HIV, Aging, and Comorbidity

1
HIV, Aging, and Comorbidity

• R. Scott Braithwaite, MD, MSc
• Yale University School of Medicine

2
Disclosure of Financial Relationships

• This speaker has no significant financial
  relationships with commercial entities to
  disclose.

This slide set has been peer-reviewed to ensure
that there areno conflicts of interest
represented in the presentation.
3
Outline

• ARV-related increased life expectancy
• The impact of increased life expectancy on
  comorbidity prevalence
• Distribution of comorbidities in HIV-infected
  individuals
• The impact of increased comorbidity on
• Timing of ARV initiation
• Appropriateness of primary care practice
  guidelines (e.g., colorectal cancer screening)

4
Definition of Comorbidity

• Any condition not included in the CDC list of
  AIDS defining conditions
• Includes conditions associated with HIV
  infection, but not an AIDS defining condition

5
Research Tool Computer Simulation

• Why use computer simulation?
• Can estimate mortality over long-periods (beyond
  clinical follow-up)
• Can weigh long-term harms (i.e. resistance)
  against short-term benefits (i.e. lower VL)
• Can integrate data from multiple sources
• Can compare scenarios unlikely to be evaluated in
  RCTs

6
Using the validated simulation to evaluate what
if scenarios

• If you followed all HIV-infected patients until
  death, how many would die of comorbidities rather
  than AIDS?
• Has HIV life expectancy become long enough to
  justify using guidelines with immediate harms and
  delayed benefits?
• Do cumulative benefits from earlier ARV
  initiation exceed cumulative harms?

7
Schematic of Simulation
8
Unobserved or rarely observed characteristics
Viral Replication
HIV Mutations
CART Resistance
CART Adherence
CART Toxicity
CART effectiveness
Patient Characteristics
Viral Load
CD4 Count
DEATH FROM HIV/AIDS
DEATH FROM COMORBIDITIES
9
Validation of Simulation
10
Results - Calibration
Proportion on 1st HAART regimen
CHORUS data
Simulation
Years after starting HAART
11
Results - Calibration
Proportion on 2nd HAART regimen
CHORUS data
Simulation
Years after starting HAART
12
Results - Calibration
Proportion on 3rd HAART regimen
CHORUS data
Simulation
Years after starting HAART
13
Results - Calibration
Proportion Surviving
CHORUS data
Simulation
Years after starting HAART
14
Results - Validation
3 year Mortality ()
Simulation
Cohort data
CD4 lt50 50-99 100-199 200-350
gt350
15
Using the validated simulation to evaluate what
if scenarios

• If you followed all HIV-infected patients until
  death, how many would die of comorbidities rather
  than AIDS?
• Do cumulative benefits from earlier ARV
  initiation exceed cumulative harms?
• Has HIV life expectancy become long enough to
  justify using guidelines with immediate harms and
  delayed benefits?

16
Simulation Results Life Expectancy with and
without ARV
Years
Age 50

without
Age 40
with
without
with
Age 30
without
with
CD4 750
CD4 500
CD4 200
17
Age 40
Life Expectancy with ARV
Years
30
20
xxxxxxxxxxxxxxxxxxxxxxx
10
HIV 104
0
HIV 105
HIV 106
CD4 800
CD4 500
CD4 200
18
Age 40
Deaths from comorbidities

80
30
60
20
xxxxxxxxxxxxxxxxxxxxxxx
xxxxxxxxxxxxxxxxxxxxxxx
40
10
20
HIV 104
0
0
HIV 105
HIV 106
CD4 800
CD4 500
CD4 200
19
Age 50
Life Expectancy with ARV
Years
30
20
xxxxxxxxxxxxxxxxxxxxxxx
10
HIV 104
0
HIV 105
HIV 106
CD4 800
CD4 500
CD4 200
20
Age 50
Deaths from comorbidities

