Disability-adjusted life years associated with chronic comorbidities among people living with and without HIV: Estimating health burden in British Columbia, Canada

Life span of people living with HIV (PLWH) has increased dramatically with the advent of modern antiretroviral therapy. As a result, comorbidities have emerged as a significant concern in this population. To describe the burden of chronic comorbidities among PLWH and HIV-negative individuals in British Columbia (BC), Canada, we estimated disability-adjusted life years (DALYs) related to these comorbidities. Based on a population-based cohort in BC, antiretroviral-treated adult PLWH and 1:4 age-sex-matched HIV-negative controls were followed for ≥1 year during 2001–2012. DALYs combined years of life lost to premature mortality (YLLs) and due to disability (YLDs), and were estimated following the Global Burden of Diseases’ approaches. DALYs associated with non-AIDS-defining cancers, diabetes, osteoarthritis, hypertension, dementia, cardiovascular (CVD), kidney, liver and chronic obstructive pulmonary diseases were each measured for 2008–2012. Among PLWH, DALYs attributed to non-AIDS-related cancers were also estimated for 2013–2020. We observed that at baseline, our matched cohort consisted of 82% males with a median age of 40 years (25th-75th percentiles: 34–47). During 2008–2012, 7042 PLWH and 30,640 HIV-negative individuals were alive, where PLWH experienced a twofold higher DALYs associated with chronic comorbidities (770.2 years/1000 people [95% credible intervals: 710.2, 831.6] vs. 359.0 [336.0, 382.2]). Non-AIDS-defining cancers and CVD contributed the highest DALYs in both populations, driven by YLLs rather than YLDs. Among PLWH, we estimated increasing DALYs attributable to non-AIDS-defining cancers with 91.7 years/1000 people (77.4, 106.0) in 2013 vs. 97.6 (81.0, 115.2) in 2020. In this study, we showed that PLWH experience a disproportionate burden of chronic comorbidities compared to HIV-negative individuals. The observed disparities may relate to differential health behaviors, residual HIV-related inflammation, and ART-related toxicities. As aging shapes future healthcare needs, our findings highlight the need to enhance prevention and management of comorbidities as part of HIV care.


Introduction
As advances in antiretroviral therapy (ART) have dramatically increased longevity among people living with HIV (PLWH), the increased burden of non-AIDS chronic comorbidities has emerged as a major concern [1][2][3][4]. Compared to the general population, PLWH had reportedly a higher risk of developing chronic comorbidities such as non-AIDS-related cancers, kidney, lung, and cardiovascular diseases [5,6], which often co-occurred [7,8]. In British Columbia (BC), Canada, where ART has been available free of charge since 1992 and more than 60% of known PLWH in 2020 were 50 years or older [9], chronic comorbidities were diagnosed up to 10 years earlier among PLWH than HIV-negative controls [10]. Chronic comorbidities have also increasingly become the predominant cause of death among PLWH [11,12], with chronic comorbidities making up to 78% of all deaths among BC's PLWH in 2017, an increase from 35% in 2004 [13].
Disability-adjusted life years (DALYs) are a comprehensive metric of disease burden, accounting for both morbidity and mortality [14]. DALYs quantify the gap between the perfect health and observed health status based on years of life lost to premature deaths (YLLs) and years of healthy life lost due to disability (YLDs). DALYs were constructed as part of the first annual Global Burden of Diseases (GBD) study in the early 1990s [15]. Under the collaboration between the Institute of Health Metrics and Evaluation and the World Health Organization, the GBD has since undergone annual updates and several methodological changes, with the latest iteration investigating the health effects of 369 diseases and injuries in 204 countries in 2019 [16,17]. Countries have also adopted DALYs to estimate their nation's overall and cause-specific disease burden [18][19][20][21][22][23][24], often projecting the future burden of diseases to aid policy making and clinical service planning [25][26][27]. Note that despite many settings adopting DALYs as a metric, methodological choices frequently differed to address disparities in data availability and quality across countries and diseases [28].
In this study, we estimated and compared the burden of nine chronic comorbidities among PLWH and a population of age-sex-matched HIV-negative controls in BC by estimating YLLs, YLDs and DALYs associated with these comorbidities during 2008-2012. Among PLWH, the study also estimated the disease burden associated with non-AIDS-defining cancers from 2013 to 2020. To achieve these aims, we used a population-based cohort with linkages to key provincial health databases. We supplemented the GBD methodology with robust methodological approaches that catered to the BC setting to address potential biases arising from comparing PLWH and HIV-negative controls, and the use of administrative data for health research.

