EMA发布预测性生物标志物检测指南草案并公开征求意见
Draft guideline on predictive biomarker assays in the context of medicinal product lifecycle
EMA人用药品委员会(CHMP)通过预测性生物标志物检测指南草案,公开征求意见自2026年11月9日起至2027年3月31日。草案面向生物标志物引导药品的上市许可及上市后申请,涉及检测的分析验证、临床验证、桥接研究以及产品信息(SmPC)中的相关表述,并说明与IVDR下伴随诊断的衔接。草案将替代EMA/CHMP/800914/2016概念文件。
草案梳理了预测性生物标志物检测在药品开发各阶段的验证要求及与IVDR的衔接,可供申报资料准备时对照。
PDF 文字版;图形和原始排版请参阅官方 PDF。
第 1 页
1 22 April 2026
2 EMA/CHMP/800914/2016
3 Committee for Medicinal Products for Human Use (CHMP)
4 Guideline on predictive biomarker assays in the context of
5 medicinal product lifecycle
6 Draft
Draft agreed by MWP September 2026
Adopted by CHMP 03 November 2026
Start of public consultation 09 November 2026
End of consultation (deadline for comments) 31 March 2027
7
8 This guideline replaces the Concept paper on predictive biomarker-based assay development in the
9 context of drug development and lifecycle (EMA/CHMP/800914/2016)
10
Comments should be provided using this EUSurvey form. For any technical issues, please contact
the EUSurvey Support.
11
Keywords Predictive biomarkers, biomarker assays, assay development,
medicinal product development, co-development, companion
diagnostics, CDx, In-Vitro Diagnostics, IVD
12
13
Official address Domenico Scarlattilaan 6 ● 1083 HS Amsterdam ● The Netherlands
Address for visits and deliveries Refer to www.ema.europa.eu/how-to-find-us
Send us a question Go to www.ema.europa.eu/contact Telephone +31 (0)88 781 6000 An agency of the European Union
© European Medicines Agency, 2026. Reproduction is authorised provided the source is acknowledged.
第 2 页
14 Guideline on predictive biomarker assays in the context of
15 medicinal product lifecycle
16 Table of contents
17 Executive summary ..................................................................................... 3
18 1. Introduction ............................................................................................ 3
19 2. Scope....................................................................................................... 4
20 3. Legal basis and procedural background................................................... 5
21 3.1. Procedural context for the assessment of predictive biomarker assays (candidate CDx) 5
22 3.2. Legal basis and relevant guidelines ......................................................................... 6
23 4. Context of use of biomarker assays during medicinal product
24 development ............................................................................................... 8
25 5. Scientific rationale for predictive biomarkers and their corresponding
26 assays ......................................................................................................... 9
27 6. Analytical validation of biomarker assays during medicinal product
28 development ............................................................................................. 10
29 6.1. General aspects ................................................................................................. 11
30 6.2. Assay considerations .......................................................................................... 11
31 6.3. Considerations for specific assays ......................................................................... 12
32 6.4. Characterisation studies ...................................................................................... 12
33 6.5. Assay adaptations .............................................................................................. 13
34 6.6. Considerations for centralised versus decentralised testing ...................................... 13
35 7. Samples for biomarker analysis during medicinal product development 14
36 7.1. Pre-screening .................................................................................................... 15
37 7.2. Baseline samples ................................................................................................ 15
38 7.3. Missing data ...................................................................................................... 16
39 8. Clinical validation of biomarkers and validation of biomarker assays .... 16
40 9. Bridging studies .................................................................................... 18
41 10. Summary of Product Characteristics (SmPC) ....................................... 20
42 11. Post-approval follow-up ...................................................................... 21
43 12. Definitions and abbreviations .............................................................. 21
44 13. References .......................................................................................... 25
45
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 2/26
第 3 页
46 Executive summary
47 Predictive biomarker-guided medicines represent one of the most widely implemented regulatory
48 concepts of personalised medicine in Europe, by linking the clinical use of a medicinal product to a
49 corresponding predictive biomarker, with the intent to tailor treatment decisions to patients most likely
50 to benefit in terms of efficacy and/or safety. The efficacy and/or safety of such a predictive biomarker-
51 guided medicinal product relies on a corresponding analytically robust, reliable and clinically useful
52 biomarker assay. For its use with a medicinal product, it may be a companion diagnostic (CDx) as
53 defined in the In Vitro Diagnostics Regulation (IVDR).
54 This guideline describes the information needed for the submission of a Marketing Authorisation
55 Application (MAA) for a biomarker-guided medicinal product. Guidance is provided on the generation of
56 a robust data package for the biomarker and corresponding biomarker assay for the purpose of the
57 marketing authorisation and post-authorisation applications. The guideline outlines best practices and
58 the key evidence expected to be generated throughout the clinical development programme to support
59 medicinal product use. Additionally, it provides general recommendations on how to reflect relevant
60 information related to such biomarkers and their associated assays in the Summary of Product
61 Characteristics (SmPC).
62 This guideline is primarily intended for medicinal product developers, but may also be useful for in vitro
63 diagnostic (IVD)/CDx manufacturers working with medicinal products developers to prepare the
64 marketing authorisation application, as it covers the interface with the IVDR. To facilitate
65 understanding, where possible, terminologies from both the medicines and the IVDR frameworks are
66 used.
67 1. Introduction
68 Technological advancements, among other factors, have resulted in an increase in prevention and
69 treatment strategies that consider the variability in characteristics of (groups of) individuals. This has
70 led to the development and definition of several closely related concepts such as precision medicine or
71 personalised medicine. While there are no universally accepted definitions of these concepts, European
72 Union (EU) Health Ministers described personalised medicine in the Council conclusions on personalised
73 medicine for patients as “a medical model using characterisation of individuals’ phenotypes and
74 genotypes (e.g. molecular profiling, medical imaging, lifestyle data) for tailoring the right therapeutic
75 strategy for the right person at the right time, and/or to determine the predisposition to disease and/or
76 to deliver timely and targeted prevention” [1].
77 Predictive biomarker-guided medicines development is among the most widespread applications of the
78 personalised medicine concept, resulting in the clinical use of a medicinal product together with its
79 corresponding predictive biomarker assay (refer to section 12 for definitions). Accordingly, patients are
80 identified by biomarkers predictive for efficacy and/or safety for a treatment decision with the
81 corresponding medicinal product. The efficacy and/or safety of such a medicinal product strictly relies
82 on having a corresponding accurate, reproducible and clinically useful assay for predictive biomarker
83 measurement. Therefore, the development of a predictive biomarker-guided medicinal product requires
84 the identification and clinical validation of a biomarker, and the (co-)development of a corresponding
85 biomarker assay.
86 If the use of predictive biomarker assays to identify patients who are most likely to benefit or be at
87 increased risk of adverse reactions as a result of treatment with the corresponding medicinal product is
88 mandatory, this has to be adequately reflected in the product information, in order to ensure the
89 effective and safe use of the associated medicinal product [2].
90 The regulatory frameworks for medicinal products and IVDs are separate. Whereas the focus of the
91 medicines sector is on the justification of the biomarker in the context of the use of the medicinal
92 product, this justification cannot be provided unless there is a specific assay supporting reliable
93 conclusions on the measurement of the biomarker.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 3/26
第 4 页
94 Where a predictive biomarker assay in the context of the therapeutic use of a medicinal product fulfils
95 the definition of a CDx, this creates an interface with the medical device legislation, notably Regulation
96 (EU) 2017/746 on IVDR, particularly Article 2 (7) [3], which defines a CDx as:
97 “a device which is essential for the safe and effective use of a corresponding medicinal product to
98 (a) identify, before and/or during treatment, patients who are most likely to benefit from the
99 corresponding medicinal product; or
100 (b) identify, before and/or during treatment, patients likely to be at increased risk of serious adverse
101 reactions as a result of treatment with the corresponding medicinal product.”
102 A CDx is essential for defining patients’ eligibility for a specific treatment with a medicinal product
103 through the quantitative or qualitative determination of (a) specific biomarker(s). An IVD is not a CDx
104 if it is used to ensure that the concentration of a medicinal product in the human body stays within the
105 therapeutic window during treatment (recital 12 of the IVDR [3]) or when it is used to establish the
106 dosage of a medicinal product for patients that are already eligible to receive that medicinal product.
107 Further considerations on when an IVD is considered to be a CDx are available in the Medical Device
108 Coordination Group (MDCG) guidance on classification rules for in vitro diagnostic medical devices
109 under Regulation (EU) 2017/746 [4].
110 Under the IVDR risk classification rules, CDx are assigned to risk class C (or higher if relevant rules
111 apply), which is the second highest risk class. To place a CDx on the market, the manufacturer must
112 receive certificates from the notified body (NB) for the quality management system and for the
113 technical documentation of the device. The NB assessment of the technical documentation involves,
114 among other elements, assessment of the data provided by the manufacturer for demonstration of the
115 scientific validity, analytical performance and clinical performance of the device. The interface is
116 addressed in the legislation through the CDx consultation procedure with Medicinal Product Authorities
117 as part of CDx certification by NB, although the approval of a biomarker-guided medicinal product is
118 independent from the CDx certification.
119 This guideline describes the interface and highlights important common ground between the
120 pharmaceutical legislation and the IVDR, from a medicinal product regulatory framework perspective.
121 Therefore, this guideline provides recommendations for the lifecycle management of predictive
122 biomarker-guided medicinal products, i.e., when a corresponding predictive biomarker assay is
123 required, which likely will be a candidate for a CDx as defined by the IVDR. It may be possible to use
124 data from the (confirmatory) clinical trial(s) as evidence to support the clinical performance of the
125 candidate CDx.
