Estimating HIV Management and Comorbidity Costs Among Aging HIV Patients in the United States: A Systematic Review

BACKGROUND: As life expectancy of patients infected with human immunodeficiency virus (HIV) approaches that of the general population, the composition of HIV management costs is likely to change. OBJECTIVES: To (a) review treatment and disease management costs in HIV, including costs of adverse events (AEs) related to antiretroviral therapy (ART) and long-term toxicities, and (b) explore the evolving cost drivers. METHODS: A targeted literature review between January 2012 and November 2017 was conducted using PubMed and major conferences. Articles reporting U.S. costs of HIV management, acquired immunodeficiency syndrome (AIDS)-defining events, end of life care, and ART-associated comorbidities such as cardiovascular disease (CVD), chronic kidney disease (CKD), and osteoporosis were included. All costs were inflated to 2017 U.S. dollars. A Markov model-based analysis was conducted to estimate the effect of increased life expectancy on costs associated with HIV treatment and management. RESULTS: 22 studies describing HIV costs in the United States were identified, comprising 16 cost-effectiveness analysis studies, 5 retrospective analyses of health care utilization, and 1 cost analysis in a resource-limited setting. Management costs per patient per month, including routine care costs (on/off ART), non-HIV medication, opportunistic infection prophylaxis, inpatient utilization, outpatient utilization, and emergency department utilization were reported as CD4+ cell-based health state costs ranging from $1,192 for patients with CD4 > 500 cells/mm3 to $2,873 for patients with CD4 < 50 cells/mm3. Event costs for AEs ranged from $0 for headache, pain, vomiting, and lipodystrophy to $31,545 for myocardial infarction. The mean monthly per-patient costs for CVD management, CKD management, and osteoporosis were $5,898, $6,108, and $4,365, respectively. Improvements in life expectancy, approaching that of the general population in 2018, are projected to increase ART-related and AE costs by 35.4% and comorbidity costs by 175.8% compared with estimated costs with HIV life expectancy observed in 1996. CONCLUSIONS: This study identified and summarized holistic cost estimates appropriate for use within U.S. HIV cost-effectiveness analyses and demonstrates an increasing contribution of comorbidity outcomes, primarily associated with aging in addition to long-term treatment with ART, not typically evaluated in contemporary HIV cost-effectiveness analyses.

Historically, U.S. health economic analyses in HIV have focused on the costs associated with HIV treatment and management, costs of AIDS-defining events, and to a lesser extent, HIV prevention costs. [9][10][11][12][13][14][15][16] However, considering the increasing prevalence of ART-related long-term toxicities and age-related HIV comorbidities, an assessment of their impact on health care costs is warranted. The objective of this study was to review costs associated with HIV management, with particular focus on ART-related long-term toxicities, including CVD, chronic kidney disease (CKD), and osteoporosis. A secondary objective was to subsequently highlight and quantify the evolving cost drivers associated with HIV management through an economic modeling analysis.
HIV (PLHIV) in 2015, of whom 162,500 were undiagnosed patients. 5 The incidence of new HIV diagnoses in the United States is steady, with 38,739 new patients diagnosed in 2017. 5 With the advent of newer, more potent ARTs, the life expectancy of patients living with HIV has substantially improved, and by 2020, an estimated 70% of PLHIV in the United States are projected to be aged 50 years and above. 6,7 HIV requires life-long treatment, and chronic use of ART has been shown to significantly increase the risk of fractures and osteoporosis, renal and metabolic disorders, central nervous system (CNS) disorders, cardiovascular disease (CVD), and liver disease. 8 These long-term ART toxicities, coupled with comorbidities, which are more prevalent in PLHIV, have the potential to substantially increase health care resource utilization and costs associated with HIV management.  Titles and abstracts identified by the search strategy were selected based on eligibility criteria. Studies in adult PLHIV reporting costs of HIV management, in particular CD4+ or viral load-based health states, adverse events (AEs), AIDSdefining events, and end of life care, were included. Studies reporting costs of common long-term ART toxicities such as CVD, CKD, and osteoporosis, including fractures, were also included. The inclusion criteria restricted studies to cost of illness studies or economic evaluations published in the English language. The costs associated with HIV testing, AEs, comorbidities, and HIV management were extracted and inflated to 2017 U.S. dollars using the U.S. Bureau of Labor Statistics Consumer Price Index before reporting in this study. 18 Where mean costs are presented, a simple mean across endpoints was utilized (i.e., means were not weighted).

