Outline
• Biomarker definition • Classical examples of biomarkers • Purposes of biomarkers (more examples) • Clinical trial programs What is a biomarker and what can it be used for? • (Regulatory aspects) DSBS meeting February 21, 2011 Philip Hougaard, Biometrics, Lundbeck • Gene expression biomarkers
H. Lundbeck A/S 10-Mar-11 1 H. Lundbeck A/S 10-Mar-11 2
Biomarker definition Properties of a biomarker?
• A characteristic that is objectively • Constant (vs. Time-dependent)?: measured and evaluated as an indicator Yes (like genes), but constant of healthy biological processes, biomarkers cannot be outcomes pathological processes, or pharmacological responses to • Binary? Yes, but outcome biomarkers therapeutic intervention are more useful if they are continuous, • Note: ”an indicator of” means I think ”reflecting” •Multivariate? Yes (but often this is used • “evaluated as” means “with the purpose to derive a univariate summary) of”, not “proven to be”
H. Lundbeck A/S 10-Mar-11 3 H. Lundbeck A/S 10-Mar-11 4
Examples that are not Surrogate endpoint definition biomarkers
• Rating scales (MADRS) • A biomarker that is intended to • PRO (patient reported outcomes) substitute for a clinical endpoint. A • Tolerability surrogate endpoint is expected to predict clinical benefit (or harm or lack •Pain of benefit or harm) based on • Imaging interpretation (picture alone is epidemiological, therapeutic, a biomarker) pathophysiological, or other scientific evidence
H. Lundbeck A/S 10-Mar-11 5 H. Lundbeck A/S 10-Mar-11 6
1 Weight Related variables
• Most classical example • BMI (weight/height2) • Doctors office: Light clothes, no shoes • LBM (lean body mass) formula • Home: very light clothes, morning •LBM concept • Daily variation in the kg-range • BSA (body surface area) formula • Does weight reflect health? Age-dependence; height adjustment; composition: fat versus • Waist circumference muscles: apple-shape (abdominal fat) •Waist hip ratio • Does weight-change reflect health? • Is weight-reducing treatments helpful? Water- reducing; metabolism-increasing
H. Lundbeck A/S 10-Mar-11 7 H. Lundbeck A/S 10-Mar-11 8
FDA (weight) Temperature
• Primary endpoint: %weight loss (mean • Circadian variation difference to placebo stat. sign. and • Core temperature versus surface ≥5%) + response criteria temperature •Inclusion: BMI≥30 or BMI≥27+comorbidity • Measurements: rectal, mouth, ear • Reasonable sample size: 4500 for 1y • Monthly variation (females) • DEXA scanning on some • Fever (increased temperature) present • Safety important: Recently a drug was with many infections taken off the market due to increased cardiovascular risk
H. Lundbeck A/S 10-Mar-11 9 H. Lundbeck A/S 10-Mar-11 10
Blood pressure Cardiovascular markers
• Systolic and diastolic • Cholesterol (high-density versus low- • Supine or sitting density) • White coat hypertension • Free fatty acids • High measurement error • Triglycerides • Office versus 24h measurements • EMA guideline (2010): 3 baseline visits; systolic primary; diastolic secondary; sufficient number of patients for mortality (1y); no suspicion of adverse target organ effect
H. Lundbeck A/S 10-Mar-11 11 H. Lundbeck A/S 10-Mar-11 12
2 Diabetes biomarker primer Diabetes biomarkers
Biomarker Type 1 Type 2 • Blood glucose sensitive to meals. Urine glucose (UT) ↑ ↑ Strategies: Blood glucose (UT) ↑ ↑ • 1. Unrestricted (high upper normal Hba1c (target) ↑ ↑ limit) Insulin (UT) ↓ ↔ • 2. Fasting (overnight) C-peptide ↓ ↔ • 3. Fasting supplemented with specific Antibodies (sel.) ↑ Normal challenge (glucose tolerance test) Insulin resistence Normal ↑ • 4. HbA1c (reflecting average blood Age Flat incidence Old glucose over 6 weeks) BMI Normal ↑ pH/HCO3 (UT) ↓ (severe) H. Lundbeck A/S 10-Mar-11 13 H. Lundbeck A/S 10-Mar-11 14
Conclusion on classical Types of biomarkers biomarkers
• Many classical biomarkers are widely •Diagnostic used and accepted, both to define • Early detection diseases and for treatment decisions • Monitoring • However, it is also well known that they •Prognostic are far from perfect •Predictive/Safety/dose
H. Lundbeck A/S 10-Mar-11 15 H. Lundbeck A/S 10-Mar-11 16
Examples-diagnostic Examples – early detection
• Disease vs. healthy or • Gene test for monogenic diseases (like disease vs. other diseases? cystic fibrosis) • PSA (still controversial) •Temperature • Breast cancer screening • Bacterial tests (streptococcal) • Origin of cancer tissue
H. Lundbeck A/S 10-Mar-11 17 H. Lundbeck A/S 10-Mar-11 18
3 Examples – Monitoring Prognostic
