All-cause mortality in the cohorts of the Spanish AIDS Research Network (RIS) compared with the general population: 1997–2010

Background Combination antiretroviral therapy (cART) has produced significant changes in mortality of HIV-infected persons. Our objective was to estimate mortality rates, standardized mortality ratios and excess mortality rates of cohorts of the AIDS Research Network (RIS) (CoRIS-MD and CoRIS) compared to the general population. Methods We analysed data of CoRIS-MD and CoRIS cohorts from 1997 to 2010. We calculated: (i) all-cause mortality rates, (ii) standardized mortality ratio (SMR) and (iii) excess mortality rates for both cohort for 100 person-years (py) of follow-up, comparing all-cause mortality with that of the general population of similar age and gender. Results Between 1997 and 2010, 8,214 HIV positive subjects were included, 2,453 (29.9%) in CoRIS-MD and 5,761 (70.1%) in CoRIS and 294 deaths were registered. All-cause mortality rate was 1.02 (95% CI 0.91-1.15) per 100 py, SMR was 6.8 (95% CI 5.9-7.9) and excess mortality rate was 0.8 (95% CI 0.7-0.9) per 100 py. Mortality was higher in patients with AIDS, hepatitis C virus (HCV) co-infection, and those from CoRIS-MD cohort (1997–2003). Conclusion Mortality among HIV-positive persons remains higher than that of the general population of similar age and sex, with significant differences depending on the history of AIDS or HCV coinfection.


Background
Mortality of HIV-infected persons in Western countries has decreased significantly due to improvements in combined antiretroviral therapy (cART) [1,2]. Nevertheless it continues to be higher than in the general population [3][4][5], even in HIV-infected patients with good initial response to cART [6]. Global reduction in mortality has been achieved thanks to a decrease of AIDSrelated deaths which has led to a greater relevance of other causes of death in relation to co-morbidities, such as hepatitis C virus (HCV) and/or hepatitis B virus (HBV) co-infections, drug abuse and cardiovascular diseases [2,7].
In Barcelona and Navarre, HIV-positive subjects were found to have a higher mortality compared to the general population [8,9] but no estimates are available for the whole country. Unlike other cohorts, in this work we have analyzed data of a cohort of persons with HIV infection recruited during a period where highly effective antiretroviral treatment is available and all patients are naïve to treatment. But we believe that even though these patients may be in a better starting point than patients in other similar studies, the risk of mortality compared with the general population is still higher.
Therefore, the objectives of this study were to calculate the overall mortality rates, standardized mortality ratios (SMR), and excess mortality rates in the cohorts of the Spanish AIDS Research Network (RIS) -CoRIS-MD and CoRIS, comparing the overall mortality rates observed in HIV positive subjects in both cohorts with the mortality rates of the general population of similar age and sex.

Patients
We analyzed data from the cohorts of HIV-infected adults of the Spanish AIDS Research Network (RIS). CoRIS-MD is a multicenter cohort including data from 1997 to 2003 from 9 hospitals of 7 Spanish Autonomous regions assembled in 2003. CoRIS is a multicenter cohort which recruits patients from 2004 onwards from 28 health-care centers and hospitals in 12 of the 17 Autonomous regions that compose Spain [10,11]. Both cohorts recruit patients newly attended in any of the participating sites. Ethics approval was obtained from all hospitals Ethics' Committees (see Appendix 1 all hospitals participants) and every patient provides written informed consent to participate in the cohorts. For this analysis, we selected subjects who were naïve to cART at cohort entry, older than 20 years, had a follow up of more than 6 months and had had at least one diagnostic test for hepatitis C virus.
To calculate mortality rates, AIDS variable was classified as "Yes" when the person had AIDS before entering the cohort, AIDS at cohort entry or AIDS during followup and "No" when the person didn't develop AIDS at any moment during the study.

