| |
Mortality in HCV/HIV Coinfected Drug Users in Amsterdam Twice that of Drug Users - Temporal trends in mortality among people who use drugs compared with the general Dutch population differ by hepatitis C virus and HIV infection status
|
| |
| |
Download the PDF here
"We investigated whether mortality among PWUD from the ACS has come closer to that of the general Dutch population in recent calendar periods. As hypothesized, we observed a decline in mortality among PWUD compared with the general Dutch population after 1996. However, despite this decline, mortality rates among PWUD are still 11 times higher than those of the general population in the most recent calendar period. Of interest, mortality due to non-natural deaths came closer to the general Dutch population over time......the benefits of cART appeared to be less pronounced for PWUD than for MSM.....The observed decline in non-natural deaths in our study could be explained by a decrease in the popularity of injecting drug use in Amsterdam over time......During 2006-2012, only 26.9% of the ACS participants injected drugs compared with 77.6% during 1985-1989. Also, PWID might have become safer injectors over time, reducing the risk of overdose. These changes in injecting risk behavior could be ascribed to harm-reduction interventions and demographical changes, as recently demonstrated in our modeling study [31]. To summarize, a combination of factors such as availability of comprehensive harm-reduction interventions, HIV and HCV therapy and changing drug patterns among PWUD probably led to the reduction in mortality observed in the present study, especially among HCV/HIV-coinfected women. In addition, HIV-infected PWUD who survived the pre-cART period might have been the PWUD exhibiting less risk behavior and the lowest risk of dying.....the HIV-related mortality in the general Dutch population probably decreased at a faster rate than it did for PWUD from the ACS [Amsterdam Cohort Studies]"
"In line with the two studies [22,34] we found that HCV/HIV-coinfected PWUD have a higher SMR than HIV-monoinfected and HCV-monoinfected PWUD. Our overall SMR (35.9) for HCV/HIV coinfection was higher than the SMR (12.8) described by Hernando et al........The CMR [crude mortality rate] for HCV/HIV-coinfected PWUD was more than twice that of the CMR for all PWUD and had a pattern similar to the CMR for all PWUD over time (Fig. 1a and c, Table 2).
"In conclusion, in line with our hypothesis, significant declines in all-cause and non-natural morality rates were observed among PWUD compared with the general Dutch population. Women with an HCV/HIV coinfection contributed to the decline in the all-cause SMR over time. However, PWUD are still at an increased risk of dying even when uninfected with HCV and HIV. Our results also suggest that, despite the availability of cART, HIV-related deaths remain an important cause of mortality among PWUD when compared with the general Dutch population. This study reinforces the importance of a high coverage of comprehensive harm-reduction interventions combined with timely HIV and HCV treatment uptake to reduce excess mortality among PWUD."
----------------------
Temporal trends in mortality among people who use drugs compared with the general Dutch population differ by hepatitis C virus and HIV infection status
AIDS Nov 13 2014
van Santen, Daniela K.a; van der Helm, Jannie J.a; Grady, Bart P.X.a,c; de Vos, Anneke S.b; Kretzschmar, Mirjam E.E.b,d; Stolte, Ineke G.a,c; Prins, Mariaa,c
aDepartment of Research, Cluster Infectious Diseases, Public Health Service Amsterdam, Amsterdam
bJulius Center, University Medical Center Utrecht, Utrecht
cDepartment of Infectious Diseases, Center for Infection and Immunology Amsterdam (CINIMA), Academic Medical Center (AMC), Amsterdam
dCentre for Infectious Disease Control, National Institute of Public Health and the Environment, Bilthoven, the Netherlands.
Abstract
Objectives: We aimed to identify temporal trends in all-cause and cause-specific mortality rates among people who use drugs (PWUD) compared with the general Dutch population and to determine whether mortality trends differed by hepatitis C virus (HCV)/HIV (co) infection status.
Design: Longitudinal cohort study.