126 Throughout this guideline, the medicines framework terminology is preferentially used. Such is the
127 case when using the terminology “assay” in a broad sense encompassing assays used in clinical trials
128 and MAA irrespective of its status according to the IVDR. When an equivalent term is found in the
129 IVDR, this term is referred to in brackets and italics. Because the classification as CDx is not in the
130 remit of the Medicinal Product Authority, the terms “candidate CDx” or IVD are used when referring to
131 predictive biomarker assays (refer to section 4 and 12 of the guideline).
132 2. Scope
133 This guidance provides the scientific requirements for the development and validation of predictive
134 biomarkers and their corresponding assays for the purpose of MAA and post-authorisation applications
135 (e.g., indication extensions) for biomarker-guided medicinal products, outlining the essential evidence
136 to be generated throughout the clinical development programme.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 4/26
第 5 页
137 Additionally, general guidance is provided on how to reflect information related to such predictive
138 biomarkers and biomarker assays in the medicinal product information (the SmPC), as well as certain
139 regulatory aspects developers should be aware of in the European regulatory framework for medicines.
140 The document also covers guidance for follow-on CDx candidates developed after the authorisation of
141 the associated medicinal product. The document specifically addresses the interface between the
142 medicines’ framework and IVDR and uses both terminologies where they are related (IVDR terminology
143 provided in italics and in brackets).
144 This guideline is primarily intended for medicinal product developers but may also be useful for
145 IVD/CDx manufacturers working with medicinal product developers to prepare a marketing
146 authorisation application, as it covers the interface with the IVDR.
147 The following aspects are out of scope for this guideline:
148 • In vivo diagnostics, as defined and covered in the Guideline on clinical evaluation of diagnostic
149 agents [6]
150 • Classification & categorisation of IVDs
151 • Requirements for IVDs in the context of clinical trial approval (which is regulated under
152 Regulation (EU) No 536/2014 on clinical trials on medicinal products for human use, also
153 referred to as Clinical Trials Regulation (CTR) [7]) by Medicinal Product National Competent
154 Authority (NCA) or requirements for CDx performance study approval by Device NCAs (under
155 IVDR [3])
156 • Recommendations for conformity assessment under the IVDR [3]
157 3. Legal basis and procedural background
158 3.1. Procedural context for the assessment of predictive biomarker assays
159 (candidate CDx)
160 In the medicinal product regulatory setting, two scenarios may occur in which an assessment of a
161 predictive biomarker assay (candidate CDx (refer to section 4 and 12 of the guideline)) by a Medicinal
162 Product Authority (European Medicines Agency (EMA) or NCA) is conducted.
163 1. During a MAA or post-authorisation application procedure (e.g., an extension of
164 indication) for a biomarker-guided medicinal product
165 2. During a CDx consultation procedure (legally mandated by the IVDR) submitted by a NB to
166 a Medicinal Product Authority.
167 (1) During a MAA and post-authorisation application procedures
168 Biomarker assays that are used for patient selection (for efficacy and/or safety) define the patient
169 population investigated in clinical trials and considered for the benefit/risk (B/R) assessment of a
170 predictive biomarker-guided medicinal product. Evaluating the biomarker assay in terms of analytical
171 data robustness and reliability to measure biomarker(s) is therefore key to understanding the B/R
172 profile of the medicinal product in the intended population. It is important to note that during the MAA
173 procedure, the Medicinal Product Authority does not assess the IVDR conformity of the biomarker
174 assay (IVD) used in the clinical (pivotal) studies, even if it is a candidate CDx. This aspect falls under
175 the legal remit of the NB performing a conformity assessment in a separate procedure.
176 The MAA assessment for a predictive biomarker-guided medicinal product evaluates the
177 appropriateness of the biomarker assay for patient selection, including whether the choice of
178 biomarker-identified populations in the pivotal study(ies) was justified (refer to section 5 of the
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 5/26
第 6 页
179 guideline). It also seeks to determine whether this supports the proposed target population described
180 in the SmPC, and to assess the study outcome(s) in relation to the claimed conditions of use.
181 The analytical validation of the biomarker assay used in the pivotal clinical study(ies) is an important
182 consideration; therefore, the analytical method is evaluated as part of the assessment report (refer to
183 section 6 of the guideline). Based on this evaluation, a conclusion is drawn regarding the suitability,
184 robustness and reliability of the data generated by the biomarker assay for defining the
185 target/intended patient population and supporting the B/R assessment.
186 Clinical validation is described either by substantiating that biomarker status, as determined by the
187 biomarker assay, is predictive of a clinical endpoint (for novel assays), or – where available - by a
188 concordance study with a clinically valid reference assay. Moreover, cut-off point selection is described
189 and discussed in detail since this is of particular importance for the B/R assessment (refer to section 8
190 of the guideline). This information is gathered from the MAA dossier or extension of indication variation
191 dossier submitted by the applicant, where information on the development and validation of the
192 predictive biomarker assay is expected to be included.
193 The description and assessment of the predictive biomarker assay are summarised in the dedicated
194 sections on "in vitro biomarker test for patient selection for efficacy / safety" in the clinical assessment
195 report and overview template [8], or, where applicable, in the corresponding assessment report
196 templates for new indication and extensions of indication. The assessment of the predictive biomarker
197 assay also supports the inclusion of information in the SmPC on the biomarker assay(s) used during
198 the clinical development of the associated medicinal product (refer to section 10 of the guideline).
199 Further recommendations on the assessment of a predictive biomarker assay during a MAA or
200 extension of indication are provided in the “Frequently asked questions on medicinal products
201 development and assessment involving companion diagnostic (CDx)" [9].
202 (2) CDx consultation procedure
203 When a CDx consultation procedure is submitted by a NB, the Medicinal Product Authority is asked to
204 provide a scientific opinion on the suitability of a CDx candidate intended for certification in relation to
205 the concerned medicinal product. The opinion is provided based on the draft Summary of Safety and
206 Performance (SSP) and the draft Instructions for Use (IFU) submitted by the NB. The opinion provided
207 by the Medicinal Product Authority is subsequently considered by the NB in their final decision on
208 issuing EU certification for the CDx candidate. The consultation procedure assessment by the Medicinal
209 Product Authority is formally restricted to the information provided in the draft SSP and draft IFU but is
210 also based on the MAA assessment for the corresponding medicinal product.
211 Detailed guidance on procedural aspects of the CDx consultation procedure is provided in the
212 “Guidance on the procedural aspects for the consultation to the European Medicines Agency by a
213 notified body on companion diagnostics” [5], and additional practical information can be found in the
214 “Questions & Answers - Practical arrangements on the companion diagnostics consultation procedure
215 to the European Medicines Agency by notified bodies” [10].
216 3.2. Legal basis and relevant guidelines
217 This guideline should be read in conjunction with the introduction and general principles of Annex I to
218 Directive 2001/83/EC, as amended, all other relevant EU legislation, and EU and International Council
219 for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) guidelines.
220 These include, but are not limited to:
221 EMA guidelines and relevant documents:
222 • Guidance on the procedural aspects for the consultation to the European Medicines Agency by
223 a notified body on companion diagnostics (EMA/198592/2022)
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 6/26
第 7 页
224 • Guideline on clinical evaluation of diagnostic agents (CPMP/EWP/1119/98/Rev. 1)
225 • Guideline on good pharmacogenomic practice (EMA/CHMP/718998/2016)
226 • NOTICES FROM EUROPEAN UNION INSTITUTIONS, BODIES, OFFICES AND AGENCIES 2015/C
227 421/03 Council conclusions on personalised medicine for patients Guideline on the clinical
228 evaluation of anticancer medicinal products Rev.6. (EMA/CHMP/205/95)
229 • Guideline on the investigation of subgroups in confirmatory clinical trials
230 (EMA/CHMP/539146/2013)
231 • Qualification of novel methodologies for drug development: guidance to applicants
232 (EMA/CHMP/SAWP/72894/2008)
233 • Scientific guidelines with SmPC recommendations EMA/813125/2012 rev. 8
234 (EMA/CHMP/483022/2019)
235 • Wording of therapeutic indication - A Guide for Assessors of Centralised Applications
236 (EMA/CHMP/483022/2019)
237 • Reflection Paper on methodological issues in confirmatory clinical trials planned with an
238 adaptive design (CHMP/EWP/2459/02)
239 • Frequently asked questions on medicinal products development and assessment involving
240 companion diagnostic (CDx) (EMA/CHMP/821321/2022)
241 • Questions & Answers - Practical arrangements on the companion diagnostics consultation
242 procedure to the European Medicines Agency by notified bodies (EMA/619893/2022)ICH M10
243 guideline on bioanalytical method validation and study sample analysis, Step5
244 (EMA/CHMP/ICH/172948/2019)
245 ICH guidelines:
246 • ICH guideline M10 on bioanalytical method validation and study sample analysis - Questions
247 and Answers (EMA/CHMP/ICH/660315/2022)
248 • ICH guideline E18 on genomic sampling and management of genomic data, Step 5
249 (EMA/CHMP/ICH/11623/2016)
250 • ICH E6 (R3) Guideline for good clinical practice (GCP) - Step 5 (EMA/CHMP/ICH/162879/2026)
251 European Commission guideline and legislation:
252 • REGULATION (EU) 2017/746 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 5 April
253 2017 on in vitro diagnostic medical devices and repealing Directive 98/79/EC and Commission
254 Decision 2010/227/EU (https://eur-lex.europa.eu/eli/reg/2017/746)
255 • EUROPEAN COMMISSION, NOTICE TO APPLICANTS, Revision 2 - A GUIDELINE ON SUMMARY
256 OF PRODUCT CHARACTERISTICS (SmPC)
257 Medical Device Coordination Group (MDCG) guidance and relevant documentation:
258 • Guidance on general principles of clinical evidence for In Vitro Diagnostic medical devices
259 (IVDs) (MDCG 2022-2)
260 • Guidance on Classification Rules for in vitro Diagnostic Medical Devices under Regulation (EU)
261 2017/746 (MDCG 2020-16)
262 • Questions & Answers regarding performance studies of in vitro diagnostic medical devices
263 under regulation (EU) 2017/746 (MDCG 2025-5)
264 • Q&A on the interface between Regulation (EU) 536/2014 on clinical trials for medicinal
265 products for human use (CTR) and Regulation (EU) 2017/746 on in vitro diagnostic medical
266 devices (IVDR) (MDCG 2022-10)
267
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 7/26
第 8 页
268 4. Context of use of biomarker assays during medicinal
269 product development
270 The context of use (CoU) of a medicinal product with a biomarker refers to the specific circumstances,
271 patient population, and purpose for which the biomarker is used, including its role in medicinal product
272 development. It includes the (proposed) indication of the medicinal product, the intended purpose of
273 the IVD, the target population, and conditions for use.
