Guideline on Enrichment Strategy and Design in

Clinical Trials

 

1.       Overview

The purpose of clinical trials is to demonstrate the efficacy and safety of an investigational drug in human subjects. However, due to the complexity of pathophysiological characteristics of patients and mechanism of action of a drug product, the therapeutic effects may not be the same among different subjects, thus affecting the efficiency of clinical trials. To improve the efficiency of clinical trials by enrolling subjects who would benefit the most from an investigational drug, the concept of enrichment strategies has emerged.

Enrichment refers to prospectively and precisely defining the target population in a clinical trial that maximizes the benefit of participants from the investigational drug based on certain characteristics of subjects (such as demographics, pathophysiology, histology, genomics and proteomics, etc.). There are many subject-selection enrichment strategies in clinical trials, for example, subjects can be enriched based on their responses to the study drug, their insensitivity to existing drugs, or their likelihood to have endpoint events.

This guideline describes the principles and methods of commonly used enrichment strategies and designs and their advantages and disadvantages, and explains the key considerations from practical applications and regulatory perspectives. In this guideline, "enrichment strategy" primarily refers to the strategies used to select subjects who may benefit from randomized controlled clinical trials, but can also be extended to single-arm trials using external (historical or parallel) controls.

The guideline is applicable to confirmatory clinical trials for the purpose of supporting drug registration and marketing authorization and can also be used as a reference for clinical trials with non-registration purposes.

 

2.    Applicability of Enrichment Strategy and Design

Broadly speaking, the concept of enrichment is used in all clinical trials, which can be reflected in the inclusion and exclusion criteria of subjects with the purpose of enrolling subjects who most likely respond to the investigational drug, so as to improve the efficiency of clinical trials. For example, when studying a cholesterol-lowering drug to reduce the incidence of cardiovascular events, a clinical trial may only enroll subjects whose total blood cholesterol concentration is higher than a threshold. In fact, different enrichment strategies and designs may be chosen, based on the disease area, mechanism of action of the drug, and the response of the subjects. Whether to use and how to choose enrichment strategies should be considered from the aspects of scientific validity, interpretability and generalizability of trial results in medical practice.

1) Scientific validity: This includes scientific rationale for screening subjects, sensitivity and specificity of screening methods that meet certain requirements, measures to avoid bias (such as randomization, blinding, etc.) in the design of the trial, and the control of type I error.

2) Interpretability of findings: This refers to that the efficacy and safety results of the investigational drug in the enriched population can be explained in terms of the pathophysiology, genomics, genetics, or drug mechanism of action of the disease; if they cannot be explained due to limited knowledge in biology, medicine, or pharmacology, the therapeutic effects of the investigational drug in the enriched population needs to be reproducible to certain extent.

3) Generalizability in medical practice: Enrichment strategies should be able to be widely used in clinical practice in order to timely and accurately identify patients who respond or are sensitive to investigational drugs. Sometimes, the generalization of a screening method is not possible due to its complexity, low sensitivity, high cost, etc., or the screening method is too time-consuming and cannot enrich patients on a timely manner, which will affect the generalizability of enrichment strategies and methods.

 

3.   Commonly used enrichment strategies and designs

According to the main research question and implementation process of clinical trials, different enrichment strategies can be used. There are five commonly used types of enrichment strategies: homogeneous enrichment, prognostic enrichment, predictive enrichment, hybrid (prognostic and predictive combined) enrichment, and adaptive enrichment.

In practice, an enrichment strategy and design are chosen based on some markers related to the mechanism of drug action. Here "markers" refer to various characteristic variables such as epidemiological factors related to subject prognosis or response to drug treatment (e.g., demographics), past medical history, family history, clinically observed variables (e.g., disease severity), laboratory tests (e.g., pathophysiology, drug metabolism), biomarkers (e.g., genomics and proteomics). According to the roles of different markers, they can be classified into prognostic, predictive, and hybrid markers. In addition, in some disease areas, there may be no clear markers, and subjects are generally enriched based on their responses to treatment during screening, or data from other clinical trials and literature reports.

3.1 Enrichment for Homogeneity

Enrichment for homogeneity refers to a strategy of reducing the heterogeneity of subjects to improve the power of clinical trials. The simplest and most practical way to reduce heterogeneity is to select subjects with stable disease as much as possible, accurately define the selected subjects, and accurately measure the status of the disease and related variables. For example, in clinical trials of hypertension drugs, subjects' blood pressure may be measured for a period of time before enrollment to exclude those with large variations in blood pressure.