Methods: Using data from the Amsterdam Cohort Studies among 1254 PWUD (1985-2012), all-cause and cause-specific standardized mortality ratios (SMRs) were calculated; SMRs were stratified by serological group (HCV/HIV-uninfected, HCV-monoinfected, and HCV/HIV-coinfected) and calendar period. Temporal trends were estimated using Poisson regression.
Results: The overall all-cause SMR was 13.9 (95% confidence interval 12.6-15.3). The SMR significantly declined after 1996, especially due to a decline among women (P < 0.001). The highest SMR was observed among HCV/HIV-coinfected individuals during 1990-1996 (SMR 61.9, 95% confidence interval 50.4-76.0), which significantly declined after this period among women (P = 0.001). In contrast, SMR for HCV-monoinfected, and HCV/HIV-uninfected PWUD did not significantly change over time. The SMR for non-natural deaths significantly declined (P = 0.007), whereas the SMR for HIV-related deaths was the highest during all calendar periods.
Conclusions: We found evidence for declining all-cause mortality among PWUD compared with the general population rates. Those with HCV/HIV-coinfection showed the highest SMR. The decline in the SMR seems to be attributable to the decline in mortality among women. Mortality rates due to non-natural deaths came closer to those of the general population over time. However, HIV-related deaths remain an important cause of mortality among PWUD when compared with the general Dutch population. This study reinforces the importance of harm-reduction interventions and HCV/HIV treatment to reduce mortality among PWUD.
Introduction
People who use drugs (PWUD) are at an increased risk of premature mortality compared with the general population. In the Netherlands, the start of a major heroin epidemic occurred during the 1970s [1,2]. The prevalence of hepatitis C virus (HCV) peaked during the 1980s, when more than 85% of ever-injectors tested HCV-positive [3,4]. The HIV epidemic commenced during the 1980s [5,6]. In 1985, the prospective Amsterdam Cohort Studies (ACS) among PWUD were initiated to track the HIV epidemic [5].
To minimize the damage PWUD inflict upon themselves and society, harm-reduction interventions were implemented in Amsterdam, the Netherlands, from 1979, making it one of the first countries to do so [7,8]. These included methadone programs and needle exchange programs (NEPs) [7,8]. Especially, the combination of methadone programs and NEP uptake has been associated with a lower risk of HIV and HCV acquisition among people who inject drugs (PWID) [9]. For HIV-infected individuals, combination antiretroviral therapy (cART) became widely available from 1996, which has been shown to substantially reduce HIV-related mortality [10-12]. However, HIV-infected PWID tend to initiate cART at a later stage than other risk groups [13], leading to a diminished probability of survival [14-16]. For HCV-infected PWUD, HCV treatment became relatively easy to access in Amsterdam from 2005, when a special unit was launched to treat HCV-infected PWUD [17], and might have reduced liver-related mortality, which generally requires two to three decades of chronic HCV infection [18].
Previous studies have shown that PWUD have a 4.4-47.6 higher risk of dying than the general population [19-24]. This increased risk of dying differs between studies and countries, which could be ascribed to differences in calendar period of follow-up, study population characteristics, drug-use practices, HIV/HCV incidence and background prevalence, and availability and uptake of harm-reduction interventions [20].
As the ACS have reached 27 years of follow-up, which is one of longest worldwide, we had the unique opportunity to identify changes in mortality rates over different calendar periods. Trends in mortality rates can serve as a proxy for changing health patterns and the effectiveness of health-related interventions that have been implemented.
We hypothesized that because of the introduction and availability of harm-reduction interventions and HCV and HIV therapy, PWUD have reached mortality rates closer to those of the general Dutch population in recent calendar periods. Our study aimed to identify temporal trends in all-cause and cause-specific mortality rates among PWUD compared with the general Dutch population; and to determine whether excess mortality trends differed by HCV/HIV status.
