Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Imaging biomarker roadmap for cancer studies


Imaging biomarkers (IBs) are integral to the routine management of patients with cancer. IBs used daily in oncology include clinical TNM stage, objective response and left ventricular ejection fraction. Other CT, MRI, PET and ultrasonography biomarkers are used extensively in cancer research and drug development. New IBs need to be established either as useful tools for testing research hypotheses in clinical trials and research studies, or as clinical decision-making tools for use in healthcare, by crossing 'translational gaps' through validation and qualification. Important differences exist between IBs and biospecimen-derived biomarkers and, therefore, the development of IBs requires a tailored 'roadmap'. Recognizing this need, Cancer Research UK (CRUK) and the European Organisation for Research and Treatment of Cancer (EORTC) assembled experts to review, debate and summarize the challenges of IB validation and qualification. This consensus group has produced 14 key recommendations for accelerating the clinical translation of IBs, which highlight the role of parallel (rather than sequential) tracks of technical (assay) validation, biological/clinical validation and assessment of cost-effectiveness; the need for IB standardization and accreditation systems; the need to continually revisit IB precision; an alternative framework for biological/clinical validation of IBs; and the essential requirements for multicentre studies to qualify IBs for clinical use.


A biomarker is a “defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes or responses to an exposure or intervention, including therapeutic interventions” (Refs 1,2). The current FDA–NIH Biomarker Working Group definition — adopted in this consensus statement — states explicitly that “molecular, histologic, radiographic or physiologic characteristics are examples of biomarkers” (Ref. 2). This approach seeks to clarify inconsistency in terminology, because some previous definitions have restricted the scope of biomarkers to describing biological molecules. Such narrow definitions regard values obtained from imaging and other techniques as measurements of an underlying biomarker, rather than being biomarkers themselves. However, the current FDA–NIH definition takes a broader view. Earlier definitions also stated that biomarkers should have 'putative' diagnostic or prognostic use3, although this requirement is no longer specified by FDA/NIH. Biomarkers must be measured, but can be numerical (quantitative) or categorical (either a quantitative value or qualitative data; for definitions, see Supplementary information S1 (table)).

The use of both imaging biomarkers (IBs) and biospecimen-derived biomarkers is widespread in oncology. In healthcare settings, biomarker uses include screening for disease; diagnosing and staging cancer; targeting surgical and radiotherapy treatments; guiding patient stratification; and predicting and monitoring therapeutic efficacy, and/or toxicity4. In research, biomarkers guide the development of investigational drugs as they progress along the pharmacological audit trail5, in which they can indicate the presence of drug targets, target inhibition, biochemical pathway modulation or pathophysiological alteration by investigational drugs; drug therapeutic efficacy in specific groups of patients; and tracking of drug resistance6. The use of biomarkers has led to the identification of potentially successful drugs early in the developmental pipeline, thereby accelerating market approval for some therapies and enabling drug developers to reduce overall costs by identifying ineffective or toxic compounds at the earliest opportunity5.

Despite some biomarkers being used extensively and others showing great potential7,8, a surprisingly limited number of biomarkers guide clinical decisions9,10,11. Some putative cancer biomarkers are not adopted because they do not measure a relevant biological feature nor enable disease diagnosis or outcome prediction. In such cases, the biomarker is appropriately devalidated12. Many other promising biomarkers, however, are neither devalidated nor qualified for use in research or healthcare settings and, instead, are confined in the academic literature without real application owing to a lack of efficient and effective strategies for biomarker translation13.

All biomarkers, including IBs, must cross two 'translational gaps' before they can be used to guide clinical decisions14,15 (Fig. 1). Biomarkers that can reliably be used to test medical hypotheses cross the first gap becoming useful 'medical research tools'; if the biomarker crosses the second gap then it becomes a 'clinical decision- making tool'. Some biomarkers that have only crossed the first translational gap are nevertheless highly useful in the development of therapies5,13.

Figure 1: Overview of the imaging biomarker roadmap.

Imaging biomarkers must cross translational gap 1 to become robust medical research tools, and translational gap 2 to be integrated into routine patient care. This goal is achieved through three parallel tracks of technical (assay) validation, biological and clinical validation, and cost effectiveness.