In general, in addition to the conventional inclusion and exclusion criteria, the following aspects should be considered in order to more accurately define an enriched population:

1) Inclusion criteria: Inclusion criteria should be more carefully defined to ensure consistent baseline characteristics among enrolled subjects.

2) Exclusion criteria: Exclude those subjects who ① are too sensitive to placebo; ② have unstable baseline test results, such as subjects with unstable conditions or symptoms during the primary screening period; ③ may die prematurely due to a concomitant disease; ④ take drugs with similar therapeutic effects to the test drug; ⑤ may not tolerate the test drug; ⑥ may withdraw from the study early due to complications.

3) Compliance: Subjects with good compliance should be included, i.e., subjects who do not withdraw for non-medical reasons (e.g., inconvenience to go to the study site), and subjects who are able to adhere to the treatment according to the trial protocol, so as to reduce differences due to excessive withdrawal of subjects or use of different treatment methods. Criteria for identification and selection of compliant patients must be set prior to randomization.

4) Training: The investigators and clinical trial coordinators shall receive relevant training to ensure that subject enrollment and trial conduct strictly follows the study protocol.

3.2 Prognostic enrichment

Prognostic enrichment refers to a strategy of increasing the power of the study by enrolling high-risk patients based on their prognostic markers. In general, high-risk patients are more likely to experience an endpoint event of interest or disease progression. This strategy mainly increases the absolute effect, not the relative effect, of a treatment. For example, in a clinical trial aiming to reduce the incidence of an endpoint events, after a period of treatment, the incidence of the endpoint events is reduced from 10% to 5% in the high-risk population and from 1% to 0.5% in the low-risk population. Although the relative effects of both trials are reduced by 50%, the former obviously requires less sample size or shorter follow-up time to observe the efficacy of the investigational drug. There are two commonly used prognostic enrichment designs.

1) Event-based enrichment design

In studies with the reduction of the incidence of endpoint events as the primary objective, the investigational drug is generally considered more effective in reducing more events in higher risk population. Therefore, enrolling high-risk subjects should be considered. In general, when the sample size is unchanged, the high-risk population is more likely to have more endpoint events than the low-risk population, and the incidence of endpoint events can be greatly reduced after treatment, leading to a more powerful trial. This strategy is often used in studies of drugs for anti-tumor and cardiovascular diseases, such as breast or ovarian cancer prevention in female population with BRCA1/2 mutation and in studies of hypolipidemic drugs in which patients with high concentration of low-density lipoprotein (LDL), low concentration of high-density lipoprotein (HDL) and high concentration of C-reactive protein (CRP) in blood are enrolled. In some disease areas, such as Alzheimer's disease and various cancer drug studies, high-risk patients can also be screened by genomic or proteomic screening.

2) Slowing disease-progression-based enrichment design

Prognostic enrichment designs can also be used to study an experimental drug that can slow disease progression, such as clinical trials for Alzheimer's disease, Parkinson's disease, rheumatoid arthritis, chronic obstructive pulmonary disease (COPD), and malignancies, where subjects with potentially faster disease progression can be enrolled. For example, rheumatoid arthritis patients with the following characteristics tend to have faster disease progression: rheumatoid factor positive, certain clinical characteristics (such as multiple joints affected, diseases other than joints, subcutaneous nodules, limited activities, etc.) and abnormal laboratory results (such as decreased hemoglobin); COPD patients with the following characteristics may have faster disease progression: a history of recent onset (at least one attack in the past year) or higher plasma fibrinogen. In the study of anti-tumor drugs, common prognostic markers include histological grade, vascular invasion, molecular subtype, and metastatic tumor nodules.

It should be noted that if there is an interaction between a prognostic marker and the trial drug, i.e., the study drug has an effect on both marker-positive and marker-negative patients, but the efficacy size is different, then the prognostic marker can also play a predictive role, and such markers are usually called hybrid markers.

3.3 Predictive enrichment

Predictive enrichment refers to an enrichment strategy in which subjects who are most likely to respond to the test drug are selected based on their physiological, response history, or disease characteristics related to the study drug’s mechanism. For example, in targeted anti-cancer therapy, subjects may respond to a treatment based on drug-related targeted genes or proteins, or physiological functions (e.g., renin hypertension/hypotension, chronic heart failure score). Adopting this strategy can increase both the absolute effect and relative effect of the test drug, so that a higher power can be achieved with a smaller sample size. This enrichment strategy is useful when only a small percentage of subjects with a disease respond to the test drug (e.g., only a subset of subjects has drug-acting receptors). In practice, subjects can be selected based on either the investigator's knowledge of the disease (e.g., various markers) or previous trial data and results.