• HbA1c (diabetes) • Mammaprint: 70 gene expression to • Immune markers (after transplantation) split breast cancer patients into high vs. • Temperature (during infection) low risk of metastases • Titre (after vaccination) • Tumour size (cancer) • Blood oxygen (heart and lung diseases)
•QTc (specific safety study)
H. Lundbeck A/S 10-Mar-11 19 H. Lundbeck A/S 10-Mar-11 20
Predictive (which treatment; Biological understanding which dose)
• “Companion diagnostic”, “personalized •Useful medicine”: Measurable before/without • But regulatory pathway does not treatment require this • Recommended to follow up on this in •Body weight parallel to other activities • P450 enzymes (normal or poor metabolizers) • Herceptin (Her2 positive breast cancer)
H. Lundbeck A/S 10-Mar-11 21 H. Lundbeck A/S 10-Mar-11 22
Practical issues Validation and qualification
• Timing: A diagnostic needs fast turn- • Validation: around time • The measurement is correct in the • Other: Inconvenience to patient; risk to technical sense: Including sources of patient; circadian variation (incl. random error meals); sex, age, weigth dependence; storage; stability; transportation; • Qualification (clinical validation): central lab. • The measurement is related to the clinical state of the patient
H. Lundbeck A/S 10-Mar-11 23 H. Lundbeck A/S 10-Mar-11 24
4 Clinical trials to qualify a Clinical trials to qualify a biomarker for monitoring biomarker as diagnostic
• Example: For HbA1c, DCCT (1993) was the • Trial in relevant population (meaning turning point similar to the one where it will be • 1441 patients; stratified by retinopathy; applied) followed for up to 9 years; price-tag 1 bill. $ • If multivariate diagnostic, formula • Intensified insulin treatment lead to both reduced HbA1c and reduced risk of derivation should be pre-specified complications (vs. conventional therapy) • Evaluate sensitivity/specificity • Relation between post-treatment HbA1c and compared to gold standard. Evaluate complications in intensive group AUC for ROC curve
H. Lundbeck A/S 10-Mar-11 25 H. Lundbeck A/S 10-Mar-11 26
Evaluation as diagnostic AUC - ROC (in targeted population!)
Subject count Diseased Healthy • Vary the threshold between Biomarker Biomarker+ A B + and – subgroups Biomarker- C D • Combined values of sensitivity and specificity gives ROC curve Diseased/healthy according to reference standard • Evaluate performance by the AUC Sensitivity: A/(A+C); Specificity: D/(B+D) (more robust evaluation than selecting PPV: A/(A+B); NPV: D/(C+D) a single threshold) Odds ratio: A D / (B C) Agreement: (A+D)/(A+B+C+D) not relevant!
H. Lundbeck A/S 10-Mar-11 27 H. Lundbeck A/S 10-Mar-11 28
Clinical trials to qualify a Potential trial designs I biomarker as predictive (FDA guideline)
• If multivariate biomarker, formula derivation should be pre-specified • Positive subgroup effect checked according to standard practice (two adequate and well-controlled studies) • Confirmed in later independent studies • Negative subgroup effect checked, but potentially not as rigorous (interaction expected) depending on biological understanding
H. Lundbeck A/S 10-Mar-11 29 H. Lundbeck A/S 10-Mar-11 30
5 Potential trial designs II Genes or proteins? (FDA guideline)
• Genes (DNA): Constant, discrete, massively parallel • Gene expresssions (mRNA): time- varying, continuous, microarray (massively parallel), qPCR (parallel) •Proteins: time-varying, continuous, difficult sample handling (freezing), separate measurements
H. Lundbeck A/S 10-Mar-11 31 H. Lundbeck A/S 10-Mar-11 32
Gene expression biomarkers Regulatory issues
• Measure RNA for treatment choice • FDA: a biomarker is a device (handled by CDRH, center for devices and radiological • Multivariate so a selection step is health) necessary • Derivation of a gene-based biomarker • Define subgroup to be treated: (selection based on many genes) require pre- • Show effect in subgroup specified gene selection method • Show/argue for no effect in • Endpoint for diagnostic biomarkers is AUC- ROC complementary subgroup • Only few have passed the regularity hurdle
H. Lundbeck A/S 10-Mar-11 33 H. Lundbeck A/S 10-Mar-11 34
Lundbeck biomarker examples Summary
• Gene expressions in depression • Classical well-known biomarkers are far • Imaging (such as PET) from perfect • Many new biomarkers are suggested • EEG • Diagnostic; Prognostic; Predictive; • Various measurements known to reflect Surrogate endpoints occupancy on specific receptors • Only few have passed the regularity hurdle • Biological understanding is key • Biomarkers are the future
H. Lundbeck A/S 10-Mar-11 35 H. Lundbeck A/S 10-Mar-11 36
6