HIV Cohorts Do they offer relevant information or second level data - PowerPoint PPT Presentation

Loading...

PPT – HIV Cohorts Do they offer relevant information or second level data PowerPoint presentation | free to download - id: 234332-ZDc1Z



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

HIV Cohorts Do they offer relevant information or second level data

Description:

Observational studies compared to clinical trials. So why bother with ... ZDV(D4T) / LAM vs. ZDV(D4T) / LAM / IDV. FHDH 9712. EuroSIDA 1367. SHCS 1011 ... – PowerPoint PPT presentation

Number of Views:39
Avg rating:3.0/5.0

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: HIV Cohorts Do they offer relevant information or second level data


1
(No Transcript)
2
HIV Cohorts Do they offer relevant information
or second level data?
  • Amanda Mocroft
  • Royal Free and University College Medical School,
    London, UK

3
Overview
  • Review of the different types of evidence
  • Confounding by indication
  • Observational studies compared to clinical trials
  • So why bother with observational studies?
  • Clinical trials versus observational studies
  • Conclusions

4
Overview
  • Review of the different types of evidence
  • Confounding by indication
  • Observational studies compared to clinical trials
  • So why bother with observational studies?
  • Clinical trials versus observational studies
  • Conclusions

5
Randomised controlled clinical trial
  • Patients are RANDOMLY assigned to 2 or more
    treatment groups
  • Experimental group receives new intervention
  • Comparison group receives standard of care or
    placebo
  • Well balanced for confounders
  • Allows direct assessment of treatment
    intervention

6
Controlled clinical trial
  • Patients are assigned to 2 or more treatment
    groups
  • Allocation to treatment is not randomised
  • More likely to suffer from bias than randomised
    controlled trials

7
Cohort (observational) studies
  • Non-experimental study design
  • Follows groups of patients and considers how
    events differ
  • Estimates whether exposure causes events
  • Prospective studies follow patients forward from
    given time point
  • Prospective more reliable than retrospective

8
Case control study
  • Examines group of people with the event to those
    without the event
  • Examines how exposure differs between the 2
    groups
  • Most useful for rare events
  • Provide weaker evidence that cohort studies

9
Case series
  • Analysis of series of people with the disease
  • No comparison group
  • Weaker evidence than case-control studies

10
How is evidence rated?
11
Overview
  • Review of the different types of evidence
  • Confounding by indication
  • Observational studies compared to clinical trials
  • So why bother with observational studies?
  • Clinical trials versus observational studies
  • Conclusions

12
Confounding by indication
  • When a drug (antiretroviral) treatment serves as
    a marker for a particular clinical characteristic
  • Underlying difference in patients taking
    treatment A versus treatment B which may be
    related to outcome

13
Confounding by indication
  • May be caused by known, measured or unmeasured
    factors
  • Known measured
  • exposure group
  • age
  • Known unmeasured
  • currently active IDU
  • smoking status

14
Confounding by indication
  • What about unmeasurable or unknown?
  • Unmeasurable
  • stability of lifestyle
  • Tolerability of toxicities
  • Unknown
  • ???????????

15
Confounding by indicationExample
  • Comparison of Single and Boosted Protease
    Inhibitor
  • Versus Nonnucleoside Reverse Transcriptase
    Inhibitor
  • Containing cART Regimens in Antiretroviral-Naïve
    Patients
  • Starting cART After January 1, 2000
  • A Mocroft et al for the EuroSIDA study group

HIV Clinical Trials 20067271-284.
16
Confounding by indicationExample
  • Odds of starting PI-based versus NNRTI-based cART
  • regimen after 1/1/2000
  • Adjusted analysis
  • patients from Northern Europe had significantly
    lower odds of starting a PI-based regimen
  • OR 0.45, 95 CI 0.300.66, plt0001
  • patients with a higher CD4 count nadir had
    significantly lower odds of starting a PI-based
    regimen
  • OR 0.71 per doubling of CD4 nadir, 95 CI
    0.640.66, p lt .0001)

Mocroft et al HIV Clinical Trials 20067271-284.
17
Confounding by indicationExample
  • Regional differences
  • availability of antiretrovirals
  • local clinical guidelines
  • something else?
  • Less sick patients more likely to start
    NNRTI-based regimen
  • Evidence from clinical trials
  • CD4 and/or VL do not capture all clinical
    information

18
Confounding by indicationWhat can we do about it?
  • Awareness of issue
  • Use more sophisticated statistical techniques
  • Propensity scores uses dataset to model the
    probability of starting a given treatment
  • Heckman models
  • Marginal structural models
  • How do we know if we have removed the bias?