274 Data supporting the context of use should be submitted at MAA including support for the reliability of
275 the biomarker assay(s), which is also an obligation from the Good Clinical Practice (GCP) and IVDR
276 perspectives:
277 • ICH E6 (R3) Guideline for GCP [11] mandates that clinical trials are designed to protect the
278 rights, safety and well-being of participants and to ensure the reliability of results. In
279 particular, it is stated in ANNEX 1 (3.16.1) that "The [medicine] sponsor should ensure that
280 data acquisition tools are fit for purpose and designed to capture the information required by
281 the protocol. They should be validated and ready for use prior to their required use in the
282 trial”.
283 • The IVDR also requires that prior to a clinical performance study of an IVD, the analytical
284 performance is specified in the clinical performance study plan (Annex XIII 2.3.2 (k) [3]). The
285 analytical performance also needs to be described in the investigator’s brochure to be prepared
286 for a clinical performance study (Annex XIV section 2.3 [3]).
287 In principle, there are three potential stages of development for (predictive) biomarker assays used in
288 clinical medicinal product development, but not all stages are sequential or mandatory:
289 Stage A) Exploratory/research setting
290 A biomarker assay is used for early exploration of a biomarker signal in association with a medicinal
291 product or a medicinal product combination (e.g., by applying retrospective analysis). In a clinical trial
292 setting, patients could be stratified on the basis of the biomarker but there would be no change in the
293 treatment that they would receive on the basis of the assay results. In the context of the clinical trial,
294 the assay would not fulfil the IVD definition according to the IVDR [9]. This setting is usually linked to
295 early-phase clinical trials to potentially guide future patient selection, dose selection, or understanding
296 the mechanism of action. This type of assay may later become an IVD – either an “early candidate
297 CDx” (stage B) or a “mature candidate CDx” (stage C) if it proves to be robust, reliable, and clinically
298 meaningful.
299 Stage B) IVD/“early candidate CDx” setting
300 A biomarker assay with an established analytical validity (analytical performance) and with a
301 hypothesis that the biomarker is important in relation to the medicinal product (for example as a result
302 of a study under the previous stage), can be used to support a clinical decision, for instance to include
303 or exclude patients into a clinical trial, or to allocate them to treatment arms. This setting likely fulfils
304 the definition of an IVD according to the IVDR [3, 4]. However, at this stage of development it is still
305 uncertain if a biomarker-guided medicinal product development is justified. In cases where the clinical
306 trial data confirm the biomarker hypothesis, the IVD can be termed an “early candidate CDx”. The
307 clinical trial data (considering patients receiving and not receiving the treatment) and the analytical
308 (analytical performance) and clinical validation (clinical performance) of the IVD, may guide whether
309 the IVD/early candidate CDx may be developed towards a ”mature candidate CDx” (stage C).
310 Stage C) IVD/“mature candidate CDx” setting
311 At this stage, a biomarker assay with an established analytical validity (analytical performance) is used
312 in the pivotal clinical trial(s) of a medicinal product in order to support a biomarker-guided use of the
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 8/26
第 9 页
313 medicinal product itself. This assay can therefore be considered as a “mature candidate CDx”. In the
314 mature candidate CDx setting, the clinical validity in the context of the medicinal product development
315 and the predictive biomarker hypothesis are to be confirmed. Co-development of the biomarker assay
316 with the concerned medicinal product in a clinical development programme may occur for an initial
317 MAA, but may also be required for extensions of indications post-approval MAA as a predictive
318 biomarker assay is treatment and disease dependent. In the recommended co-development setting,
319 the biomarker assay used in the pivotal clinical trial(s) (i.e., the mature candidate CDx), may be the
320 subject of the conformity assessment according to the IVDR for its certification as a CDx device.
321 The three different stages illustrate that decisions on predictive biomarker assay/candidate CDx
322 development may evolve during medicinal product development. Additionally, the suitability of the
323 predictive biomarker assay/candidate CDx as a CDx device will be assessed at a later stage during the
324 conformity assessment under the IVDR provisions.
325 In cases, where a candidate CDx has not been co-developed with the medicinal product, it is essential
326 that a CDx candidate proposed for conformity assessment has a similar performance as the biomarker
327 assay used in the pivotal clinical trial(s). This requires the bridging of analytical data to assure that the
328 patient population identified by the proposed CDx candidate is comparable to the patient population
329 originally selected by the pivotal biomarker assay. This ensures that the safe and effective use of the
330 predictive biomarker-guided medicinal product as described in the SmPC is not impacted. For further
331 information regarding bridging studies and requirements, please refer to section 9 “Bridging studies” of
332 this guideline.
333 5. Scientific rationale for predictive biomarkers and their
334 corresponding assays
335 In the context of medicinal product regulation, the concept of scientific rationale is similar to scientific
336 validity in the IVDR. The successful development of a predictive biomarker assay depends on multiple
337 factors. One that should be critically considered is the scientific rationale for the predictive role of the
338 biomarker measured by the corresponding assay to determine the likelihood of benefit (or risk) from a
339 treatment with a specific medicinal product in a specific target population. In this respect, the
340 characteristics of the biomarker, such as its prevalence in the target patient population, relationship to
341 disease pathogenesis and treatment effect, as well as the mechanistic rationale for its use, should be
342 thoroughly explored in the context of the associated medicinal product development.
343 Generally, a biomarker may have a simultaneous diagnostic, prognostic, and, in some cases predictive
344 role (refer to section 12 for terminology). Therefore, biomarker measurement can potentially serve
345 different purposes at different timepoints of patient management. In the context of treatment with a
346 specific biomarker-guided medicinal product, the predictive role of a biomarker is of particular
347 relevance.
348 It is important to distinguish between the different roles that the same biomarker may have. For
349 example, a biomarker may initially be used to detect or confirm the presence of a disease or condition,
350 or to identify individuals with a specific disease subtype. At a later stage, the same biomarker may also
351 have a predictive role by identifying patients who are more likely to respond to, or benefit from, a
352 specific medicinal product for that disease or disease subtype.
353 For the development of predictive biomarker assays, the scientific rationale needs to distinguish
354 between the prognostic and predictive roles for the same biomarker, and the rationale for medicinal
355 product development in a specific patient population.
356 The link between the assumed therapeutic action of the medicinal product and the role of the
357 biomarker(s) in relation to the targeted therapeutic indication should be clearly outlined (refer to
358 section 4 and 12 of the guideline). For any single biomarker or multi-marker panel, mechanistic and
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 9/26
第 10 页
359 biological plausibility should be justified. In most cases, clinical data are necessary to substantiate the
360 clinical impact of biomarker measurement.
361 In the case of novel biomarkers not previously used for patient selection, or where a scientific rationale
362 exists but the current level of knowledge is insufficient to consider the biomarker as established, the
363 patient selection strategy used in confirmatory trials should be adequately justified in the MAA dossier.
364 This should be based on the available non-clinical/translational data, the mechanistic/biologic rationale,
365 and, where available, clinical data. The scientific rationale (scientific validity) always need to be well
366 justified when a biomarker is used to include or exclude patients from confirmatory clinical trial(s). In
367 such cases, no data are collected for the subpopulation(s) excluded from the clinical trial(s), and this
368 may hamper the MAA assessment of the biomarker’s predictive value.
369 Choice of assay
370 As a chosen biomarker can be measured by different assay types (e.g. genomic/genetic versus
371 proteomic formats), the rationale for the selected assay needs to be justified.
372 Cut-offs
373 Biomarker assays often provide a qualitative result classifying subjects into two or more groups.
374 However, qualitative results are based on an underlying quantitative variable, which plays a crucial role
375 in establishing the cut-offs between qualitative classifications (e.g. the gene variant allele frequency,
376 the number of ligands (soluble proteins), or the number of cells expressing a certain protein on a
377 tissue slide). The underlying rationale for the selected cut-off that defines a patient as having a
378 positive biomarker status needs to be provided. Particularly, it should be clarified whether the cut-off is
379 determined by the analytical sensitivity of the assay (e.g., the limit of detection) or by the quantitative
380 relationship between the biomarker and clinical outcomes. The correlation of the chosen cut-off with
381 clinical outcome(s) needs to be shown (for details refer to section 8 of the guideline).
382 The cut-off may also be based on approaches that include multiple parameters. For example, complex
383 multi-parameter scoring algorithms are typically applied for semi-quantitative tissue-cut-slide assays
384 like immunohistochemistry (IHC)-based technologies.
385 Considerations for multi-marker panels
386 In clinical studies, biomarker status may be determined by a single marker or by multiple markers
387 (e.g. multiple markers that encompass a panel of genetic or genomic aberrations within one or several
388 genes). Multiple markers can be measured using distinct assays or a single assay. The chosen multi-
389 marker approach used to determine the biomarker status of the patient should be defined and
390 scientifically justified.
