Title: Endpoints in HIVAIDS Clinical Trials Qiming Liao, Ph.D. GlaxoSmithKline
1Endpoints in HIV/AIDS Clinical Trials Qiming
Liao, Ph.D.GlaxoSmithKline
2Outline
- AIDS epidemic and survival
- HIV/AIDS drugs and treatment
- Surrogate markers and historic change in primary
endpoint in HIV/AIDS clinical trials - Reduction in HIV-1 RNA levels
- Time to virologic failure and time to loss of
virologic response (TLOVR) - Early virologic non-response
- Issues on choice of endpoints
- Summary
3(No Transcript)
4Estimated proportion of persons surviving, by
months after AIDS diagnosis during 19942001 and
by year of diagnosisUnited States
Source CDC
5Classes of HIV Drugs
- Nucleoside reverse transcriptase inhibitors (ZDV,
3TC, Combivir, Abacavir, Trizivir, ddI, d4T, TDF,
ddC) - Protease inhibitors (APV, IDV, LPV, NFV, RTV,
SQV, ATV, FPV) - Non-nucleoside reverse transcriptase inhibitors
(EFV, DLV, NVP) - Fusion inhibitors
6Approved Antiretroviral Agents
3TC/ZDV
TDF
ABC/3TC
FTC/TDF
DLV
ddC
3TC
ABC
EFV
NVP
ddI
ZDV
d4T
FTC
ABC/3TC/ZDV
87
91
92
94
95
96
97
98
99
00
88
89
90
01
04
03
02
93
T-20
NRTI
PI
NNRTI
FI
7HIV Treatment
- Monotherapy (late 1980s, ZDV)
- Mono and Dual therapy (early 1990s)
- Highly active antiretroviral therapy (HAART,
since 1996) at least 3 drugs - Complications of HAART
- viral resistance to drug
- fat redistribution
- hyperlipidemia
8HIV Care and Clinical Trials
- Standard of care and first line treatment
regimens (IAS HIV Treatment Guidelines) - Dynamic therapeutic area (treatment changes over
time) - Active control and relatively a large number of
tries with open label - Superiority and non-inferiority design
- Naïve population symptom rare
9Clinical Endpoint and Surrogate Markers
- First ZDV trial death
- Disease progression or death
- Surrogate markers CD4 cell count and HIV-1 RNA
levels - Endpoints based on HIV-1 RNA
- reduction from baseline
- proportion of subjects lt 400 (or 50) copies/mL
(lower limit of the assay) - time to virologic failure
10Reduction in HIV-1 RNA
- Surrogate Endpoint Reduction of HIV-1 RNA from
baseline - Assay limitation (e.g. HIV-1 RNA lt400 copies/mL,
or lt50 copies/mL (lower limit), HIV RNA gt750,00
copies/mL (upper limit)) - Statistical problem What to do with censored
HIV-1 RNA measurements?
11Reduction in HIV-1 RNA
- Simplify the problem
- assigning the lower or upper limit (or lower
limit -1 or upper limit 1) to those censored
measurements - analysis of variance (log10 transformation) or
non-parametric methods - Take into account of censoring and use
time-to-event techniques
12Reduction in HIV-1 RNA
- Assumptions
- no censoring at baseline
- only possible left censoring post baseline
- Reduction of HIV-1 RNA is observed or right
censored - Use Kaplan-Meier, logrank test etc.
- Some Issues
- reduction in HIV-1 RNA could be negative (easy to
deal with shifting) - reduction depends on Baseline HIV RNA (use Cox
proportional hazards model)
13Reduction in HIV-1 RNA
- Other Methods
- Check assumption of normality of HIV-1 RNA
reductions (in log10 transformation) - Use parametric analysis analogous to the
Kaplan-Meier based analysis - Fit censored analysis of covariance using the
LIFEREG procedure in SAS - Accelerated failure time model without worrying
about the assumption of normality
14Endpoint of Percent Undetectable on HIV-1 RNA
- Proportion of HIV-1 RNA lt 400 copies/mL (assay
limit 400) and lt200, 50 copies/mL (assay limit
50) at Weeks 16, 24, 48 - Rates vary from studies to studies
- Depend on study population, entry criteria
- Analysis methods on handling missing data,
treatment switch and intensification
15Endpoint of Percent Undetectable on HIV-1 RNA
- Proportion of HIV-1 RNA lt 50 at Week 24
- ITT MF (missing considered as failure, switch
included) - ITT MF, switchF
- Observed (missing ignored, switch included)
- As treated (missing ignored and switch not
included) - Model-based approach
- Confusing clinicians and others in the field?
