1. Surrogate endpoints in cancer randomized controlled trials:
Application to sarcoma and breast cancer
Marion Savina, INSERM U1219 et Institut Bergonié (Bordeaux)
Yassine Laghzali, Institut du Cancer de Montpellier
Simone Mathoulin-Pélissier, INSERM U1219 et Institut Bergonié (Bordeaux)
Carine Bellera, INSERM U1219 et Institut Bergonié (Bordeaux)
Sophie Gourgou , Institut du Cancer de Montpellier
Journées du club SMAC 2017 –

2. Primary endpoints in phase III Randomized Controlled Trials (RCTs)
Overall Survival (OS)
Gold standard in cancer RCTs
Easily measurable
Objectively defined: time from randomization to death
Validated by health authorities
Require a large number of patients and long follow-up
Journées du club SMAC 2017 –

3. Primary endpoints in phase III Randomized Controlled Trails (RCTs)
Overall Survival (OS)
Gold standard in cancer RCTs
Easily measurable
Objectively defined: time from randomization to death
Validated by health authorities
Require a large number of patients and long follow-up
Alternative endpoints
Progression-free survival (PFS)
Include events other than death: disease progression, toxicity…
Observed earlier and more frequently
↘ number of patients, duration and cost of the trials
Require rigorous validation as surrogate endpoint for OS
Journées du club SMAC 2017 –

4. Evaluation of surrogate endpoints
Final outcome
Treatment effect on the surrogate endpoint
Intervention
Treatment effect on the final outcome
Definition (Biomarker Definition Working Group, 2001)
To replace the final outcome
At the patient level: the surrogate correlates and predicts the final outcome
At the trial level: the treatment effect on the surrogate correlates and predicts the treatment effect on the final outcome
Specific to the therapeutic situation: disease, type of treatment, population,…
Meta-analysis of RCTs using individual-patient data
Journées du club SMAC 2017 –

5. Issues specific to the therapeutic situations
Soft-tissue sarcoma (STS)
Rare and heterogeneous disease
Median survival (metastatic setting) ~24 months
→ Difficulty to recruit a large number of patient
1 meta-analysis (Zer et al. 2016)
PFS as surrogate for OS
Aggregate data
Journées du club SMAC 2017 –

6. Issues specific to the therapeutic situations
Soft-tissue sarcoma (STS)
Rare and heterogeneous disease
Median survival (metastatic setting) ~24 months
→ Difficulty to recruit a large number of patient
1 meta-analysis (Zer et al. 2016)
PFS as surrogate for OS
Aggregate data
Breast cancer
33% of new cancer cases
5-year survival rate ~90%
Longer survival in adjuvant setting
→ Very long follow-up to observe a tt effect on OS
1 meta-analysis (Ng et al., 2008)
DFS as surrogate for OS
Aggregate data
Journées du club SMAC 2017 –

7. Methods – Meta-analyses of individual-patient data
Identification of eligible RCTs
Published or soon-to be published
Evaluating at least 1 time-to-event endpoint other than OS
Contacts (PI / sponsors) and authorizations
Data management / merging
Common definition of endpoints (International DATECAN guidelines)
Common follow-up duration
Statistical analyses
Individual-level association
Trial-level association
Journées du club SMAC 2017 –

8. Methods – Surrogate properties
Individual-level association
Association between the endpoints: unit = patient
Joint modeling of the candidate surrogate and OS
1-parameter copula function
0 (null association) < ρSpearman < 1 (perfect association)
Trial-level association
Association between the treatment effects: unit = trial
Method 1: Weighted linear regression (WLR)
Treatment effects (log(HR)) estimated independently (Cox models)
Linear regression model weighted by trial size: 0 < R²WLR < 1
Method 2: 2-step model (2SM)
Treatment effects (log(HR)) estimated simultaneously (copula function)
Error-in-variable model: 0 < R²2SM < 1
Marion SAVINA – Journées du club SMAC 2017 –

9. Results – Metastatic STS: Data
14 RCTs
N = 2 846
Median follow-up: 9.4 months months (median = 35.5 months)
Treatment
Metastatic setting
Chemotherapy-based treatments
Journées du club SMAC 2017 –

10. Results – Metastatic STS: Data
14 RCTs
2 RCTs with 2 experimental arms (same drug): combined into 1 experimental arm
1 RCT with 2 distinct analyses (1st line and 2nd line): considered as 2 distinct trials
15 trial units
Journées du club SMAC 2017 –

11. Results – Metastatic STS: Endpoints
Endpoints (International DATECAN guidelines):
Progression-free survival (PFS): time to progression or death
Time-to progression (TTP): time to progression or death from cancer
Time-to treatment failure (TTF): time to progression or death from cancer or death from treatment toxicity
Censoring
2-year OS
6-month and 1-year surrogate endpoints
Journées du club SMAC 2017 –