Discussion
We investigated whether mortality among PWUD from the ACS has come closer to that of the general Dutch population in recent calendar periods. As hypothesized, we observed a decline in mortality among PWUD compared with the general Dutch population after 1996. However, despite this decline, mortality rates among PWUD are still 11 times higher than those of the general population in the most recent calendar period. Of interest, mortality due to non-natural deaths came closer to the general Dutch population over time.
The decline in the SMR among PWUD seems to be mainly attributable to the decline in mortality after 1996 among HCV/HIV-coinfected women; this is in line with a study among PWID in Norway, which showed that compared to men, women had a lower risk of mortality in the long term, although in the short term - within 3 years of inclusion in the study -, women had a higher risk of mortality [26]. One explanation for the different mortality trends among men and women might be that women sought HIV and/or drug treatment earlier than men and were less likely to be imprisoned [26]. Furthermore, the decline among HCV/HIV-coinfected PWUD might be explained by the availability of cART from 1996 onwards. In line with our findings, a study among HIV seroconverters showed that overall and cause-specific mortality rates decreased after the introduction of cART [27]. However, the benefits of cART appeared to be less pronounced for PWUD than for MSM [27]. Nonetheless, cART availability alone cannot explain the decrease of the SMR observed, as a more constant SMR would be expected because the positive effects of cART have also reduced mortality among HIV-infected individuals in the general Dutch population. Hence, the decline of the SMR after 1996 is also likely to be attributable to the significant decline in non-natural deaths over time. This is in contrast to the findings from the 2008 Annual Report by the European Monitoring Centre for Drugs and Drug Addiction, in which a rebound of overdose mortality was observed in many European countries from 2003 to 2005 [28]. The observed decline in non-natural deaths in our study could be explained by a decrease in the popularity of injecting drug use in Amsterdam over time [29,30]. During 2006-2012, only 26.9% of the ACS participants injected drugs compared with 77.6% during 1985-1989. Also, PWID might have become safer injectors over time, reducing the risk of overdose. These changes in injecting risk behavior could be ascribed to harm-reduction interventions and demographical changes, as recently demonstrated in our modeling study [31]. To summarize, a combination of factors such as availability of comprehensive harm-reduction interventions, HIV and HCV therapy and changing drug patterns among PWUD probably led to the reduction in mortality observed in the present study, especially among HCV/HIV-coinfected women. In addition, HIV-infected PWUD who survived the pre-cART period might have been the PWUD exhibiting less risk behavior and the lowest risk of dying.
Of interest, the HIV-related CMR among PWUD significantly and substantially decreased after 1996, whereas the HIV-related SMR did not significantly change over time. HIV-related mortality is dependent on the incidence and prevalence of HIV in a population. Therefore, given our higher proportion of HIV-infected PWUD compared with the general Dutch population, as expected, we observed a high HIV-related SMR. However, even though the proportion of HIV-infected PWUD in our study population decreased over calendar time, no decrease was observed in the HIV-related SMR. Therefore, the HIV-related mortality in the general Dutch population probably decreased at a faster rate than it did for PWUD from the ACS. HCV-coinfection, which is common among our study population, might play a role as it has been shown that HCV/HIV-coinfected individuals have a higher risk of death from HIV/AIDS than HIV-monoinfected individuals [12,32]. Also, even though the benefits of cART have been observed in all risk groups [33], the HIV-positive individuals from exposure groups other than PWUD might have easier access to care, be more adherent to cART, have a better socioeconomic status, and a healthier lifestyle.
In line with the two studies [22,34] we found that HCV/HIV-coinfected PWUD have a higher SMR than HIV-monoinfected and HCV-monoinfected PWUD. Our overall SMR (35.9) for HCV/HIV coinfection was higher than the SMR (12.8) described by Hernando et al.[22], but was very similar to the findings by McDonald et al. (34.0) [34]. The difference in SMR between the studies might be explained by differences in the study period. Hernando et al.'s study comprises data from the cART era (1997-2010), whereas both our and McDonald et al.'s study had follow-up both during the pre-cART and the cART era [22,34].