Modeling
A modeling analysis was undertaken to estimate the impact of increased life expectancy among PLHIV on costs associated with HIV treatment and management. A previously presented and validated economic model was used. [19][20][21][22] Model inputs were derived from U.S.-focused publications where possible, and U.S.-specific costs identified by the literature review were used (Table 1). Our model used published values of CD4 health state costs and AIDS-defining events estimated using simulation techniques by Farnham et al. (2013) to predict HIV-related events and associated costs over patient lifetimes. 23  Annual HIV-related costs: Juusola et al., 47 Vergis et al., 68 Schacker et al., 69 Daar et al., 70 Long et al., 71 Bozette et al., 14 Schackman et al., 15 Barnett et al. 61 Cost of AIDS: Long et al., 71 Bozzette et al., 14 Schackman et al., 15 Barnett et al., 61 Gebo et al., 12 Hutchinson et al., 72 Sanders et al. 73 HIV testing: CMS 67 Morris et al. 35 •   66 : ARV and non-ARV drug costs, lab utilization, and outpatient care costs were derived directly from the SAC pharmacy, Calgary Laboratory Services, and the SAC costing database. Inpatient costs (i.e., unit service costs) were supplied by the regional health service providers. The unit costs used are market values charged to the regional payer (Calgary Health Region). All costs were obtained directly from the costing agencies, reported in Canadian dollars, and adjusted for inflation to 2006  [24][25][26] Costs and benefits were discounted at a rate of 3% annually. The model has been described in detail elsewhere. [19][20][21][22] In brief, modeled health states included CD4 cell count-based health states nested within 3 strata of viral load and death. Patients initiated their first treatment (ART 1) in one of the CD4 and viral load states based on their baseline parameters. During each monthly cycle, patients' viral status can improve, decline, or remain constant, driven by treatment-specific efficacy estimates. Four treatment lines are incorporated: 3 defined ART regimens and 1 unspecified salvage therapy line composed of pooled ART regimens. Patients are assumed to be at risk of experiencing treatment failure due to AEs, virologic failure (inability to achieve HIV RNA level < 50 copies/mL), and viral rebound (HIV RNA ≥ 50 copies/mL after virologic suppression). Upon discontinuation, patients moved to the next treatment line. In a typical cycle length of 1 month, patients were exposed to risks of treatment-related AEs, AIDS-defining events, CVD events, CKD progression, bone fractures, and death.

Analyses
The current analysis focused on the potential impact of improving HIV survival on the incidence and costs of CVD, endstage renal disease (ESRD), and fracture incidence. Utilizing published estimates, the model was calibrated to HIV-specific life expectancy predicted at different time points between 1996 and 2008. 27 Mortality in the cost-effectiveness model is estimated based on country-specific life table estimates adjusted based on relative risks stratified by CD4 cell counts; calibration was undertaken through adjustment of virologic suppression rates for PLHIV, with historic cohorts less likely to remain suppressed. The incidence of CVD was estimated using Framingham risk equations and the incidence of fractures was estimated by the Sheffield FRAX tool. 28,29 The incidence of CKD stage 5 was defined as an estimated glomerular filtration rate (eGFR) of < 15 mL/min/1.73 m². An annual decline in eGFR (mean annual decline: −1.23 mL/min/1.73 m²) and bone mineral density (mean annual decline: −0.0029 g/cm³) was applied based on published literature. 30

Summary of Modeling Studies for CEAs or Economic Evaluations (continued)
continued on next page and fracture incidence, with all other parameters remaining unchanged.

■■ Results
The index searches identified 1,347 references resulting in 59 full-text articles. In total, 33 publications that described costs or resource use in the U.S. setting were included in the qualitative synthesis of this study (Appendix B, available in online article). Of these, 5 conference proceedings that did not present the costing year were excluded from the final selection.
Of the 22 full-text publications that fully described costs, 16 studies used an economic model to conduct cost-effectiveness analyses (CEAs) or economic evaluations, 5 were retrospective analyses of health care utilization, and 1 study undertook a cost analysis in a resource-limited setting (Table 2). 23,25,26,[34][35][36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52] A comparative analysis was performed utilizing 6 hypothetical cohorts, representing the life expectancy of PLHIV at different periods in time, with a baseline age of 20 years modeled over a lifetime. These included patients that do not achieve viral suppression (assumed equivalent to survival prior to the introduction of multiclass combination ART [approved by the U.S Food and Drug Administration in 1997]; cohort 1); patients with survival rates observed in years 1996, 2000, 2004, and 2008 (cohorts 2-5); and patients achieving a life expectancy equivalent to the general population in the United States in 2018 (cohort 6). 32, 33 The predicted cumulative incidence of CVD; CKD stage 5; and hip, spine, shoulder, and arm fractures and total lifetime HIV-related costs were compared. Analyses were varied probabilistically to reflect the heterogeneity of the population by varying clinical inputs related to CVD, ESRD,