- Cautious on cross study comparisons
16Time to Virologic Failure
- Virologic failure
- two successive measurements of HIV-1 RNA gt 400
(50, 200) copies/mL after HIV RNA lt400 (50)
copies/mL or never get below 400 (50) copies/ml
by Week 24 (16, 36) - definition varies
- Two types of virologic failure
- Lack of initial response or non-responder
- Early rebound and later relapse
17Time to Virologic Failure
- Issues on how to define time to virologic failure
- Non-responder or virologic rebounder, which does
better? - Artificial nature of clinical trial (e.g. allow
24 weeks window for response) creates issues with
definition of events, censoring, etc -- see
accompanying examples
18(No Transcript)
19(No Transcript)
20Time to Virologic Failure
- Issues on defining time to virologic failure
- PT 3 PT 4 PT 5
- Method 1 0 0 20
- Method 2 24 24 24
- Method 3 (0,24) (0,24) 20
- (0,24) means censored between 0 and 24
- Which method is more reasonable?
- Method 3 with some statistical challenge due to
both right and left censoring
21More Time to Event Endpoints
- Time-to-event endpoints
- time to virologic failure (TVF)
- time to virologic response (TVR)
- duration of viral suppression (DVS)
- TVFTVR DVS
- composite endpoint
- Time to loss of virologic response (TLOVR)
(regimen termination end point (RT))
22Composite Endpoint
- Time to loss of virologic response (regimen
termination), defined as the first occurrence of - virologic failure
- lack of initial response, early rebound, or later
relapse - permanent treatment discontinuation
- AIDS-defining illness
- death
- study withdrawal
23Time to Virologic Failure VS TLOVR
- Time to virologic failure versus time to regimen
termination endpoint (TLOVR)? - NIH/FDA Workshop on HIV Endpoints (February,
2001) - Gilbert, DeGruttola et al. Virologic and regimen
termination surrogate end points in AIDS clinical
trials, JAMA 2001 - How to monitor the study? If co-primary endpoints
are chosen, how to define stopping rules?
24Time to Virologic Failure VS TLOVR
- Both endpoints are imperfect surrogates for
clinical endpoints - large uncertainties regarding how well treatment
effects on the surrogate endpoints predict
treatment effects on clinical outcomes - e.g., a meta-analysis of 16 clinical trials of
dual nucleoside regimens showed limited surrogacy
of the viral load change endpoint - virologic information does not necessarily
measure problems associated with treatment - e.g., toxic effects, adherence difficulties,
resistance
25Relative Advantages of a TLOVR Endpoint
- In studies of sequences of regimens, it more
directly counts the costs of expending regimens - toxicity costs
- non-adherence costs
- drug resistance costs
- Higher event rate compared to a purely virologic
endpoint - disseminate study results more quickly
- caution more events dont necessarily result
more statistical power
26Relative Advantages of a TLOVR Endpoint
- HIV-1 RNA alone (purely virologic endpoint) can
be misleading when patients with tolerability
problems on an inferior regimen are salvaged with
a superior regimen - TLOVR reflects the current belief that the longer
the first regimen effectively suppresses HIV RNA
with tolerable side effects the more patients
will benefit Blinding may be compromised during
the first treatment regimen termination for side
effects - Patient management is no longer uniform after the
first treatment regimen termination - Rebound may occur during regimen modification
- leading to a HIV-1 RNA endpoint that is not
related to the virological efficacy of the
regimen under study
27Relative Advantages of a Purely Virologic Failure
Endpoint
- Allows separate investigation of efficacy and
safety - Defined more objectively than a regimen
termination endpoint - less dependent on physician/patient choice
- less subject to bias in open-label studies
- Its use as a primary endpoint assures that a
secondary analysis of a regimen termination