12. Results – Metastatic STS: 6-months PFS vs 2-year OS
Journées du club SMAC 2017 –

13. Results – Metastatic STS: 6-months PFS vs 2-year OS
Individual-level association: ρSpearman= 0.60 [0.56; 0.63]
Journées du club SMAC 2017 –

14. Results – Metastatic STS: 6-months PFS vs 2-year OS
Individual-level association: ρSpearman= 0.60 [0.56; 0.63]
Trial-level association:
Method 1: Weighted linear regression
Method 2: 2-step model
Log(HR[6-months PFS])
log(HR[2-year OS])
R²WLR = 0.27 [0.00; 0.55]
R²2SM = 0.08 [0.00; 0.98]
Journées du club SMAC 2017 –

15. Results – Metastatic STS
Endpoint
Follow-up
Individual-level association
ρSpearman [95%CI]
Trial-level association
R²WLR [95%CI]
R²2SM [95%CI]
PFS
6 months
0.60 [0.56; 0.63]
0.27 [0.00; 0.55]
0.08 [0.00; 0.98]
12 months
0.62 [0.59; 0.65]
0.33 [0.00; 0.59]
0.00 [0.00; 0.16]
TTP
0.57 [0.54; 0.60]
0.25 [0.00; 0.54]
0.17 [0.00; 1.00]
0.60 [0.57; 0.63]
0.30 [0.00; 0.57]
0.01 [0.00; 0.41]
TTF
0.58 [0.54; 0.61]
0.26 [0.00; 0.54]
0.16 [0.00; 1.00]
0.61 [0.58; 0.63]
0.31 [0.00; 0.58]
0.01 [0.00; 0.39]
Journées du club SMAC 2017 –

16. Results – Adjuvant breast cancer: Data
5 RCTs
N =
Median follow-up: 53 months - 96 months (median = 74 months)
Treatment
Adjuvant setting
Chemotherapy alone or in combination with hormonotherapy
Journées du club SMAC 2017 –

17. Results – Adjuvant breast cancer: Data
5 RCTs → subdivided into groups of centers (trial unit)
≥ 200 patients / treatment arm for each trial unit
≥ 2 deaths / treatment arm for each trial unit
21 trial units
Journées du club SMAC 2017 –

18. Results – Adjuvant breast cancer: Endpoints
Endpoints (International DATECAN guidelines):
Relapse-free survival (RFS)
Invasive disease-free survival (iDFS)
Locoregional relapse-free survival (LRFS)
Distant disease-free survival (DDFS)
Death from any cause
 Local events
 Regional events
 Metastatic events
Contralateral breast cancer
Second cancer
Ipsilateral DCIS
Contralateral DCIS
RFS X  X
iDFS
LRFS
DDFS
Journées du club SMAC 2017 –

19. Results – Adjuvant breast cancer: Endpoints
Endpoints (International DATECAN guidelines):
Relapse-free survival (RFS)
Invasive disease-free survival (iDFS)
Locoregional relapse-free survival (LRFS)
Distant disease-free survival (DDFS)
Censoring
7-year OS
2-year and 3-year surrogate endpoints
Journées du club SMAC 2017 –

20. Results – Adjuvant breast cancer: 2-year RFS vs 7-year OS
Journées du club SMAC 2017 –

21. Results – Adjuvant breast cancer: 2-year RFS vs 7-year OS
Individual-level association: ρSpearman= 0.98 [0.98; 0.99]
Journées du club SMAC 2017 –

22. Results – Adjuvant breast cancer: 2-year RFS vs 7-year OS
Individual-level association: ρSpearman= 0.98 [0.98; 0.99]
Trial-level association:
Method 1: Weighted linear regression
Method 2: 2-step model
Log(HR[2-year RFS])
log(HR[7-year OS])
R²WLR = 0.71 [0.41; 0.82]
R²2SM = 0.72 [0.24; 1.00]
Journées du club SMAC 2017 –