1) Enrichment design based on pathophysiological characteristics

Subjects whose pathophysiological characteristics of the disease could suggest a better response to the test drug. Pathophysiology-based enrichment indicators can be biomarkers (such as gene mutation affecting tumor growth, gene/protein expression level), imaging characteristics, and clinical characteristics related to disease phenotype (such as disease staging, typing, etc.). Depending on the nature of the enrichment markers, the strategies can be classified into the following categories:

① Gene or protein marker-based: Anti-tumor drugs usually target relevant receptors, enzymes, hormones, or other endogenous active substances on or inside the tumor cell surface, for which an enriched population can be selected based on one or more corresponding gene or protein markers. For example, trastuzumab is mainly used to treat breast cancer patients who are human epidermal growth factor receptor 2 (HER-2) protein-positive. Some cell receptors that initially act as protein markers but are later confirmed as tumor gene markers (such as EGFR and BRAF gene mutations) have been used to define the pathophysiological status and to select the subjects who may benefit from target therapy.

The accuracy and precision of marker detection is essential when using a gene or protein marker in an enrichment design. If the diagnostic test is not accurate, it will not only affect the effect of enrichment design that may reduce the study power, but may also increase the type I error in non-inferior trials.

② Drug metabolite-based: The metabolic capacity of different subjects to the same investigational drug could be different. Enrolling subjects who can produce a sufficient number of active metabolites can improve the efficiency of clinical trials. In some cases, higher doses are given to patients who can produce less active substance, helping them to produce enough amount of active substance so that the efficacy of the drug is more likely to be observed. Patients who are completely unable to produce the active ingredients should be excluded from the trial.

③ Tumor metabolites-based: A clinical trial of antineoplastic drug may select subjects by measuring the amount of tumor metabolites in the tissue or blood. For example, only those subjects with metabolic reactions are enrolled, or grouped by the degree of metabolic reaction in cancer patients, and the primary analysis can be performed on subjects with strong metabolic reactions.

2) Enrichment design based on evidence of response to study drug

Such an enrichment design may allow selection of potentially suitable subjects based on their responses to the study drug (or similar drugs in the past) during the screening period.

① Screening subjects for responders: For clinical trials in which the subjects responding to the investigational drug cannot be identified based on the markers prior to the study, a reasonable screening period should be set during which all subjects are given the investigational drug. Subjects who respond to treatment are selected based on a predetermined primary or surrogate endpoint and then are randomized to different groups. Selecting responders can be performed using a two-stage randomized withdrawal design. In the first stage, subjects are tested to determine whether they can respond to the study drug (which can be done in a single-arm or randomized trial); in the second stage, responders are randomized to receive the test drug (continue to use the experimental drug) or placebo (withdraw the experimental drug) and non-responders are excluded from the study. For example, a study investigating a cholesterol lowering drug can use the randomized withdrawal design, in which subjects with high cholesterol are enrolled in the first stage and responders (based on cholesterol reduction) to the test drug in the first stage are selected to be randomized to receive the test drug or control drug in the second stage of the study.

The randomized withdrawal design has the advantage of improving the efficiency of clinical trials by selecting subjects who respond to the test drug. At the same time, the design can be used to investigate the long-term efficacy or safety of the study drug among subjects who continue using the drug and the withdrawal effect among subjects who withdraw from the trial in the second stage. On the other hand, this design is more ethical, that is, the trial can be terminated in a timely manner once treatment has failed and therefore can be used in pediatric drug research. This design by screening subjects in the first stage and randomizing them in the second stage based on their prior responses may screen more subjects and stratified randomization can be done according to the degree of their responses. The trial can first study the drug effect among subjects with strong responses and if positive outcome is observed, then the drug effect among subjects with weak responses. However, the design may not be suitable to study investigational drugs with long-term residual effect or lethal or harmful withdrawal effect, or drugs with a long duration between treatment initiation and subject responses.

② Selecting subjects based on historical data or literature reports: Subjects can be enrolled based on the characteristics identified in previous studies, i.e., if little or no significant treatment effect was observed in the overall population, but a significant effect may likely be achieved in a subpopulation, enrollment can be restricted only to the sub-population. For example, the combination isosorbide dinitrate/hydralazine hydrochloride is a drug for treatment of severe heart failure, and previous studies have found that its therapeutic effect on African-Americans is significantly better than that on Caucasians, then the subsequent randomized placebo-controlled trial enrolled 1050 African-American patients with heart failure which demonstrates the effectiveness of the combination drug in this sub-population of patients with heart failure.

3) Enrichment by selecting non-responders to existing drugs

In addition to selecting subjects who respond to the test drug, subjects who do not respond to an existing drug may be considered for a trial in order to better show the treatment effect of the test drug that has a different mechanism of action from the existing drug.