Ethics statement
Ethics approval for the COAST study was granted from the University of British Columbia/ Providence Health Care Research Ethics Board (H09-02905; H16-02036) and Simon Fraser University Office of Research Ethics (#2013 s0566). The study complies with the BC Freedom of Information and Protection of Privacy Act (FIPPA) and did not require informed consent as it is conducted retrospectively for research and statistical purposes only using anonymized data.

Data sources
We obtained longitudinal de-identified individual-level data from the Comparative Outcomes And Service Utilization Trends (COAST) study, which comprises all diagnosed adult PLWH and a 10% random sample of the general population in BC followed from 1996 to 2013 [29]. For the period after 2012, we considered PLWH aged �19 years who received treatment through the DTP between January 1, 2013 and December 31, 2020. For these PLWH, baseline was the latest between the first treatment date and January 1, 2013. They were then followed for �1 year until the earliest among death, last contact date and December 31, 2020.

Outcome variables
Our outcomes were YLLs, YLDs and DALYs associated with nine chronic comorbidities highly prevalent among PLWH and BC's general population [10,37]: cardiovascular diseases (CVD), kidney disease, liver disease, chronic obstructive pulmonary disease (COPD), non-AIDS-defining cancers (excluding Kaposi sarcoma, non-Hodgkin's lymphoma and cervical cancers; hereinafter referred to as cancers), diabetes, osteoarthritis, hypertension, and Alzheimer's and/or non-HIV-related dementia (Alzheimer's/dementia). Following age-restrictions in the case-finding algorithms, individuals considered for hypertension, COPD and Alzheimer's/dementia analyses were older than 20, 35 and 40 years, respectively. Whenever possible, we estimated YLLs, YLDs and DALYs following the methodology used in the GBD 2019 [16,17].
YLDs. YLDs quantified time lived in states of suboptimal health as the sum of the number of prevalent cases at a time period multiplied by the corresponding disability weight [16,17]. The BC Cancer Agency Registry identified cancer cases [35]. The remaining comorbidity diagnoses were ascertained by case-finding algorithms using the International Classification of Disease Ninth and Tenth Revisions (ICD 9/10) codes and Canada-wide Drug Identification Numbers (Table A in S1 Text), which have been utilized in previous publications [10,38]. These algorithms were applied to provincial administrative datasets including hospitalizations and day-surgeries [32], physician visits, laboratory and diagnostic procedures [33], and prescription drug dispensation [36]. For each year, we measured period prevalence; a comorbidity was prevalent if an individual was alive for �1 day in a given year and an existing diagnosis was identified within a 5-year observation window (i.e., at any point that year and/or within four years prior) [10]. Each comorbidity was associated with severity-specific disability weights between 0 and 1, representing a numerical assessment of non-fatal health loss, where 0 equaled full health and 1 equaled death [16,17]. We used disability weights from the GBD 2019 (Table B in S1 Text), derived from largescale surveys of the general population, as opposed to health experts, across different cultural contexts [39,40]. Given administrative data's inability to ascertain disease severity, to approximate comorbidity-specific severity distributions in our study populations, we conducted a literature review to obtain distributions from BC or comparable settings ( Table B in S1 Text).
YLLs. YLLs measured time lost to premature death as the number of deaths during a time period multiplied by a standard life expectancy at the age of death, which represents the maximum life span of healthy individuals receiving appropriate health services [16,17]. We used the theoretical minimum risk life table from the GBD 2019 (Table C in S1 Text), constructed based on the lowest observed age-specific mortality rates for both sexes from locations with populations over five million in 2016 [41]. The underlying cause of death, defined as the disease or injury that initiated the series of events leading directly to death, was collected through BC Vital Statistic Agency's mortality database. For consistency between YLLs and YLDs, we captured comorbidity-specific deaths using the same ICD 10 codes used to ascertain prevalent cases for YLD calculations (Table A in S1 Text).
DALYs. DALYs associated with each comorbidity were calculated as the sum of corresponding YLLs and YLDs, where one DALY depicted one lost year of healthy life due to a specific cause [16,17].

