Veterans Aging Cohort Studies (VACS) Update 2008 Amy Justice, MD, PhD For the VACS Project Team
The VACS Project Team (in West Haven, CT)
VACS Enrollment Sites Cleveland, OH (VACS 3 only) Pittsburgh, PA-Butt A. Houston, TX-Rodriguez-Barradas M. Los Angeles, CA-Goetz M. Manhattan/Brooklyn, NY-Simberkoff M. Bronx, NY-Brown S. Washington, DC-Gibert C. Baltimore, MD-Oursler KA. Atlanta, GA-Rimland D.
Virtual Cohort Subjects Scope of Study 33,000 HIV infected Veterans 99,000 Age, Race/Ethnicity, Region 2:1 Matched Controls Scope of Study 1998 to present Baseline HIV infected patients at initiation of HIV care Controls selected and followed in same calendar year Administrative, laboratory, and pharmacy data
Full Study 8 Sites HIV+ 1:1 Group matched to HIV- Baseline 2002 age, race/ethnicity, and site primary care pts. Baseline 2002 Ongoing annual follow up surveys Have access to all VA databases Local electronic medical records National VA databases (ICR, PBM, BIRLS, Patient treatment files, DSS, etc.) Requesting access to Medicare data Blood and DNA bank
International Multi Cohort Collaborations ART-CC NA-ACCORD Harvard Causal Modeling
VACS Career Development Awards Awardee Institution Funder Amy Justice CWRU NIA, RWJGFS Matt Freiberg U of Pittsburgh NIAAA, NIH Adeel Butt NIDA, NIH Shawn Fultz West Haven VA VA HSR&D Scott Braithwaite Yale NIAAA,RWJFS Kristina Crothers CTSA, NIH Vin Lore U of Pennsylvania NIAAA Kris Ann Oursler U of Baltimore NIA, NIH Nancy Kim VA HSR&D* *Pending
Current NIH Support (Excluding CDAs) NIAAA U10 Main Grant (Justice) R01 Braithwaite R21s: Papas, Lim (Fultz), Braithwaite NHLBI R01 Crothers R01 Berliner (NHLBI and NIA) R01 Freiberg/Justice pending NCI Site contract, NA-ACCORD Justice
Current VA Support (Excluding CDAs) VA HSR&D HERA Cohort: Brandt Informatics Consortium (Brandt, site contract) REAP: Kearns pending VA Clinical Support Smoking cessation study: Crothers
Foundations, etc. Robert Wood Johnson Foundation, Braithwaite Medical Research Council, Great Britain, Justice AMFAR pending, Mattocks
Publications Study # Publications Virtual Cohort 8 VACS 3 12 VACS 5 16 4 VACS Infrastructure 30 International Collaborations 9 Total 79
VACS Center Infrastructure Staff 3.5 FT Programmers 2 FT Biostatisticians 2 FT Operations Research Support Faculty 3 FT Informatics Faculty 1 FT Biostatistics/Epidemiology 1 FT Operations Research Faculty 1 FT Behavioral Sciences Faculty 1 FT Health Care Policy Faculty 3 FT HSR&D/Clinical Epidemiology
Current Major Initiatives Directly accessing VA EMR data for Studying comorbid illness in HIV Drug safety research Quality of care Intervention research Validating derivative VA databases for same Designing EMR intervention that optimizes utilization of resources to achieve patient and provider behavior change
Comorbidities and Toxicities in Older HIV Infected Patients
Mortality Rates (/1000 PY) Among HIV Infected and Uninfected Controls Study Age HIV + <1998 HIV+ 2000> HIV- Reference Denmark 25+ 124 25 11 Ann Intern Med 2007; 146:87-95 New York “adj” na 41 Ann Intern Med 2006;145:397-406 Barcelona 16-65 45 10 5 HIV Medicine 2007;8:251-8 10 US sites (HOPS) adult 70 20 J Acquir Immune Decif Syndr 2006;43:27-34 NYC IVDU 114 54 15 CID 2005:41;864-72 Nat’l VA 35 ~18 Med Care 2006:44;s13-24
“Non AIDS” Causes of Death Since ~2000 Source Of Known Leading Causes (%) Reference NY State Death Certificates 26% Alcohol/drug abuse (31%), CVD (24%), Cancer (21%) Ann Intern Med 2006;145:397-406 Barcelona 60% Liver ( 23%), Infection (14%), Cancer (11%), CVD (6%) HIV Med 2007:8;251-8 HOPS Ascertainment 63% Liver (18%), CVD (18%), Pulmonary (16%), Renal (12%), GI (11%), Infection (10%) Cancer (8%) J Acquir Immune Defic Syndr 2006;43:27-34 Cascade Liver (20%), Infections (24%), Unintentional (33%), Cancer (10%), CVD (9%) AIDS 2006; 20;741-9