391 Changes of the cut-off point and/or multi-marker panel
392 During clinical development, changes may be made to the assay or IVD. For example, the selected cut-
393 off may need to be changed, or a multi-marker panel may be modified (e.g., by the addition or
394 deletion of specific genetic or genomic aberrations). Changes require a sound scientific justification,
395 together with an analysis and discussion of the reliability of the conclusions that can be drawn from the
396 modified approach. Of note, from the IVDR perspective, such changes would also constitute changes to
397 the IVD/candidate CDx.
398 6. Analytical validation of biomarker assays during medicinal
399 product development
400 In section 6, only the term “samples”, which is equivalent to specimens in the IVDR, is used
401 throughout [3].
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 10/26
第 11 页
402 6.1. General aspects
403 Analytically validated assays should be used in clinical trials to fulfil the requirements by GCP ICH E6
404 (R3) Annex 1 (3.16.1) [11]. Beyond GCP requirements, more stringency on analytical assay validation
405 is expected in situations where the reliability of the biomarker assay may impact patient safety, e.g., if
406 the biomarker assay is used to select patients to be enrolled, guides the assignment of treatments
407 (stage B in section 4 of the guideline), or where a clinical trial is designed to validate the predictivity of
408 a biomarker (stage C).
409 Expectations regarding the extent and stringency of analytical validation for the biomarker assay
410 increase from the research stage (stage A) to the IVD stages (stages B and C). Analytical validation
411 acceptance criteria should be predefined and appropriately justified, proportionate to the stage of
412 development and to the risks and benefits of the assay. Analytical validation at the research stage
413 (stage A) focuses on establishing the suitability of the biomarker assay’s use based on preliminary
414 acceptance criteria. In contrast, when the biomarker assay has a medical purpose (i.e., fulfils the
415 definition of an IVD), typically used in the later or pivotal stages of medicinal product development
416 (stages B and C), analytical validation involves confirmation of the assay’s suitability for the predefined
417 validation attributes and acceptance criteria.
418 6.2. Assay considerations
419 Specific recommendations for analytical validation of different biomarker assay formats are beyond the
420 scope of this guideline, given their diversity (e.g., genomic, genetic, proteomic, other omic platforms)
421 as well as the different grades of complexity of assays (e.g., more simple ligand binding assays versus
422 highly complex IHC assays) currently used in biomarker-guided medicinal product development.
423 Moreover, these technologies are constantly being improved (e.g., genomic variant and mutation
424 analysis evolved rapidly from Sanger sequencing and polymerase chain reaction (PCR)-based
425 approaches to next-generation sequencing (NGS) technologies). Therefore, it is recommended to
426 consider general guidelines for common scenarios instead of specific recommendations that might
427 become outdated. In particular, the principles stated in the ICH M10 guideline on bioanalytical method
428 validation and on study sample analysis [12], together with the corresponding Questions and Answers
429 document [13], should be considered where applicable. For NGS-based assay platforms used for the
430 detection of germline mutations, recommendations in the guideline on good pharmacogenomic practice
431 should be followed [14]. For IVD/candidate CDx, the IVDR requirements of Annex I section 9.1(a) shall
432 be considered [3].
433 With respect to the use of an analytically validated biomarker assay (IVD/candidate CDx stages B and
434 C) in the context of a clinical trial, several general analytical rules apply. These requirements are
435 comparable in both the medicinal product framework and the IVDR, and include, for example, the
436 known analytical validation attributes of sensitivity, specificity, precision (repeatability and
437 reproducibility, e.g. inter-/intra-assay, operator-to-operator, lot-to-lot) and cut-offs. For quantitative
438 assays, the derivation of the linear range, including controlling for known relevant interference, and
439 limits of detection, are applicable [3, 12, 13]. When available, internationally recognised standard
440 materials, should be included in the characterisation of the assay. For analytical assay validation
441 purposes, typically biobank samples are used, and these samples should be sufficiently representative
442 of the future target population and should target the cut-off, if applicable. It is considered essential
443 that the development of the assay (e.g., cut-off selection or scoring rule for combined biomarkers) is
444 based on a different dataset from that used for final analytical validation. Otherwise, analytical
445 performance may be overestimated. A cut-off which is not based only on analytical assay sensitivity
446 (e.g., limit of detection (LOD)) may also be selected. In such cases, the clinical cut-off which optimises
447 the predictive clinical performance should be derived from a provisional cut-off from early clinical trials
448 (e.g., stages A or B) and subsequently be confirmed in the confirmatory clinical trial(s) (stage C).
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 11/26
第 12 页
449 6.3. Considerations for specific assays
450 If tissue-based assays (cut-slide assays, e.g., IHC assays, or fluorescence in situ hybridisation (FISH)
451 assays) are used, separate analytical validation studies should be performed to confirm the suitability
452 of each intended tissue type as specified in the (proposed) clinical indication (e.g., to evaluate tissue
453 histology-dependent matrix effects).
454 The use of innovative biomarker assay technologies or formats in medicinal product development may
455 require confirmation of the suitability of the method by a recognised and analytically validated state-
456 of-the art method. In such cases it is recommended to seek scientific advice from an NCA or EMA, or to
457 consider an EMA Qualification Opinion on the suitability of novel methodologies for medicinal product
458 development [15].
459 6.4. Characterisation studies
460 Additional characterisation studies may be needed to confirm the CoU of the biomarker assay. These
461 are illustrated via the oncology setting and may be applicable to other contexts.
462 (1) Tissue heterogeneity is often a critical feature of most solid cancers. Samples may consist of a
463 mixture of tumour tissue and accompanying benign or normal tissue which impacts the reliable
464 detection of the biomarker (panel) and may lead to false negative testing results. Therefore,
465 especially for cut-slide tissue assays like IHC or fluorescence in situ hybridisation (ISH)/FISH,
466 tissue heterogeneity should be analysed to confirm the biomarker status. Equally, for the
467 detection of genetic aberrations such as somatic mutations, e.g. by NGS or PCR assays, high
468 levels of tumour tissue heterogeneity may undermine the validity of pooled results, and the
469 combined estimate may not accurately represent individual outcomes. Characterisation studies
470 to evaluate the biomarker mutation status in heterogeneous tumour tissue samples may follow
471 different approaches. Characterising the full range of molecular heterogeneity within a tumour
472 tissue sample is essential to ensure that the genetic biomarker profile of a patient is accurately
473 reflected. Depending on the outcome of such studies, it may be required to define a minimum
474 threshold of tumour content in the sample or to include initial tumour cell isolation (e.g., by
475 microdissection technology) in the assay platform strategy.
476
477 (2) Another scenario is present in the metastatic tumour setting which may require the
478 assessment of the biomarker status between the primary tumour versus metastatic tumour
479 samples.
480
481 (3) In case of liquid biopsy (e.g., measuring the biomarker(s) in circulating tumour cells,
482 circulating free deoxyribonucleic acid (DNA), or cell free micro ribonucleic acid (microRNA)),
483 the concordance of the biomarker status between the primary tumour tissue versus liquid
484 biopsy samples should be evaluated.
485
486 (4) For some very rare rapidly growing tumours, such as glioblastomas, a stringent
487 characterisation of the suitable window of measurement is required to counteract the high risk
488 of false testing results. Generally, the acceptable window for patient testing should be justified
489 based on the persistence of the biomarker tested (e.g., human leukocyte antigen (HLA) on
490 normal tissue versus markers of infection or tumour markers).
491
492 (5) Where archival samples are used (e.g., for evaluation of the relationship of the biomarker
493 status in archival tumour tissue samples versus fresh liquid biopsy samples (see (2)) or
494 compared to fresh metastatic tumour samples (see (3))), justification that these samples
495 match the biomarker samples of patients in the intended clinical indication should be provided.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 12/26
第 13 页
496 6.5. Assay adaptations
497 A biomarker assay might be further adapted during clinical medicinal product development. In case of
498 substantial changes to the biomarker assay, the maintenance of analytical (and clinical) performance
499 in the context of the use of the medicinal product must be ensured, and respective data need to be
500 established to confirm this. Analytical performance differences, also in cases of assay improvement,
501 need to be justified in terms of maintained performance with regards to the use of the medicinal
502 product.
503 Specifically, a candidate CDx might be further adapted during clinical development to optimise patient
504 selection (e.g., changing the exact composition of a biomarker panel by adding and/or deleting
505 genomic aberrations, or changing the cut-off). It is also essential to ensure that datasets used to
506 derive an adapted assay are distinct from the dataset(s) used for its analytical validation. If a single
507 dataset is split into a training set for the derivation of the optimised assay and a test set for the
508 analytical validation, this approach must be pre-specified and transparently documented.
509 For the MAA of a biomarker-guided medicinal product, it is essential to connect all clinical data
510 collected. Therefore, if the methodological approach or the candidate CDx assay platform has
511 undergone changes, adequate justification on the continuity of earlier data relevance needs to be
512 provided in the dossier. This exercise may also require supportive bridging data for the different assay
513 versions used throughout clinical development. In cases where clinical samples are limited or not
514 available, the alternative use of biobank samples and their suitability must be justified.
515 For further details, refer also to section 9 “Bridging studies” of the guideline.
516 6.6. Considerations for centralised versus decentralised testing
517 In the clinical development setting, two options are generally available, i.e., sample testing with a
518 single assay performed in a single testing laboratory (centralised testing) and sample testing using the
519 same or different assays in different laboratories (decentralised or local testing). Each approach has
520 specific benefits and limitations with regards to practicality, regulatory implications and potential
521 impact on data quality and reliability. The decision for one or the other approach will be largely
522 determined by the CoU and the available analytical options for measuring the biomarker. On one end
523 of the spectrum is a setting where a Conformité Européenne (CE)-marked IVD/CDx for local patient
524 testing is available, which would facilitate the decentralised approach. On the other end of the
525 spectrum is a highly complex or novel analytical approach which requires specialised expertise and
526 thus fosters the preference for central testing. And in between, there is a wide range of possible
527 scenarios.
