Trends and causes of mortality in a population-based cohort of HIV-infected adults in Spain: comparison with the general population

Combination antiretroviral therapy reduces mortality of HIV-infected persons. In Spain, where this therapy is widely available, we aim to evaluate mortality trends and causes of death in HIV-infected adults, and to estimate the excess mortality compared to the general population. From 1999 to 2018 mortality by causes was analyzed in a population-based cohort of adults aged 25 to 74 years diagnosed with HIV infection in Spain. Observed deaths and expected deaths according mortality in the general population of the same sex and age were compared using standardized mortality ratios (SMRs). HIV-infected people increased from 839 in 1999–2003 to 1059 in 2014–2018, median age increased from 37 to 47 years, the annual mortality rate decreased from 33.5 to 20.7 per 1000 person-years and the proportion of HIV-related deaths declined from 64% to 35%. HIV-related mortality declined from 21.4 to 7.3 (p < 0.001), while non-HIV-related mortality remained stable: 12.1 and 13.4 per 1000, respectively. Mortality decreased principally in persons diagnosed with AIDS-defining events. In the last decade, 2009–2018, mortality was still 8.1 times higher among HIV-infected people than in the general population, and even after excluding HIV-related deaths, remained 4.8 times higher. Excess mortality was observed in non-AIDS cancer (SMR = 3.7), cardiovascular disease (SMR = 4.2), respiratory diseases (SMR = 7.9), liver diseases (SMR = 8.8), drug abuse (SMR = 47), suicide (SMR = 5.3) and other external causes (SMR = 6). In conclusion, HIV-related mortality continued to decline, while non-HIV-related mortality remained stable. HIV-infected people maintained important excess mortality. Prevention of HIV infections in the population and promotion of healthy life styles in HIV-infected people must be a priority.

trends in mortality. All-cause crude mortality fell by 38% from 33.5 in 1999-2003 to 20.7 per 1000 person-years (PY) in 2014-2018. Crude mortality declined by 3% annually and mortality adjusted for sex, age, foreign-born and transmission category declined by 5% (adjusted rate ratio [aRR] = 0.95; p < 0.001). In 1999-2003, mortality rate was higher in males than females (38.8 vs 22.9 per 1000 PY; p = 0.013); however, the more pronounced decline in mortality in males (aRR = 0.95, p < 0.001) than in females (aRR = 0.98, p = 0.323) resulted in similar mortality rates in 2014-2018 (21.0 and 20.1 per 1000 PY, respectively; p = 0.859). Declines in adjusted mortality rates were similar regardless the age group, history of injecting drug use or the foreign-born. All-cause mortality decreased regardless if HIV diagnosis had been before or in or after 1997 (aRR = 0.95, p = 0.0068 and aRR = 0.96, p = 0.013, respectively) ( Table 2).
In the 2009-2018 decade, mortality in HIV-infected population was statistically significant higher than the expected for all non-HIV-related causes of death considered. The most pronounced SMRs were observed in drug addiction or overdose (46.7 times higher), liver disease (8.8 times higher) and respiratory disease (7.9 times higher); however, in absolute terms, the highest excess mortality rates were observed in non-AIDS cancer (3.3 deaths per 1000 PY) and cardiovascular disease (1.5 per 1000 PY). Excess mortality rates were quite similar in males (19.0 per 1000 PY) and females (17.4 per 1000 PY); however, this resulted in a higher SMR for females (16.1) than for males (6.7). Excess mortality was more pronounced in HIV-infected persons with history of injecting drug use. In this group, all-cause mortality was 11 times higher and non-HIV related mortality was 6.9 times higher than in the general population, while they were 5.7 and 3.1 times higher, respectively, among people without history of injecting drug use (Table 5).