80
60
xxxxxxxxxxxxxxxxxxxxxxx
40
20
HIV 104
0
HIV 105
HIV 106
CD4 800
CD4 500
CD4 200
21
Patterns of Comorbidity Among HIV Uninfected and
Infected Veterans
A brief respite from the hypothetical. Back to
real data
Goulet J., et al. CID, 2007
22
Objectives

• Describe pattern of comorbid disease experienced
  among HIV-infected veterans across their lifespan
• Compare with demographically similar HIV
  uninfected veterans

23
Methods VACS Virtual Cohort

• 21 Matching of HIV - HIV by race/ethnicity,
  age, and gender
• 66,840 uninfected 33,420 HIV-infected veterans
• Results based on baseline demographic, ICD-9
  diagnostic code, and laboratory data

24
Comorbidity Groups

• Medical Diagnosis

- Hypertension - Diabetes - Vascular Disease
- Pulmonary Disease - Renal Disease - Liver
Disease

• Substance Abuse/Dependence
• Alcohol Abuse and Dependence
• Drug Abuse and Dependence
• No tobacco (not accurately measured by ICD-9)
• Psychiatric Disorders
• Schizophrenia
• Major depression or Bipolar
• Post traumatic stress disorder (PTSD)

25
Timing and Definitions

• Baseline defined as
• Time of presentation into HIV care at VA (HIV)
• VA care same calendar year (HIV-)
• HIV largely untreated at baseline
• Comorbidity prevalence defined as
• 12 months pre- and 6 months post- baseline
• 2 outpatient or 1 inpatient ICD-9 code

26
Validation

• Compared with full chart extraction for
• Agreement, kappa
• Prevalence
• Final codes selected based on those that best
  mimic chart extraction
• ICD-9 codes
• are specific, but not sensitive
• prevalence estimates are conservative

27
Comorbidity by HIV Status Age
Substance Use Disorders
Medical Disease
ANY ONE CONDITION
Psychiatric Disorders
28
Multi Morbidity by HIV Status Age

MULTI MORBIDITY
29
Common Comorbidities of Aging
Hypertension
Diabetes
Vascular Disease
Pulmonary Disease
30
Comorbidities Associated with HIV
Liver Disease
Renal Disease
Note change in axis Goulet J. CID, in press,
December 2007
31
Comorbidities associated with HIV COPD
Age in years
Pack years of smoking
Crothers, K. Chest, 2006
32
Hypertension, Vascular Disease, and Diabetes

• Lower prevalence in HIV vs. HIV -
• Why?
• Differential detection
• HIV uninfected into care for conditions
• Associated with obesity and HIV leaner

33
BMI By HIV Status in VACS
Justice, Medical Care 2006, 44(8 Suppl 2)S13-24
34
Conclusion

• As HIV-infected patients live longer prevalence
  of comorbidites is rising
• 40-70 of patients have gt1 comorbidity
• Patterns of vary by HIV status
• HIV-infected patients more likely to have liver
  disease, renal disease, and multi-morbidity
• Obesity related diseases less prevalent in
  HIV-infected patients.

35
Now. Incorporating this real data to into the
computer simulation

36
Using the validated simulation to evaluate what
if scenarios

• If you followed all HIV-infected patients until
  death, how many would die of comorbidities rather
  than AIDS?
• Do cumulative benefits from earlier ARV
  initiation exceed cumulative harms?
• Has HIV life expectancy become long enough to
  justify using guidelines with immediate harms and
  delayed benefits?