19
Overview
  • Review of the different types of evidence
  • Confounding by indication
  • Observational studies compared to clinical trials
  • So why bother with observational studies?
  • Clinical trials versus observational studies
  • Conclusions

20
Aims To assess the degree of bias present when
evaluating clinical effectiveness of
antiretroviral regimens in observational
databases by comparing results with those from
randomised clinical trials
21
Comparison 1ZDV monotherapy versus dual
nucleosides
  • FHDH 11102
  • EuroSIDA 1099
  • SHCS 656
  • Trials
  • Top line ZDV / DDC
  • Bottom line ZDV / DDI

Phillips et al. AIDS 1999132075-2082
22
Comparison 2ZDV / DDI vs ZDV / DDC
  • FHDH 13201
  • EuroSIDA 1544
  • SHCS 750

Phillips et al. AIDS 1999132075-2082
23
Comparison 3ZDV(D4T) / LAM vs ZDV(D4T) / LAM /
IDV
  • FHDH 9712
  • EuroSIDA 1367
  • SHCS 1011

Phillips et al. AIDS 1999132075-2082
24
What can you conclude?
  • Comparisons can be misleading when there are
    substantial imbalances in prognostic variables
    between treatment groups
  • Bias is not necessarily present and in most
    circumstances, the results from observational
    studies were similar to those of the clinical
    trials

25
Comparison of efavirenz and nevirapine RCTs
26
2NN Study with treatment failure at/before
week 48
Randomised clinical trials
P 0.994
P 0.091
N 220
387 400
209
Adapted from Van Leth et al Lancet
20043631253-63
27
2NN Study with VL lt 50 copies/ml
Randomised clinical trials
Van Leth et al Lancet 20043631253-63
28
2NN StudyIncrease in CD4 cells
Randomised clinical trials
Van Leth et al Lancet 20043631253-63
29
Comparison of efavirenz and nevirapine
Observational studies
1st author Year NVP / EFA ARV-naïve ARV-naïve ARV-naïve ARV-naïve
Phillips 2001 1325 / 878 Mix Mix Mix Mix
Matthews 2002 167 / 237 Yes Yes Yes Yes
Cozzi Lepri 2002 460 / 234 Yes Yes Yes Yes
Kaiser 2002 523 / 555 Yes Yes Yes Yes
Bannister 2007 389 / 370 Mix Mix Mix Mix

Adjusted additionally for resistance to started regimen Adjusted additionally for resistance to started regimen Adjusted additionally for resistance to started regimen Adjusted additionally for resistance to started regimen
30
Comparison of efavirenz and nevirapine
Observational studies
  • How was failure defined?
  • In patients who responded to treatment and
    achieved VL lt 500
  • VL gt 500
  • In patients who did not respond to treatment
  • VL gt 500 after 6 months of treatment

Kaiser et al used 400 copies/ml
31
Comparison of efavirenz and nevirapine
Observational studies
Plt0.0001 0.02 0.0006
lt0.0001 lt0.0001
Increased risk EFV vs NVP
Risk of failure EFV vs NVP (RH or OR)
Decreased risk EFV vs NVP
0.33
32
Overview
  • Review of the different types of evidence
  • Confounding by indication
  • Observational studies compared to clinical trials
  • So why bother with observational studies?
  • Clinical trials versus observational studies
  • Conclusions

33
Observational studiesSo why bother?
  • Natural history studies and comparative
    populations
  • Adverse events
  • Discontinuation of disease-specific prophylaxis
  • Complementing the design of clinical trials

34
Natural historyWithout treatment
Egger et al. Lancet 2002360119-129
35
Natural historyWith treatment
Egger et al. Lancet 2002360119-129
36
Natural historyPredicted 6 month risk of AIDS or
death
Phillips et al. AIDS 20041851-58
37
Natural historyImpact of cART AIDS and death
Incidence per 100 PYFU (95 CI)
Calendar year of follow-up
Updated from Mocroft et al. Lancet 200336222-29
38
  • N3825 patients starting cART from ART-naive
  • Started cART between 1996-2002
  • Virologic failure if 1st VL at 6-12 months after
    starting cART gt 500 copies/ml
  • with virologic failure decreased from 38.9 for
  • patients starting cART in 1996 to 24.8 in 2003

