Steps to stratified medicine: r n toxicity e s r an example - - PDF document

Steps to stratified medicine: an example from lung cancer

Kinga Malottki k.malottki@bham.ac.uk

Using a treatment

Xinghua Hu S, Foster T, Kieffaber A. Pharmacogenomics and personalized medicine: mapping future value creation. BioTechniques 2005; 39 (4)

r e s p

• n

d e r s n

• n

‐ r e s p

• n

d e r s toxicity

Using a treatment – predictive biomarkers

r e s p

• n

d e r s n

• n

‐ r e s p

• n

d e r s

Biomarker 1 Biomarker 2 BM 3

toxicity

Erlotinib for Non‐Small‐Cell Lung Cancer (NSCLC)

• Population: patients with NSCLC after or unsuitable

for chemotherapy

• Intervention: erlotinib at any dose
• Comparator: any, including no
• Outcomes: overall survival, response
• Design: any, except case reports
• Biomarkers: EGFR expression, EGFR copy number,

EGFR mutation, KRAS mutation

Erlotinib for NSCLC

• Not a systematic review
• Search in MEDLINE up to March 2010
• Internet searches
• Checking references of included studies
• 10 completed studies (1 RCT: BR.21)
• 2 ongoing RCTs (MARVEL, SATURN)

Identified studies

EGFR expression EGFR copy number EGFR mutation KRAS mutation Milton 2006 Rosell 2009 Lee 2010 SATURN uncontrolled controlled study type Perez‐Soler 2004 Hesketh 2008 TRUST BR.21 MARVEL study Miller 2008 Felip 2008 Ahn 2008

EGFR mutations in NSCLC

• Found in:

– ~30‐50% East‐Asian – ~10% Western Europe and North America

• More frequent in:

– Never smokers – Females – Adenocarcinomas – East‐Asian

• Prognosis unclear (some studies seem to show increased

survival in patients with mutations)

• Exons 18‐21, most common deletions in exon 19 and

missense mutation in exon 21 (L858R) – ~85%

Reference list available on request

Using a treatment – predictive biomarkers

r e s p

• n

d e r s n

• n

‐ r e s p

• n

d e r s

EGFR mutation

toxicity

EGFR mutations in included studies

• 8 studies (N: 21‐731) assessed EGFR mutations
• Mutation status available for 19‐100% of patients
• 7 studies used polymerase chain reactions (PCR), 1

study did not report the method

• Mutations assessed:

– 5 studies: exons 19 and 21, – 2 studies: exon 18 , 19 and 21 – 1 study: exon 18 ‐ 21

EGFR mutation – overall survival

group N ERL vs. PL all patients 731 HR 0.7 (0.58, 0.85) wild type 137 HR 0.73 (0.49, 1.10) mutation 40 HR 0.77 (0.40, 1.50) study N HR TRUST 6 mutation 85 wild type 0.33 (0.12, 0.91) Felip 2008 5 mutation 34 wild type 0.902 (p=0.8468) Ahn 2008: mutation assessed in 92 of 120; OS longer in mutation (p=0.023 vs. wild type) – possibly long OS in patients in whom mutation could not be assessed

BR.21 survival HR: erlotinib vs. placebo Survival HR: EGFR mutation vs. wild type Median survival (months): all patients, EGFR mutation and wild type

EGFR mutation – response OR (RECIST)

Study or Subgroup Ahn 2008 BR.21 Felip 2008 Lee 2010 Miller 2008 Milton 2006 TRUST Total (95% CI) Total events Heterogeneity: Tau² = 0.37; Chi² = 8.44, df = 6 (P = 0.21); I² = 29% Test for overall effect: Z = 6.06 (P < 0.00001) Events 14 4 2 2 15 1 2 40 Total 24 15 5 3 18 1 4 70 Events 11 7 1 1 4 2 26 Total 68 101 34 9 63 3 68 346 Weight 29.8% 22.4% 8.7% 6.5% 18.7% 3.6% 10.3% 100.0% M-H, Random, 95% CI 7.25 [2.57, 20.46] 4.88 [1.23, 19.37] 22.00 [1.51, 319.48] 16.00 [0.67, 383.02] 73.75 [14.88, 365.52] 21.00 [0.27, 1646.18] 33.00 [2.96, 368.36] 14.41 [6.08, 34.13] EGFR mutation EGFR wild type Odds Ratio Odds Ratio M-H, Random, 95% CI 0.01 0.1 1 10 100 Favours EGFR wild type Favours EGFR mutation

EGFR mutation – predicting response

Index test: EGFR mutation Reference test: response

Linardou H, Dahabreh IJ, Kanaloupiti D, Siannis F, Bafaloukos D, Kosmidis P, Papadimitriou CA, Murray S. Assessment of somatic k‐RAS mutations as a mechanism associated with resistance to EGFR‐targeted agents: a systematic review and meta‐analysis of studies in advanced non‐small‐cell lung cancer and metastatic colorectal cancer. Lancet Oncol 2008; 9 (10):962‐972

Study Ahn 2008 BR.21 Felip 2008 Lee 2010 Miller 2008 Milton 2006 TRUST TP 14 4 2 2 15 1 2 FP 10 10 3 1 3 2 FN 11 7 1 1 4 2 TN 57 94 33 8 59 3 66 Sensitivity 0.56 [0.35, 0.76] 0.36 [0.11, 0.69] 0.67 [0.09, 0.99] 0.67 [0.09, 0.99] 0.79 [0.54, 0.94] 1.00 [0.03, 1.00] 0.50 [0.07, 0.93] Specificity 0.85 [0.74, 0.93] 0.90 [0.83, 0.95] 0.92 [0.78, 0.98] 0.89 [0.52, 1.00] 0.95 [0.87, 0.99] 1.00 [0.29, 1.00] 0.97 [0.90, 1.00] Sensitivity 0 0.2 0.4 0.6 0.8 1 Specificity 0 0.2 0.4 0.6 0.8 1

Could EGFR mutation be used in clinical practice?

