Download time to progression ratio

The time to progression ratio for
phase II trials of personalized medicine
Marc Buyse, ScD
IDDI, Louvain-la-Neuve, and
I-BioStat, Hasselt University, Belgium
[email protected]
Outline
•
Definition of TTPR
•
TTPR in gastro-intestinal stromal tumors
•
TTPR to design a trial in advanced colorectal cancer
•
TTPR for trials of personalized medicine
•
Tentative conclusions
The TTP ratio (TTPR)
Start of Rx
for advanced
disease
TTP1
TTP2
TTP3
Rx1
Rx2
Rx3
First
progression
Second
progression
TTPR = TTP2 / TTP1
…
Death
Use of TTPR
Cytostatics are not expected to induce tumor shrinkage, but it is
hoped that they can stabilize the tumor (delay progression).
For second-line therapies, the « time to progression ratio » (or
« growth modulation index »), is defined as
TTPR = TTP2 / TTP1
Given the natural history of most tumors, TTPR generally does not
exceed 1 (i.e. TTP2 tends to be shorter than TTP1).
Von Hoff suggested that TTPR > 1.33 reflects activity of the
second-line therapy.
Ref: Mick et al. Controlled Clinical Trials 21:343-59, 2000.
TTPR-based designs : second-line treatment only
Start of Rx
for advanced
disease
TTP1
TTP2
Rx1
Rx2
First
progression,
entry on trial
Second
progression
TTPR-based designs : two lines of treatment
Start of Rx
for advanced
disease,
entry on trial
TTP1
TTP2
Rx1
Rx2
First
progression
Second
progression
TTPR when doubling dose of imatinib for
progressing gastro-intestinal stromal tumors
Proportion of patients with TTPR > 1.33 after cross-over
from 400 mg to 800 mg of imatinib daily, broken down
by response to 400mg
Partial response
2/3 (67%)
Stable disease
13/36 (36%)
Progressive disease
12/72 (17%)
All patients
Ref: Zalcberg et al, Eur J Cancer 41:1751-7, 2005.
27/110 (25%)
Trial comparing two imatinib doses in patients with
gastro-intestinal stromal tumors
R
181 *
400 mg imatinib daily
196 *
800 mg imatinib daily
* Nr of patients with adequate DNA for KIT genotype analysis
Ref: Debiec-Rychter et al, Eur J Cancer 42:1093-1103, 2006.
Cumulative incidence of response by KIT mutation
Ref: Debiec-Rychter et al, Eur J Cancer 42:1093-1103, 2006.
TTPR after cross-over by KIT mutation
Proportion of patients with TTPR > 1.25 after cross-over
from 400 mg to 800 mg
Exon 11 mutation (N=248)
83%
Exon 9 mutation (N=58)
57%
Wild type (N=52)
P= 0.0017
7%
Ref: Debiec-Rychter et al, Eur J Cancer 42:1093-1103, 2006.
P= 0.0012
Trial comparing two chemotherapy sequences in
patients with advanced colorectal tumors
R
109
FOLFIRI – FOLFOX
111
FOLFOX – FOLFIRI
Primary endpoint = Time to second progression (TTP1 + TTP2)
Secondary endpoints = TTP1, TTP2, survival
Ref: Tournigand et al, J Clin Oncol 22:229-37, 2004.
1.0
Distribution of TTPR in advanced colorectal cancer
Patients Events
137
0.6
0.4
0.2
0.0
Probability
0.8
147
0
0.4
0.8
2
3
TTP Ratio
4
5
6
1.0
Distribution of TTPR in advanced colorectal cancer
Patients Events
137
0.6
0.2
0.4
  of the patients had a TTPR > 0.5
0.0
50%
Probability
0.8
147
0
0.4
0.8
0.5
2
3
TTP Ratio
4
5
6
1.0
Distribution of TTPR in advanced colorectal cancer
Patients Events
137
0.6
  of the patients had a TTPR > 0.8
0.4
Probability
0.8
147
0.0
0.2
33%
0
0.4
0.8
0.8
2
3
TTP Ratio
4
5
6
1.0
Distribution of TTPR in advanced colorectal cancer
Patients Events
137
0.6
  of the patients had a TTPR > 1
0.4
Probability
0.8
147
0.0
0.2
25%
0
0.4
0.8
1
2
3
TTP Ratio
4
5
6
1.0
Distribution of TTPR in advanced colorectal cancer
Patients Events
137
0.6
0.4
0.2
20%
  of the patients had a TTPR > 1.25
0.0
Probability
0.8
147
0
0.4
0.8
1.25
2
3
TTP Ratio
4
5
6
TTPR – test of hypothesis
A possible null hypothesis is:
H0: TTPR = TTP2 / TTP1  HR0
versus the alternative hypothesis:
HA: TTPR = TTP2 / TTP1 > HR0
1.0
Test of hypothesis in advanced colorectal cancer
Patients Events
137
0.2
0.4
0.6
H0 : TTPR ≤ 0.75
0.0
50%
Probability
0.8
147
0
0.4
0.8
0.75
2
3
TTP Ratio
4
5
6
A sign test statistic
For the ith patient, let ri be equal to
+1
if
TTP2 > TTP1  HR0
–1
if
TTP2  TTP1  HR0 and TTP2 is uncensored
The test statistic (equivalent to a sign test statistic)
X² = (i ri)² / i ri²
has a ² distribution with 1 d.f.