Lampe et al, Arch Intern Med 2006166521-528
39
Unadjusted and adjusted risk ratios of
virological failure by year of starting cART
1999 is reference category. Unadjustedadjusted
for cohort only Adjusted 1adjusted for cohort,
age, risk group, pre-HAART VL and CD4 count,
previous AIDS Adjusted 2 adjusted for all
above factors plus starting regimen as defined by
3rd drug and nucleoside combination.
Lampe et al, Arch Intern Med 2006166521-528
40
N2488 patients with triple-class failure 276
deaths during follow-up of deaths with known
cause due to HIV 66
dead by years from triple class
failure 1 2 3 4 4.5 10.0 15.3 21.3
Ledergerber et al, Lancet 2004 36951-62
41
CD4-cell count slopes for patients whose
off-treatment viral load is known
Ledergerber et al, Lancet 2004 36951-62
42
Predictors of death in people with triple class
failure
Relative hazard
Univariable
Multivariable
1 2.21 (1.51-3.22) 5.75 (4.09-8.10)
1 1.02 (0.68-1.55) 0.97 (0.64-1.48)
Current viral load lt4 4-5 gt5
Viral load ever lt 500 copies/mL before baseline
0.37 (0.27-0.50)
0.84 (0.59-1.19)
1 3.64 (2.13-6.22) 27.0 (16.9-43.3)
1 2.87 (1.65-5.00) 15.8 (9.28-27.0)
Current CD4 gt200 50-200 lt50
Stage CDC C at baseline
1.53 (1.11-2.10)
2.38 (1.79-3.17)
Age (per 10 years older)
1.16 (1.01-1.34)
1.24 (1.06-1.44)
Female gender
0.98 (0.64-1.51)
0.94 (0.59-1.51)
Infection via injection drug use
1.18 (0.80-1.74)
1.63 (1.05-2.54)
Total drugs currently on 0 (off-treatment) 1-2
3 4 gt5
5.64 (4.09-7.80) 1.73 (1.12-2.67) 1 0.71
(0.46-1.11) 0.85 (0.52-1.38)
2.85 (1.98-4.10) 1.20 (0.77-1.89) 1 0.65
(0.41-1.03) 0.86 (0.51-1.45)
Ledergerber et al, Lancet 2004 36951-62
43
N1160 patients starting antiretrovirals 47
with clinical adverse event (9 severe) 27 with
laboratory adverse event (16 severe)
44
Clinical adverse events
Fellay et al Lancet 20013581322-132
45
Laboratory adverse events
Fellay et al Lancet 20013581322-132
46
  • N23,468 patients
  • 36,199 PYFU, 126 MI
  • Adjusted RR per years exposure to cART 1.26
  • (95 CI 1.12 1.41), plt0.0001

DAD Study Group N Engl J Med 2003349
1993-2003
47
MI by CART exposure
MIs per 1,000 PY (95 CI)
Test for trend plt0.00001
Years on CART
Total 126 36,199
No. MIs
3 9 14 22
31 47
No. PY
5,714 4,140 4,801 5,847
7,220 8,477
DAD Study Group N Engl J Med 2003349
1993-2003
48
  • N23,437 patients
  • 94,469 PYFU, 345 MI

49
Adjusted risk of MI and exposure to NNRTIs and PIs
IRR 1.16 (1.10 1.23), plt0.001
IRR 1.05 (0.98 1.13), p0.17
DAD Study Group NEJM 20073561723-1735
50
Other adverse events and observational studies
  • Pancreatitis
  • Lactic acidosis
  • Chronic kidney disease
  • End stage renal disease
  • Non-AIDS defining malignancies
  • Liver-related disease
  • Metabolic syndrome

51
  • N7333 patients
  • 378 discontinued primary chemoprophylaxis
  • 59 discontinued secondary chemoprophylaxis
  • No relapses of PCP during 247 PYFU
  • Incidence 0 95 CI 0 1.5

Van Weverling et al, Lancet 19993531293-1298
52
  • N325 patients discontinued secondary PCP
  • chemoprophylaxis after starting cART and with a
  • CD4 gt 200 at discontinuation
  • No relapses of PCP during 325 PYFU
  • Incidence 0 99 CI 0 1.3

Ledergerber et al, NEJM 2001344168-174
53
N358 patients on cART interrupting maintenance
chemoprophylaxis for CMV, MAC, toxoplasmosis
and cryptococcosis whilst CD4 gt 50 5 relapses
during 781 PYFU
Kirk et al, Ann Intern Med 2002137 239-250
54
Incidence of relapse after discontinuation of
maintenance prophylaxis
  • Disease Events PYFU Rate (95 CI)
  • per 100 PYFU
  • CMV 2 370 0.54 (0.07 1.95)
  • MAC 2 222 0.90 (0.11 3.25)
  • Toxo 1 119 0.84 (0.02 4.68)
  • Crypto 0 70 0.0 (0.0 5.27)

Kirk et al, Ann Intern Med 2002137 239-250
55
(No Transcript)
56
The ESPRIT Study (1)
  • Multicenter, international, randomised clinical
    trial
  • Compare rates of new AIDS/deaths in patients on
    ARVs with or without IL-2
  • Ongoing trial
  • Currently blinded to treatment differences

57
The ESPRIT Study (2)
  • Data from EuroSIDA used in design of ESPRIT in
    1998 to predict event rates
  • Only initial 2 years could be predicted
  • Sample size and planned follow-up in ESPRIT
  • 4000 patients
  • 320 endpoints (new AIDS/death)
  • 80 power to detect 27 reduction in clinical
    progression in IL-2 arm

