit method. Survival probabilities were expressed as percentages with 95% confidence intervals (CI). A survival analysis was carried out using the log-rank test for two samples and Gehan's and Wilcoxon's test for multiple samples. To assess the independent influence of variables on survival (multivariate analysis), logistic regression analysis was performed for in-ICU survival and Cox proportional hazard models were constructed for long-term survival.

Results

During the study period, 2028 ICU admissions were recorded, including 426 (21%) related to HIV infection.

ICU utilization

No significant decrease was found by ARIMA analysis after the introduction of HAART in the number of HIV-related ICU admissions per month during the study periods.

Comparison between pre-HAART and HAART eras

The characteristics of the patients admitted to the ICU during the pre-HAART era (n = 196) and the HAART era (n = 230) are reported in Table 1. There were more admissions for sepsis and fewer admissions for heart failure during the HAART era. The proportion of AIDS-related admissions to the ICU was significantly higher in the pre-HAART era than in the HAART era. Up to 40% of patients in both periods were unaware of their HIV infection before the ICU admission and more than 50% were not under anti-retroviral treatment in both study periods.

Table 1

In-ICU survival

Overall, 98 (23%) of the 426 patients died in the ICU. As compared to patients who died, survivors were more likely to have been on antiretroviral treatment prior to ICU admission (P = 0.02), they have been no AIDS prior to ICU admission; they had lower SAPS II (≤ 40 points, P = 0.00001) and Omega score (≤ 75 points, P = 0.001) values and were less likely to have required mechanical ventilation in the ICU (P < 0.00001). By multivariable analysis, only SAPS II [odds ratio (OR), 4.8; 95% CI, 2.8-8.2, P < 0.000001), Omega score (OR, 1.9; 95% CI, 1.1-3.3; P = 0.003), and mechanical ventilation (OR, 3; 95% CI, 1.7-5.4; P = 0.002) were independently associated with in-ICU death.

Long-term survival

Three-hundred sixty six patients (366/426, 86%) had no history of previous ICU admission. Among them, 277 patients were discharged alive from ICU. Sixteen patients were lost to follow-up 3-16 months after ICU discharge.

The cumulative proportions of patients who survived after ICU discharge were 85.3% (95% CI, 79-91) at 12 months and 70.8% (95% CI, 63-78) at 24 months. Overall median and mean survival times were 565 days and 680 ± 486 days, respectively.

By univariable analysis, the following variables significantly predicted long-term survival (mean survival time in days ± SD): HIV disease stage (no AIDS, 752 ± 433; first AIDS-defining condition, 655 ± 522; AIDS before ICU admission, 570 ± 528; P = 0.0003); HAART availability (pre-HAART era, 496 ± 300; HAART era, 78 ± 592; P = 0.01); AIDS-related admission (no, 750 ± 466; yes, 575 ± 522; P = 0.02); and admission group (severe sepsis, 656 ± 588; cardiac, 688 ± 575; respiratory, 725 ± 516; neurological, 589 ± 455; miscellaneous, 568 ± 502; P = 0.005).

Mean survival times in days were 688 ± 562 for Pneumocystis carinii pneumonia, 799 ± 502 for bacterial pneumonia, 575 ± 372 for Toxoplasma gondii encephalitis, 438 ± 480 for tuberculosis, and 73 ± 61 for lymphoma.

Fig. 1 presents overall survival curves of HIV-infected patients discharged alive from ICU as a function of HAART availability eras (Log-Rank test, P = 0.001).

Fig. 1

By multivariable analysis, two factors were independently associated with long-term mortality, namely, HAART availability (OR, 0.45; 95% CI, 0.25-0.82) and HIV disease stage at ICU admission (non AIDS: OR, 1; first AIDS-defining condition: OR, 1.57 and 95% CI, 1.12-2.2; AIDS before ICU admission: OR, 2.46 and 95% CI, 1.75-3.45).

Discussion

Significant reductions in the need for medical ward admission of HIV-infected patients after the advent of HAART have been reported [^1,19,20]. In contrast, we found in the present study that both the number of HIV-related admissions to our ICU and the proportion of HIV-infected patients in our ICU patient population were unchanged by the advent of HAART. Decreases in the incidences of opportunistic and bacterial infections have been reported since the introduction of HAART [^1,21-26]. Whereas the proportion of deaths related to bacterial pneumonia has increased [^23-25], P. carinii pneumonia and nearly all AIDS-defining opportunistic infections have decreased significantly since the advent of HAART [²⁶]. It has also been reported that sepsis was present in 75% of HIV-infected patients admitted to the ICU and that the lungs were the most common site of infection (59% of cases) [⁵]. Similarly, 60% of our patients were admitted for severe sepsis associated with bacterial pneumonia. We found that sepsis was more common and AIDS-related disease less common as reasons for ICU admission in the HAART era as compared to the pre-HAART era, although the proportions of patients admitted for a first AIDS-defining condition were similar.

Our study identified some patients' characteristics that may explain these results. First, 40% of HIV-infected patients admitted to the ICU were unaware of their HIV infection or were not receiving medical attention for it. Second, only 50% of the patients were taking anti-retroviral treatment at the time of ICU admission. Third, these proportions were similar in the pre-HAART and HAART eras. In previous series of HIV-infected patients admitted to the ICU, the percentages of patients without anti-retroviral treatment at ICU admission were closely similar in the pre-HAART era [^6,8,11] and HAART eras [^5,7]. Fourth, only 28% of patients were under HAART before ICU admission after HAART advent. The percentage of patients receiving combination anti-retroviral treatment including a protease inhibitor increased from 0% in 1993 to 40-65% in 1997 in the USA [^20,21]. These findings are of particular concern, because they may indicate some degree of HIV preventive and screening measures failure in France. However, our study population does not represent the overall HIV population in France, and may be considered to be a group at high-risk of complications needing ICU admission.

The short-term mortality rate of HIV-infected patients in the present study was similar to the rates reported in the pre-HAART era [^5,8-15]. In-ICU mortality did not change between the pre-HAART and HAART era. In keeping with previous data [^6,-8], we found by multivariable analysis that severity of the acute illness, as assessed in the present study by the SAPS II, Omega score, and the need for mechanical ventilation, strongly predicted short-term mortality. We found 1- and 2-year survival rates of 85% and 71%, respectively, with an overall median survival time of 30 months (565 days). Previously reported 1-year survival rates in HIV-patients discharged alive from the ICU in the pre-HAART era ranged from 28% to 44% [^5,7,9,10]. The present study provides the first HAART-era data on long-term survival of HIV-infected patients after ICU discharge. Clearly, the median and mean survival times reported here, as well as the cumulative survival after ICU discharge, were far better than in earlier studies [^{5,7,9,10,13,14}]. We found significant improvements in survival times between the pre-HAART and HAART eras. Similar improvements have recently been reported in the general population of AIDS patients [²⁷], as well as in HIV-infected patients after ICU admission [^6,7]. It was recently reported that the survival time of HIV-infected patients with severe P. carinii pneumonia admitted to ICU increases from 25% to 69% and was associated with HAART use [⁶]. In the present study, only 28% of HIV-infected patients were on HAART before ICU admission, however some patients was rapidly started on HAART over the first weeks following ICU discharge.

We conclude that in our institution ICU utilization for HIV-infected patients has not been changed by the advent of HAART. Although most HIV-infected patients admitted to the ICU were not taking potent antiretroviral treatment, the median survival times improved after the advent of HAART. Our data indicate that the short-term survival of HIV-infected patients admitted to the ICU is primarily dependent on acute illness severity but that long-term survival is closely dependent on HIV disease-related variables, including the availability of HAART.