endpoint can be carried out
28Relative Advantages of a Purely Virologic Failure
Endpoint
- Gives equal weight to
- mild treatment-limiting side effects and serious
irreversible toxicities - virologic endpoints regardless of the remaining
therapeutic options - treatment-related failures and loss-to-follow-up
- Encourages less diligent follow-up after initial
regimen termination - Leads to limited ability to assess the extent of
remaining therapeutic options
29Clinical Beliefs Underlying the Use of Each
Endpoint
- Purely virologic endpoint
- the effect of the investigated therapies on
plasma HIV-1 RNA levels captures the essential
information needed to define the role of the
therapies in clinical practice for the target
population - TLOVR endpoint
- expending a regimen more closely measures
tangible harm
30Recommendations from FDA GuidanceClinical
Considerations for Accelerated and Traditional
Approval of ART Drugs Using Plasma HIV RNA
Measurements
- Blinded comparisons with well-characterized
controls are preferable - Studies that clearly show the investigational
drugs contribution to decreases in HIV RNA as
part of a combination regimen (Accelerated
Approval 24 weeks) - Studies that show the drugs contribution toward
sustained suppression of HIV RNA (Traditional
Approval 48 weeks) - Critical that management decisions be made in
uniform manner - TLOVR endpoint analysis is required
- Sensitivity analyses are encouraged
- Pure Virological ITT as well as any other
analysis thought useful by sponsor may also be
performed
31Abacavir/Combivir versus Indinavir/Combivir in
Therapy Naïve Adults at 48 Weeks (CNAAB3005)
- Multi-center, double-blind, controlled study in
which 562 HIV-infected, therapy-naive adults with
a pre-entry plasma HIV RNA gt10,000 copies/mL were
randomized to receive either - ZIAGEN (300 mg twice daily) plus COMBIVIR
(lamivudine 150 mg/zidovudine 300 mg twice
daily), or - Indinavir (800 mg 3 times a day) plus COMBIVIR
twice daily. - Study participants were male (87), Caucasian
(73), African-American (15), and Hispanic (9) - Median age was 36 years, the median pretreatment
CD4 cell count was 360 cells/mm3, and median
plasma HIV RNA was 4.8 log10 copies/mL
32CNAAB3005Virologic Endpoint and Options
- Virologic endpoint confirmed HIV RNA gt 400
copies/mL at Week 16 or later - Patients who met virologic endpoint had 3 options
- Continue randomized therapy
- Discontinue randomized therapy and receive
open-label ABC (300mg BID) and/or IDV (800mg q8h)
and/or 3TC/ZDV (150/300mg BID) and/or add any
licensed ART - Discontinue all study medication and withdraw
from study - Patients who permanently discontinued study med
prior to Week 16 were to be followed by
investigator
33CNAAB3005Subject Accountability at Week 48
IDV/COM
ABC/COM
34(No Transcript)
35(No Transcript)
36ACTG 364 Study Design Phase II, Randomized,
Partially Double - Blind
ACTG 364
ACTG 302
R A N D O M I Z E
NFV EFV placebo NRTIs
ACTG 175
NFV placebo EFV NRTIs
NFV EFV NRTIs
ACTG 303
1 or 2 new NRTIs assigned as part of dual NRTI
components (d4T ddI, d4T3TC, or ddI3TC)
Source Presentation by Bosch and Gilbert on HIV
Endpoint Workshop, 2001
37ACTG 364
38ACTG 364
39Issues on Salvage Therapy for Experienced Patients
- Differences relative to naïve setting
- More heterogeneity of drug histories and
treatments - More advance disease stage
- Patients may be more willing to tolerate
mild/moderate events to derive therapeutic
benefit - Patients may have fewer future treatment options
- Primary endpoint and definition of VF may be
different - CD4 cell count may be more predictive of
clinical benefit
40Endpoint on Salvage Therapy for Experienced
Patients
- Endpoints of proportion of HIV-1 RNA lt50 (400)
copies/mL or virologic failure may not be the
best in populations where a moderate suppression
in plasma HIV-1 RNA may confer clinical benefit - Change from baseline may be more appropriate