37
HIV life expectancy with alternative CD4
thresholds for initiating antiretroviral therapy
A cohort simulation based on observational data
and Markov modeling R. Scott Braithwaite, MD,
MSc,1 Mark S. Roberts, MD, MPP,2,3 Chung Chou H.
Chang, PhD,2,3 Matthew Bidwell Goetz, MD,4
Cynthia Gibert, MD,5 Maria Rodriguez-Barradas,
MD,6 Steven Shechter, PhD,7 Andrew Schaefer,
PhD,8 Robert Koppenhaver, BS,8 Amy C. Justice,
MD, PhD1
Ann Intern Med. 2008 148 178-185.
38
Background

• When to start combination antiretroviral therapy
  (CART) is unclear
• Benefits with earlier treatment
• Reduce risk of AIDS/death
• Harms with earlier treatment
• Toxicity, particularly with increasing
  comorbidity and/or age
• Decreased Quality of Life
• Accumulation of genotypic resistance

39
Objective

• To estimate life expectancy and quality-adjusted
  life-years (QALY) with alternative CD4 thresholds
  for initiating antiretroviral therapy

40
Methods

• Integration of Markov modeling and observational
  cohort analysis
• Idea
• Use Markov modeling to weigh harm/benefit
  tradeoffs over long time periods
• Use observational analysis to estimate magnitude
  of harms and benefits

41
Methods Estimating harms

• CART-related toxicity (Virtual Cohort)
• Upper bound estimate
• Differential mortality rates between 5742
  HIV-infected patients on CART and 11484 matched
  uninfected controls
• Limited analysis to those with low probability of
  AIDS/HIV deaths
• CD4gt500
• Controlled for comorbidity, other factors

42
Methods Estimating Harms

• CART-related ?Quality of Life (VACS)
• Compared utility scores of patients
• Reported symptoms attributable to CART
• Did not report symptoms attributable to CART
• Utility scores estimated from SF 12 (SF 6-D)
• Controlled for other quality of life predictors
• Disaggregated from impact on HIV progression

43
Methods Estimating harms

• Genotypic mutation accumulation
• Estimated via previous calibration of simulation
  
• Validated via replication of data from disparate
  cohorts
• Approx 20 of CART-naïve patients with typical
  adherence develop gt1 mutations in first year
• Higher rates in subsequent years

44
Methods Estimating benefits

• Reduction in likelihood of AIDS/HIV-death
• Estimated via previous calibration of simulation
  
• Validated via replication of data from disparate
  clinical cohorts
• Survival increased 5 to 20 years by CART
• Reduction in likelihood AIDS/death greater with
  lower CD4 counts

45
Results

• CART Toxicity
• Based on virtual cohort analysis, may increase
  non-HIV-related mortality by up to 3.6-fold
• Worst case scenario or upper bound estimate
• If earlier treatment is favored using this
  estimate, it will definitely be favored
  considering true effect

46
Results

• CART quality of life impact
• Decreases utility by 0.08 if symptomatic
• 0.053 averaged across all patients on CART
• Utility is preference-weighted quality of life
  measure on scale of 0 to 1
• Effect size clinically meaningful but not
  overwhelming
• Similar to partial impotence or mild angina
• Less than complete impotence or moderate angina

47
Results Life Expectancy
48
Results - QALY
49
Cumulative incidence of resistance mutations at 5
years (mean)
7
6
5
4
3
2
1
0
200 350 500 200 350 500
200 350 500 200 350 500
CD4 treat (cells/ul)
Viral Load (copies/ml)
10,000 30,000
100,000
300,000
50
Cumulative CART regimens used at 5 years (mean)
4
3
2
1
0
200 350 500 200 350 500
200 350 500 200 350 500
CD4 treat (cells/ul)
Viral Load (copies/ml)
10,000 30,000
100,000
300,000
51
Treat CD4200 favored
CART-related toxicity (X non-HIV mortality)
Treat CD4350 favored
Treat CD4500 favored
Plausible Upper bound
10 30 100 300 10 30
100 300 10 30 100 300
Viral Load (X103 cells/ml)
30
40 50
Age (years)
52
Limitations

• Toxicity estimate is upper bound
• When simulation suggests that earlier treatment
  is favored, it really is
• When simulation suggests that later treatment is
  favored, we dont really know
• Does not consider newer CART regimens or baseline
  resistance
• However both would bias model towards later
  treatment so inferences still valid

53
Conclusions Treat earlier but use caution in
elderly
54
Using the validated simulation to evaluate what
if scenarios

• If you followed all HIV-infected patients until
  death, how many would die of comorbidities rather
  than AIDS?
• Do cumulative benefits from earlier ARV
  initiation exceed cumulative harms?
• Has HIV life expectancy become long enough to
  justify using guidelines with immediate harms and
  delayed benefits?