Enrichment for selecting non-responders is appropriate for clinical trials with the following conditions: the investigational drug has a different mechanism of action from an existing drug, or the investigational drug is at least slightly more effective than an existing drug. If no selection of subjects is performed, a larger sample size may be required to show efficacy of the investigational drug; on the contrary, if only subjects who do not respond to the existing drug are selected, because the response rate of the control group is very low, a smaller sample size may be necessary to demonstrate that the experimental group is superior to the control group. It must be noted that for certain potentially life-threatening or progressive diseases, randomizing subjects who do not respond to the control drug may be unethical.

3.4 Mixed enrichment

Mixed enrichment refers to an enrichment strategy using hybrid markers (e.g., prognostic and predictive markers) simultaneously to reduce subject heterogeneity. For some disease areas, the mechanisms for disease occurrence, development and prognosis could be complex, leading to a highly heterogeneous population of subjects. Therefore, it is unlikely to enrich the subjects using a single marker, while the use of multiple markers or a composite marker (such as a comprehensive score) for patient enrichment can effectively reduce the heterogeneity of subjects and thus improve the study efficiency.

It should be noted that individual markers that compose of the composite marker should be listed when using the composite marker score and their roles and relationships should be elucidated. If different individual markers are given different weights, the biological principle should be described in detail.

3.5 Adaptive enrichment

Adaptive enrichment strategy refers to a strategy of mid-course modification on the target study population (e.g., changing the inclusion and exclusion criteria, adjusting sample size of the overall population and/or enriched sub-population) based on the results of interim analysis as pre-defined in the study protocol, on the premise of ensuring the validity and integrity of the trial.

When the efficacy of the investigational drug is uncertain in subjects with positive and negative markers, the trial can enroll both marker-positive and -negative subjects before the interim analysis and adaptative enrollment can be made according to the results of interim analysis. If the interim analysis shows that the efficacy in marker-negative subjects is much lower than that in marker-positive subjects, then the enrollment can be restricted only to marker-positive subjects. If evidence is insufficient to show that the efficacy of marker-positive subjects is higher than that of marker-negative subjects, the trial can also first enroll marker-positive subjects and then marker-negative subjects if the interim analysis shows therapeutic effects in marker-positive subjects; otherwise, the trial should stop.

In general, if the relationship of a marker to the therapeutic effect is uncertain, then it is generally recommended to enroll marker-negative subjects, which can help assess the benefits and risks of the drug in the full population. When the predictability of a marker is uncertain, the primary analysis can be performed in the full population; if the marker-positive population and the full population are primary analysis populations, the test level α shall be split according to certain rules. In either case, the testing hypothesis should be pre-specified in the protocol and the type I error needs to be controlled.

 

4. Other Considerations of Enrichment Strategy and Design

4.1 Sensitivity and specificity of marker detection

When screening tests are used to select subjects, the reliability of the screening method must be taken into account in order to more accurately select subjects who are at high risk or most likely respond to the test drug. Ideally, a screening test should have a high sensitivity for selecting subjects at high risk or who respond to test drug and a high specificity for identifying subjects at low risk or who do not respond to test drug.

When biomarkers are used to screen subjects, if the threshold value of predictive markers cannot be accurately given, the sensitivity and specificity of different threshold points of markers can be analyzed by Receiver Operating Curve (ROC) analysis, and the screening effect can be measured by the area under ROC curve. With regard to the determination of predictive marker thresholds, it is generally possible to first give a preliminary threshold in the early research stage and then adjust it during the trials with a larger sample size to obtain a more reliable threshold.

4.2 Whether the subjects with positive and negative markers are included

The enrichment design can enroll either only marker-positive subjects or both marker-positive and marker-negative subjects. However, one of the key issues in the enrichment design is the proportion of marker-positive and marker-negative subjects to be enrolled. In general, the following enrichment strategies can be considered:

1) Enrolling only marker-positive subjects

If the mechanism of action or available data show that the investigational drug has significant efficacy in marker-positive subjects but has less or no efficacy in marker-negative subjects, then the trial should not enroll marker-negative subjects.

2) Enrolling both marker-positive and -negative subjects

If the mechanism of action or available data suggest that marker-positive subjects may have better efficacy than those with marker-negative subjects, the trial can enroll both marker-positive and -negative subjects if the test drug is less toxic. This strategy has the advantage of providing a reasonable benefit-risk estimate in a non-enriched population.