The SMR for liver-related deaths did not significantly change over time. This could be attributed to an increase in liver-related deaths in the general Dutch population [35]. Furthermore, HCV/HIV-coinfected PWUD might have died of HIV-related causes before experiencing the consequences of HCV [4]. However, the burden of HCV-related disease among PWUD in our study was made visible by the increased liver-related CMR during the latest calendar period.
We found that the CMR for natural causes significantly increased over time. This can be partly explained by ageing of our cohort participants and the lower risk of HIV-related mortality among PWUD in the cART era. However, if CODs are misclassified (i.e. underlying HCV and HIV-related cause of death is not recognized), this might contribute to the increasing CMR for natural deaths.
Nevertheless, the SMR for natural causes did not significantly change over time.
Several limitations of our study should be mentioned. First, even though we adjusted for age group, sex, calendar period, and serological group, other determinants, such as active drug use, alcohol consumption, hepatitis B, or smoking, were not taken into account. Second, PWUD who were ever positive for HCV antibodies were considered HCV-positive for the whole study period, thus spontaneous or treatment-induced HCV clearance was not taken into account. Although the minority, especially in HCV/HIV-coinfected individuals, clears the virus spontaneously, this definition of HCV positivity might have led to an underestimation of the SMR for PWUD with a chronic HCV infection. However, active PWID who clear the virus are at risk of HCV re-infection. Third, although coverage and components of harm-reduction interventions changed during the study period, they were already implemented before the ACS started; therefore we do not have a proper comparison with a period without such interventions. Furthermore, our sample of PWUD might not be representative for the general PWUD Dutch population.
In conclusion, in line with our hypothesis, significant declines in all-cause and non-natural morality rates were observed among PWUD compared with the general Dutch population. Women with an HCV/HIV coinfection contributed to the decline in the all-cause SMR over time. However, PWUD are still at an increased risk of dying even when uninfected with HCV and HIV. Our results also suggest that, despite the availability of cART, HIV-related deaths remain an important cause of mortality among PWUD when compared with the general Dutch population. This study reinforces the importance of a high coverage of comprehensive harm-reduction interventions combined with timely HIV and HCV treatment uptake to reduce excess mortality among PWUD.
Methods
Study population
The ACS among (injecting and non-injecting) PWUD is an open, prospective cohort study initiated in 1985, aiming to investigate the epidemiology, natural history, and pathogenesis of HIV, and to evaluate the effect of interventions. Participation in the ACS is voluntary and written informed consent is obtained at intake [25]. Recruitment is by means of local methadone outposts, a sexually transmitted diseases clinic, and word of mouth. PWUD visit the Public Health Service of Amsterdam every 4-6 months; they give blood and complete a standardized questionnaire about their health and sexual and drug use behavior during each cohort visit. To encourage continued participation, 12 Euros were paid per follow-up visit. Participants, aged between 20 and 64 years with at least two cohort visits and with both longitudinal HCV and HIV test results (1254/1661), were included in our study. This study was approved by the Medical Ethical Committee of the Amsterdam Medical Centre, the Netherlands.
Laboratory methods
At entry, the ACS participants were tested for HIV antibodies by ELISA and at every follow-up visit if previously negative. In 2005, participants with at least two cohort visits between December 1985 and January 2005 were retrospectively tested for HCV antibodies with a third-generation ELISA (AxSym HCV version 3.0; Abbott, Wiesbaden, Germany). Individuals who were anti-HCV-negative at entry were tested for HCV antibodies at the most recent visit. On finding HCV seroconversion (defined as the presence of HCV antibodies in a previously HCV-seronegative individual), we tested samples between the first and the most recent visit to obtain the most exact seroconversion interval. From 2005 onwards, HCV testing occurred prospectively.