528 The centralised approach has the benefit of data generation within a consistent testing environment;
529 however, it requires the transfer of samples from local sites to the central testing facility. The transfer
530 may impact sample stability and storage conditions requiring additional sample transport validation.
531 The timeframe of the results availability is influenced by both the sample shipping time and processing
532 time at the central facility.
533 The most important advantage for decentralised testing is the potentially faster availability of results
534 for both the treating physicians and the patients, and it may be crucial in cases of unstable samples.
535 However, this approach also presents risks related to the consistency of assay performance across
536 different laboratories. For instance, where different assays are used locally, comparable assay
537 performance needs to be ensured for the consistent determination of biomarker status. This may
538 include bridging studies and/or consideration of round-robin tests. Therefore, in decentralised settings,
539 the use of the same assay across all sites is strongly recommended, or if not operationally feasible,
540 similar assay technology with comparable performance characteristics must be considered.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 13/26
第 14 页
541 If a panel where multiple markers are used to determine the biomarker status of a patient, then a
542 uniform definition of the set of markers across all sites is required.
543 A complex analytical assay platform (e.g. cut-slide assays, like IHC) is associated with specific
544 challenges in ensuring adequate technical expertise at local sites and hence achieving reproducible
545 results across the local laboratories. Therefore, in such cases, centralised testing is generally
546 recommended to ensure consistency and reliability of study data.
547 Candidate CDx aspects
548 If it is decided to develop a candidate CDx (stages B and C), it might be more challenging to (co-
549 )develop a suitable candidate CDx when testing is performed in a decentralised rather than a
550 centralised setting, as the assays used across the different testing laboratories would need to be
551 compared to ensure equal performance. Therefore, in a decentralised scenario the use of the same
552 assay as candidate CDx across the different testing laboratories is the recommended approach. In
553 cases where different local assays are used, a candidate CDx status may be assigned to one of them,
554 or, alternatively, to an additional (novel) assay platform. For both scenarios, a prospectively defined
555 strategy for assuring comparable performance characteristics between the candidate CDx and different
556 local assay platforms should be in place. For example, testing results of different local assays may be
557 confirmed by a (central) candidate CDx.
558 At the time of the marketing authorisation, it is essential to ensure that reliable and robust data are
559 available to support the B/R assessment of the corresponding medicinal product, especially when
560 decentralised testing was applied. Due to the complexity and variety of decentralised testing
561 approaches, developers may also consider seeking scientific advice from EMA.
562 Additional considerations for analytical suitability of biomarker assays during medicinal
563 product development beyond the EU
564 During a global development programme of a biomarker-guided medicinal product and corresponding
565 biomarker assay(s), clinical studies may be performed within the EU, partially within and outside the
566 EU, or entirely outside the EU. While EU legislation applies to clinical studies conducted in the EU, GCP
567 compliance and assays being fit for purpose are generally required. Challenges in consistency of safety
568 and efficacy data may arise from variability in (patient) populations, analytical assay performance,
569 and/or testing laboratory accreditation standards.
570 Therefore, in order to ensure consistency of the testing results during clinical development, it is
571 strongly recommended that patient samples are centrally tested, regardless of their origin (EU or non-
572 EU). Where testing is conducted outside the EU, analytically validated assays should be used in a
573 laboratory compliant with accepted standards, such as ISO15189 or CAP/CLIA.
574 In situations where centralised testing may not be feasible, it is recommended that the same assay, or
575 at least a similar assay technology with comparable performance characteristics, is used across all
576 concerned local accredited testing laboratories.
577 Generally, the recommendations provided in section 6.6 “Considerations for centralised versus
578 decentralised testing” of the guideline also apply in a global testing setting, particularly with respect to
579 aspects relevant to a candidate CDx. Also, in a global setting, the assurance of reliable and robust
580 testing data generated in clinical studies within and outside the EU must be confirmed at the time of
581 MAA to support the B/R assessment of the associated biomarker-guided medicinal product.
582 7. Samples for biomarker analysis during medicinal product
583 development
584 In section 7, the term “samples”, which is equivalent to specimens in the IVDR, is used throughout [3].
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 14/26
第 15 页
585 Samples for biomarker analyses are collected at all stages of medicinal product development. Different
586 sample types may be collected for analysis: blood and blood fractions (plasma, serum, buffy coat, red
587 blood cells), tissue biopsies (from surgery, autopsy, transplant), urine, saliva/buccal cells, and any
588 other types of biospecimens (e.g. breast milk, cell lines, semen, ascites, hair). Sample types relevant to
589 the context of use should be collected, processed, and stored under conditions that preserve their
590 stability and reliability with respect to the intended downstream analyses. To ensure sample quality
591 that supports decision-making, the general principles for sample collection, processing, transport,
592 storage, and disposition, as outlined in the ICH E18 guideline, should be followed [16]. According to
593 ICH E6(R3) on GCP [11], biological samples may not be collected or stored from any subject without
594 first obtaining informed consent based on adequate information regarding the purpose, potential risks,
595 and intended use of the samples (including information on potential biobanking).
596 The source and quality of the sample for biomarker analyses performed in confirmatory studies should
597 be appropriately justified in relation to type and purpose, because they may be obtained under variable
598 conditions (leading to uncertain sample quality) and the time between sample collection and study
599 treatment initiation can vary widely. Freshly obtained samples collected using standardised procedures
600 for collection and processing are generally preferred.
601 7.1. Pre-screening
602 Screening procedures (i.e., relevant procedures actively evaluating potential participants for enrolment
603 in clinical trials) are an integral part of clinical development and are described in the clinical trial
604 protocols. Generally, pre-screening refers to the pre-selection of potential trial participants based on
605 archival data that were collected independently of the clinical trial data and are already available to the
606 investigator, sometimes prior to obtaining informed consent for participation in the trial.
607 In the context of biomarker assay development, pre-screening can lead to so-called pre-screening bias.
608 Pre-screening bias can occur, for example, when a local test result is used to include or exclude
609 patients from the planned screening procedures. Patients whose samples are considered biomarker-
610 positive by a local test are usually retested using a central test (which might be the candidate CDx) and
611 will be included in the trial only if biomarker positivity is confirmed by this central test. However,
612 patients with biomarker-negative results based on the local test will be excluded from the trial. This
613 may lead to the underrepresentation of patients whose samples are negative by the local test but
614 positive when assessed using the central test. Therefore, biomarker-based pre-screening (selection)
615 should generally be avoided.
616 Furthermore, it is encouraged to thoroughly analyse the prevalence of the biomarker based on
617 epidemiological data, to document pre-screening procedures, and to test (to the extent possible) a
618 proportion of local test-negative samples using the candidate CDx to estimate the magnitude of the
619 potential bias. Pre-screening bias may be estimated by comparing the treatment effect in patients with
620 and without available pre-screening tests. While pre-screening may be inevitable in instances of very
621 low prevalence of biomarker-positive populations in the context of large-scale genome profiling (e.g.
622 tumour samples), estimating the proportion of potentially underrepresented populations is highly
623 recommended. In any scenario, the pre-screening processes and the number of patients screened in a
624 study need to be accurately and thoroughly documented to understand the potential for bias.
625 7.2. Baseline samples
626 A baseline sample refers to the initial sample taken for measurement from a participant before the start
627 of the intervention or treatment in a clinical trial. Baseline sample collection and analysis is crucial for
628 the investigation of the impact and value of the biomarker(s) of interest. The timing of baseline sample
629 collection (e.g., tissue biopsies) should generally occur as close as possible to the initiation of study
630 treatment, preferably during the screening period, following an appropriate wash-out period after any
631 prior therapy, in order to minimise residual treatment effects and avoid confounding the interpretation
632 of baseline data. Additional aspects to consider regarding source and timing of sampling include intra-
633 patient variability in biomarker expression, which can impact on the performance of the biomarker, as
634 well as the temporal variation in biomarker levels over time.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 15/26
第 16 页
635 Consideration should be given to the need to retain all samples collected which may be required for
636 future use during the course of product development. In addition to samples from patients enrolled in
637 the trial, samples of patients who were excluded from the trial due to their biomarker status may be
638 valuable for future bridging studies. For further information regarding bridging, refer to section 9
639 “Bridging studies” of the guideline.
640 7.3. Missing data
641 During clinical development, collection of biomarker samples (specimens) should be as complete as
642 possible, according to the planned sampling schedule. Missing samples are more commonly
643 encountered when biomarker testing is performed at a timepoint distant from baseline sample
644 collection, as is often the case in retrospective analyses. Reasons for missing samples should be
645 adequately documented, and any potential bias should be investigated [17]. There can be causes of
646 missing data other than missing samples.
647 In general, the eligibility criteria for patient enrolment, should outline an inclusion requirement of
648 baseline biomarker data. Exceptions (where this might not be possible) must be well justified. In any
649 case, the population with a biomarker result needs to be representative of the overall study population
650 and this should be demonstrated considering that missing data may not occur at random but may be
651 related to other factors such as disease state (e.g., tumour size) or operational reasons (e.g., non-
652 compliant study site) with a large potential for study results bias. Since biomarker data is used to
653 select/exclude patients from a treatment (sub)group, statistical methods to handle missing biomarker
654 data and addressing potential bias need to be pre-specified in the analysis plan.
655 8. Clinical validation of biomarkers and validation of
656 biomarker assays
657 In the guidance below, the term B/R is related to the B/R of medicinal products and is used as applied
658 in the medicines’ framework.
659 A biomarker may be predictive of the efficacy or safety outcomes of a medicinal product. This implies
660 that the indication of the medicinal product should cover the biomarker-defined target population with
661 a positive B/R balance. The predictive ability of the biomarker(s) for patient clinical outcome may
662 initially be hypothesised prior to clinical development based on the mechanism of action and may
663 progressively evolve, becoming more evident as clinical development advances.