Ref: Mick et al. Controlled Clinical Trials 21:343-59, 2000.
A sign test statistic
HR0 = 0.7
 = 0.05
Correlation = 0.7
90%
85%
80%
Ref: Mick et al. Controlled Clinical Trials 21:343-59, 2000.
A sign test statistic
HR0 = 0.7
 = 0.05
Correlation = 0.5
Ref: Mick et al. Controlled Clinical Trials 21:343-59, 2000.
A sign test statistic
HR0 = 0.7
 = 0.05
Correlation = 0.3
Ref: Mick et al. Controlled Clinical Trials 21:343-59, 2000.
TTP1 vs. TTP2 in advanced colorectal cancer
R² = 0.03
TTP1 vs. TTP2 in advanced colorectal cancer
Statistics for correlated survival times
In the absence of censoring, TTP1 and TTP2 can be compared
using a paired t-test or a non-parametric test for paired
observations.
If TTP2 is censored, TTP1 and TTP2 are paired survival times. The
ordinary rank test statistics (e.g. logrank or Gehan-Wilcoxon)
can be used with variance corrected to account for the
correlation between TTP1 and TTP2.
Ref: Jung, Lifetime Data Analysis 5:67-79, 1999.
TTPR – another test of hypothesis
Let p be the proportion of patients for whom TTPR > HR0.
A possible null hypothesis is:
H 0 : p  p0
versus the alternative hypothesis:
HA: p > p0
which leads to Flemings’ one-stage or Simon’ two-stage designs.
1.0
Tests of hypothesis in advanced colorectal cancer
Patients Events
137
0.6
0.4
0.2
22%
H0 : p0 ≤ 22%
0.0
Probability
0.8
147
0
0.4
0.8
1.33
2
3
TTP Ratio
4
5
6
Trial of molecular profiling
TTP1
TTP2
At least two prior
therapies for advanced
disease, no further
therapy available
Molecular profiling of
tumor biopsy by IHC,
FISH or micro-array
to identify target
Last
progression,
entry on trial
Progression
on targeted
therapy
Ref: Von Hoff, AACR 100th Annual Meeting, Denver, CO, April 18-22, 2009.
Trial of molecular profiling
Trial designed to test p0 (proportion of patients with TTPR > 1.3):
H0: p  p0 = 15%
Primary analysis: proportion of patients with TTPR > 1.3:
18 / 66 (27%, 95% C.I. 17% - 38%, P = 0.007)
Breast
Colorectal
Ovarian
Others
8 / 18 (44%)
4 / 11 (36%)
1 / 5 (20%)
5 / 32 (16%)
Ref: Von Hoff, AACR 100th Annual Meeting, Denver, CO, April 18-22, 2009.
Trial of molecular profiling
Promising results, and amongst the 18 patients with TTPR > 1.3, none
would have received same drug through plysician’s choice.
However,
• Is TTPR > 1.3 good enough?
• Trial was not randomized, therefore no evidence that physician’s
choice could have yielded similar results
• Only 66 patients of 106 could have molecular profiling
Trial designs using TTPR – pros and cons
+ Test time to progression rather than response; hence well suited
to test cytostatic agents
+ Patients serve as their own control, a desirable feature to control
between-patient variability
+ Efficient if substantial correlation between TTP1 and TTP2
-
Choice of appropriate value for HR0
-
TTP1 difficult to estimate retrospectively, and potentially biased
downwards if standard first-line treatment included in design and
new agent is promising
-
Inefficient if poor correlation between TTP1 and TTP2