Braithwaite Model Computer simulation of natural history of HIV Probabilistic Mimics heterogeneity in clinical populations Risk for death based on age (HIV uninfected veterans), CD4, viral load 3 major advances over other HIV models Represents mutations in HIV genome Represents nonadherence to HAART Calibrated and validated using clinical data We created a computer simulation of the natural history of HIV in the era of HAART. The simulation is probabilistic, and it therefore can mimic some of the heterogeneity seen in clinical populations. All patients were followed until death, and the risk for death was based on characteristics including age, CD4 count, and viral load. This simulation offers 3 major advances over other HIV models. First, it represents mutations in the HIV genome, and therefore can predict time to treatment failure of HAART regimens. It represents nonadherence to HAART, which has been shown to be the primary cause of virological failure in numerous patient groups and health care settings. Lastly, it has been calibrated and validated using clinical data, which increases its generalizability. Braithwaite R. et al, Am J Med 2005;118:890-898
“Background” Deaths on HAART Percent Non-AIDS Deaths Age 70 Age 60 Age 50 Age 40 Age 30 CD4 >350 CD4 201-350 CD4 51-200 CD4 <50 Braithwaite R, Am J Med 2005: 118 890-898
Incident “Comorbidity” in HIV Aging Treatment Toxicity HIV Progression Substance Use and Other Behaviors
Working Definition of Comorbidity Any condition not included in the CDC list of AIDS defining conditions “Non AIDS” Includes HIV independent conditions HIV associated, but not AIDS defining, conditions Toxicity masquerading as comorbidity One condition may have multiple causes
Attributable Risk Background : event rate among uninfected individuals otherwise like those with HIV HIV : event rate among those with HIV infection and/or disease progression Attributable HIV Risk = HIV – Background ARV : event rate among those on treatment Attributable ARV Risk= ARV-HIV
Why Differentiate? HIV Risk: Provide mechanistic insight into HIV and comorbidity (targets of intervention) If mechanism similar as among uninfected, then general approach may be adapted If mechanism atypical, may offer new insight ARV Risk: Separate toxicity from background risk due to HIV, aging, and substance use Consider risk of toxicity in choice of Rx Mechanism of condition may be atypical
Do we have reason to believe that mechanisms of comorbidity differ among those in treatment for HIV compared to those without infection?
Strategies for Management of ARV Therapy (SMART) RCT of interrupted ARV treatment based on immune reconstitution to minimize toxicity Found that non AIDS events were higher in those randomized to interrupted therapy HR 1.7 More liver and renal injury, and cardiovascular events Concluded that at least some “non AIDS” events may be associated with HIV disease progression
[The SMART] finding suggests that elucidating the effect of uncontrolled viremia on cytokine expression, inflammation, and immune activation may also have implications for infected individuals not yet on ART. — Rajesh T. Gandhi, MD Published in Journal Watch Infectious Diseases November 29, 2006
Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) Secondary data analysis, surprise finding 11 Cohort, >33000 pts., European Increased risk for myocardial infarction with DDI (RR 1.5) and abacavir (RR 1.9) Evident only while on drug or within 6 months of stopping DAD Study Group, Lancet , published online April 2, 2008
…This finding was unexpected because neither drug is thought to have substantial effects on metabolic factors… DAD Study Group, Lancet , published online April 2, 2008
Hayflick Limit on T Cell Response Cytotoxic T cells (CD8 cells) HIV and age associated with terminal “replicative senescence” after chronic antigen driven proliferation Irreversible cell cycle arrest Shortened telomeres Inability to up regulate telomerase Loss of CD28 expression Apoptosis resistance Reduced control of viral infection and cancer Effros, Mechanisms of Ageing and Development. 2004:125;103-106
Major Questions Does HIV increase risk of non AIDS events? Cardiovascular disease Liver disease Renal disease Cancer What is the ARV toxicity risk for these events? What are the likely mechanisms?