664 The aim of predictive biomarker-guided medicine development is to (co-)develop a medicinal product
665 along with (an) appropriate IVD(s) able to identify the patient population that would benefit from the
666 medicinal product, or that has a lower risk of serious adverse reactions. Consequently, a predictive
667 biomarker IVD can be considered clinically valid if it separates patients with a favourable B/R profile
668 from patients with an unfavourable or clinically meaningfully worse B/R profile.
669 Consequently, in order to clinically validate a predictive biomarker IVD, it needs to be shown that the
670 biomarker status, as measured by the IVD, is predictive of a clinical outcome relevant for the B/R
671 assessment. This is demonstrated by evidence of a differential treatment effect according to biomarker
672 status, i.e. by showing that a biomarker-by-treatment interaction exists. Generally, such an
673 investigation requires data from the overall patient population, rather than from a population pre-
674 selected based on biomarker status, as this allows comparison of the treatment effect between
675 biomarker-positive and biomarker-negative patients. Without inclusion of biomarker-negative patients,
676 the predictive effect of the biomarker may not be elucidated. Therefore, biomarker-negative patients
677 should be included in the clinical development programme whenever possible.
678 The treatment-by-biomarker interaction may be investigated during different development steps of the
679 medicinal product, serving different purposes (exploratory or confirmatory). During early clinical
680 development, the interaction can be investigated in an exploratory way to identify potential signals, or
681 to propose cut-off value(s) for the biomarker assay that serve to generate hypotheses for the next
682 development steps. (Potential) surrogate endpoints (e.g. pharmacodynamic endpoints or endpoints
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 16/26
第 17 页
683 indicating clinical activity) that may allow prediction of the interaction for later clinically relevant
684 outcomes might be investigated at this stage. The findings regarding the predictive value of the
685 biomarker and the proposed cut-off would be usually considered exploratory and need to be validated
686 (or confirmed) in the subsequent development step(s). In late clinical development, the investigation of
687 the treatment-by-biomarker interaction should serve to clinically validate the biomarker assay. When
688 designing the pivotal trial(s) for the medicinal product, which also serves to generate the pivotal
689 evidence for the IVD, the following scenarios are of relevance:
690 (1) Biomarker-stratified
691
692 There is a prior rationale that the B/R ratio may be improved in the biomarker-defined target
693 population, but it cannot be excluded that the B/R may also be positive in a broader patient
694 population, or there is uncertainty whether the selected cut-off is appropriate. Therefore, both
695 biomarker-positive and biomarker-negative patients are selected for the study. This allows for the
696 investigation of the clinical validity of the biomarker assay by analysing the treatment-by-biomarker
697 interaction according to a pre-specified plan. Randomisation of patients should generally be
698 stratified by (appropriately defined) biomarker status aiming to ensure adequate balance in the
699 experimental and control arms, which is particularly important for small studies and/or rare
700 biomarkers. The results may (or may not) support that biomarker status is predictive of a
701 favourable B/R balance, i.e. that the biomarker assay is (or is not) clinically valid for patient
702 selection. Of note, the absence of an interaction cannot be concluded based on a non-significant
703 interaction. And vice versa, even if there is clear evidence of an interaction, this does not
704 necessarily imply a negative B/R in patients with a negative biomarker status (who may simply
705 have a smaller overall benefit). Indeed, as a treatment effect in the broader patient population may
706 be driven by the effect in the patients with positive biomarker status, subgroup analyses can help
707 clarify the treatment effect in the disjoint subgroups of patients with positive and negative
708 biomarker status. For continuous biomarker assays, this means that subgroup analyses based on
709 the selected cut-off - or disjunct biomarker subgroups in case of multiple pre-specified cut-offs
710 (e.g., biomarker below cut-off A, biomarker between cut-off A and cut-off B, and biomarker above
711 cut-off B) - should be provided.
712
713 (2) Biomarker-targeted
714
715 There is a strong prior rationale that the proposed treatment has a positive B/R ratio only in the
716 biomarker-defined target population. Therefore, only patients with positive biomarker status are
717 included in the study that serves as the basis to support clinical validity. For continuous biomarkers,
718 these would be patients above (or below) a specific cut-off. While it is possible to confirm a positive
719 B/R balance in patients with positive biomarker status, the biomarker-by-treatment interaction
720 cannot be investigated but assumes no (sufficient) treatment effect in the patients with negative
721 biomarker status. This approach requires a strong scientific rationale, usually based on previous
722 clinical or non-clinical data. As only patients above or below a specific cut-off are included, a clinical
723 justification of the cut-off is particularly important.
724 The assumption at the planning phase of the study to investigate the B/R profile of the medicinal
725 product can also be that the B/R in the target population is independent of the biomarker status.
726 However, exploratory subgroup analysis to investigate the consistency of the treatment effect across
727 the overall population could show a strong treatment-by-biomarker interaction which suggests that B/R
728 may be positive only in a biomarker-defined target population. The claim of a biomarker assay being
729 predictive on this basis may be acceptable only after careful assessment of the credibility of such a
730 finding, particularly if the B/R balance in the overall population is unconvincing and the intention is to
731 identify a subpopulation with a convincing B/R (refer to Guideline on the investigation of subgroups in
732 confirmatory clinical trials [18]). In many cases, replication of such findings will be required for
733 confirmation of clinical validation. However, if there is a concern that the B/R may be negative in
734 patients with a specific biomarker status, the possibility that the biomarker is predictive should not be
735 disregarded even if uncertainties remain (precautionary principle).
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 17/26
第 18 页
736 In case there are sub-entities of the biomarker assay (e.g., different mutations in multi-marker
737 panels), explorative subgroup analyses should be provided to support that the biomarker assay is
738 predictive across the entire population of patients with positive biomarker status. Adaptive two-stage
739 designs of a clinical trial that include an interim analysis can allow for timely restriction of the study to
740 only biomarker-positive patients while generating data in biomarker-negative patients, thereby limiting
741 exposure of patients to a potentially ineffective or unsafe treatment in case of scenario (1) above. A
742 decision on the cut-off could be made in the first stage and confirmed in the second stage. However,
743 specific considerations that apply for design and analysis of adaptive studies need to be taken into
744 account, including appropriate pre-specification and multiplicity control, which can be read in the
745 Reflection Paper on methodological issues in confirmatory clinical trials planned with an adaptive design
746 [19].
747 Single arm trials are usually not suitable to establish that a biomarker is predictive because, in such a
748 setting, predictive and prognostic roles cannot be disentangled without additional assumptions. When a
749 randomised controlled trial is not feasible, seeking scientific advice is recommended.
750 Validation of cut-offs
751 The cut-off of the biomarker assay that differentiates between patients with a positive biomarker status
752 and negative biomarker status needs to be provided. Even if the biomarker assay provides a qualitative
753 result (usually binary, i.e. biomarker present or absent), there is usually an underlying quantitative
754 variable (refer to section 5 of the guideline). The clinical validity of the chosen cut-off needs to be
755 justified.
756 The choice of the cut-off for a biomarker assay should be guided by its predictive effect, i.e. the aim
757 should be to select it in a way that it differentiates patients with a positive B/R from those with a
758 negative B/R. This could be based on an exploration of the quantitative relationship between the
759 biomarker and clinical outcomes, which may be confirmed in subsequent steps. Note, when the
760 biomarker is also prognostic, the cut-off for a prognostic effect will usually not be identical to the cut-
761 off for a predictive effect. If the cut-off is exclusively based on analytical sensitivity of the assay (e.g.
762 the LOD), but is not associated with clinical validity, it remains uncertain whether all patients with a
763 positive biomarker status can be expected to benefit from treatment.
764 In scenario (1) above, in case of uncertainty regarding the cut-off prior to the study, the study might
765 be used to explore and define the cut-off. In that case, the study should allow for a subsequent
766 evaluation and confirmation of the cut-off in a separate dataset to avoid overfitting. This could be
767 achieved, for example, by using a smaller proportion of patients to derive the cut-off and the remaining
768 patients to validate it. This should be clearly pre-specified and preferably be temporally separated (i.e.,
769 the first enrolled patients are used for the cut-off derivation, the later enrolled patients for the cut-off
770 confirmation).
771 In scenario (2) above, the validity of the chosen cut-off for patient selection could be supported by
772 investigating the interaction of the biomarker in subjects above the cut-off. It should be shown that the
773 treatment effect is decreased for biomarker values close to the cut-off, but, at the same time, that
774 subjects whose biomarker value is close to the cut-off still have a positive B/R.
775 In the setting where the biomarker-by-treatment interaction is evident only after the results are
776 available, unless there is a strong rationale for a cut-off, further validation in independent data is
777 usually required.
778 9. Bridging studies
779 In section 9, the term “samples”, which is equivalent to specimens in the IVDR, is used throughout [3].
780 This guideline primarily addresses considerations related to the co-development of a single biomarker
781 assay with a corresponding medicinal product. However, bridging studies may be required in the
782 following scenarios where different biomarker assays are used.
783 “Internal” bridging within the same development:
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 18/26
第 19 页
784 • Co-development but major assay changes: the early/mature candidate CDx undergoes
785 significant changes (i.e., with potential impact on assay performance) during the clinical
786 development programme of the medicinal product (e.g., during the pivotal clinical trial).
787
788 • Different assays used for a biomarker: different biomarker assays (measuring the same
789 biomarker) may be used during the clinical development programme of the medicinal product,
790 either concurrently or sequentially (e.g., when decentralised testing is used, refer to section
791 6.6 of the guideline, specifically paragraph “candidate CDx aspects"). Several scenarios can
792 apply. In the absence of a co-developed assay, a CDx candidate may differ from the assay
793 used in the medicinal product development (i.e., was not co-developed) but will be the first
794 CDx to be certified after medicinal product authorisation.
795 “External” bridging, e.g., across different IVD developments:
796 • Follow-on CDx candidates: The device seeks the same clinical indication and intended purpose
797 as a Conformité Européenne (CE) certified predecessor CDx device, and targets the same
798 biomarker, but it may use the same or a different technology [9].