55
Tailoring Clinical Guidelines to Comorbidity The
Case of Cancer Screening in HIV

• R. Scott Braithwaite, MD, MSc
• John Concato, MD
• Chung Chou Chang, PhD
• Mark S. Roberts, MD
• Amy C. Justice, MD, PhD

Arch Intern Med. 2007 167(21)2361-5.
56
Introduction

• Individuals with HIV living longer
• Increasingly likely to die of non-HIV illnesses
• However, life expectancies still shorter than for
  general population
• Especially for low CD4 and/or salvage regimens
• No systematic method to predict whether
  guidelines developed on general population should
  apply to individuals with HIV

57
Introduction

• Payoff Time Minimum time until incremental
  benefits gt incremental harms
• Applies to any guideline where harms are
  short-term and benefits are long-term
• Colorectal cancer screening (CRC)
• Will vary by guideline and by patient population
• Payoff time can be compared to life expectancy
• If death likely before payoff time, guideline not
  advised
• If death unlikely before payoff time, guideline
  advised

58
Objective

• To predict which HIV patients would benefit from
  colorectal cancer screening.

59
Illustrative Cases

• 1. 60 year-old HIV male on salvage ARV, CD4
  count 46
• Comorbidities COPD (severe), hepatitis C
• 2. 60 year-old HIV female on 1st line ARV, CD4
  count 392
• Comorbidities diabetes

60
(No Transcript)
61
Compare payoff time to life expectancy
Case 1 60 year-old HIV male on salvage ARV,
CD4 count 46

• Payoff time for Case 1 is 7.3 years
• Life Expectancy for Case 1 is 5.1 years
• Because life expectancy is less than payoff time
  (minimum time until benefits exceed harms), Case
  1 is unlikely to benefit from colorectal cancer
  screening

62
Compare payoff time to life expectancy

• Case 2 60 year-old HIV female on 1st line ARV,
  CD4 count 392

• Payoff time for Case 2 is 5.7 years
• Life Expectancy for Case 2 is 15.1 years
• Because life expectancy is less than payoff time
  (minimum time until benefits exceed harms), Case
  2 is likely to benefit from colorectal cancer
  screening

63
Limitations

• Analyses do not consider rate of developing new
  ARVs
• Simulation can be modified to address this
• Framework will not be applicable to every HIV
  patient
• Requires EMR/informatics capability to integrate
  easily into care

64
Conclusion

• Payoff time is quantitative objective framework
  for predicting who will benefit
• CRC screening may not always be appropriate for
  HIV individuals
• Low CD4
• Salvage ARV
• May simultaneously improve quality of care and
  reduce resource expenditures
• May impact quality measures and P4P rules

65
Outline

• Amount by which ARV has increased life expectancy
• The impact of increased life expectancy on
  comorbidity prevalence
• Distribution of comorbidities in HIV-infected
  individuals
• The impact of increased comorbidity on
• Timing of ARV initiation
• Appropriateness of primary care practice
  guidelines (e.g., colorectal cancer screening)

66
Summary

• ARV has dramatically increased survival
• HIV just another chronic disease, like diabetes?
• Increased survival has increased prevalence of
  non-HIV-related comorbidities
• Increasing evidence favors starting HAART earlier
• Benefit may be lower with ? age or comorbidity
• Primary care screening guidelines are often
  applicable to HIV patients
• Payoff time may help to determine when particular
  guidelines are applicable