| |
All-cause and liver-related mortality in HIV positive subjects compared to the general population: Differences by HCV co-infection......'HCV/HIV Coinfected Death Rates 11 Times Higher than in the general population in Spain'
|
| |
| |
Download the PDF here
Journal of Hepatology October 2012
Victoria Hernando1,2, Santiago Perez-Cachafeiro3, Charlotte Lewden4, Juan Gonzalez5, Ferran Segura6, Jose Antonio Oteo7, Rafael Rubio8, David Dalmau9, Santiago Moreno10, Julia del Amo1,2, CoRIS
1Red de Investigacion en Sida, Centro Nacional de Epidemiologia, Instituto de Salud Carlos III, Madrid, Spain; 2CIBER de Epidemiologia y Salud Publica (CIBERESP), Spain; 3Fundacion IDI Complexo Hospitalario de Pontevedra, Pontevedra, Spain; 4INSERM, U897, Bordeaux, France; 5Hospital Universitario La Paz, Madrid, Spain; 6Corporacio Sanitaria Parc Tauli, Sabadell, Spain; 7Hospital San Pedro-CIBIR, Logrono, Spain; 8Hospital Universitario Doce de Octubre, Madrid, Spain; 9Hospital Universitari Mutua de Terrassa, Terrassa, Spain; 10Hospital Universitario Ramon y Cajal, Madrid, Spain
"Data on HIV positive adults from two multicenter cohort studies in Spain (CoRIS-MD and CoRIS) were analyzed......We estimated standardized mortality ratio (SMR) and excess mortality, comparing death rates from our cohort (globally and by HCV co-infection) with death rates from the general population
standardized.....we have shown an excess of all-cause and liver-related mortality in HIV positive people compared to the general population in the last decade, despite improvements in the management of both HIV and HCV infections......Rapid changes in the epidemiology of HCV co-infection and major advances in the treatment of these infections are likely to shape future patterns of excess mortality in the coming years.....mortality in HCV negative subjects is two and a half times higher than in the general population while death rates in HCV positive individuals are eleven times higher......The SMR for all cause-mortality for all subjects was 5.6 (95% CI: 4.9-6.4). However, marked differences by HCV status were seen: the SMR was 2.4 (95% CI: 1.9-3.1) for HCV negative subjects and 11.5 (95% CI: 9.9-13.4) for those HCV positive (Table 2).......Having HCV co-infection and AIDS yielded an SMR of 20.8 (16.5-26.1) and having AIDS and being HCV negative had an SMR of 4.8 (3.5-6.7).......all-cause mortality in HIV positive HCV negative subjects who were AIDS-free was 1.4 times higher than the general population and increased to 20.8 times higher in HCV positive subjects who had an AIDS defining condition.....Overall, both mortality rates as SMR and excess mortality rates were higher for injecting drug users (IDUs) than men having sex with men (MSM) and heterosexuals, patients with AIDS, with and without cART and for subjects included between 1997 and 2003.....Excess mortality for HCV negative subjects was of similar magnitude in men and women, and was higher in people over 40years of age, IDUs, people with a previous AIDS diagnosis, individuals with CD4 cell counts below 200 cells, those with over 100,000 copies of HIV VL, and those on cART"
IDUs: have highest death rates among coinfected for all transmission categories, CD4 <200 has higher death rate, having a viral load <20,000 has a much lower death rate compared to high viral load (>100,000)
"We estimated standardized mortality ratio (SMR) and excess mortality, comparing death rates from our cohort (globally and by HCV co-infection) with death rates from the general population standardized by sex......In this study, we compare the overall and liver-related mortality rates observed in HIV positive subjects followed-up in the cohorts of the Spanish Network on HIV/AIDS Research - CoRIS-MD and CoRIS, stratified by HCV co-infection status, with the expected mortality in the general population of the same age and sex in Spain from 1997 to 2008. We calculate both SMRs and excess mortality rates."
"The all-cause mortality in HIV positive subjects in the Cohorts of the Spanish AIDS Research Network of Excellence between 1997 and 2008 is nearly six times higher than that of the general population of the same age and sex. However, remarkable differences are observed according to the HCV co-infection status: mortality in HCV negative subjects is two and a half times higher than in the general population while death rates in HCV positive individuals are eleven times higher. Liver-related mortality in HIV positive subjects in CoRIS is nearly ten times higher than in the general population of the same age and sex, and again, notable differences by HCV co-infection status are seen. While non-significant increases in liver-related mortality are observed in HCV negative people, liver-related mortality is twenty-two times higher in HCV infected subjects compared to the general population of the same age and sex."