Veteran Non AIDS (ICD9) Events Medical Disease Substance Use Disorders ANY ONE Psychiatric Disorders Goulet J., et al. Clin Infect Dis. 2007 Dec15;45:1593-601
Veteran Multi Morbidity (ICD9) Goulet J., et al. Clin Infect Dis. 2007 Dec15;45:1593-601
Behavioral Risk Factors Unpublished VACS Data
Metabolic and Vascular Disease
ICD9 codes for: mycardial infarction, stroke, transient ischemic attack and/or peripheral vascular disease
Lung Disease
Obstructive Lung Disease Pack years of Smoking Age in Years Crothers, et al. Chest, 2006
Liver Disease
Cirrhosis, HIV, Complexity, and Age
Renal Disease
Work in progress VACS 8
Cancer
McGinnis KA, et al. Pre and post HAART cancer incidence among HIV positive veterans. XIV International AIDS Conference, Barcelona, Spain, July 7-12, 2002.
Hepatocellular Carcinoma Standardized IRRs Comparing HIV Infected Patients With HIV Negative Controls (n 42,037) Model 1 Model 2 Characteristic IRR 95% CI IRR 95% CI HIV 1.68 1.02 to 2.77 0.96 0.56 to 1.6 HCV — — 12.54 6.46 to 24.3 Alcohol ab/depend — — 1.85 1.03 to 3.35 Age 1.05 1.03 to 1.06 1.08 1.05 to 1.10 Race African American 2.06 1.05 to 4.06 1.36 0.70 to 2.65 Hispanic 4.88 2.20 to 10.81 3.98 1.81 to 8.76 Unknown/other 1.33 0.58 to 3.00 1.74 0.75 to 4.04 McGinnis et al. Hepatocellular Carcinoma and Non-Hodgkin’s Lymphoma: The Role of HIV, Hepatitis C Infection, and Alcohol Abuse J Clin Oncol 2006 24:5005-5009.
Non-Hodgkin’s Lymphoma Standardized IRRs Comparing HIV-Positive Subjects With HIV-Negative Controls (n 41,906) Model 1 Model 2 Characteristic IRR 95% CI IRR 95% CI HIV 9.71 6.99 to 13.49 10.03 7.19 to 13.97 HCV — — 0.73 0.50 to 1.06 Alcohol ab/depend — — 0.69 0.49 to .96 Age 1.02 1.01 to 1.03 1.02 1.01 to 1.03 Race African American 0.86 0.64 to 1.16 0.94 0.69 to 1.29 Hispanic 1.11 0.67 to 1.83 1.15 0.69 to 1.90 Unknown/other 0.81 0.57 to 1.15 0.77 0.54 to 1.09 McGinnis et al. Hepatocellular Carcinoma and Non-Hodgkin’s Lymphoma: The Role of HIV, Hepatitis C Infection, and Alcohol Abuse J Clin Oncol 2006 24:5005-5009.