PowerPoint slide

Several publications have described strategies for developing and evaluating cancer biomarkers, focusing mainly on biospecimen-derived biomarkers — that is, those derived from patient tissue or biofluids4,16,17,18,19,20,21. The processes of initial discovery, validation and qualification share many similarities for IBs and biospecimen-derived biomarkers; however, substantial inherent differences exist between both biomarker types (Supplementary information S2 (table)). The FDA and NIH have recognized this distinction and have outlined specific recommendations for image acquisition and analysis in IB development22,23. Questions of how IB acquisition and analysis should be standardized, and how terminology should be harmonized have been addressed by numerous academic, clinical, industrial and regulatory groups. These groups include the FDA2,24, the US National Cancer Institute (NCI) through the Quantitative Imaging Network (QIN)25 and the Cancer Imaging Program phase I and II Imaging trials initiative26, the Quantitative Imaging Biomarkers Alliance (QIBA)27,28, the American College of Radiology Imaging Network (ACRIN)29, the European Society of Radiology (ESR)30, the European Organisation for Research and Treatment of Cancer (EORTC) through the QuIC-ConCePT consortium13,31, the European Association of Nuclear Medicine (EANM)32, the International Society for Magnetic Resonance in Medicine (ISMRM)33 and Cancer Research UK (CRUK)34. Their efforts have produced consensus guidelines for the acquisition and analysis of several IBs33,34,35,36,37. Many of these organizations also have highlighted the need for a detailed validation and qualification roadmap to improve IB translation38,39. Recognizing this need, representatives from CRUK and the EORTC, together with other assembled experts in radiology, cancer imaging sciences, oncology, biomarker development, biostatistics and health economics, have formulated an Imaging Biomarker Roadmap for Cancer Studies (Fig. 2). In this Consensus Statement, we outline this roadmap and identify specific considerations for IB validation and qualification, providing illustrative examples of IBs in various stages of development. From this framework, we provide 14 recommendations to accelerate the successful clinical translation of effective IBs, as well as the devalidation40,41 of IBs that lack utility.

Figure 2: The imaging biomarker roadmap.

A detailed schematic roadmap is depicted. The imaging biomarker (IB) roadmap differs from those described for biospecimen- derived biomarkers. For imaging, the technical and biological/clinical validation occur in parallel rather than sequentially. Of note, essential technical validation occurs late in the roadmap in many cases (such as full multicentre and multivendor reproducibility). Definitive clinical validation studies (IB measured against outcome) are deferred until technical validation is adequate for large trials. In the absence of definitive outcome studies, early biological validation can rely on a platform of very diverse graded evidence linking the IB to the underlying pathophysiology. Cost-effectiveness impacts on the roadmap at every stage, owing to the equipment and personnel costs of performing imaging studies. Technical validation and cost-effectiveness are important for IBs after crossing the translational gaps because hardware and software updates occur frequently. Therefore, technical performance and economic viability must be re-evaluated continuously. SOP, standard operating procedure. Image reproduced from

PowerPoint slide

Current uses of imaging biomarkers

An IB is a measurement derived from one or more medical images. Many IBs are used routinely in healthcare (Table 1). IBs provide readily available, cost-effective, non-invasive tools for screening, detecting tumours and serial monitoring of patients, including assessments of response to therapy and identification of therapeutic complications. IBs can enable tracking of a particular tumour repeatedly over time, can map the spatial heterogeneity within tumours, and can evaluate multiple different lesions independently within an individual.

Table 1 Selected list of imaging biomarkers used in clinical oncology decision-making

Applications for IBs include the American College of Radiology Breast Imaging-Reporting and Data System (ACR BI-RADS) mammographic breast morphology42; clinical tumour, node, metastasis (TNM) stage43 (Box 1); objective response as defined in RECIST version 1.0 (Ref. 44) or 1.1 (Ref. 45) criteria (Box 2); bone mineral density T-score measured by dual energy X-ray absorptiometry46; and left ventricular ejection fraction47 (LVEF; Box 3). These IBs underpin patient care worldwide, although they are subject to continuing research to improve their performance. In addition, IBs could achieve companion diagnostic status, as seen with the IB [99mTc]-etarfolatide folate receptor positivity48 (FR+; Box 4), demonstrating that imaging can be necessary to guide the choice of therapy for individual patients.

IBs frequently add value in cancer research (Table 1). The use of IBs can enable the measurement of patient response to treatment before a survival benefit is observed, which can subsequently lead to early regulatory approval of new drugs49 (Box 2). IBs can indicate the presence of drug targets and target inhibition, for example by proving receptor occupancy50 (Supplementary information S3 (box)). IBs have the unique potential to provide serial non-invasive mapping of tumour status during treatment. For example, absolute values of 18F-FDG-PET maximum standardized uptake value (SUVmax) at baseline, or changes in this value observed early during treatment have been used for proof-of-mechanism and proof-of-principle51 in drug development (Box 5), or to demonstrate nonspecific responses to treatment52 (Box 5). Dynamic contrast-enhanced (DCE) CT or MRI-derived K trans (Ref. 53) (Box 6) and dynamic contrast- enhanced ultrasonography area under the curve (AUC) values (Supplementary information S4 (box)) have been used as IBs of pharmacodynamic (PD) changes and response to treatment54. The use of IBs has led to improved margins of radiotherapy dose delivery55 (Box 5) and surgical margins56.