799 In all these scenarios, bridging studies need to be conducted. The aim of a bridging study is to show
800 that two assays (original and new) are equivalent, meaning that the biomarker-selected populations are
801 highly similar for the two assays. In other words, it needs to be confirmed that the clinical benefit of the
802 medicinal product is maintained if a new IVD is used instead of the one used in the original clinical trial.
803 In a bridging study, the primary focus is to assess the concordance between the original and the new
804 assay, by using a set of samples analysed by both assays. The concordance should be evaluated using
805 positive percentage agreement (PPA) and negative percentage agreement (NPA). Overall percentage
806 agreement (OPA) is only considered relevant if the distribution of the samples used to establish OPA
807 mirrors the actual frequency in the real-world population (e.g., if a mutation occurs in 30% of the
808 patients in the indication, then approximately 30% of the samples used to calculate OPA should be true
809 positives with respect to the mutation). In any case, acceptance criteria for analytical equivalence
810 between the original and new assay should be pre-defined and appropriately justified in terms of PPA
811 and NPA. Acceptance criteria should be expressed in terms of lower bounds of the 95% confidence
812 intervals. They should be justified by clinical and analytical considerations and not by the available
813 sample size. For example, the concordance range should consider the extent of variability of the assay
814 as determined in the analytical validation studies. Ranges are expected to be narrower for genetic-
815 based assay (e.g. NGS) versus targeted proteomic assays (e.g. IHC).
816 The choice of samples for a bridging study should allow for a robust evaluation of PPA and NPA (i.e.
817 both positive and negative samples are needed) and be appropriately justified with respect to the
818 target indication. The samples should be representative of the target population in which the assay will
819 be used and should cover a wide range of different biomarker values, especially around and close to the
820 threshold. For bridging studies performed within the clinical development programme of the medicinal
821 product, a representative subset of the retained (left-over) samples (i.e. samples originating from the
822 initial clinical study) should be used. If samples collected outside the clinical development of the drug
823 are used (e.g. biobank samples), justification that these are representative for the context in which the
824 assay will be used should be provided.
825 If the samples from pivotal clinical trial(s) are available for bridging, the concordance should be
826 evaluated by analysing the primary and secondary endpoints as per the original clinical trial, but with
827 samples categorised according to the new assay instead of the original one and then comparing the
828 results from this re-analysis to the original clinical trial results. Generally, concordance should always
829 be evaluated whenever possible, but it is acknowledged that this may not be practically feasible in
830 many cases (e.g. for follow-on CDx candidate). Of note, modelling and simulation approaches to
831 extrapolate concordance in the absence of access to the original samples are currently discouraged due
832 to the high uncertainty and variability associated with these approaches.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 19/26
第 20 页
833 When reporting the outcome of a concordance analysis, the following information should be presented
834 in the MAA submission: the reference assay that was used, the original data/clinical trials that were
835 used as a reference, and a description of the samples analysed (origin, comparability to the original
836 population and a flowchart of sample inclusion/exclusion). Both the point estimate and the two-sided
837 95% confidence intervals of PPA and NPA should be provided, and the lower boundary of the
838 confidence interval should exceed the threshold that was pre-specified as success criterion.
839 Specifically, in the case of continuous markers, additional analyses such as the correlation of the assay
840 outcomes or a more granular concordance analysis across a range of cut-offs may deliver relevant
841 supportive information.
842 10. Summary of Product Characteristics (SmPC)
843 The SmPC is the basis of information for healthcare professionals on how to use medicinal products
844 safely and effectively. The Public Assessment Report contains comprehensive information on the
845 scientific assessment of the medicinal product. Information on the correct use of the IVD device is
846 available in the IFU.
847 Information on specific aspects of treatment with an approved predictive biomarker-guided medicinal
848 product related to its use with the validated assay for the purpose of patient selection should be
849 described in the relevant sections of the SmPC to a sufficient extent to ensure the safe and effective
850 use of the medicinal product. Information on the biomarker(s) and biomarker assay(s) should be
851 included and/or updated in relevant sections of the SmPC in line with the SmPC guideline [8, 20, 21,
852 22] and considering the examples given by the SmPC advisory group on pharmacogenomic information
853 in the SmPC [23].
854 In line with established guiding principles, inclusion of information about the biomarker and the
855 biomarker assay (including CDx) is foreseen in sections 4.1, 4.2, 4.3, 4.4, 4.8 and 5.1 of the SmPC, as
856 relevant [9].
857 SmPC sections 4.1, 4.2, 4.3
858 When the indication or contraindication is dependent on the predictive biomarker and its detection or
859 measurement by a corresponding predictive biomarker assay, inclusion of the biomarker reference in
860 section 4.1 (Indication) or in section 4.3 (Contraindication) is mandatory. For example, the indication
861 in section 4.1 of the SmPC can refer to a particular genotype and/or phenotype (e.g. the expression of
862 a gene or protein). Situations where the medicinal product should not be prescribed for safety reasons,
863 i.e. contraindications, are in the scope of the section 4.3. Such circumstances could also include a
864 particular genotype or phenotype.
865 Description of the predictive biomarker should be included with as much precision as possible
866 (including cut-off values where applicable, e.g. for continuous biomarkers) in order to adequately
867 define the intended patient population with a positive B/R profile. Sections 4.1 and 4.3 of the SmPC
868 reflect the essential and mandatory use of predictive biomarker assays. More detailed information may
869 be included in other SmPC sections, such as section 4.2. The extent of information provided in sections
870 4.1 and 4.2 may depend on the simultaneous diagnostic role of the predictive biomarker, with the
871 objective of clearly specifying the patient population with a positive B/R profile and, consequently,
872 excluding patients with a negative B/R profile from treatment. Where the patient population defined by
873 a biomarker used for disease diagnosis overlaps with that defined by a predictive biomarker, the level
874 of detail included in sections 4.1 and 4.2 should be determined on a case-by-case basis, provided that
875 it is sufficient to ensure the safe and effective use of the medicinal product and a clear definition of the
876 patient population with a positive B/R profile.
877 For indications with multiple or composite biomarkers where it is not reasonable to specify the cut-off
878 or other detailed information in section 4.1, a reference to section 4.2, where the biomarker status will
879 be further characterised, should be included. The following sentence will be included in section 4.1 “For
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 20/26
第 21 页
880 biomarker-defined patient selection, see section 4.2”. Similarly, when the patient population is defined
881 by other biomarkers not predictive for treatment with a specific medicinal product, the extent of
882 information in sections 4.1 and 4.2 can be decided on a case-by-case basis.
883 Where the use of specific methods, samples (specimens) (e.g. blood vs tissue), or analytes (e.g. DNA
884 vs RNA) is mandatory to ensure the selection of patients with a positive B/R profile, this can be
885 outlined in section 4.2. of the SmPC.
886 It is recommended that the wording in section 4.2 (Posology) of the SmPC refers to a CE marked IVD
887 or a validated test, i.e. “… should be assessed by a CE marked IVD with the corresponding intended
888 purpose. If the CE marked IVD is not available, an alternative validated test should be used.”.
889 SmPC sections 4.4 and 5.1
890 The biomarker and methods for its detection or measurement in clinical trials (e.g. clinical trial assay
891 or CDx with its brand name if applicable) should be described in section 5.1. The use of brand names
892 in this section can be required as an additional measure for safe and/or effective use of the medicinal
893 product (e.g. when distinctive information on concurrently used tests/assays is needed). Section 5.1
894 needs to clearly state which types of samples (specimens) were used in the clinical trial as the basis of
895 the presented data. Similarly, the type of analyte is to be specified (e.g. DNA, RNA, circulating tumour
896 DNA (ctDNA)), unless this information is already reflected in section 4.2 as a result of the assessment
897 on the mandatory scope of the information. When information on the predictive biomarkers tested in
898 clinical trials is extensive (e.g. numerous genetic alterations or composite biomarkers) and could thus
899 dilute the description of the clinical trials results in section 5.1, such information can be included under
900 subsection ‘Mechanism of action’ describing predictive biomarker(s).
901 In cases where the indication/target population is restricted to a subpopulation of trial population(s)
902 (e.g., defined by a predictive biomarker), it should be noted in the text whether the primary analysis
903 was positive in all enrolled patients. Nevertheless, a description of treatment effect estimates in all
904 enrolled patients is not necessary as it is not aligned with the restricted indication. All results should be
905 subsequently included for the restricted indication only.
906 While the main descriptive information on the use of biomarkers and biomarker assays is included in
907 section 5.1, additional description may need to be provided in section 4.4, particularly in case where
908 there are uncertainties related to biomarker detection or measurement for the identification of patients
909 who are likely to benefit, or where specific precautions are required. Section 4.4 may also include
910 special warning and precautions for use related to biomarkers that are relevant to safety (with cross-
911 reference to section 4.8), for example in cases of deficiency of certain enzymes.
912 11. Post-approval follow-up
913 If, at the time of the marketing authorisation, uncertainties remain on predictive biomarkers used to
914 select patients and/or certain performance characteristics of the corresponding predictive biomarker
915 assays, these aspects may need to be addressed post-approval depending on their potential impact on
916 the B/R profile of the selected patient population.
917 12. Definitions and abbreviations
918 This section provides definitions of key terms and abbreviations which are used throughout the
919 document for clarity and consistency. Where available, the definitions align with official definitions.
Analytical validation Demonstration that a bioanalytical method is suitable for its
intended purpose [12]
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 21/26
第 22 页
B/R Benefit/Risk
Biobank samples Samples (specimens) which have been collected, processed
and stored in a biobank for use in current or future clinical
studies; collected under clinical protocols.