Definitions and statistics
Information about vital status was obtained by matching the ACS data at regular intervals against the municipal and national population registries in the Netherlands. Causes of death (CODs) were systematically obtained from hospital records, general practitioners, the national HIV Monitoring Foundation, or coroners. The CODs were divided into four categories: HIV-related deaths; liver-related deaths; natural causes other than HIV or liver-related deaths (e.g. cardiovascular-related deaths, non-AIDS-related cancers); and non-natural causes (i.e. overdose, accidents, homicide, and suicide).
Follow-up time was calculated from the first cohort visit until: loss to follow-up, last known date to be alive, death, or cohort censoring date (i.e. 31 December 2012). Calendar periods were defined as follows: period 1 (1985-1989) - pre-cART and before adequate methadone dose (≥60 mg/day) was generally available in Amsterdam; period 2 (1990-1996) - pre-cART with adequate methadone substitution therapy; period 3 (1997-2000) - early cART era; period 4 (2001-2005) - late cART era; period 5 (2006-2012) - late cART and HCV treatment era.
Four HCV/HIV serological groups were defined: HCV/HIV-uninfected, HIV-monoinfected, HCV-monoinfected, and HCV/HIV-coinfected. HIV and HCV status were treated as time-dependent variables. For HIV and HCV seroconverters, the midpoint between the last negative and the first positive antibody test was used to estimate the moment of seroconversion. After the last cohort visit, the serological group was carried forward until the end of follow-up. In a sensitivity analysis, a more stringent censoring strategy was applied (i.e. censoring took place at the end of the calendar period of the last cohort visit) to check whether results were robust.
Crude mortality rates (CMRs) per 1000 person-years, including 95% confidence intervals (CIs), were calculated. Standardized mortality ratios (SMRs) were used to compare the mortality rate among the PWUD with the mortality rate in the general Dutch population. SMR is the ratio of the observed deaths among PWUD from the ACS and the expected number of deaths. Expected deaths were calculated by multiplying the person-years accrued from our study population by the mortality rate of the general Dutch population matched by age group, sex, and calendar period.
Mortality rates of the general Dutch population were obtained from the Human Mortality Database (HMD) (www.mortality.org), and the numbers of deaths per COD were obtained from Statistics Netherlands (www.cbs.nl). To calculate the mortality rate of the general Dutch population per calendar period, age group, and sex, the following formula was used:
where MR is the mortality rate; GP the general population; and μ the average.
The SMRs for each calendar period, serological group, and COD were calculated using univariable Poisson regression models with the natural logarithm of the expected deaths as offset term; the exponential of the coefficient of the Poisson model is the SMR. Multivariable Poisson models, with the same offset term, were used to obtain the effect of calendar period and serological group on the SMR and are expressed in SMR ratios; SMR ratios can be interpreted as a relative SMR. P values for trends were obtained from the multivariable models. Multivariable models were corrected for age group, sex and serological group. We additionally checked for interactions between calendar period and the other covariates. For the calculation of the SMR per cause of death, PWUD were included in the analysis irrespective of their HCV/HIV status. Stata version 11.2 was used (Stata Statistical Software: Release 11; Stata Corp LP, College Station, Texas, USA).
Results
Out of 1254 PWUD, at entry, 63.9% (n = 801) were men, their median age was 30 years [interquartile range (IQR) 26-36], and 72.4% (n = 908) had ever injected drugs (Table 1). The median follow-up time was 15.0 years (IQR 9.5-20.9). At the study entry, the proportion of those who ever injected was similar among those included (i.e. at least two cohort visits) and excluded in this study. However, individuals included in our study were older, more often of Dutch nationality, less often homeless and a higher proportion of them were men compared to the excluded PWUD.