If a marker can be identified before the start of the trial, stratified randomization can be implemented within stratum, and the primary analysis can be restricted to marker-positive subjects. In practice, the primary analysis can also be performed in the full population, or simultaneously in the full population and in marker-positive subpopulation, with appropriate control of type I error.

In general, if the threshold for a marker or the magnitude of response for marker-negative subjects is uncertain, it is recommended to include marker-negative subjects.

4.3 Inclusion population and analysis set

The main concerns of using enrichment strategies are the applicability and extrapolability of the findings, that is, when using an enrichment design, it is important to consider whether this enrichment strategy can be used in medical practice to identify subjects who respond to the study drug and whether the drug has similar efficacy in a wider patient population. Note that it is equally important to study patient populations who do not meet the enrichment criteria. It should also be noted that the enrolled subjects and the primary analysis set identified in the trial may be different (the latter may be a subset of the former), but these must be clearly defined in the study protocol. When genetic or other test results are not immediately available and patients need timely treatment, the overall population can be selected to provide more safety information, but the primary efficacy analysis can be a subset of the study population.

4.4 Different effects of enrichment strategies on superiority and non-inferiority trials

Using a marker to select subjects has a different impact on superiority and non-inferiority trials. For superiority trials using prognostic enrichment, if the screening method is insensitive, more subjects need to be screened in order to enroll a pre-specified sample size of enriched subjects; if the screening method has a lower specificity, the sample size of enriched subjects may be large or the trial may last longer in order to obtain a sufficient number of endpoint events. For superiority trials using predictive enrichment strategies, if the sensitivity of screening method is not high, there will not be enough subjects eligible for criterion of inclusion, and if the specificity is not high, more ineligible subjects would be enrolled. Nevertheless, neither prognostic nor predictive enrichment strategy increased the type I error in superiority trials.

However, for non-inferiority trials, the accuracy of screening method will not only affect the sample size or duration of the trial, but may also inflate the type I error rate. For example, a non-inferiority trial using a prognostic enrichment strategy may result in a lower efficacy estimate of the positive control group than in previous studies if the screening method of the positive control is different from that in previous studies, thereby increasing the type I error. The impact on type I error using predictive enrichment strategies is more complex in non-inferiority trials, depending on whether the marker is related to the efficacy of the test drug and the active comparator, or to the efficacy of only one of the treatments. Therefore, the screening method for selecting subjects in non-inferiority trials should be consistent with the screening method of positive control in previous studies, or both screening methods have similar in terms of operating characteristics (e.g., sensitivity and specificity).

4.5 Control of type I errors

If a trial enrolls both enriched and non-enriched populations, different hypothesis testing strategies may be considered according to the accuracy of screening method and subject responses to the treatment. If multiple hypotheses are specified, such as hypotheses relating to marker-positive and overall populations, multiplicity adjustment should be carried out. The distribution of type I error α under different assumptions can be set according to the degree of response of the marker-positive population to the drug, the proportion of the marker-positive population in the overall population, and the sample size required according to the prespecified power of the test. When hypothesis testing is performed for overall population and enriched subpopulation, independent or sequential testing strategy may be adopted for hypothesis testing.

 

5. Regulatory Considerations

5.1 Clearly define the enriched population

Whether, when, and what enrichment strategy is used in clinical trials depends mainly on whether the enriched population can be accurately identified, which has a clear impact on the specification of product label and subsequent medical practice. The labelling should describe the indicated population strictly according to the enriched population studied. If the enriched population cannot be accurately identified using the enrichment strategy and design, it may not accurately define the patient populations who respond to the treatment, thus failing to accurately guide drug use in clinical practice.

5.2 Do not neglect the efficacy in non-enriched populations

After the efficacy and safety of the investigational drug in the enriched population have been confirmed, the corresponding information of the investigational drug in the non-enriched populations should also be considered. Further studies in non-enriched populations may provide a more comprehensive picture of the benefit-risk profile of the test drug and thus provide a basis for its use in a broader patient population.

For drugs approved based on prognostic enrichment in high-risk populations, different outcome measures may be used in subsequent trials among low-risk populations, such as case fatality rate in high-risk populations and a clinically significant composite outcome measure in low-risk populations, which can help improve trial efficiency for the later patient populations.

5.3 Predetermine the study protocol and communicate with the regulatory authorities

In general, subject selection should be preplanned and determined prior to trial start. If characteristic variables or markers are known, enrichment can be implemented by screening subjects. When the effect or distribution of characteristic variables or markers in the study population is uncertain, adaptive enrichment may be considered, that is, enrichment can be adjusted during the course of the trial according to the interim analysis of the accumulated data. Regardless of the strategy and design used, the adjustment methods and processes should be described in advance in the study protocol to ensure study integrity and validity, and adequately communicated to the regulatory authorities.