Biomarker assay The test or device used to detect/measure the biomarker.
Biomarker status The positive/negative outcome of the biomarker assay,
defined for all biomarkers to be included in a medicinal
product indication. The status can simply refer to the
detection of the biomarker in the sample for qualitative
biomarkers (i.e., being above or below the detection limit),
or it can be based on the dichotomisation of a continuous
value for continuous biomarkers (i.e., being above or below
an established cut-off).
Biomarker-defined target The population where the medicinal product is expected to
population exhibit a positive B/R based on the biomarker status.
The co-development of the medicinal product and
biomarker assay aims to identify such a population as close
as possible.
Biomarker-selected population The population selected based on the biomarker status
(positive or negative), in which the medicinal product has
shown a positive B/R and is hence reflected in the SmPC.
Can be either the population with a positive biomarker
status or the one with a negative biomarker status.
Candidate CDx A biomarker assay used in pivotal clinical trials for the
development of a predictive biomarker-guided medicinal
product in order to establish the data for B/R assessment;
and that is (co-)developed towards a CDx as defined in the
IVDR. Depending on the stage of clinical development, this
can either be an “early” or a “mature” candidate CDx.
CE Conformité Européenne (European Conformity)
Clinical validation of a predictive The process to establish the ability of an assay to
biomarker assay differentiate the efficacy or safety of a medicinal product
based on biomarker status.
Co-developed CDx An IVD device that is developed in a clinical development
programme together with the concerned medicinal product,
either in view of an initial marketing authorisation or post-
approval extension of the indication. A co-developed CDx is
usually the device used in a pivotal clinical trial for the
concerned medicinal product. Alternatively, the CDx may
be developed in the context of a bridging study.
Companion diagnostic (CDx) Defined in Article 2(7) of Regulation (EU) 2017/746 [3] as
follows: a device which is essential for the safe and
effective use of a corresponding medicinal product to: (a)
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 22/26
第 23 页
identify before and/or during treatment patients who are
most likely to benefit from the corresponding medicinal
product; or (b) identify before and/or during treatment
patients likely to be at increased risk of serious adverse
reactions as a result of treatment with the corresponding
medicinal product.
CoU (Context of Use) A commonly used statement which clearly describes the
way in which the predictive biomarker (assay) is to be used
in relation to the medicinal product development. This
terminology (although linked) is broader in meaning than
the “intended purpose” of the IVD according to the IVDR.
The context of use (CoU) of a medicinal product or tool is a
formal, concise statement defining the specific
circumstances, patient population, and purpose for which it
is used, often focusing on its role in drug development,
clinical trials, or regulatory approval. It includes the
intended use, target population, and conditions for use.
For biomarkers, the CoU specifically outlines the category
(e.g., prognostic, predictive) and its role in drug
development. It ensures that the product is "fit for
purpose" and provides relevant data to support regulatory
or scientific decisions.
The context of use (CoU) of a medicinal product with a
biomarker refers to the specific circumstances, patient
population, and purpose for which the biomarker is used,
often focusing on its role in drug development, clinical
trials, or regulatory approval. It includes the intended use,
target population, and conditions for use.
ctDNA Circulating tumour DNA
CTR Clinical Trials Regulation [7]
DNA Deoxyribonucleic acid
EMA European Medicines Agency
EU European Union
FISH Fluorescence in situ hybridisation
GCP Good Clinical Practice
HLA Human leukocyte antigen
ICH International Council for Harmonisation of Technical
Requirements for Pharmaceuticals for Human Use
IHC Immunohistochemistry
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 23/26
第 24 页
IFU (instructions for use) According to the IVDR Art 2 (14) [3], “instructions for use”
means the information provided by the manufacturer to
inform the user of a device's intended purpose and proper
use and of any precautions to be taken.
Intended purpose According to the IVDR Art2 (12) [3], “intended purpose”
means the use for which a device is intended according to
the data supplied by the manufacturer on the label, in the
instructions for use or in promotional or sales materials or
statements or as specified by the manufacturer in the
performance evaluation.
ISH In situ hybridisation
IVD In vitro diagnostic medical device
IVDR Regulation (EU) 2017/746 on in vitro diagnostic medical
devices [3]
Limit of Detection
LOD
MAA Marketing Authorisation Application
MDCG Medical Device Coordination Group
NB Notified Body
NCA National Competent Authority
NGS Next-generation sequencing
NPA Negative percentage agreement
Omic Referred as multiple omic technologies (e.g., genomic,
proteomic, transcriptomic)
OPA Overall percentage agreement
PCR Polymerase Chain Reaction
PPA Positive percentage agreement
Predictive biomarker A biomarker, which in association with a specific treatment,
identifies patients that are more likely to experience a
certain effect on the clinical outcome.
Predictive biomarker assay An assay to detect or/and quantify a predictive biomarker.
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 24/26
第 25 页
Prognostic biomarker A biomarker which identifies patients that are more likely
to experience a certain clinical outcome, regardless of
treatment.
RNA Ribonucleic acid
Round-robin test A procedure in which multiple laboratories analyse identical
samples using the same method to evaluate the
reproducibility and performance of that method across
different settings [24].
SmPC Summary of Product Characteristics
SSP Summary of safety and performance (IVDR terminology)
920
921 13. References
922 01. NOTICES FROM EUROPEAN UNION INSTITUTIONS, BODIES, OFFICES AND AGENCIES 2015/C
923 421/03 Council conclusions on personalised medicine for patients (https://eur-lex.europa.eu/legal-
924 content/EN/TXT/PDF/?uri=CELEX:52015XG1217(01))
925 02. Verbaanderd, C., Trullás Jimeno, A., Engelbergs, J., Zander, H. Reischl, I., Moreno Oliver, A.,
926 Vamvakas, S., Vleminckx, C., Bouygues, C., Girard, T., Day, F., and Frias, Z. (2023) Biomarker-Driven
927 Developments in the Context of the New Regulatory Framework for Companion Diagnostics in the
928 European Union, CLINICAL PHARMACOLOGY & THERAPEUTICS, 114 (2)
929 (https://doi.org/10.1002/cpt.2928)
930 03. REGULATION (EU) 2017/746 OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 5 April
931 2017 on in vitro diagnostic medical devices and repealing Directive 98/79/EC and Commission Decision
932 2010/227/EU (https://eur-lex.europa.eu/eli/reg/2017/746)
933 04. Guidance on Classification Rules for in vitro Diagnostic Medical Devices under Regulation (EU)
934 2017/746 (MDCG 2020-16)
935 05. Guidance on the procedural aspects for the consultation to the European Medicines Agency by a
936 notified body on companion diagnostics (EMA/198592/2022)
937 06. Guideline on clinical evaluation of diagnostic agents (CPMP/EWP/1119/98/Rev. 1)
938 07. Regulation (EU) No 536/2014 of the European Parliament and of the Council of 16 April 2014 on
939 clinical trials on medicinal products for human use, and repealing Directive 2001/20/EC Text with EEA
940 relevance (Regulation - 536/2014 - EN - EUR-Lex)
941 08. <Co>Rapporteur day <60*><80> critical assessment report - Clinical aspects
942 (https://www.ema.europa.eu/en/human-regulatory-overview/marketing-authorisation/assessment-
943 templates-guidance)
944 09. Frequently asked questions on medicinal products development and assessment involving
945 companion diagnostic (CDx) (EMA/CHMP/821321/2022)
946 10. Questions & Answers - Practical arrangements on the companion diagnostics consultation
947 procedure to the European Medicines Agency by notified bodies (EMA/619893/2022)
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 25/26
第 26 页
948 11. ICH E6 (R3) Guideline for good clinical practice (GCP) - Step 5 (EMA/CHMP/ICH/162879/2026)
949 12. ICH guideline M10 on bioanalytical method validation and study sample analysis, Step5
950 (EMA/CHMP/ICH/172948/2019)
951 13. ICH guideline M10 on bioanalytical method validation and study sample analysis - Questions and
952 Answers (EMA/CHMP/ICH/660315/2022)
953 14. Guideline on good pharmacogenomic practice (EMA/CHMP/718998/2016)
954 15. Qualification of novel methodologies for drug development: guidance to applicants
955 (EMA/CHMP/SAWP/72894/20)
956 16. ICH guideline E18 on genomic sampling and management of genomic data, Step 5
957 (EMA/CHMP/ICH/11623/2016)
958 17. Guideline on Missing Data in Confirmatory Clinical Trials (EMA/CPMP/EWP/1776/99 Rev. 1)
959 18. Guideline on the investigation of subgroups in confirmatory clinical trials (EMA/CHMP/539146/2013)
960 19. Reflection Paper on methodological issues in confirmatory clinical trials planned with an adaptive
961 design (CHMP/EWP/2459/02)
962 20. Scientific guidelines with SmPC recommendations (EMA/813125/2012 rev. 8)
963 21. Wording of therapeutic indication - A Guide for Assessors of Centralised Applications
964 (EMA/CHMP/483022/2019)
965 22. A GUIDELINE ON SUMMARY OF PRODUCT CHARACTERISTICS (SmPC) (EUROPEAN COMMISSION,
966 NOTICE TO APPLICANTS, Revision 2)
967 23. https://www.ema.europa.eu/en/documents/presentation/presentation-pharmacogenomics-
968 information-summary-product-characteristics_en.pdf
969 24. Thompson, Michael, Ellison, Stephen L. R. and Wood, Roger. "The International Harmonized
970 Protocol for the proficiency testing of analytical chemistry laboratories (IUPAC Technical Report)" Pure
971 and Applied Chemistry, vol. 78, no. 1, 2006, pp. 145-196.
972 (https://doi.org/10.1351/pac200678010145)
Guideline on predictive biomarker assay development in the context of medicinal product lifecycle
EMA/CHMP/800914/2016 Page 26/26
来源:EMA Scientific Guidelines · ema.europa.eu