Summary HIV is associated with comorbidities of aging with substance use COPD, Liver and Renal disease, HCV, Cancer HIV is not associated with common cardiovascular risk factors other than smoking Lower risk of obesity, hyperlipidemia, diabetes, and hypertension ARV therapy is associated with the cardiovascular risk factors listed above (of note HCV is associated with diabetes, decreased lipids, wasting)
Implications Comorbid disease patterns are distinct in HIV+ / HIV- veterans Some associated with HIV itself Partially due to substance use May also be due to decreased obesity and HCV infection Guidelines for screening and treatment of comorbid disease require adaptation to HIV+ patients Appropriate control groups (like those with HIV with respect to demographics, substance use, and HCV infection) will be essential to explore Risk of ARV toxicity Theories of early immune senescence Alternative mechanisms of disease
How Do We Prioritize Care? “Aggressive” HIV treatment HAART to slow HIV disease progression Prevention for positives Comorbidity and toxicity Prevention Screening and management Goals Minimize HIV transmission and cumulative frailty Maximize survival and quality of life As efficiently as possible
Developing and Validating a Preclinical Frailty Index for HIV Amy Justice, MD, PhD for the VACS Project Team
Frailty is evident over time through an excess vulnerability to stressors, with reduced ability to maintain or regain homeostasis after a destabilizing event…[it] is currently identified through characteristics that are directly related to physical function and that at the same time are consequences of the accumulation of subclinical conditions, acute and chronic disease, and behavioral and social risk factors. —AGS/NIA Conference on Frailty in Older Adults Walston et al. J Am Geriatr Soc 54:991-1001, 2006
Pre Clinical Frailty is The sum total “accumulation of subclinical conditions, acute and chronic disease, and behavioral and social risk factors” that provide an pre clinical indication of vulnerability. Pre clinical frailty should be predictive of: Mortality Hospitalizations and urgent visits Functional status Progression of pre clinical frailty
Likely Markers of Pre Clinical Frailty among HIV Infected Individuals Chronic viral infection/ inflammation (HIV, HCV, HBV) Immune dysfunction/ reduced function Bacterial and fungal infection Immune “reconstitution” syndrome Cancer Aging/ substance use organ system dysfunction Obesity and wasting Pulmonary dysfunction Vascular dysfunction Renal dysfunction Liver dysfunction Bone marrow dysfunction CNS dysfunction
Hypotheses Markers of preclinical frailty will add substantially to the ART CC model These will modify the association of ART-CC variables with mortality
1st Step: Develop a Preclinical Frailty Index for HIV Begin with established prognostic variables in HIV and mortality outcome Add indicators of preclinical frailty Begin with consistently measured and widely available measures Add other important measures to determine marginal improvement in accuracy
Antiretroviral Treatment Cohort Collaboration (ART-CC) Model 13 cohorts of HIV infected starting combination ART (CART) 12,574 patients 344 deaths, 750 AIDS events Median follow up 1.75 years (longest 5.5 years) State of the art methods (Weibull model) Based on CD4 cells (<50; 50-99; 100-199; 200-349;>350)/ml3 HIV-1 viral load >5 log/ ml3 Age>50 years IV Drug Use Stratified by cohort and CD4 count C statistic 0.73 in development paper
Not Included –Preclinical Frailty Age>65 Organ system (Bone Marrow; Renal; Liver; GI-Wasting) Behaviors (non IV Drug use; Alcohol; Tobacco; Adherence) High risk diagnoses (ICD-9, labs, pharmacy/ self report) Cancer (nonsquamous cell) Viral hepatitis (HCV, HBV) Other infectious diseases (fungal, mycobacterium, pneumonia, sepsis) Diabetes Pulmonary Disease (COPD) Vascular Disease Dementia Depression Functional status
Methods Development: Virtual Cohort- at CART initiation 7516 started CART 1/1/1999-8/1/2002 6439 (86%) CD4, VL, and hemoglobin levels 6 months 4,170 (56%) not missing any variables at baseline 1168 deaths within first 6 yrs Censored at 6 yrs to comply with modeling assumptions Validation and extension: VACS baseline 3,300 HIV+ and 3,300 HIV- Approximately 600 deaths within first 4 years (2/3rds in the HIV+)
Analyses Consider covariance and relative bivariate associations in grouping variables Nested Models (log linear, Poisson, Cox) ART-CC: cd4, vl, age 50+, IV drug use ART-CC + labs: add FIB 4, eGFR, anemia, age 65+ ART-CC + labs + diagnoses: add cancer, copd or pneumonia, vascular disease, diabetes Compare C statistics Calibration in the small Discrimination Short term and long term prediction
Etiologic Analyses Consider association with CD4, age, and substance use In strata of high and low CD4 determine whether variables still discriminate mortality Stratify models by calendar year of CART initiation to see if this alters association with mortality
Validation Analyses in VACS Evaluate missing data by imputing values Validate in separate data (VACS) Determine impact of variables in HIV- sample Extend by testing additional variables in HIV+/- Patient reported diagnoses Wasting, Obesity Functional Status Smoking IV drug use
Further Etiologic Analyses in VACS: CART Exposure Divide HIV infected patients into 3 strata of CART exposure Test whether, after adjustment for CD4, VL, adherence, and age, variables are associated with levels of CART exposure
Do Other Factors Contribute to Prognosis Do Other Factors Contribute to Prognosis? And Do These Modify Association between ART CC and Mortality?