 

 

References

[1]    Amur S, LaVange L, Zineh I, et al. Biomarker Qualification: Toward a Multiple Stakeholder Framework for Biomarker Development, Regulatory Acceptance, and Utilization. Clinical Pharmacology and Therapeutics, 2015; 98 (1): 34-46.

[2]    Barker AD, Sigman CC, Kelloff GJ, et al. I-SPY 2: an adaptive breast cancer trial design in the setting of neoadjuvant chemotherapy. Clinical pharmacology and therapeutics, 2009; 86 (1): 97-100.

[3]    Bibbins-Domingo K, Fernandez A. BiDil for heart failure in black patients: implications of the U.S. Food and Drug Administration approval. Annals of Internal Medicine, 2007; 146 (1): 57-62.

[4]    Chapman PB, Hauschild A, Robert C, et al. Improved survival with vemurafenib in melanoma with BRAF V600E mutation. The New England Journal of Medicine, 2011; 364 (26): 2507-2516.

[5]    D'Agostino RB Sr. The Delayed-Start Study Design. The New England Journal of Medicine, 2009; 361 (13): 1304-1306.

[6]    D'Amico AV, Chen MH, Roehl KA, et al. Preoperative PSA velocity and the risk of death from prostate cancer after radical prostatectomy. The New England Journal of Medicine, 2004;19 351 (2): 125-135.

[7]    Early Breast Cancer Trialists' Collaborative Group (EBCTCG). Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15-year survival: an overview of the randomised trials. Lancet, 2005; 365 (9472): 1687-1717.

[8]    European Medicines Agency (EMA). Points to consider on multiplicity issues in clinical trials.

[9]    European Medicines Agency (EMA). Guideline on the investigation of subgroups in confirmatory clinical trials.

[10]  European Medicines Agency (EMA). Qualification Opinion of Alzheimer’s Disease Novel Methodologies/biomarkers for BMS-708163. 2011.

[11]  European Medicines Agency (EMA). Qualification Opinion of Low Hippocampal Volume (Atrophy) By MRI for Use in Regulatory Clinical Trials - in Pre-Dementia Stage of Alzheimer’s Disease. 2011.

[12]  Fan C, Oh DS, Wessels L, et al. Concordance among GeneExpression–Based Predictors for Breast Cancer. The New England Journal of Medicine, 2006; 355 (6): 560-569.

[13]  Freidlin B, Korn EL. Biomarker enrichment strategies: matching trial design to biomarker credentials. Nature20 Reviews Clinical Oncology, 2014; 11 (2): 81-90.

[14]  Havel JJ, Chowell D, Chan TA. The evolving landscape of biomarkers for checkpoint inhibitor immunotherapy. Nature Reviews Cancer, 2019; 19 (3): 133-150.

[15]  Hughes RA, Donofrio P, Bril V, et al. Intravenous immune globulin (10% caprylate-chromatography purified) for the treatment of chronic inflammatory demyelinating polyradiculoneuropathy (ICE study): a randomised placebo-controlled trial. The Lancet Neurology, 2008; 7 (2): 136-144.

[16]  Institute of Medicine (US) Committee on Strategies for Small-Number-Participant Clinical Research Trials. Small Clinical Trials: Issues and Challenges. Washington, DC. The National Academies Press, 2001.

[17]  International Conference on Harmonization (ICH). E5 guideline “Ethnic Factors in the Acceptability of foreign Clinical Data E5 (R1)”.

[18]  International Conference on Harmonization (ICH). E5 guideline “Implementation Working Group Questions & Answers (R1)”.

[19]  Jiang W, Freidlin B, Simon R. Biomarker-adaptive threshold design: a procedure for evaluating treatment with possible biomarker-defined subset effect. Journal of the National21 Cancer Institute, 2007; 99 (13): 1036-1043.

[20]  Kowanetz M, Zou W, Gettinger SN, et al. Differential regulation of PD-L1 expression by immune and tumor cells in NSCLC and the response to treatment with atezolizumab (anti-PD-L1). Proceedings of the National Academy of Sciences of the United States of America, 2018; 115 (43): E10119-E10126.

[21]  Liu A, Liu C, Li Q, et al. A threshold sample-enrichment approach in a clinical trial with heterogeneous subpopulations. Clinical trials, 2010; 7(5): 537-545.

[22]  Loo E, Khalili P, Beuhler K, et al. BRAF V600E Mutation Across Multiple Tumor Types: Correlation Between DNAbased Sequencing and Mutation-specific Immunohistochemistry. Applied immunohistochemistry & molecular morphology, 2018; 26 (10): 709-713.