Many more IBs used to measure tumour anatomy, morphology, pathophysiology, metabolism or molecular profiles are being developed in order to study the hallmarks of cancer57. Between 2004–2014, approximately 10,000 studies reported new or established IBs (Supplementary information S5 (box)), including IBs derived from new modalities (such as photoacoustic imaging58) and new techniques (such as MRI dynamic nuclear polarization59; Supplementary information S6 (box)), and new analytical approaches (such as radiomic profiling of tumours to extract multiple features60,61; Supplementary information S7 (box)), contributing to this ever-increasing number of IBs.

IBs have four key attributes. First, they are a subset of all biomarkers. Second, they can be quantitative or qualitative. Quantitative IBs (measured on an interval or ratio scale28) are used in patient care, but other measurements that fall outside this definition (for example, the ACR BI-RADS category, clinical TNM stage, or objective response) are categorical measurements and are also important IBs. In this Consensus Statement, we have deliberately included all image measurements (quantitative or categorical) that satisfy the definition of an IB, according to the current FDA–NIH Biomarker Working Group definition2. Third, IBs are derived from imaging modalities, techniques or signals, but are distinct entities; for example, the change in median apparent diffusion coefficient (ADC) is a biomarker, and is distinct from the modality (MRI), the technique (diffusion-weighted imaging) or the measured signal (free induction decay) required to generate the IB (Supplementary information S8 (figure)). Fourth, one imaging measurement can support multiple distinct IBs. For example, 18F-FDG-PET has improved disease staging in patients with non-small-cell lung cancer, lymphoma or melanoma by facilitating the identification of nodal and distant metastases62. In this case, the IB is clinical TNM stage, which is defined using CT and PET data rather than using CT data alone (Box 1). Similarly, systems such as PERCIST (for all solid tumours)63 and the Deauville five point system (for lymphoma)64 aim to improve the assessment of response criteria by incorporating PET data, but produce an ordered categorical IB — namely objective response (Box 2). This approach is conceptually different from quantifying absolute values of 18F-FDG-PET data to derive putative cut-off points subsequently used to identify patients with poor prognosis65 or evidence of specific pathway modulation in a clinical trial of an investigational new drug66,67 (Box 5).

Many quantitative image measurements comprise continuous data, which must then be categorized to facilitate clinical decision-making. Clinicians often decide between two or more alternative treatment options for each patient, informed by whether a biomarker value is above or below a cut-off point. For example, cancer therapy-related cardiac dysfunction has been defined as LVEF reduction of ≥10 percentage points to a value below normal (53% for adults; Box 3). Clear guidelines detailing how scintigraphy or 2D echocardiography should be performed have been established for this IB of toxicity for use in healthcare47. In distinction, other healthcare-related IBs (such as clinical TNM stage and objective response) are measured as ordered categorical variables. In this case, the boundary between several categories is defined by cut-off points; alternatively, categories can be combined to create a single cut-off point (for example, to select, continue or stop therapy).

In research applications, data are often interpreted as a continuous rather than used in a categorized way. For example, in early phase trials, continuous variable PD biomarkers (including percentage change in tumour size68, 18F-FDG-PET SUVmax (Ref. 52), ultrasonography AUC69 and MRI median Ktrans (Ref. 53) or ADC70 (Boxes 5,6; Supplementary information S6 (box)) can indicate antitumour activity of therapeutic agents used at different doses and time points. These studies have led to the demonstration of proof-of-concept and proof-of-mechanism, the definition of pharmacokinetic (PK)–PD relationships and informed on dose selection for novel therapeutic approaches53.

Specific considerations for IBs

IBs and biospecimen-derived biomarkers differ in several important aspects, limiting the relevance to IBs of previous roadmaps (designed for biospecimen-derived biomarkers13,71; Supplementary information S2 (table)). The performance of imaging devices of different makes and models installed in different clinical centres can vary considerably. These devices are designed, approved, maintained, and operated to provide images38 that diagnostic radiologists, nuclear medicine physicians and other clinicians interpret, often with little need to quantify the data obtained. Innovation is largely driven by competition to improve image quality and the user interface; vendors and purchasers often have only secondary interest in how improvements in image quality will affect quantification or standardization of IBs. The measurement of many IBs requires the administration of tracers or contrast agents (for example, diagnostic drugs, usually investigational or off-label), the development and availability of which are uncertain. Many steps in IB validation require a new prospective clinical trial; whenever investigational tracers or contrast agents are involved, the high burden of regulation can cause significant delays in this process.