"As previously reported, HIV-infected populations have higher mortality rates than the general population of the same age and sex. Our data are consistent with these findings, but they also highlight the remarkable excess mortality in subjects with HCV co-infection [3], [17], [26], [27]. We observe that the all-cause mortality in HIV positive HCV negative subjects who were AIDS-free was 1.4 times higher than the general population and increased to 20.8 times higher in HCV positive subjects who had an AIDS defining condition."
"To conclude, we have shown an excess of all-cause and liver-related mortality in HIV positive people compared to the general population in the last decade, despite improvements in the management of both HIV and HCV infections. Co-infection with HCV seems to play a very important role in all-cause and liver-related excess mortality. Rapid changes in the epidemiology of HCV co-infection and major advances in the treatment of these infections are likely to shape future patterns of excess mortality in the coming years."
"As far as we know, only one previous study has estimated SMRs in HIV positive populations by the HCV infection status [6]. Lewden et al. found that all-cause mortality in HIV positive subjects co-infected with HCV was 14 times higher than in the general population, whereas the SMR for those not co-infected was 4.4. Our results are similar"
"various studies have estimated the SMR for all-cause mortality in cohorts of HIV-infected subjects and have found that their mortality rates are 3-14 times higher than in the general population [3], [4], [5], [7]"
"Treatment for HCV infection in co-infected individuals has been more recent in our setting [38]. In a short survey of 24 European countries, Salmon et al. showed that this treatment has been prescribed to only 10% of the subjects who need it [39]. Nowadays, this treatment is widely recommended since successful treatment of chronic hepatitis C, besides preventing the development of end-stage liver disease, may also reduce the risk of subsequent liver toxicity during antiretroviral therapy in HIV/HCV-co-infected patients [40]."
"Regarding liver-related mortality, we observed an important difference between men and women. In women, liver-related SMR was 38 times higher than for women of the general population of the same age. These differences may be explained by the relative low liver-related death in women from the general population [35], together with the special characteristics of the HIV positive women; a higher prevalence of co-morbidities (29.5% of women in this study were current or former injecting drug users and 40.2% were co-infected with HCV) and their lower socio-economic level [36]."
Abstract
Background & Aims
We aimed at comparing overall and liver-related mortality rates, observed in HIV positive subjects followed-up in the Cohorts of Spanish Network on HIV/AIDS Research stratified by HCV co-infection status, with the expected mortality of the general population of same age and sex in Spain, for the period 1997 - 2008.
Methods
We estimated standardized mortality ratio (SMR) and excess mortality, comparing death rates from our cohort (globally and by HCV co-infection) with death rates from the general population standardized by sex in 5year-age bands.
Results
Overall, 5914 HIV positive subjects were included, 37.3% of which were co-infected with HCV; 231 deaths occurred, 10.4% of which were liver-related. SMR for all causes mortality for the HIV positive subjects was 5.6 (CI 95% 4.9-6.4), 2.4 (1.9-3.1) for HCV negative subjects and 11.5 (9.9-13.4) for HCV positive ones. Having HCV co-infection and AIDS yielded an SMR of 20.8 (16.5-26.1) and having AIDS and being HCV negative had an SMR of 4.8 (3.5-6.7). SMR for liver-related mortality was 1.8 (0.6-5.7) for HCV negative subjects vs. 22.4 (14.6-34.3) for HCV positive ones. Overall, both mortality rates as SMR and excess mortality rates were higher for injecting drug users (IDUs) than men having sex with men (MSM) and heterosexuals, patients with AIDS, with and without cART and for subjects included between 1997 and 2003.
Conclusions
There was an excess of all-cause and liver-related mortality in our cohorts compared with the general population. Furthermore, HCV co-infection in HIV positive patients increased the risk of death for both all causes and liver-related causes.
Introduction
Since the introduction of combined antiretroviral treatment (cART), important reductions in all cause-mortality in HIV positive subjects have been observed [1], [2]. Nevertheless, HIV positive subjects still have higher mortality rates than the general population standardized for age and sex [3], [4]. These standardized mortality ratios (SMRs) have been reported to be higher in women, injecting drug users (IDUs), people with AIDS and in those with lower CD4 cell counts [5], [6], [7]. In HIV positive subjects within 5years of seroconversion, no difference in mortality is observed as compared to the general population, and in individuals on cART with high CD4 counts, excess HIV-related mortality seems to be similar to other chronic conditions [4], [8].