The characteristics and demographical distribution of PWUD changed across the calendar periods (Table 1). During the first calendar period, 82.4% of PWUD were aged between 20 and 34 years compared with only 11.9% in the most recent calendar period. The percentage of participants who ever injected drugs or injected drugs during the calendar period decreased over time. During the first calendar period, 20.2% were HCV/HIV-uninfected, 1.3% HIV-monoinfected, 52.3% HCV-monoinfected, and 26.3% HCV/HIV-coinfected. During the latest calendar period, 38.8% were HCV/HIV-uninfected, 1.8% HIV-monoinfected, 45.6% HCV-monoinfected, and 13.9% HCV/HIV-coinfected. During 18 575 person-years of total follow-up time, 96 PWUD seroconverted for HIV and 54 for HCV.
Overall mortality rates
Among 1254 PWUD, 406 deaths were observed during the study period. The overall CMR for all-cause mortality was 21.9 per 1000 person-years (95% CI 19.8-24.1). The overall effect of calendar period on the CMR was significant (P = 0.003) (Table 2). The highest CMR was observed between 1990 and 1996 (27.5 per 1000 person-years; 95% CI 23.1-32.6) (Table 2). The CMR declined from 27.5 in 1990-1996 to 17.4 in 2001-2005, and was followed by a slight increase to 23.7 in the period 2006-2012 (Fig. 1a).
The overall SMR was 13.9 (95% CI 12.6-15.3), meaning that drug users from the ACS had a 13.9 times higher mortality rate compared with the age, sex, and calender-matched general Dutch population. During 1985-1990, the SMR was 15.8 (95% CI 9.7-25.9), it increased to 25.5 (95% CI 21.5-30.3) in 1990-1996, followed by a decline to 14.1 (95% CI 11.2-17.8) in 1997-2000, 9.9 (95% CI 7.9-12.3) in 2001-2005, and 10.8 (95% CI 9.0-13.1) in 2006-2012 (Fig. 1b, Table 3). The SMR decreased with increasing age (SMR20-34 = 22.9, 95% CI 18.7-28.2; SMR35-49 = 15.5, 95% CI 13.7-17.5; SMR50-64 = 7.3, 95% CI 5.8-9.3) and was higher among women (SMRmen = 12.8, 95% CI 11.3-14.4; SMRwomen = 16.6, 95% CI 14.0-19.6).
In multivariable analysis, a significant interaction between calendar period and sex (P = 0.03) was observed. The SMR significantly decreased in later calendar periods among women [adjusted SMR ratio (aSMRr)2006-2012 = 0.6, 95% CI 0.4-0.9, compared to 1990-1996; P < 0.001] (Table 3), while remaining stable among men. HCV-monoinfected (aSMRr = 1.9, 95% CI 1.4-2.8) and HCV/HIV-coinfected (aSMRr = 7.0, 95% CI 5.0-9.8) PWUD had a significantly higher SMR compared with HCV/HIV-uninfected PWUD (P < 0.001) (Table 3).
All-cause mortality per serological group
Hepatitis C virus/HIV-uninfected people who use drugs
Although the overall effect of calendar period was not significant (P = 0.16), the CMR for HCV/HIV-uninfected PWUD slightly increased from 2001 to 2005 onwards (Fig. 1c, Table 2). The overall SMR was 4.7 (95% CI 3.5-6.5) (Table 3). As illustrated in Fig. 1d, the SMR slightly increased after 2000, although the overall effect of calendar period was not statistically significant in multivariable analysis (P = 0.54) (Table 3).
HIV-monoinfected people who use drugs
The overall CMR for HIV-monoinfected PWUD was 37.9 (95% CI 19.0-75.9) (Table 2) and the overall SMR was 26.3 (95% CI 13.2-52.6). Given the small number of HIV-monoinfected PWUD and deaths among them, the SMRs per calendar period were not calculated for them.