- Appropriate data handling and average methods
(e.g., LOCF, AAUCMB etc.) are required in these
settings to minimize bias
41Endpoint on Early Virologic Non-Response
- Time to Virologic failure and TLOVR endpoints not
valid - Not enough time to allow patients to have viral
decline below assay limit (lt400, 50 copies/mL) - Patients with different follow-up time
- Unexpected early clinical concern over randomized
treatment regimen (e.g. ESS30009)
42ESS30009 Study Design Open-label, randomized,
multicenter trial
- Entry criteria
- ? 18 years old
- ART-naïve (?14 days ART)
- VL ? 5,000 copies/mL
- No CD4 restrictions
48 weeks
EFV 600mg QD ABC/3TC FDC QD n180 planned
Screening
Randomize
TDF 300mg QD ABC/3TC FDC QD n180 planned
- Primary endpoint
- Proportion with HIV-1 RNA lt50 c/mL at week 48
Randomization (11) stratified by HIV-1 RNA
lt100K or ?100K c/mL
43Unplanned Interim Analysis
- Following rapid study accrual, several cases of
early virologic non-response in the TDF ABC/3TC
arm were reported - Urgent unplanned interim analysis performed for
subjects with ?8 weeks HIV-1 RNA data (N194 of
345 randomized patients) - Virologic non-response defined as
- 1) lt2.0 log decline in HIV-1 RNA by week 8 or
- 2) ?1.0 log rebound from nadir at any subsequent
visit or - 3) Confirmed HIV-1 RNA gt400 c/mL following two
consecutive HIV-1 RNA lt50 c/mL
44Baseline Characteristics For subjects with at
least 8 weeks HIV-1 RNA data
- EFV ABC/3TC TDF ABC/3TC (N92) (N102)
- Median age, years 36.0 38.5
- Male Sex 90 92
- Race Black / Hispanic / White 37 / 9 / 53
29 / 9 / 62CDC Class C 9 lt1 - Median HIV-1 RNA, log10 c/mL 4.71 4.53
Median CD4 count, cells/mm3 280.5 251.5
45EFV ABC/3TC HIV-1 RNA Response For subjects
with at least 8 weeks HIV-1 RNA data (N92)
HIV-1 RNA c/mL
46TDF ABC/3TC HIV-1 RNA Response For subjects
with at least 8 weeks HIV-1 RNA data (N102)
HIV-1 RNA c/mL
47Virologic Non-Response For subjects with at
least 8 weeks HIV-1 RNA data
- Criteria EFV ABC/3TC TDF ABC/3TC
(N92) (N102) - 1 lt2.0 log decline from 3/92 (3.3) 32/102
(31.4) baseline by week 8 - 2 ?1.0 log rebound 0/92 (0) 8/102
(7.8) from nadir - Both criteria 1 and 2 met 2/92
(2.2) 10/102 (9.8) - Total 5/92 (5.4) 50/102 (49)
No subjects met virologic non-response criteria
3Difference between arms plt0.001, 95 CI
-54.3, -32.8
48ESS30009 Discussion
- TDF ABC/3TC resulted in an unexpectedly high
rate of early virologic non-response - A large proportion of patients with non-response
to TDF ABC/3TC had K65R and M184V - At early follow-up, EFV ABC/3TC resulted in a
rapid decline in viral load, consistent with
previous trials
49Structured Treatment Interruption (STI)
- STI strategies (e.g. one week on, one week off 2
months on, 1 months off CD4 count guided (lt350
on, gt350 off), etc. ) - Long term studies to evaluate
- Endpoint (taking into resistance cost, drug cost,
etc.)
50Choice of Endpoints
- Which endpoint to choose as the primary endpoint?
- all endpoints are not created equal
- vary with patient population (naïve or
experienced) and study objectives - driven by the medical/scientific questions of
interest - power and accommodation of missing data
- which is best surrogate for a long-term outcome?
51Choice of Endpoints
- How to incorporate dropout/missing data into
analysis - treat dropouts as censored observations
- treat dropouts as failures
- model dropout process
- Which to choose as primary analysis method for
primary endpoint? - Check sensitivity of analysis results
52Summary
- Study endpoints using surrogate markers
- Time-to-event techniques to analyze censored
measurement data - Statistical methods on time to failure data with
fixed allowable time for response - Issues on choosing primary endpoint for HIV
clinical trials (pure time to virologic failure
or composite endpoint) - Endpoint on detecting early virologic
non-response - Endpoint on strategic trials