This increase in life expectancy, together with the co-morbidities often associated with HIV infection such as co-infection by hepatitis B and C, alcohol, tobacco, and drug use, are responsible for increases in non-AIDS defining causes of death [2], [9], [10]. Liver-related deaths are one of the commonest causes of death in the post cART era [11], [12], [13]. HIV and HCV co-infection have detrimental effects on the natural history of each virus: HIV infection has been reported to accelerate HCV progression [14], [15], and higher mortality is described in HIV positive subjects with HCV co-infection [16], [17]. In Spain, despite the continuous decrease of HCV co-infection due to declines in the proportion of injecting drug users (IDUs), HCV prevalence in HIV positive subjects remains high, and mortality due to liver diseases in HIV populations partly replaces mortality from AIDS prior to cART [18], [19], [20], [21]. Liver-related deaths were the second commonest cause of death in CoRIS, the Cohort of the Spanish Research Network on AIDS, accounting for 10% of all events [22], [23].
Higher SMRs in HIV positive people co-infected with HCV have been reported by Lewden et al. [6]. In HIV positive populations with high HCV co-infection rates, as is the case in Spain, stratification by HCV infection would help understand whether the elevated mortality ratios, observed in HIV positive men and women of different age-groups and transmission categories as compared to the general population, differ significantly by HCV status. In this study, we compare the overall and liver-related mortality rates observed in HIV positive subjects followed-up in the cohorts of the Spanish Network on HIV/AIDS Research - CoRIS-MD and CoRIS, stratified by HCV co-infection status, with the expected mortality in the general population of the same age and sex in Spain from 1997 to 2008. We calculate both SMRs and excess mortality rates.
Discussion
The all-cause mortality in HIV positive subjects in the Cohorts of the Spanish AIDS Research Network of Excellence between 1997 and 2008 is nearly six times higher than that of the general population of the same age and sex. However, remarkable differences are observed according to the HCV co-infection status: mortality in HCV negative subjects is two and a half times higher than in the general population while death rates in HCV positive individuals are eleven times higher. Liver-related mortality in HIV positive subjects in CoRIS is nearly ten times higher than in the general population of the same age and sex, and again, notable differences by HCV co-infection status are seen. While non-significant increases in liver-related mortality are observed in HCV negative people, liver-related mortality is twenty-two times higher in HCV infected subjects compared to the general population of the same age and sex.
As previously reported, HIV-infected populations have higher mortality rates than the general population of the same age and sex. Our data are consistent with these findings, but they also highlight the remarkable excess mortality in subjects with HCV co-infection [3], [17], [26], [27]. We observe that the all-cause mortality in HIV positive HCV negative subjects who were AIDS-free was 1.4 times higher than the general population and increased to 20.8 times higher in HCV positive subjects who had an AIDS defining condition.
Despite the sustained reduction of HCV serial prevalence from 1997 onwards in our cohort of HIV positive subjects, overall HCV prevalence was 37.3%, similar to figures reported by EuroSIDA for Southern European countries [17]. HIV and HCV are well established causes for death from all causes [17], [22], [28], [29]. In addition to this, people infected with HIV and/or HCV have been reported to have higher rates of drug, alcohol, and tobacco use than the general population [26], [30], [31]. Data from a subset of CoRIS members shows that 42% of the men and 54% of the women reported current smoking [32], while these proportions were 31% and 21% for the general population between 16 and 64years of age, according to the National Health Survey [33]. Therefore, it is likely that higher smoking rates are responsible for part of the excess mortality observed in our population. Regarding alcohol use, preliminary data in our cohort show that 75% of the men and 44% of the women had drunk alcohol in the preceding 12months [32]. This is slightly lower than the figures for the general population, (80% and 57% [33]), which may be due to sick people drinking less alcohol than those in the general population. A recently published study on the cardiovascular risk in the patients from our cohort has shown a prevalence of smoking of 47% as well as of other co-morbidities, such as diabetes and hypertension (3% and 9%, respectively [34]).
SMRs for all-cause mortality were higher for people less than 40years old, and this was observed both in subjects co-infected with HCV and those not co-infected. As mortality rates in the general population rise, mortality in our cohort increasingly resembles that of the general population. As expected, excess mortality rates increased with age, highlighting the value of using both relative and absolute measures to estimate excess mortality.