Hepatitis C virus-monoinfected people who use drugs
The CMR for HCV-monoinfected PWUD was borderline significantly different across the calendar periods (P = 0.05) and was the highest between 2006 and 2012 (CMR = 22.7, 95% CI 17.1-30.0) (Fig. 1c, Table 2). The overall SMR was 9.5 (95% CI 8.1-11.3) (Table 3). As illustrated in Fig. 1d, the SMR decreased until 2001-2005 and was followed by a slight increase in 2006-2012, although the overall effect of calendar period was not statistically significant in the multivariable analysis (P = 0.23) (Table 3).
Hepatitis C virus/HIV-coinfected people who use drugs
The CMR for HCV/HIV-coinfected PWUD significantly differed across the calendar periods (P = 0.03) and was the highest between 1990 and 1996 (CMR = 67.5, 95% CI 55.0-82.9), which is also the highest CMR observed in this study (Fig. 1c, Table 2). The CMR for HCV/HIV-coinfected PWUD was more than twice that of the CMR for all PWUD and had a pattern similar to the CMR for all PWUD over time (Fig. 1a and c, Table 2). The overall SMR was 35.9 (95% CI 31.5-41.0) (Table 3). HCV/HIV-coinfected PWUD had the highest SMR during all calendar periods compared with the other serological groups and especially during the period 1990-1996 (SMR = 61.9, 95% CI 50.4-76.0) (Fig. 1d). As illustrated in Fig. 1d, the SMR declined after 1990-1996 and remained stable in the two most recent periods. In multivariable analysis, the effect of calendar period was significantly different among women and men (P = 0.03). Among women, the SMR was significantly lower in the last three calendar periods compared with 1990-1996 (P < 0.001), whereas the SMR remained relatively stable over time among men (Table 3).
Mortality per cause of death
The most common CODs were natural deaths (n = 111), followed by non-natural (n = 107), HIV-related (n = 87), and liver-related deaths (n = 30).
Natural causes
The CMR for natural deaths significantly increased over time (P = 0.001) - from 1.7 (95% CI 0.4-6.7) between 1985 and 1989 to 10.0 (95% CI 7.5-13.4) between 2006 and 2012 (Fig. 1e, Table 2). The overall SMR was 5.0 (95% CI 4.1-6.0) and was the lowest of the cause-specific SMR. As illustrated in Fig. 1f, the SMR for natural causes decreased after 1990-1996 (Fig. 1f), although the overall effect of calendar period was not significant (P = 0.38) (Table 4).
Non-natural causes
In contrast to the CMR for natural causes, the CMR for non-natural deaths significantly decreased over time (P = 0.001) - from 9.2 (95% CI 5.1-16.6) in 1985-1989 to 3.0 (95% CI 1.8-5.1) in 2006-2012 (Fig. 1e, Table 2). The overall SMR for non-natural deaths was 21.3 (95% CI 17.6-25.7), and we observed a steady decline of the SMR over time (P = 0.007) (Fig. 1f, Table 4).
HIV-related causes
The CMR for HIV-related deaths reached its peak in the period 1990-1996 (CMR = 11.0, 95% CI 8.4-14.4), followed by a statistically significant decline after 1996 (P < 0.001) (Fig. 1e, Table 2). The overall SMR was 798.2 (95% CI 647.0-984.9). As illustrated in Fig. 1f, we observed a higher HIV-related SMR after 1990, although the overall effect of calendar period was not significant in the multivariable analysis for men (P = 0.30) and bordeline signficant for women (P = 0.06) (Table 4).
Liver-related causes
The overall CMR for liver-related deaths was 1.6 (95% CI 1.1-2.3) and did not significantly differ across the calendar periods (P = 0.38) (Table 2). The overall SMR for liver-related deaths was 72.4 (95% CI 50.6-103.5). As illustrated in Fig. 1f, the SMR for liver-related deaths decreased after 2000, although the overall effect of calendar period was not significant in the multivariable analysis (P = 0.11) (Table 4).
Finally, in a sensitivity analysis with a more stringent censoring strategy, comparable results were observed.
|
|
| |
| |
|
|
|