Author
Cost/Resource Description Cost/Resource Source and Description Gallant et al. 25 CVD, CKD, and fracture per-patient per-month costs: • Commercial (private payer) (CVD, CKD, and fracture), Medicare (public payer) (CVD and CKD):  51 • Other costs by CD4 (unadjusted total costs per patients per month)

AE and AIDS-Defining Event Costs
Three studies estimated costs of AEs, reporting them as per-event costs or annual event-related costs (Table 1). 34 S. claims databases calculated costs for several AEs ranging from $0 for AEs such as headache, pain, vomiting, and lipodystrophy to $23,964 for intracranial hemorrhage, $25,912 for stroke, and $31,545 for myocardial infarction. 34,48 Costs for the majority of AEs were less than $5,000 per event. Four studies estimated costs stratified by types of opportunistic infections (OIs) into bacterial OIs, viral OIs, fungal OIs, protozoal OIs, cancer, and others (Table 1). 23,34,35,41 The mean per-event costs for OIs ranged from $8,495 for viral OIs to $13,036 for protozoal OIs. The overall mean cost per event across all AIDS-defining events was $10,465 and the corresponding monthly mean cost was $4,464.

Long-Term Toxicity Costs
Outcomes associated with CVD, CKD, and fractures/osteoporosis were explored as long-term toxicities in HIV-infected patients. Three studies reported the cost of CVD, 2 studies reported the cost of CKD, and a single study reported the cost in CVD incidence between 40 and 80 years, after which it plateaued. The incidence of CKD stage 5 remained relatively low until patients reached the age of 75 and increased significantly thereafter. Analysis showed a 33.4% increase in life expectancy in cohort 6 compared to cohort 5. Such an improvement in survival represented an increase of 71.3%, 58.3%, and > 1,000% in the incidence of bone fractures, CVD, and CKD stage 5, respectively. The economic impact associated with the increased incidence of CVD, CKD stage 5, and bone fractures was reflected in the total lifetime cost outcomes predicted by the model (Figure 3). The total HIV lifetime cost per patient was estimated to be $1,246,810, $1,420,869, and $1,673,510 for cohorts 2, 5, and 6, respectively. The cost increases were driven by treatment and AE costs (35.4% increase between cohort 2 and cohort 6) but also by costs of treatment for comorbidities such as CVD (179.9% increase) and CKD (174.3% increase). HIV management costs, including costs of inpatient care, emergency department and outpatient visits, OI prophylaxis, HIV testing, and non-HIV medication, however, decreased as HIV patients approached general population survival rates due to improvements in CD4 cell counts and viral loads. of fracture/osteoporosis. 25,26,41 The study by Peng et al. (2015) reported the total per-event cost of CVD to be $28,728, while studies by Smit et al. and Gallant et al. (2018) reported further cost breakdowns with inpatient, outpatient, emergency care, and prescription costs for non-HIV prescriptions (Table 1). 25,41 The mean monthly cost of CVD management across all studies was estimated to be $5,898 per patient. The costs of CKD and osteoporosis were also further divided into cost components, with average monthly costs of $6,108 and $4,365, respectively.