Statistical analyses
Descriptive analysis of patients' characteristics was carried out using frequency distribution for categorical variables and median (interquartile range -IQR) for continuous variables.
Individuals were followed up from study entry to death, last study contact or the administrative censoring date (31/12/2003 in CoRIS-MD and 31/12/2010 in CoRIS) whichever arose first. We calculated mortality rates, overall and according to socio-demographic and clinical characteristics, as the number of deaths by 100 persons-year (py) of follow-up with 95% confidence intervals (95% CI) calculated using the exact Poisson method.
Standardized mortality ratios (SMR) were estimated for all-cause mortality in CoRIS-MD and CoRIS, comparing with the overall mortality rates of the general population standardized by sex and age. SMR were estimated as the ratio of observed deaths to expected deaths, had our patients had the same distribution of mortality as the general population. SMR were calculated through Poisson models offsetting expected mortality rates, and adjusted for gender, age, category of transmission and HCV test. Mortality rates for general population, between 1997 and 2010, were obtained from the National Statistics Institute (www.ine.es), stratified by sex and age at 5 year intervals. A constant mortality rate within each 5 year stratum was assumed.
A sensitivity analysis was performed to assess a possible selection bias. The SMR was calculated for the first 12 months after cohort entry separately for all patients together. This was to determinate whether it is necessary to include a lag time to avoid an overestimation of SMR.
Excess Mortality Rates were calculated as the difference between observed and expected deaths according to mortality in the general population, divided by the number of persons-year (py) of follow-up. Confidence intervals for Excess Mortality Rates were estimated using Poisson's exact method.
In the sensitivity analysis, considering only the first 12 month of follow-up, SMR is lower than in the complete analysis (4.0; 95% CI 2.4 -5.6).
Finally, regarding excess mortality rate, as an absolute estimator, results are similar to those observed for crude mortality rates (Figure 1).

Discussion and conclusion
Our results show that all-cause mortality in CoRIS-MD and CoRIS cohorts, between 1997 and 2010, is close to seven times higher than that of the general population of the same age and sex. Significant differences have been found depending on the history of AIDS and HCV co-infection.
A previously published study, carried out in similar cohorts in Europe and North America, found a lower global SMR, of 3.36 (95% CI: 3.16 -3.56), but with a notable heterogeneity between cohorts depending on participant-specific characteristics, and being higher for cohorts with a greater representation of IDUs [12]. For example, Aldaz et al. found mortality of HIV-infected persons in Navarre (Spain) to be 14 times higher than mortality in general population; 63% of this cohort had been infected through the use of injected drugs [8].
These differences could also be related to the higher prevalence of HCV-co-infection as the standardized mortality in HCV co-infected subjects in our study was 9.2 times higher than the general population's. Similar results were found by Lewden et al., where SMR for HCV co-infected persons were 13.9 compared to 4.4 for the HCV negative subjects [4]. In a previous study of CoRIS-MD and CoRIS cohorts, an important increase of the risk of both all cause mortality and liver-related mortality was observed for HIV patients coinfected with HCV [13]. Berenguer et al. also found a decrease in overall mortality in HIV patients in cART era, but only in HCV negative subjects [14] and Chen et al. in a meta-analysis found that the risk of mortality was increased in HCV/HIV coinfected patients in HAART era [15]. In our study, we found a similar SMR for patients recruited in CoRIS, from 2004 onwards, and those recruited in CoRIS-MD, from 1997 to 2003, after adjustment for gender, age, transmission category and HCV infection. That is, the difference in the subject's characteristics along these years, the decrease in the representation of IDUs and the percentage of HCV co-infected subjects [11,16,17] were corrected after adjustment. Others studies observed a lower mortality in recent years with the improvement in antiretroviral therapies [18][19][20][21], although when specific groups were analyzed, for example: IDUs, found that mortality risk remain elevated [21].
We found non-statistically significant, lower mortality rates in women compared to men. Eventhough the women in our study showed a mortality ratio 10.5 times higher than women of the same age from the general population, and almost doubled the one from men in the cohorts. This higher relative mortality in women could be explained by the fact that women in the general population have a higher life expectancy than men, and specifically, mortality in the general population is very low in women between ages 30 to 40, where we find the majority of HIV-infected women [22]. The lower excess mortality rate in women is consistent with the higher proportion of HIV-infected men in the Spanish epidemic, and in our cohorts [23].
A possible limitation in the calculation of SMR could be using mortality rates in the general population to calculate the expected deaths, because this population contains HIV-related deaths. In our analysis, HIV-related mortality represents a small proportion of all-cause mortality in the general population of Spain, so therefore we consider correct to use the general population mortality rates to calculate the mortality rates in a non-HIV infected population.
The sensitivity analysis shows that when we establish as inclusion criteria to have at least 6 months of followup, we are introducing a time window to avoid the selection bias indirectly and overestimate SMRs.
To conclude, mortality in HIV-infected persons continues to be higher than that of the general population, although it has decreased in recent years. For future studies, we would highly recommend to consider, along with global mortality, excess mortality rate for specific causes of death, such as hepatic, non-aids related malignancies or drug-related, especially among IDUs.
Fieldwork, data management and analysis: Paz