[23]  Parkinson Study Group. A controlled, randomized, delayedstart study of rasagiline in early Parkinson disease. Archives of neurology, 2004; 61 (4): 561-566.

[24]  Priscilla Velengtas, Penny Mohr, Messner DA. Making informed decisions: Assessing the strengths and weaknesses of study designs and analytic methods for comparative effectiveness research. National Pharmaceutical Council ed.22 Washington, DC. 2012.

[25]  Ridker PM, Danielson E, Fonseca FA, et al. Rosuvastatin to prevent vascular events in men and women with elevated Creactive protein. The New England Journal of Medicine, 2008; 359 (21): 2195-2207.

[26]  Scandinavian Simvastatin Survival Study Group. Randomised trial of cholesterol lowering in 4444 patients with coronary heart disease: The Scandinavian Simvastatin Survival Study (4S). Lancet, 1994; 344 (8934): 1383-1389.

[27]  Schrock AB, Ouyang C, Sandhu J, et al. Tumor mutational burden is predictive of response to immune checkpoint inhibitors in MSI-high metastatic colorectal cancer. Annals of oncology, 2019; 30 (7): 1096-1103.

[28]  Singh BN. Comparative efficacy and safety of bepridil and diltiazem in chronic stable angina pectoris refractory to diltiazem. The Bepridil Collaborative Study Group. The American journal of cardiology, 1991; 68 (4): 306-312.

[29]  Simon R. Clinical trial designs for evaluating the medical utility of prognostic and predictive biomarkers in oncology. Personalized medicine, 2010; 7 (1): 33-47.

[30]  Taylor AL, Ziesche S, Yancy C, et al. Combination of isosorbide dinitrate and hydralazine in blacks with heart23 failure. The New England Journal of Medicine, 2004; 351 (20): 2049-2057.

[31]  Temple R. Enrichment of clinical study populations. Clinical Pharmacology & Therapeutics, 2010; 88 (6): 774-778.

[32]  Temple RJ. Special study designs: early escape, enrichment, studies in non-responders. Communications in Statistics -Theory and Methods, 1994; 23 (2): 499-531.

[33]  U.S. Food and Drug Administration (FDA). Enrichment Strategies for Clinical Trials to Support Determination of Effectiveness of Human Drugs and Biological Products – Guidance for Industry.

[34]  U.S. Food and Drug Administration (FDA). Qualification of Biomarker–Plasma Fibrinogen in Studies Examining Exacerbations and/or All-Cause Mortality in Patients with Chronic Obstructive Pulmonary Disease. 2016.

[35]  U.S. Food and Drug Administration (FDA). Qualification of Biomarker–Total Kidney Volume in Studies for Treatment of Autosomal Dominant Polycystic Kidney Disease. 2016.

[36]  Wang S, Hung HMJ, O Neill RT. Genomic Classifier for Patient Enrichment: Misclassification and Type I Error Issues in Pharmacogenomics Noninferiority Trial. Statistics in Biopharmaceutical Research, 2011; 3 (2): 310-319.24

[37]  Wang SJ, Hung HM, O'Neill RT. Adaptive patient enrichment designs in therapeutic trials. Biometrical Journal, 2009; 51 (2): 358-374.

[38]  Wang SJ, O'Neill RT, Hung HM. Approaches to evaluation of treatment effect in randomized clinical trials with genomic subset. Pharm Stat, 2007; 6 (3): 227-244.

 

 

 

Appendix 1: Glossary

Sensitivity: one of the basic indicators to evaluate the accuracy of diagnostic test and screening test. In an enrichment study for a drug clinical trial, sensitivity indicates the probability that a subject who is at high risk for an endpoint event or who responds to the drug can be correctly identified.

Adaptive Enrichment Design: According to the predetermined plan and based on the interim analysis results of clinical trial data, on the premise of ensuring the rationality and integrity of the trial, it is allowed to adaptively update the inclusion and exclusion criteria during the trial and select the adaptive design for the subjects who may benefit from the treatment.

Randomized Withdrawal Design: In such a design, all subjects receive the investigational drug during the initial open-label period, then subjects who do not respond to the drug withdraw from the trial, and subjects who respond (enriched) are randomized to receive the investigational drug or placebo in the second phase of the trial.

Specificity: It refers to one of the basic indexes for evaluating the accuracy of diagnostic test and screening test. In an enrichment study for a drug clinical trial, specificity indicates the probability of being able to correctly identify subjects who are at low risk for an endpoint event or who do not respond to the drug.

