Randomized clinical trials with biomarkers: design issues

Abstract

Clinical biomarker tests that aid in making treatment decisions will play an important role in achieving personalized medicine for cancer patients. Definitive evaluation of the clinical utility of these biomarkers requires conducting large randomized clinical trials (RCTs). Efficient RCT design is therefore crucial for timely introduction of these medical advances into clinical practice, and a variety of designs have been proposed for this purpose. To guide design and interpretation of RCTs evaluating biomarkers, we present an in-depth comparison of advantages and disadvantages of the commonly used designs. Key aspects of the discussion include efficiency comparisons and special interim monitoring issues that arise because of the complexity of these RCTs. Important ongoing and completed trials are used as examples. We conclude that, in most settings, randomized biomarker-stratified designs (ie, designs that use the biomarker to guide analysis but not treatment assignment) should be used to obtain a rigorous assessment of biomarker clinical utility.

Figure 1

Biomarker designs. A) Biomarker-stratified design. All patients are randomly assigned regardless of biomarker status with the random assignment and analysis plan stratified by the biomarker status. Sometimes, a standard (nonstratified) randomization can be used (with the analysis plan stratified by the biomarker) when postrandomization biomarker evaluation is feasible. B) Enrichment design. The biomarker is evaluated on all patients, but random assignment is restricted to patients with specific biomarker values. C) Biomarker-strategy design. Patients are randomly assigned to an experimental treatment arm that uses the biomarker to direct therapy or to a control arm that does not. Some biomarker-strategy designs evaluate biomarkers only in patients randomly assigned to the biomarker-directed arm.

Figure 2

Examples of biomarker-stratified designs. A) The Marker Validation for Erlotinib in Lung Cancer (MARVEL) trial (10). Second-line advanced non–small cell lung cancer (NSCLC) patients were randomly assigned to erlotinib or pemetrexed with random assignment stratified by epidermal growth factor receptor gene (EGFR) status as measured by fluorescent in situ hybridization (FISH). B) The Cancer and Leukemia Group B (CALGB)-30506 trial (http://www.cancer.gov/clinicaltrials/CALGB-30506). Stage I NSCLC patients are randomly assigned to either chemotherapy or observation with random assignment stratified by risk group (high vs low) as defined by the Lung Metagene Score (11). Chemotherapy-arm patients receive physician choice of one of three prespecified chemotherapy regimens.

Figure 3

Examples of biomarker-strategy designs. A) The excision repair cross-complementing 1 (ERCC1) trial (14). Non–small cell lung cancer (NSCLC) patients were randomly assigned to the control arm that received cisplatin+docetaxel or the biomarker-strategy arm that switched patients classified as cisplatin-resistant to the gemcitabine+docetaxel regimen while treating those who were sensitive with cisplatin+docetaxel. B) The Tumor Chemosensitivity Assay (TCA) ovarian cancer study (15). Patients were randomly assigned to the biomarker-strategy arm that used a chemosensitivity assay that measured ATP levels in drug-treated cancer cells to choose from a panel of 12 chemotherapy regimens or to the control arm that received the physician's choice of chemotherapy.

Figure 4

Examples of combination designs. A) The Spanish Lung Cancer Group (SLCG) 0601 trial (17). This trial uses a combination of enrichment and biomarker-strategy designs. First, enrichment is used to restrict the stage IV non–small cell lung cancer (NSCLC) population to patients who have mutated epidermal growth factor receptor (EGFR) genes (exon 19 or 21) in their tumors. Eligible patients are then randomly assigned to a control arm (erlotinib) or to a biomarker-strategy arm in which patients are assigned either to gemcitabine+cisplatin or to docetaxel+cisplatin depending on 14-3-3σ gene methylation status. B) The Tarceva Italian Lung Optimization (TAILOR) study (http://www.cipomo.it/membri/documenti/protocolli/TAILORsinossi.doc). NSCLC patients are assessed at the time of registration for 1) exon 19 or 21 EGFR mutations, 2) EGFR gene copy number by fluorescent in situ hybridization (FISH), 3) EGFR protein expression by immunohistochemistry (IHC), and 4) KRAS mutation. Patients with EGFR exon 19 or 21 mutations are treated with erlotinib; patients without the mutation are randomly assigned to the erlotinib or docetaxel arms, with random assignment stratified by EGFR gene copy number, EGFR protein expression, and KRAS mutation status.