Analytical approaches
Categorical variables were compared using the Fisher's exact test and continuous variables were compared using the Kruskal-Wallis test [42]. Statistical analyses were performed using R software version 3.2.2 (R Core Team, Vienna, Austria) or SAS software version 9.4 (SAS, Cary, North Carolina, United States), while data simulations were coded in Python and conducted using the NUMPY library [43]. This study consisted of a two-pronged analysis. Of note, our data were exhaustive (i.e., no missing values), with all eligible individuals included in each analysis.
Comparison of burden of chronic comorbidities among PLWH and HIV-negative controls, 2008-2012. We compared YLLs, YLDs and DALYs, overall and per 1000 people, associated with the abovementioned chronic comorbidities across HIV status for a combined 2008-2012 period. The combined estimates were chosen to ensure sufficient observations, particularly of mortality cases. Further, to ensure combined estimates that are most contemporary, we restricted our analyses to the last five years of COAST data (i.e., 2008-2012). To measure uncertainties around the point estimate of each outcome, the Monte Carlo simulation approach was used to estimate the 95% credible interval (CrI) [44]. Unlike a confidence interval which derived from hypothesized repeats of an experiment, a CrI illustrates a range containing a percentage (e.g., 95%) of probable estimates based on observed data [45]. Hence, the lower and upper limits of a 95% CrI are the 2.5th and 97.5th percentiles, within which there is a 95% probability that the true estimate may lie given the evidence provided by our observed data. For YLLs, we conducted non-parametric bootstrapping, comprising 10,000 bootstrap samples with sample lengths equaled comorbidity-specific death counts [46]. For YLDs, data simulation was also repeated 10,000 times. For each simulation, disease severity was randomly assigned based on literature-derived probabilistic distributions. Subsequently, a number within the range of corresponding severity-specific disability weights was randomly assigned [28]. YLDs were then adjusted for the presence of other comorbidities at the individual level, assuming multiplicative combined disability weights (Note A in S1 Text) [16,26]. The 10,000 simulated values from YLLs and YLDs were then randomly paired and summed to estimate the 95% CrI of DALYs [28]. Additionally, given an ample number of prevalent cases, we subanalyzed YLD estimates for 2008-2012 stratified by ethnicity (White, non-White, unknown), history of injection drug use (people who have ever injected drugs [PWID], non-PWID, unknown) and sex at birth (male, female).
Estimation of burden of cancers among PLWH, 2013-2020. With COAST data limited until 2012, we used additional data from the DTP to estimate YLLs, YLDs and DALYs associated with cancers (non-AIDS-defining only) among PLWH for 2013-2020. The DTP dataset provides demographic information on PLWH receiving ART in BC, but, unlike COAST, is not linked to healthcare utilization datasets (e.g., hospitalization, physician visit and prescription datasets). Given the absence of information on HIV-negative individuals in the DTP dataset, our analysis was restricted to PLWH. Cancers were chosen due to its adequate number of observed mortality cases among PLWH in the COAST study, where, due to the required �1 year follow-up, no deaths were observed in 2001 (i.e., the first calendar year of the study follow-up period). Accordingly, we employed generalized linear mixed-effects regression to model the number of deaths on age, sex, and observation year for 2002-2012, assuming a negative binomial distribution and log link (Note C in S1 Text) [47,48]. The resulted model was then applied to the DTP dataset to estimate the annual number of cancers-specific deaths in each age-sex subgroup for 2013-2020. To estimate annual total YLLs along with the 95% CrI, we conducted parametric bootstrapping with normal approximation using standard errors associated with the projected number of deaths in each subgroup [49]. Annual age-sex-standardized YLLs per 1000 people were estimated by an identical bootstrapping process using the estimates of age-sex-specific deaths in Canada's 2011 census population, the reference population [50]. We adapted the original GBD projections which estimated YLDs and DALYs based on the last observed YLLs-to-YLDs ratio [51,52]. Given the fluctuation in our observed YLLsto-YLDs ratios over the years, however, we estimated annual total and age-sex-standardized YLDs and DALYs, alongside their respective 95% CrI, based on the mean of all observed YLLs-to-YLDs ratios instead.

Comparison of burden of chronic comorbidities among PLWH and HIVnegative controls, 2008-2012
During 2008-2012, 7042 PLWH and 30,640 HIV-negative controls were alive, leading to 5356.5 and 10,945.7 years in estimated DALYs associated with the selected chronic comorbidities, respectively. Overall DALYs were two times higher among PLWH compared to HIV-negative controls (770.2 years/1000 people [95% CrI: 710.2, 831.6] vs. 359.0 [336.0, 382.2], respectively), with PLWH experiencing higher DALYs associated with most comorbidities, except hypertension, diabetes, and osteoarthritis (Fig 1). Except for hypertension and osteoarthritis (in both populations) and Alzheimer's/dementia (among PLWH only), the burden of most comorbidities was driven by YLLs (Fig 2) rather than YLDs (Fig 3). Consequently, the rankings of DALYs were largely identical to those of YLLs, with cancers and CVD as predominant contributors constituting over 70% of the total estimated YLLs and DALYs in both populations. Among PLWH, COPD and liver diseases contributed the third and fourth highest YLLs and DALYs, as did diabetes and COPD among HIV-negative controls. Note, however, that DALYs related to COPD were 12 times higher among PLWH.
Sub-analyses of YLDs. Among PLWH, during 2008-2012, YLDs associated with the selected chronic comorbidities did not vary significantly by ethnicity but varied greatly by history of injection drug use and sex at birth (Fig 4). YLDs associated with Alzheimer's/dementia, for instance, were substantially elevated among PWID, while COPD and liver diseases were substantially elevated among PWID and female PLWH. YLDs associated with cancers, CVD and diabetes were higher among male PLWH.