C Statistics: Artcc: 0.66 +Labs: 0.71 +Diagnoses: 0.73
What if We Omit All AIDS Associated Variables? (Omitting CD4 cell count, HIV-1 viral load, and AIDS Defining Conditions)
Observed Mortality Rates by Deciles of Risk Deaths/100 PY
Prediction Deciles and Median CD4 Count CD4 Cell Count (median) per mm3
Major Contributors to Mortality Risk Advancing age Low CD4 cell counts Likely liver fibrosis Likely renal failure Anemia Viral hepatitis Cancer Drug or alcohol abuse/addiction
Less Impact, But Significant HIV-1 Viral Load AIDS Defining Conditions Vascular disease Pulmonary disease Diabetes
Next Steps Finalize variables in models Impute missing and run with and without Do nesting in the opposite direction as well Finalize content of 1st, 2nd, 3rd papers Adds information for mortality and associated with functional status (dev VC and val VACS) 6 month changes in markers important (VC and VACS) Predicts renal and liver toxicity after drug exposure
How Do We Prioritize Care? “Aggressive” HIV treatment HAART to slow HIV disease progression Prevention for positives Comorbidity and toxicity Prevention Screening and management Goals Minimize HIV transmission and cumulative frailty Maximize survival and quality of life As efficiently as possible
Veterans Aging Cohort Study PI and Co-PI: AC Justice, DA Fiellin Scientific Officer (NIAAA): K Bryant Participating VA Medical Centers: Atlanta (D. Rimland, C Jones-Taylor), Baltimore (KA Oursler, R Titanji), Bronx (S Brown, S Garrison), Houston (M Rodriguez-Barradas, N Masozera), Los Angeles (M Goetz, D Leaf), Manhattan-Brooklyn (M Simberkoff, D Blumenthal, J Leung), Pittsburgh (A Butt, E Hoffman), and Washington DC (C Gibert, R Peck) Core Faculty: K Mattocks (Deputy Director), S Braithwaite, C Brandt, K Bryant, R Cook, J Conigliaro, K Crothers, J Chang, S Crystal, N Day, J Erdos, M Freiberg, M Kozal, M Gaziano, M Gerschenson, B Good, A Gordon, J Goulet, M Hernan, K Kraemer, J Lim, S Maisto, P Miller, L Mole, P O’Connor, R Papas, H Paek, J Robins, C Rinaldo, M Roberts, J Samet, B Tierney, J Whittle Staff: D Cohen, A Consorte, K Gordon, F Kidwai, F Levin, K McGinnis, M Rambo, J Rogers, M Skanderson, F Whitsett Major Collaborators: Immunology Case Registry, Pharmacy Benefits Management, Framingham Heart Study, Women’s Interagency HIV Study, Massachusetts Veterans Epidemiology Research and Information Center (MAVERIC), Health Economics Research Center (HERC), Center for Health Equity Research and Promotion (CHERP), ART-CC, NA-ACCORD Funded by: National Institute on Alcohol Abuse and Alcoholism (2U10 AA 13566); National Institute on Aging (K23 G00826); Robert Wood Johnson Generalist Faculty Scholar Award; an Inter-Agency Agreement between National Institute on Aging, National Institute of Mental Health, and the Veterans Health Administration; the VHA Office of Research and Development; and, VHA Public Health Strategic Health Care Group.
Biometric Proxies for Comorbidity FIB4>3.25 for liver cirrhosis =age X AST /(platelets X Sqrt of ALT) eGFR<30 for renal insufficiency =186.3 X creatinine-1.154 X age-0.203 X (1-(0.258 X gender)) X (1+(0.21 X black)) BMI>30 for obesity =703 X weight (lbs) / height2 (in2)