Discussion
From 1999 to 2018, the adjusted mortality rate in a population-based cohort of people with a diagnosis of HIV infection declined by 5% annually, following the trend that had begun with the introduction of highly active antiretroviral therapies 3,10 . This trend has contributed to reduce the excess mortality in the HIV-infected population. Nevertheless, in the 2014-2018 period, mortality still was 7.4 times higher than in the general population of the same sex and age. HIV-infected men who had sex with men showed the lowest excess mortality as compared to male general population, probably due to the earlier diagnoses and lower probability of other risk factors and comorbidities than in other transmission categories; however their mortality was still 4.3 times higher than expected in the last period. Although persons diagnosed with HIV infection in 1997 or after, period of use of combined antiretroviral therapy, had lower mortality than those diagnosed before, they still had 6.1 times higher mortality than the general population. These findings suggest that one part of the excess mortality associated to HIV-infection may be very difficult to control even with wide availability of antiretroviral treatments, and that promotion of healthy lifestyles in these patients is also very important. These findings also highlight the great advantage in terms of life expectancy that primary prevention of HIV transmission still has over the antiretroviral treatment.
The excess mortality observed in the current study was greater than that reported from clinical cohorts of HIV-infected patients, which may be explained by selection of patients in this kind of cohorts [8][9][10][11][12] ; however, our results are close to those of another population-based study [18][19][20] .
HIV-related mortality decreased from 2 to 1 out of 3 deaths during the study period. The HIV-related mortality rate declined 8% annually, being the main component of the reduction in excess overall mortality among HIV-infected population. Although highly active antiretroviral therapy was already available before the study period, since then safer antiretroviral drug combinations that favour adherence have been progressively introduced 16 , and initiation of these treatments turned from a threshold based on CD4 count to as soon as possible after diagnosis may have contributed to maintain the reduction in HIV-related mortality 6,14,15 . The main reduction in mortality was observed in patients with diagnosis of AIDS-defining events, in whom the risk of mortality 21 and the potential benefit by a successful antiretroviral therapy are higher. However, mortality reduction was small or null in HIV-infected persons without prior AIDS-defining events, since mortality among them is mainly related to other comorbidities and risk factors, and its prevention requires other kind of preventive and healthcare interventions.
Non-HIV-related mortality remained stable during the study period, with an important excess as compared to the general population. Excess mortality in relative terms (SMR) was more pronounced in drug addiction or overdose and liver disease, since these causes were strongly associated to history of injecting drug use, which was present in many of HIV-infected people 11,13 . Nevertheless, the highest excess mortality rates were observed in cancer and cardiovascular disease, since they were also leader causes of death in HIV-infected population as they are in the general population. This finding could be explained by the higher frequency of tobacco, alcohol and recreational drug consumption, the faster aging and the premature presentation of comorbidities in HIV-infected people 22 .
History of injecting drug use was associated to higher mortality by most causes. It was present in the majority of deaths by drug overdoses or addiction, was a risk factor for coinfections by hepatitis B and C viruses and was frequently associated to injuries, suicide and to unhealthy life styles as alcohol abuse and smoking 23,24 .
During the study period, excess mortality rates declined more in HIV-infected males than in females, and by the last study period the differences by sexes had disappeared. Females presented higher excess mortality in relative terms, since in the general population women have lower mortality than men. This is partly explained because women assume less risk and take more care of their health, which may not be true in HIV-infected females, who have been associated with situations of disadvantage for the care of their health [24][25][26] .
Since excess mortality rates were higher in people older than 45 years, the aging of the cohort and the premature presentation of comorbidities 22 favor to maintain the excess mortality. However the higher SMRs in people under 45 years of age indicate the important relative impact of HIV infection on mortality since young ages.
In the studied cohort, people with history of injecting drug use predominated, although there was a progressive increase in the proportion of people probably infected by sexual transmission, coinciding with the (2020) 10   www.nature.com/scientificreports www.nature.com/scientificreports/ epidemiological change described in Spain 27 . Median age and time since the diagnosis of HIV increased due to the aging of the cohort. The percentage of patients who had been diagnosed with AIDS-defining events decreased throughout the study, due to the lower incorporation of new cases and the higher mortality associated with this situation [18][19][20][21]23,28,29 . Although the proportion of people born out of Spain increased, they did not show higher mortality than Spain-born people as it has been reported in other countries 26 . This can be due to the less time of HIV-infection and the similar access to the HIV-diagnosis and treatment.
The average percentage of patients who had started antiretroviral therapy increased up to 87.9% in the 2014-2018 period, when the indication of this therapy was extended to all HIV diagnosed patients. Patients who had not started therapy may be explained by the transition process from old to new guidelines, patients deceased before starting of clinical follow-up, those who refuse treatments or who do not contact the healthcare system. This study was conducted in a country with universal and free of charge health-care coverage for the entire resident population, including HIV diagnosis and antiretroviral treatment. This fact discards socioeconomic reasons for not receiving medical follow-up and antiretroviral treatment; and therefore, differences in mortality should be explained by other causes. In Spain, it has been estimated that 29% and 48% of HIV diagnoses were done in patients with advanced disease and late diagnosis, respectively 30 . These figures show a potential for improvement according the 2020 UNAIDS strategy, which goals are 90% of HIV-infected people diagnosed, 90% of them receiving antiretrovirals, and 90% of those in treatment achieving viral suppression 31 .
The present work has some limitations. This descriptive population-based study is not adequate to evaluate survival or to conclude about causal factors since many participants did not enter at the seroconversion or HIV diagnosis, and variables such as antiretroviral therapy, viral load, CD4 cell count, smoking status and drug consumption were not monitored during the follow-up. The study has been carried out in a single region and the study size could be small for some analyses. The analysis was limited to the population with stable residence, and non-resident patients were not included. It has been described that populations with high mobility could have worse follow-up of their health problems 32 . Losses of follow-up are a relevant source of bias in cohort studies; however, people were included only during the period with residence and healthcare coverage in Navarra and a variety of information sources were used that virtually rules out losses of follow-up 33 . HIV-related mortality includes more causes than AIDS-mortality, therefore, care should be paid in comparison with studies that analysed AIDS mortality 29,34 . Since important sociodemographic changes happened in the study population, time-trend analyses were adjusted not only by sex and age, but also by country of birth and transmission category to maintain the comparability over time. The impact of direct-acting antiviral therapy of HCV infection in mortality could not be detected since this therapy was just being introduced in the last study period.
Only the main cause of death was considered, but not other relevant causes that could also contribute. These results are conditioned by the HIV epidemiology in Spain, where most HIV infections were acquired by risk practices related with injecting drug use. Undiagnosed HIV infections and mortality that could occur before diagnosis were not represented in this study.