Regarding liver-related mortality, we observed an important difference between men and women. In women, liver-related SMR was 38 times higher than for women of the general population of the same age. These differences may be explained by the relative low liver-related death in women from the general population [35], together with the special characteristics of the HIV positive women; a higher prevalence of co-morbidities (29.5% of women in this study were current or former injecting drug users and 40.2% were co-infected with HCV) and their lower socio-economic level [36].
As far as we know, only one previous study has estimated SMRs in HIV positive populations by the HCV infection status [6]. Lewden et al. found that all-cause mortality in HIV positive subjects co-infected with HCV was 14 times higher than in the general population, whereas the SMR for those not co-infected was 4.4. Our results are similar, although slightly lower for the HCV negative population. Various studies have estimated the SMR for all-cause mortality in cohorts of HIV-infected subjects and have found that their mortality rates are 3-14 times higher than in the general population [3], [4], [5], [7]. In interpreting these findings, the composition of the cohort, the background HCV prevalence, and the mortality rates of the general population have to be taken into account. Lower SMRs have been reported in large international cohorts made up of heterogeneous populations, whereas higher SMRs are derived from smaller cohorts with larger proportions of injecting drug users [3], [4]. Regarding excess mortality, our rate of 1.0 per 100 p-y, is slightly higher than the 0.6 per 100 p-y for the period 2004-2006, reported by the CASCADE Collaboration [8]; overall HCV prevalence in that study was lower than ours, and it was composed exclusively of seroconverters with longer periods of disease-free follow-up. Our data are consistent with previous publications reporting important reductions in mortality in the cART era in settings, such as Spain, where access to cART is universal [6], [37].
Treatment for HCV infection in co-infected individuals has been more recent in our setting [38]. In a short survey of 24 European countries, Salmon et al. showed that this treatment has been prescribed to only 10% of the subjects who need it [39]. Nowadays, this treatment is widely recommended since successful treatment of chronic hepatitis C, besides preventing the development of end-stage liver disease, may also reduce the risk of subsequent liver toxicity during antiretroviral therapy in HIV/HCV-co-infected patients [40].
There are some study limitations that merit discussion. We are aware that variables other than age and sex may account for the higher mortality rates found in our cohort, such as current use of injecting drugs. Unfortunately, this information was not available, but adjustment for this factor would probably have resulted in lower SMRs. Nevertheless, the SMRs for transmission categories other than IDUs also showed an excess mortality. Similarly, information on socio-economic status and alcohol and tobacco use is missing from both numerators and denominators, which means the SMRs found in this study may be an overestimation. Approximately 15% of the subjects with HCV infection will clear the infection spontaneously, but this can only be detected by PCR, which was available for only a minority of our patients [41]. Therefore, the assumption that these HCV antibody positive but PCR negative subjects are HCV positive, underestimates the magnitude of the association. We are aware that additional data on chronic hepatitis C and HCV treatment data would be very important, but those data were not uniformly collected during all the study period and were available only for few patients.
Deaths were coded in the same way for both numerators and denominators, based on information provided by the National Basic Death File using ICD-9 and ICD-10 coding algorithms. This is an important asset of this study as the majority of previous studies have coded deaths in cohort members using algorithms different from those used for denominators. Concordance between different death coding algorithms has been insufficiently studied [23]. With respect to subjects excluded from the analysis, we believe that they do not introduce important biases. In the first place, if we perform a sensitivity analysis considering all unknowns as positive, the SMRs are similar to the results obtained. Second, by eliminating those with less than 6months of follow-up, we are losing subjects who die shortly after cohort entry. Since people included in the cohort are new cases in hospitals, these early deaths during follow-up would be represented by people dying from AIDS who have been diagnosed very late, and these represent only a small percentage of the cohort. Finally, the association between HCV infection and liver-related death could be overestimated if the physician who certified the cause of death knew the HCV status of the subject; those with positive HCV serology would be more likely assigned a liver-related death [42]. While this bias may, indeed, account for some of the observed associations, it has to be highlighted that we obtained the cause of death for all cohort members from the National Basic Death File and that it is unlikely that all doctors certifying deaths had the information on the HCV status of the subject.