Modeling Analyses
Predicted survival for each of the modeled patient cohorts is presented in Figure 2. In line with published estimates, the model assumed improved survival between 1996 and 2008. Patients failing to achieve viral suppression (cohort 1), who were assumed representative of patients before the widespread use of highly active ART, were estimated to be associated with a life expectancy of 22.4 years. The model estimated life expectancies of 36.2, 37.9, 40.5, and 45.1 years for a typical 20-yearold patient at the time of diagnosis for cohorts 2-5, respectively.
The incidence of bone fractures remained relatively constant over the modeled horizon, while there was a gradual increase in AIDS-defining event incidence and virological failure. The most significant estimated changes were the rapid increase in costs of comorbidities associated with aging and long-term use of ART, such as CVD, CKD stage 5, and osteoporosis. As life expectancy improved beyond 50 years, the estimated incidence of these ART-and age-associated comorbidities significantly increased, resulting in their increased contribution to total HIV costs.
This study focused on 3 of the most common ART-specific comorbidities. However, there is evidence to suggest that several other chronic conditions such as depression, diabetes, and respiratory diseases are more prevalent in HIV patients, and their contribution to lifetime HIV costs is likely to increase in aging patients. 54 Interventions that improve communication and potentially regular contact between patients and health care professionals are likely to lead to early detection and better management of these comorbidities. There is a large amount of literature relating ART to long-term toxicities, some of which are highly prevalent in aging HIV patients. Interventions that reduce the cumulative exposure of antiretrovirals may also help in reducing prevalence of these toxicities. Future research should focus on assessing the impact of reduced antiretroviral exposure on lifetime HIV costs.

■■ Discussion
With the advent of newer, more efficacious ARTs, HIV patients are expected to live longer, with a mean life expectancy approaching that of the general population. 6,7 This aging population requires chronic care to manage HIV disease progression, and with extension of life comes the likelihood of additional resource-intensive age-related comorbidities, which are not currently a focus of contemporary HIV CEAs. 54 Long-term ART exposure has the potential to lead to the early onset of such comorbidities due to the impact of certain HIV therapies on long-term toxicity risk factors. [55][56][57] As a result, the composition of care needed for HIV patients has evolved over the last 2 decades and is expected to change further in the future.
Our study reviewed published literature on costs of HIV management in the United States and modeled the potential future costs of aging HIV patients. Modeled results estimated that, as the survival of HIV patients approached that of the general population, lifetime HIV costs increased, with the distribution of costs across different contributors varying significantly. HIV treatment and related AE and toxicity costs were estimated to increase as patients received treatment for a longer duration, while HIV management costs showed an estimated decrease, as improved survival is inherently associated with improved viral load and CD4 cell count profiles and associated reductions

DISCLOSURES
This analysis was sponsored by ViiV Healthcare, which had no role in the analyses and interpretation of study results. Ward, Sugrue, Hayward, and McEwan are employees of HEOR Ltd, which received funding from ViiV Healthcare to conduct this study. Anderson is an employee of GlaxoSmithKline and holds shares in the company. Punekar and Oglesby are employees of ViiV Healthcare and hold shares in GlaxoSmithKline. Lopes was employed by ViiV Healthcare at the time of the study and holds shares in GlaxoSmithKline.
The costs of specific comorbidities identified in our review for HIV patients were higher than those reported in the literature for non-HIV patients. The mean annual cost of CKD stage 5 in our study was $73,296. A study by Ozieh et 60 The corresponding annual costs in our study, which included medical and prescription costs, were $52,380. This suggests an additional cost burden among HIV patients for these comorbidities, which may be due to the effect of long-term treatment with ART. Given the substantial costs associated with such comorbidities, there is real potential for the incidence of such events to significantly affect cost-effectiveness outcomes and, subsequently, reimbursement decisions.

Limitations
This study has some limitations. First, all HIV-and ARTassociated potential comorbidities were not considered. Second, although systematic literature reviews are the preferred method for sourcing parameters for cost-effectiveness models, our approach was limited to a targeted literature review. We would not anticipate a systematic review to identify much or any additional relevant sources, however, since the evidence base in this specific area is not extensive. Third, those studies captured by this review commonly obtained cost data from one of a number of key cost studies, due to the limited evidence base, which represents a key limitation of this study.

■■ Conclusions
This study demonstrated that as life expectancy of HIV patients improved, a significantly increased contribution of ART-associated comorbidities, including CKD, CVD, and osteoporosis, is observed. This contrasts with the overall estimated decrease in HIV management costs, likely due to improvements in patients' viral load and CD4 cell counts and subsequent reduced likelihood of experiencing AIDS-defining event incidence and virological failure. Future CEAs should consider the complete range of clinical outcomes for which HIV treatments may have an impact, rather than relying on historical precedent when informing modeled endpoints. Future research should also focus on assessing the effect of reduced antiretroviral exposure on lifetime HIV costs, in addition to how the resultant costs compare to that of a general population.