conclusions
Advances in the antiretroviral therapy and improvements in the management of HIV-infected patients have coincided with a progressive decline in HIV-related mortality in a population-based cohort of HIV-infected persons, mainly among those who had had diagnosis of AIDS-defined events, while non-HIV-related mortality remained stable. Despite the wide availability of antiretroviral therapies in Spain, HIV infection was still associated to important excess mortality, overall and by many specific non-HIV-related causes. These causes of death have surpassed HIV-related causes, and this trend probably will increase due to the progressive aging of the HIV infected population. Since antiretroviral therapy does not seem enough to correct the excess mortality, prevention of HIV infections in the population and promotion of healthy life styles in HIV-infected people must be a priority.

Methods
Design and population. This study analysed the population-based dynamic cohort of adults between 25 and 74 years old diagnosed with HIV infection and residing in Navarra, a northern Spanish region of 650,000 inhabitants approximately. The resident population in the region had easy and free access to medical care and to the HIV test in primary healthcare, specialized outpatient consultations and hospitals. Combined antiretroviral therapies have been available free of charge for the patient in 3 HIV-centres during the study period. HCV antiviral treatments have been offered to HCV co-infected patients and since 2014 direct-acting antiviral therapies were progressively offered to all co-infected patients.
information sources and variables. HIV infection reporting is mandatory for healthcare professionals according to the Navarra Decree 383/1997, and patient consent is not required. This study analysed data from the enhanced epidemiological surveillance of HIV-infection that includes all cases confirmed in laboratories of the region since the beginning of the epidemic 35 . It has been prospectively updated by an automatic reporting of all new laboratory-confirmed cases and by the active search of HIV-infected patients attended in medical centres. AIDS-defining events were obtained from the AIDS case reporting system and the hospital discharge database, vital status was updated according the regional mortality register, and changes in the residence were updated from administrative databases of the Navarra Health Service 35 .
The variables considered were sex, birth date, transmission category (history of injecting drug use, men who had sex with men, heterosexual risk exposure, and other risk or unknown), country of birth (Spain or other), date of HIV diagnosis, hepatitis C antibody positive result, date of AIDS diagnosis, date of death and cause of death.
All data were treated in a strictly confidential manner according to the ethical principles of the Helsinki Declaration of 1964 revised by the World Medical Organization in Fortaleza 2013, and the Spanish and European personal data protection regulations. The study protocol was approved by the Navarra Public Health Authority and the Navarra Ethical Committee for Clinical Research. Patient information was anonymized prior to analysis. Statistical analysis. The study period started on first January 1999 for previously diagnosed infections, on the date when HIV infection was diagnosed for new detected cases, and on the date of registration in the Navarra Health Service for previously diagnosed HIV-infected patients who came to reside in the region. The follow-up ceased at the date of death, the date of deregistration as a resident or by 31 December 2018, whichever came first.
Percentages were calculated for categorical variables and were compared by χ 2 test. Median and interquartile range were obtained for quantitative variables which were compared by the Kruskall Wallis test.
The number of person-years of follow-up was used as the denominator for the calculation of average annual mortality rates for five-year periods. Poisson regression models were used to estimate average annual change in mortality rates, overall, by sex, age, history of injecting drug use, country of birth, diagnosis of an AIDS-defining event, HIV diagnoses before 1997 (year of extended use combined antiretroviral therapy) and groups of causes of death. Rate ratios of annual trend unadjusted and adjusted by sex, age-group (25-44, 45-54 and 55-74 years), transmission category and country of birth were calculated.
The number of observed deaths in the cohort of persons diagnosed with HIV infection was compared with expected mortality of the Navarra general population of the same sex and ten-year group of age. Relative excess mortality was estimated by standardized mortality ratios with their 95% confidence intervals obtained by the exact Mid-P method. Excess mortality rates were calculated as the difference of observed minus expected deaths and divided by the number of PY of follow-up. These analyses were repeated for each five-year period, sex, age group, transmission category and HIV diagnoses before or in or after 1997.
For the last ten-year period (2009-2018) observed mortality in the cohort of HIV-infected persons by causes was compared with the expected mortality estimated from the general population.