• Test varies across studies
• Possible confounding
• Mostly uncontrolled studies or relatively small

subgroups of patients

• Prognostic rather than predictive? Both?
• Patients in whom assessment could not be carried
• ut – selection bias?
• Alternative treatments?

KRAS mutations in NSCLC

• Found in:
• ~ 12‐30% Caucasian
• ~ 5‐10% East‐Asian
• More common in:

– History of smoking – Adenocarcinomas – Caucasian

• Prognostic value unclear (some studies seem to

show shorter survival in patients with mutations)

• Exons 2‐3, most common in exon 2

Reference list available on request

Using a treatment – predictive biomarkers

r e s p

• n

d e r s n

• n

‐ r e s p

• n

d e r s

KRAS mutation

toxicity

KRAS mutations in included studies

• 5 studies (N: 83‐731) assessed KRAS mutations
• Status known in 29‐79%
• 4 studies used PCR and 1 direct sequencing
• Mutations assessed:

– 2 studies: exon 2 – 3 studies: exon 2 and 3

KRAS mutation – overall survival

median lci uci Felip 2008 all patients 4 3.4 6.4 KRAS wild type 3.7 2.9 6.8 KRAS mutation 4.5 2 7.3 Miller 2008 all patients 17 NR NR KRAS wild type 21 NR NR KRAS mutation 13 NR NR 5 10 15 20 25 group N ERL vs. PL all patients 731 HR 0.7 (0.58, 0.85) wild type 137 HR 0.69 (0.49, 0.97) mutation 40 HR 1.67 (0.62, 4.50)

BR.21 survival HR: erlotinib vs. placebo Survival HR: mutation vs. wild type Median survival (months): all patients, KRAS mutation and wild type

study N HR 11 mutation 78 wild type 7 mutation 32 wild type TRUST Felip 2008 0.81 (NR) 1.64 (0.97, 2.80)

KRAS mutation – response OR (RECIST)

Study or Subgroup BR.21 Miller 2008 TRUST Total (95% CI) Total events Heterogeneity: Tau² = 0.00; Chi² = 1.63, df = 2 (P = 0.44); I² = 0% Test for overall effect: Z = 1.81 (P = 0.07) Events 1 1 Total 20 11 18 49 Events 10 7 20 37 Total 98 78 62 238 Weight 48.3% 25.2% 26.5% 100.0% M-H, Random, 95% CI 0.46 [0.06, 3.84] 0.41 [0.02, 7.76] 0.06 [0.00, 0.98] 0.26 [0.06, 1.12] favours KRAS wild type KRAS wild type Odds Ratio Odds Ratio M-H, Random, 95% CI 0.01 0.1 1 10 100 favours KRAS wild type favours KRAS mutation

KRAS mutation – predicting response

Index test: KRAS mutation Reference test: lack of response

Linardou H, Dahabreh IJ, Kanaloupiti D, Siannis F, Bafaloukos D, Kosmidis P, Papadimitriou CA, Murray S. Assessment of somatic k‐RAS mutations as a mechanism associated with resistance to EGFR‐targeted agents: a systematic review and meta‐analysis of studies in advanced non‐small‐cell lung cancer and metastatic colorectal

• cancer. Lancet Oncol 2008; 9 (10):962‐972

Study BR.21 Miller 2008 TRUST TP 19 11 18 FP 1 FN 88 71 42 TN 10 7 20 Sensitivity 0.18 [0.11, 0.26] 0.13 [0.07, 0.23] 0.30 [0.19, 0.43] Specificity 0.91 [0.59, 1.00] 1.00 [0.59, 1.00] 1.00 [0.83, 1.00] Sensitivity 0.2 0.4 0.6 0.8 1 Specificity 0.2 0.4 0.6 0.8 1

Could KRAS mutations be used in clinical practice?

• Test varies across studies
• Possible confounding
• Small number of studies
• Mostly uncontrolled studies or relatively small

subgroups of patients

• Patients in whom assessment could not be carried
• ut – selection bias?
• Prognostic rather than predictive? Both?
• Patients with KRAS mutation should not be treated?

Conclusions

• Different properties of biomarkers depending on purpose and

context

• Might only identify a subgroup of patients with an increased

probability of (not) benefiting from treatment

• Predictive vs. prognostic biomarkers
• Possible confounding
• Possible bias: patients who cannot be tested
• Need for RCTs designed to evaluate both biomarker and

treatment

• Assessment across similar indications and/or drugs with

similar mechanism of action?