To conclude, we have shown an excess of all-cause and liver-related mortality in HIV positive people compared to the general population in the last decade, despite improvements in the management of both HIV and HCV infections. Co-infection with HCV seems to play a very important role in all-cause and liver-related excess mortality. Rapid changes in the epidemiology of HCV co-infection and major advances in the treatment of these infections are likely to shape future patterns of excess mortality in the coming years.
Results
Overall, 5914 subjects, 18,951 p-y of follow-up and 231 deaths, 24 (10.4%) of which liver-related, were observed in our cohorts. Over three quarters (76.0%) were men, 71.6% were under 40years, 31.3% were or had been IDUs, 32.4% were MSM, 30.6% were heterosexuals, and 37.3% were co-infected with HCV (Table 1). Information on baseline CD4 cell count was available for 5,229 subjects, and the median was 334 (IQR 150-540): 344 (IQR 152-547) for HCV negative and 319 (IQR 147-525) for HCV positive individuals (p=0.02). Although all patients included in the cohorts have to be naïve for antiretroviral treatment at entry, 74.7% began cART anytime during the follow-up.
Death rates, SMR, and excess mortality rates for all-cause mortality in HCV negative and HCV positive subjects
All-cause mortality was 1.22 per 100 p-y (95% CI: 1.07-1.39): 0.61 per 100 p-y (95% CI: 0.47-0.78) in HCV negative subjects and 1.96 per 100 p-y (95% CI: 1.68-2.28) in those who were HCV positive. For both HCV negative and HCV positive subjects, death rates were higher in men, subjects over 40years of age, IDUs, those with an AIDS diagnosis, baseline CD4 counts below 200 cells, HIV VL values over 100,000 copies and those on cART (Table 2).
The SMR for all cause-mortality for all subjects was 5.6 (95% CI: 4.9-6.4). However, marked differences by HCV status were seen: the SMR was 2.4 (95% CI: 1.9-3.1) for HCV negative subjects and 11.5 (95% CI: 9.9-13.4) for those HCV positive (Table 2).
Compared to the general population, higher SMRs were observed in both HCV negative and HCV positive cohort members for men and women, people over and under 40years of age, different transmission categories, with different baseline CD4 cell counts and HIV VL, with or without cART, in both cohorts. Nearly all SMRs showed statistically significant increases except those categories with a low number of observed deaths (Table 2). As expected, SMRs were markedly higher in HCV positive people than in HCV negative ones. Fig. 1 shows the SMRs observed in subjects according to HCV infection and AIDS diagnosis. Having HCV co-infection and AIDS yielded an SMR of 20.8 (16.5-26.1) and having AIDS and being HCV negative had an SMR of 4.8 (3.5-6.7).
The overall excess mortality rate for cohort members was 1.0 (95% CI: 0.8-1.1) per 100 p-y: 0.35 (95% CI: 0.26-0.49) in HCV negative subjects and 1.79 (95% CI: 1.53-2.10) in HCV positive ones. Excess mortality for HCV negative subjects was of similar magnitude in men and women, and was higher in people over 40years of age, IDUs, people with a previous AIDS diagnosis, individuals with CD4 cell counts below 200 cells, those with over 100,000 copies of HIV VL, and those on cART. The excess mortality rate was also higher in CoRIS-MD than in CoRIS. Excess mortality rates for HCV positive subjects were higher in men, people over 40years of age, IDUs, people with an AIDS diagnosis, individuals with CD4 cell counts below 200 cells, those with over 100,000 copies of HIV VL, and those on cART. Moreover, the excess mortality rate was higher in CoRIS than in CoRIS-MD (Table 2).
Death rates, SMR, and excess mortality rates for liver-related deaths
There were 24 liver-related deaths during 18,951 p-y of follow-up, leading to a mortality rate of 0.13 per 100 p-y (95% CI: 0.08-0.19), which was 0.03 per 100 p-y (95% CI: 0.01-0.09) in HCV negative subjects compared to 0.24 (95% CI: 0.16-0.38) in HCV positive ones (Table 3). Similar liver-related mortality rates were seen in men and women. Liver-related mortality was higher in subjects over 40years of age, IDUs and heterosexuals, individuals with an AIDS diagnosis, subjects with CD4 cell counts below 200, those with over 100,000 copies HIV VL, and those on cART.