Heterogeneity: In clinical trials, heterogeneity is reflected in two levels: individual and group. The former usually refers to that there are different characteristics among subjects. Different nature or status of individual subjects may lead to different responses to treatment; the latter usually refers to that subjects from different centers, ethnicities and regions have different characteristics, which may lead to different responses to treatment for different subjects.

Predictive Enrichment: refers to a study strategy or design that selectively includes the subjects who may respond to the treatment. These subjects have common biological and histopathological characteristics with predictive significance and can more sensitively display the investigational drug.

Prognostic Enrichment: A research strategy or design that selectively includes subjects who are more likely to experience an endpoint event, such as death or disease worsening, thereby reducing the sample size required to achieve a statistically significant effect.

 

 

Appendix 2: Chinese-English Vocabulary

Appendix 3: Study cases for enrichment design

Example 1: Prognostic Enrichment – Cardiovascular Study

In cardiovascular studies, outcome events may be more easily observed in high-risk subjects (such as those with AMI, stroke, cholesterol level, very severe CHF and undergoing angioplasty, etc.). The Scandinavian Simvastatin Survival Study (4S) is a trial of lipid-lowering drugs with the primary aim of assessing whether simvastatin can improve survival in patients with coronary heart disease by lowering serum cholesterol. The study was a randomized double-blind placebo-controlled multicenter clinical trial that enrolled 4444 patients with angina or previous myocardial infarction (MI), all of whom had high total cholesterol (TC) levels. During a mean follow-up of 5.4 years, cardiovascular mortality was significantly reduced with simvastatin as compared with placebo (relative risk RR 0.70, 95% CI: 0.58 – 0.85).

Example 2: Predictive enrichment – melanoma study

BRAF kinase inhibitors are a type of targeted drugs for the treatment of melanoma, and exon 15 (V600E) of the BRAF gene can be used as a predictive biomarker. The BRAF gene is known to encode a cytoplasmic serine/threonine kinase, an enzyme that regulates the mitogen-activated protein kinase signal transduction pathway that controls several important cellular functions including cell growth and division (proliferation). It has been found that BRAF V600E is mutated in a variety of tumors, such as melanoma, colorectal cancer, papillary thyroid carcinoma, hairy cell leukemia and Langerhans cell hyperplasia. In a phase III clinical trial of melanoma that enrolled 675 subjects with metastatic or unresectable BRAFV600E mutation who were treated with the BRAF kinase inhibitor vemurafenib or the chemotherapeutic drug dacarbazine, the response rate was 48% in subjects treated with vemurafenib targeted agents and only 5% in subjects treated with dacarbazine chemotherapy; the relative risk of death was reduced by 63% in subjects treated with vemurafenib.

Example 3: Predictive enrichment – MSI study

Microsatellite instability (MSI) is a biomarker that responds to immune checkpoint inhibitors. PD-1/PD-L1 pathway is a signaling pathway that regulates T cell activation and plays an important role in tumorigenesis and progression. In practice, the expression level of PD-L1 protein is usually detected by immunohistochemical method, which is used as a predictive marker and selects subjects with high expression, but its response rate to PD-1/PD-L1 inhibitors is only 10-20%. However, a 50% response rate was achieved in subjects with tumors with high-grade microsatellite instability (MSI-HIGH). Based on this, the FDA approved pembrolizumab for the treatment of subjects with MSI-HIGH-type or mismatch-repair deficient colorectal and endometrial cancers.

Example 4: Randomized withdrawal design - study of pregabalin for fibromyalgia

A clinical trial investigating the efficacy of pregabalin in the treatment of subjects with fibromyalgia used a two-stage randomized withdrawal design to compare the difference in time to loss of therapeutic response (TLTR) between pregabalin and placebo. The first phase was an open-label trial in which subjects with fibromyalgia were all treated with pregabalin and observed for 6 weeks. At 1 – 3 weeks, subjects received escalating doses of pregabalin to decide their optimal dose; at 4 – 6 weeks, subjects were maintained at this optimal dose. After completion of the open-label treatment in Stage I, subjects had to have at least a 50% pain reduction and at least a "marked improvement" in their self-evaluation on the PGIC scale in order to enter the double-blind, placebo-controlled trial in Stage II. Of the 1051 subjects, 566 entered the second phase after the first phase of treatment, of whom 287 were randomly assigned to placebo and 279 to pregabalin. After 26 weeks of treatment in the second stage, a significant difference in the time to loss of therapeutic response (LTR) was observed between the two groups (p < 0.0001). At the end of the trial, 61% (178) of placebo and 32% (90) of pregabalin achieved LTR.

 

点击此处，查看原文附件

Guideline on Enrichment Strategies and Designs in Clinical Trials（final）