The Causes of Death in HIV Infection: The Key Determinant to Define the Clinical Response to Anti-HIV Therapy

Studies have shown an increased risk of new AIDS/death among injecting drug users (IDU) starting HAART. Of 3872 patients starting HAART in the EuroSIDA study, 819 were IDU (21.2%).

During 14 769 person-years of follow-up, 499 patients progressed to new AIDS/death. Compared with homosexual individuals, IDU had an increased incidence of new AIDS/death, but only for non-HIV deaths. There is an urgent need to define and standardize causes of death in observational studies.

Since the introduction of HAART, morbidity and mortality has fallen to one-fifteenth of the level seen before HAART. In an early study of 6645 patients from the EuroSIDA study, after adjustment there were no differences in the risk of death according to the exposure group.

The Antiretroviral Treatment Cohort Collaboration studied approximately 12 000 treatment-naive patients starting HAART, of whom 17% were injecting drug users (IDU). Recent findings demonstrated an increased risk of new AIDS/death or death among IDU after starting HAART, although the study was not able to differentiate between HIV-related and other causes of death. It is therefore unclear to what extent this finding is the result of a poor response to HAART in IDU compared with other patient groups.

EuroSIDA is a longitudinal pan-European (including sites in Israel and Argentina) observational study of 9802 patients with HIV. The cause of death has been recorded since the beginning of the study.

The study investigators sought to describe differences in clinical progression after starting HAART according to exposure group and for different events (new AIDS/death, HIV-related death or non-HIV-related death).

Patients were classified as dying from an HIV-related event when the cause of death was indicated as HIV related, resulting from an AIDS-defining event or caused by invasive bacterial infection. All other causes of death were classified as non-HIV deaths. A sensitivity analysis categorized deaths from invasive bacterial infections as non-HIV related.

Patients included in this analysis had started HAART (two or more nucleoside reverse transcriptase inhibitors plus at least one protease inhibitor, non-nucleoside reverse transcriptase inhibitor or abacavir), had a CD4 cell count and viral load measured in the 6 months before starting HAART and some prospective follow-up. Patients may have received non-HAART antiretroviral treatment before starting HAART.

Those who were treatment naive at starting HAART were included in the Antiretroviral Treatment Cohort Collaboration. Patients were followed to their first event or their last follow-up, at latest December 2003. Poisson regression was used to determine the incidence rate ratio (IRR) of clinical progression after adjustment for confounding variables. These included age, previous AIDS diagnosis, previous antiretroviral treatment, HAART regimen started, hepatitis C status, date started HAART, and both CD4 cell count and viral load as time-updated variables.

A total of 3872 patients were included; 1767 were homosexual (45.6%), 819 were IDU (21.2%), 991 were heterosexual (25.6%) and 295 (7.6%) belonged to other exposure groups.

A total of 1514 patients were treatment naive (39.1%), and the most common HAART regimen was a single-protease inhibitor-containing regimen (n = 2758, 71.2%). The median CD4 cell count at starting HAART was 220 and viral load was 4.52 log₁₀ copies/ml). The median date of starting HAART was June 1997..

During 14 769 person-years of follow-up 499 patients progressed to new AIDS/death; 296 patients died, and of these, 194 (66.5%) were non-HIV related.

Compared with homosexual individual, after adjustment, there was a significantly increased incidence of new AIDS/deaths among IDU (P = 0.039]. However, there was no increased incidence of AIDS (P = 0.45) or HIV-related death (P = 0.99). The increase was confined to non-HIV deaths, in which IDU had over a twofold increased incidence (P < 0.0001).

The increased incidence in IDU was more pronounced when deaths from invasive bacterial infections were not included as HIV-related deaths (P < 0.0001). A significantly lower proportion of IDU were known to have died from HIV-related causes (24.0% versus 39.1% in all other exposure groups combined, P = 0.010, chi-squared test), and a significantly higher proportion were reported to have died from liver-related events (20.8 versus 4.0% in all other exposure groups combined, P < 0.0001, chi-squared test).

There were no significant differences between IDU and all other exposure groups combined in the proportion of patients who died from suicide or drug overdose, after a cardiovascular event, unknown cause or any other cause (P > 0.3 all comparisons, chi-squared test).

The EuroSIDA study has demonstrated an increased clinical progression among IDU after starting HAART. IDU are known to have an increased risk of chronic liver disease, bacterial infections, sepsis, drug overdoses and suicides, particularly among active IDU.

The increased incidence was only found for non-HIV deaths, and was significant after adjustment for relevant confounding variables.

In particular, the increased incidence could not be explained by co-infection with hepatitis C, although a higher proportion of IDU died from liver-related disease [emphasis added—Ed]. The introduction of HAART, more specific information on the cause of death and longer follow-up are likely explanations for the differences compared with earlier findings in which the researchers did not find an increased risk of death in IDU.

IDU had no increased incidence of AIDS or HIV-related death. This suggests, on average, that IDU who participated in EuroSIDA and started HAART had a comparable response to other exposure groups, were able to adhere to complex regimens, and benefited from a similar reduction in the risk of AIDS or HIV-related death.

EuroSIDA patients may not be wholly representative of those with HIV across Europe, but do represent a heterogeneous population who have been treated according to local guidelines, which will differ between centers and over time.

Conclusions

In conclusion, the authors write, “In summary, these results show the importance of collecting data on the precise cause of death. Without this information results could have been misinterpreted as IDU having an inferior response to HAART compared with other exposure groups.”

“As the proportion of non-HIV-related deaths increases in observational cohorts, there is a need to define and standardize causes of death more effectively, particularly as observational studies will increasingly be used for pharmaco-vigilance and the monitoring of serious adverse events.”

01-10-04

Reference
A Mocroft and others (or the EuroSIDA study group). Causes of death in HIV infection: the key determinant to define the clinical response to anti-HIV therapy (Research Letter). AIDS 18(17): 2333-2337. November 19, 2004.