Estimation of burden of cancers among PLWH, 2013-2020
Of the 9601 eligible PLWH enrolled in the DTP during 2013-2020, 83% were male with median age at baseline of 47 years, (25 th -75 th percentile: 38-54) and median follow-up of 8 years (5-8) (Table E in S1 Text). Population pyramids in Fig B in S1 Text illustrated changes in the age and sex distribution over the years highlighting that PLWH in BC were aging. Through 2020, we estimated an increasing trend in cancers-associated YLLs and YLDs (Fig C in S1 Text) as well as DALYs (Fig 5), overall and per 1000 PLWH. The trends in age-sex-

Discussion
To our knowledge, our study is the first to use DALYs to compare the burden of diseases among PLWH and age-sex-matched HIV-negative individuals. Using a large provincial administrative database which spanned over a decade, our results showed that ART-treated PLWH experienced more than double the burden of selected chronic comorbidities than their HIV-negative counterparts. In both populations, cancers and CVD contributed the most DALYs, predominantly driven by premature mortality. Diabetes and hypertension played major roles in the health burden of HIV-negative individuals, as did COPD and liver diseases in PLWH, particularly PLWH who were female or had a history of injection drug use. Additionally, among PLWH, we estimated that DALYs attributable to non-AIDS-defining cancers would continue to increase in the short term, although the age-sex standardized DALYs would decrease. Together, our findings underscore the role of different chronic comorbidities in the health burden of PLWH and the disparities of this burden compared to HIV-negative   individuals, which should be considered in the prioritization of resources and interventions that address the healthcare needs of this aging population.
Consistent with observations in BC's PLWH and HIV-negative controls, cancers and CVD contributed the highest number of DALYs in Canada and other high-resource countries [23, 53,54]. In both study populations, YLLs contributed more than 90% of cancers-associated DALYs, congruent with findings from Japan, the Netherlands and Australia [23,24,54]. Similarly, the dominance of YLLs in DALYs attributable to CVD, COPD, liver, and kidney diseases have also been examined elsewhere [21-23, [55][56][57]. This evidence stresses the importance of strategies to prolong life and reduce preventable deaths, such as secondary and tertiary prevention focusing on early detection and minimizing adverse outcomes [58], in alleviating the burden related to these chronic comorbidities. Conversely, the burden of hypertension, which is a known risk factor for other prominent chronic comorbidities, and osteoarthritis, to which no deaths were attributed in our populations and globally [59], was driven by YLDs. In mitigating the impact of these diseases on population health, the role of primary prevention, aimed at improving awareness of risk factors and altering health risk behaviors, should thus be highlighted.
The lack of prior studies examining the disparities in DALYs across HIV status precludes any direct comparisons of our findings. However, the observed disparities, including the twofold higher burden of chronic comorbidities among PLWH, were aligned with previous findings regarding prevalence [4,60,61] and mortality [62][63][64][65][66]. While cancers and CVD contributed significantly to the health burden of both study populations, COPD and liver diseases were also prominent among PLWH, as were diabetes and hypertension among HIV-negative controls. This discrepancy emphasizes the presence of differential health behaviors in both populations and the role of factors related to HIV infection and treatment, which should be considered when addressing the disproportionate disease burden among PLWH. First-generation nucleoside reverse transcriptase inhibitors, for example, have been associated with an increased risk of liver fibrosis [67]. Alcohol use, which is associated with a higher level of cirrhosis biomarkers in HIV-hepatitis C co-infected individuals, and smoking, a major risk factor of COPD, have also been found more prevalent among PLWH [68][69][70][71]. Additionally, among PLWH, our findings bolster the growing body of literature [72][73][74][75], and underline the merit of considering sex differences and history of injection drug use in formulating appropriate chronic disease interventions in this population. Among BC's PLWH, we estimated an increasing trend in DALYs associated with cancers through 2020, where annual DALY estimates per 1000 people in our population were consistently double those in Canada's general population [76]. Concurrently, we also estimated a decreasing trend in corresponding age-sex-standardized DALYs. These trends agree with the 2012-2019 trends in the global all-cause and Canada's cancers-associated DALYs [17,76], and demonstrate the impact of population growth and aging in the progression of the disease burden, reinforcing the need to better understand how aging shapes a population's healthcare needs. Particularly, the increasing DALYs represent the magnitude of the burden imposed on the healthcare system [17], which should be met with adequate infrastructure, human resources, and services. Our findings thus reaffirm present policies and interventions surrounding prevention, early screening, and management of cancers in BC, while simultaneously promoting the evolvement of these efforts, anticipating changes in population structure and needs. Future studies should project DALYs associated with other comorbidities to obtain a comprehensive outlook on the future overall health burden of PLWH, which will further elucidate optimal allocations and prioritizations of healthcare resources.