23. Results – Advanced breast cancer
Endpoint
Follow-up
Individual-level association
ρSpearman [95%CI]
Trial-level association
R²WLR [95%CI]
R²2SM [95%CI]
RFS
2 years
0.98 [0.98; 0.99]
0.75 [0.48; 0.84]
0.72 [0.24; 1.00]
3 years
0.99 [0.98; 0.99]
0.71 [0.41; 0.82]
0.85 [0.33; 1.00]
iDFS*
0.98 [0.98; 0.98]
0.78 [0.48; 0.86]
0.87 [0.46; 1.00]
0.72 [0.37; 0.83]
0.72 [0.21; 1.00]
LRFS
0.73 [0.45; 0.83]
0.55** [0.27; 0.84]
0.76 [0.51; 0.85]
0.74** [0.55; 0.93]
DDFS
0.99 [0.99; 0.99]
0.66 [0.34; 0.79]
0.76 [0.37; 1.00]
0.70 [0.40; 0.81]
0.82 [0.51; 1.00]
* Estimated on 4 trials only (N = 9922)
** R² unadjusted on measurement errors
Journées du club SMAC 2017 –

24. Conclusion Metastatic STS Breast cancer
Individual-level association: similar and moderate for the 3 endpoints
Trial-level association: low with very wide confidence intervals that include 0
→ OS should remain the primary endpoint for RCTs in metastatic STS
Breast cancer
Individual-level association: very high for the 4 endpoints
Trial-level association: high but wide confidence intervals
→ Promising first results
→ Would require further evaluation with a larger dataset
Journées du club SMAC 2017 –

25. Discussion Statistical methodology: 2-step model
Most recognized and statistically rigorous approach
Require large amount of data to provide precise estimations
Need for recommendations:
Statistical methodology
Surrogacy measures
Validated grid to interpret the results
Journées du club SMAC 2017 –

26. Marion SAVINA:

27. Back-up slides

28. Metastatic STS: Data Study Inclusion period N Treatment line
Control arm
Experimental arm
Median follow-up
62091
≥ 2011
133
1st line
Doxorubicin
Trabectedin
9.4 months
62072
369
2nd to 5th line
Placebo
Pazopanib
14.6 months
Taxogem* 70
2nd line
Gemcitabine
Gemcitabine + Docetaxel + Lenograstime
32.5 months
62061
118
Brostallicin
21.3 months
62012
455
Intensified Doxorubicin + Ifosfamide
56.4 months
GEIS9
132
22.5 months
Palsar 2 87
MAID**
MAID** + MICE**
55.7 months
62971
326
Ifosfamide
51.7 months
62962
≥ 1997 95
Doxorubicin pegylated liposomal
35.2 months
62941
≥ 1995 86
1st and 2nd line
Docetaxel
35.9 months
Palsar 1
145
Intensified MAID**
93.0 months
62912 78
2nd-line
30.6 months
103
35.5 months
62903
315
Doxorubicin + Ifosfamide
Doxorubicin + Ifosfamide + GM-GSF**
91.4 months
62901
≥ 1991
334
Epirubicin
50.2 months
* Only patients with leiomyosarcoma included
** MAID: Doxorubicin, Ifosfamide and Dacarbazine; MICE: Mesna, Ifosfamide, Carboplatin and Etoposide; GM-GSF: Recombinant human granulocyte-macrophage colony-stimulating factor.

29. Adjuvant breast cancer: Data
Study
Inclusion period N
Control arm
Experimental arm
Median follow-up
EORTC-10901
1724
6 cycles of adjuvant combination chemotherapy + …
78.59 months
Patient monitoring
Tamoxifen
PACS-01
1999
FEC*
FEC* + Docetaxel
63.47 months
HERA
1999 -
5081
Herceptin (1 year)
Herceptin (2 year)
95.80 months
PACS-04
3010
Epirubicin + Docetaxel
53.36 months
PACS-05
1515
FEC* (6 cycles)
FEC* (4 cycles)
73.69 months
* FEC: Fluorouracile + Epirubicine + Cyclophosphamide

30. Results – Advanced breast cancer
Endpoint
Follow-up
Individual-level association
ρSpearman [95%CI]
Trial-level association
R²WLR [95%CI]
R²2SM [95%CI]
RFS
2 years
0.98 [0.98; 0.99]
0.69 [0.00; 0.85]
0.63 [0.02; 1.00]
3 years
0.99 [0.98; 0.99]
0.63 [0.00; 0.82]
0.63 [0.13; 1.00]
iDFS*
0.98 [0.98; 0.98]
0.74 [0.00; 0.87]
0.86 [0.61; 1.00]**
0.65 [0.00; 0.83]
0.66 [0.00; 1.00]
LRFS
0.64 [0.00; 0.83]
0.64 [0.13; 1.00]**
0.91 [0.14; 0.95]
0.96 [0.88; 1.00]**
DDFS
0.99 [0.99; 0.99]
0.54 [0.00; 0.78]
0.45 [0.00; 1.00]
0.61 [0.00; 0.81]
0.54 [0.00; 1.00]
* Estimated on 4 trials only (N = 9922)
** R² unadjusted on measurement errors
Journées du club SMAC 2017 –