The SMR for liver-related mortality for all subjects was 9.4 (95% CI: 6.3-13-9). However, marked differences by HCV status were observed: the SMR for HCV negative subjects was 1.8 (95% CI: 0.6-5.7) vs. 22.4 (95% CI: 14.6-34.3) for those HCV positive (Table 2). The SMRs for liver-related mortality were higher in women, people under 40years of age, IDUs and heterosexuals compared to MSM, those on cART, and for CoRIS-MD.
Excess liver-mortality rates were 0.11 per 100 p-y (95% CI: 0.07-0.17): 0.01 per 100 p-y (95% CI 0.00-0.07) for HCV negative subjects and 0.23 per 100 p-y (95% CI 0.15-0.36) for HCV positive ones (Table 3).
Materials and methods
Study design, setting, and participants
Data on HIV positive adults from two multicenter cohort studies in Spain (CoRIS-MD and CoRIS) were analyzed. CoRIS-MD is a retrospectively assembled multicenter cohort, from January 1997 to December 2003, and CoRIS is an ongoing prospective cohort from January 2004 to present; for these analyses, however, only subjects followed until December 31st, 2008 were included. A detailed description of both cohorts has been published previously [24], [25]. Both cohorts recruited subjects seen for the first time at any of the participating HIV care units. To be eligible for these analyses, subjects had to be cART naïve at study entry, less than 20years old, and to have had at least one HCV test and at least 6months of follow-up.
Patient characteristics
For the purpose of this study, we considered the following variables: (1) age at entry; (2) gender; (3) year of cohort entry; (4) HIV transmission category classified as heterosexual contact, men having sex with men (MSM), IDU and other or unknown risk pattern; (5) HCV serological status classified as positive or negative antibodies; (6) CD4 count at entry (measured over a period of six months from inclusion in the cohort); (7) HIV viral load at entry (measured over a period of six months from inclusion in the cohort); (8) changes in AIDS status during follow-up (Yes/No); (9) cART initiation during follow-up (Yes/No); and (10) cohort - CoRIS-MD (patients included between 1997 and 2003) and CoRIS (patients included between 2004 and 2008).
Ascertainment and classification of deaths in the cohorts and in the general population
Vital status in CoRIS-MD and CoRIS was reported by the study sites. For CoRIS-MD a cross-check with the National Death Index was performed, and for CoRIS, active surveillance for deaths is conducted at cohort level. Causes of death for all deceased subjects were obtained from the National Basic Death File (NBDF) provided by the National Statistics Institute in 2010 for the period 1997-2008. For 1997 and 1998, cause of death was coded according to the 9th revision of the International Classification of Diseases (ICD 9) and was converted to ICD 10 codes.
Death rates in the general population for all causes and liver-related causes of death were downloaded in December 2010 from the NIS webpage (www.ine.es), stratified in 5-year age groups for men and women, for deaths occurring between January 1st, 2004 and December 31st, 2008. The causes of death were also coded according to ICD 10.
We used the same procedure to classify deaths at numerators and denominators. For deaths that occurred in 1997 and 1998, we classified as liver-related mortality all causes that were provided in an aggregated form as 'viral hepatitis', 'malignant neoplasm of the liver', 'cirrhosis', and 'other chronic diseases of the liver'. For the period from 1999 to 2008, we classified as liver-related mortality all causes of death with the following ICD-10 codes: (B15-B19) viral hepatitis, (C22) malignant neoplasm of the liver and (K70-K76) diseases of the liver.
Statistical analyses
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. Individuals were followed-up from study entry to death or the administrative censoring date (31/12/2003 in CoRIS-MD and 31/12/2008 in CoRIS), whichever arose first.
We estimated SMRs by comparing all-cause mortality rates in our cohorts stratified by HCV co-infection status with all-cause mortality rates in the general population standardized by sex in 5-year age bands. To calculate death rates and SMR, AIDS, and cART were treated as time-dependent variables. We also estimated SMRs for liver-related deaths compared to liver-related mortality in the general population. We assumed a constant death rate within each 5-year age stratum. SMRs were estimated as the ratio of the observed number of deaths divided by the expected number of deaths; the expected number of deaths was calculated by applying the mortality rates of the general population to the p-y distribution of the HIV cohort. Confidence intervals at 95% level were also calculated.
Excess mortality was calculated as the difference between observed deaths in our cohorts and expected deaths according to mortality in the general population, divided by the number of p-y of follow-up. Confidence intervals at 95% level were also calculated using the exact Poisson method.
All analyses were performed using STATA 11 (StataCorp LP, College Station TX, USA).
| |
| |
| |
|
|
|