This study has some limitations. First, administrative health data, while important for clinical and public health policy-making, were not collected for research purposes. To mitigate administrative data's susceptibility to coding errors, whenever possible, we utilized published BC Ministry of Health's case-finding algorithms to identify prevalent comorbidity diagnoses from hospitalizations and day-surgeries, physician visits, laboratory and diagnostic procedures, and prescription drug dispensation datasets, considering BC-specific claims-related practices. Moreover, with over 10 years of follow-up and a judicious 5-year lookback window [77], we are confident in our ability to capture most prevalent cases in both populations, upholding the validity of our YLD estimations. Second, the scarcity of information on disease severity distribution is a widely recognized problem [18, 78,79]. With administrative data lacking information on the severity of comorbidity diagnoses, we employed literature-derived disease severity distribution, using data from BC or Canada whenever available and applying the same distribution to both PLWH and HIV-negative controls. Further, the lack of empirical data on alternative severity distributions of each comorbidity or how these distributions may vary within and between the two study populations prevented us from conducting meaningful sensitivity analyses. Instead, to account for the uncertainty surrounding the severity distribution, the probabilistic resampling of disease severity status and random assignment of the corresponding disability weight were conducted for each of the 10,000 simulations. Notwithstanding, given that DALYs associated with most diseases were predominantly driven by YLLs, the observed disparities in DALYs between PLWH and HIV-negative individuals should persist despite this limitation. Third, administrative data rendered us unable to consider socioeconomic and lifestyle differences (e.g., alcohol use and smoking) in our analyses, although our sub-analyses unveiled substantial differences in YLDs of major comorbidities among PLWH by sex and history of injection drug use. Fourth, low mortality cases prevented further age-or sex-specific measurements of YLLs and DALYs, although, for YLDs, we were able to highlight key differential burdens across sex groups. Fifth, the estimated burden of non-AIDS-defining cancers among PLWH for 2013-2020 should be interpreted in light of methodological constraints and assumptions. For instance, the lack of socioeconomic and lifestyle data prevented us from conducting more sophisticated modelling to estimate cancersspecific YLLs [80]. Additionally, adapting earlier GBD projections [51,52], cancers-specific YLDs were estimated based on the mean of all observed YLLs-to-YLDs ratios. Lastly, COAST data are currently limited to December 2012. Note, however, that during 2012-2019, the relative DALY contributions of non-communicable diseases in Canada's general population were largely unchanged [53]. Furthermore, the methodology established in this study can be readily applied to other settings to compare the contemporary burden of chronic comorbidities between PLWH and HIV-negative controls.
In summary, our study highlights that, compared to age-sex-matched HIV-negative controls from the same geographical setting, PLWH experienced excess DALYs attributable to chronic comorbidities, with the largest contribution from premature mortality associated with non-AIDS-defining cancers and CVD. Our study also underscores the role of aging in shaping the upward trajectories of disease burden and, consequently, determining the future healthcare needs among PLWH. Overall, these findings advocate for intervention strategies that integrate primary, secondary, and tertiary prevention of chronic comorbidities into a comprehensive HIV care model, with the goals to reduce preventable deaths and morbidity and, ultimately, facilitate healthy and successful aging in this population. Factors that likely influence the observed disparities both within PLWH and between PLWH and HIV-negative individuals, such as health behaviors, sex differences, residual HIV-related inflammation, and ART-related toxicities, should also be considered in these strategies. Further research on the current and projected overall disease burden should be prioritized to ensure timely implementation of evidence-based policies and investments to ensure adequate healthcare personnel and infrastructure.