58
Patient characteristics (1)
EuroSIDA ESPRIT N 4482 4150
Male 78 76 White 86 76 IDU
18 10 HBV ve 6 6 HCV ve
20 13 Prior AIDS 26 24 Baseline VL
lt400 cp/ml 61 65
59
Patient characteristics (2)
EuroSIDA ESPRIT N 4482 4150
Median (interquartile range) Age 39 (34
46) 40 (34 46) Baseline CD4 (/mm3) 404 (342
500) 458 ( (373 586) Nadir CD4 (/mm3) 160
(67 254) 199 (92 310) Months since starting
ARVs 39 (21 68) 50 (26 77) Baseline 1/99
(3/98 7/00) 9/01 (10/975/03)
60
Observed event rates in EuroSIDA and predicted
event rates in control arm of ESPRIT
Incidence rate (95 confidence interval)
Events 55 61 52
46 44
258 PYFU 4376 3891 3171
2782 3123 17433

61
Derivation of score for clinical progression
  • Based on Cox proportional hazards using baseline
    variables
  • RH 95 CI p
  • IDU 2.33 (1.75 3.11) lt0.0001
  • Age 1.05 (1.04 1.07) lt0.0001
  • Prior AIDS 1.30 (1.00 1.70) 0.055
  • VL 1.33 (1.18 1.50) lt0.0001
  • Used to derive risk score for clinical
    progression

62
Event rates according to score decile in the
EuroSIDA study
Incidence rate (95 CI)
Score decile Score 0.84713IDU 0.05227age
0.26313AIDS 0.28596log10VL
Group 1 lt2.28 Group 4 3.21 3.38 Group 2
2.28 3.02 Group 5 3.39 3.85 Group 3 3.03
3.20 Group 6 gt3.85
63
Application to ESPRIT
  • EuroSIDA incidence rates in 6 groups reduced by
    13.5 (IL-2 effect)
  • Assumed LTFU of 2 (uniform occurrence)
  • Each patient in ESPRIT had score calculated at
    baseline
  • Number of ESPRIT patients in each of 6 score
    groups calculated
  • Incidence rate applied to estimate number of
    events

64
Accumulation of events
Score group
Number of events
Years since initiation of ESPRIT
46 91 133
175 215 253 290 320
65
Useful exercise?
  • Similar methodology used from Framingham in
    cardiovascular trials
  • Can inform on expected clinical event rate in
    randomised trials
  • Observational studies can be useful at design
    stage and to check trial assumptions during
    follow-up

66
Overview
  • Review of the different types of evidence
  • Confounding by indication
  • Observational studies compared to clinical trials
  • So why bother with observational studies?
  • Clinical trials versus observational studies
  • Conclusions

67
Issues to consider (1)
  • Loss to follow-up
  • LTFU lower in clinical trials
  • Patients well motivated in clinical trials
  • Patients in observational studies may move
    between cohorts or clinics
  • Standardised visit schedules
  • Data quality assurance
  • Probably underused in observational studies
  • Monitoring in clinical trials
  • Verification of clinical events
  • Variability on lab measurements

68
Issues to consider (2)
  • Bias
  • Confounding by indication
  • Representativeness of patients in cohorts or RCTs
  • Comparability of results from cohorts
  • Large studies not necessarily less biased
  • Collection of confounding data
  • Dont always know what will be confounders
  • Difficult to do retrospectively
  • Standardised definitions of confounders
  • Different frequency of lab measurements

69
Issues to consider (3)
  • Under-reporting of events
  • More common in observational studies
  • Some events may be more likely to be reported
    than others
  • Problems with trend data and reporting bias
  • Methods of data collection and diagnosis
    definitions
  • Heterogeneity in cohort definitions
  • Passive versus active reporting
  • Use of external data registries

70
Clinical Trials
Pros and Cons
Cohort Studies
  • Randomised unbiased comparison of patients
  • Better ascertainment of endpoints and lower LTFU
  • Standardised diagnostic criteria and data
    collection
  • Larger number of patients under FU
  • Cheaper
  • Useful for informing clinical trials
  • Generalisability

71
HIV Cohorts Do they offer relevant information
or second level data?
72
HIV Cohorts Do they offer relevant information
or second level data?
YES
YES
YES
YES
YES
YES
YES
YES
YES
73
Conclusions
  • Of course observational studies have relevant
    information to offer
  • Unfair to call information second level data
    because they can inform where no data from
    clinical trials will ever be available
  • Always interpret data carefully and consider
    potential for biases
  • Randomised clinical trials should remain the
    gold-standard for comparison of treatment
    efficacy
About PowerShow.com