Many IBs do not purport, even in principle, to measure any underlying analyte, making traditional approaches to assay validation seen with biospecimen-derived biomarkers problematic. For example, dilution linearity and reagent stability — important performance characteristics for biospecimen-derived biomarker assessment12 — cannot be assessed for biomarkers based on analysing medical images, such as texture-based IBs. The measurement validity of such IBs depends substantially on events that occur while the patient is coupled to the imaging device. Thus, the role of central reading laboratories in data processing is minor in comparison with biospecimen-derived biomarkers4.

The imaging biomarker roadmap

Biomarkers cross the two translational gaps by passing through a series of domains (discovery, or domain 1; validation, or domain 2; and qualification with ongoing technical validation, or domain 3) that address different research questions. This process follows two parallel and complementary tracks: in one track, technical (assay) performance is examined by addressing whether the biomarker is measurable precisely and accurately, and whether it is widely available in all geographical territories. In the other track, biological and clinical performances are examined by addressing whether the biomarker can be used to measure a relevant aspect of biology or predict clinical outcome. In reality, no biomarker is perfectly validated. Instead, strategies must be defined to identify, mitigate and quantify the uncertainty, risk and cost associated with any given biomarker in making research or clinical decisions72,73,74. The nature of these activities and the sequence in which they are combined constitutes a 'biomarker roadmap'.

In biospecimen-derived biomarker roadmaps, technical (assay) validation commonly occurs early in order to produce a 'locked-down biomarker' — that is, the data acquisition and analysis pathways used to measure the biomarker are fixed. In many cases, this stage is followed rapidly by biological and clinical validation because the locked-down biomarker enables widespread evaluation across multiple sites4,18. For IBs, several key aspects of technical validation (such as multicentre reproducibility) must be addressed at a relatively late stage, unlike biospecimen-derived biomarkers. Studies might address both technical and biological validation concurrently, but progress down each track might be quite independent. As evidence for validation accumulates, a third track of cost-effectiveness must be considered, because IBs must not only demonstrate an association with health benefits, but also demonstrate 'value for money' when compared with the use of clinical information alone or with alternative biofluid-based in vitro diagnostics75.

Imaging biomarker discovery — domain 1

Most biospecimen-derived biomarkers are molecular features found in the genome, transcriptome, proteome or metabolome that can be chosen rationally to address unmet needs in cancer medicine. This selection approach has also inspired the rational design of several novel targeted nuclear medicine or optical imaging tracers76. Many of the most valuable IBs, however, have had a convoluted genesis through unanticipated discoveries in the physical sciences (chemistry, computing, engineering, mathematics or physics) being matched to unmet medical needs after initial discovery, and then developed further.

Technical validation — domains 2 and 3

Complete technical validation is achieved when an IB measurement can be performed in any geographical location, whenever needed, and give comparable data. Technical validation does not address whether the IB measures underlying biology, or relates to clinical outcome and/or utility, but it can have a major influence on subsequent attempts at qualification because IBs with limited availability or poor reproducibility cannot sensibly undergo further clinical evaluation in large multicentre trials.

Precision. Repeatability and reproducibility are related measures of assay precision. Repeatability refers to measurements performed multiple times in the same subject (in vitro and/or in vivo) using the same equipment, software and operators over a short timeframe, whereas reproducibility refers to measurements performed using different equipment, different software or operators, or at different sites and times77, either in the same or in different subjects.

Different intended applications require different levels of evidence for precision3. For example, single- centre repeatability might be sufficient to validate a PD or monitoring IB in an early phase clinical trial restricted to that site53. Conversely, screening, diagnostic, prognostic and predictive IBs with putative use in whole populations require evidence of reproducibility across multiple expert and non-expert centres before they can be considered technically valid. Such multicentre validation requires complex and costly studies13.

Repeatability analysis should be performed early in IB development. Repeatability estimated from studies using rodents can seem unfavourable because, in some models, tumours can grow considerably within a few hours78, but useful data can be gained in slow-growing rodent models. Definitive repeatability analysis is best assessed in studies with humans, and should be performed at each centre that evaluates the IB anew, because reliance on historical or literature values is a source of error. Test performance is known to vary between sites, and is influenced by scanner performance, and organ site studied (for example, precision is better in brain lesions79 compared with liver or bowel lesions80,81,82). Physiological variations between individuals (depending on factors such as caffeine, nicotine, alcohol or concomitant medication) can also alter IB values83.

Multicentre studies involve different research institutions that usually utilize devices supplied by different vendors. These devices are broadly equivalent for clinical radiology purposes, although they often have important hidden differences that affect IB acquisition and analysis (Supplementary information S2 (table)). These factors do not preclude multicentre technical assessment of IB precision, but values might not be as reproducible as those derived in few-centre or single-centre studies84 (Box 6; Supplementary information S4 (box)).

Unfortunately, despite the scientific benefits associated with double baseline studies, the time and financial cost of performing such studies often deters investigators and funders from incorporating repeatability or reproducibility evaluations into study protocols. For example, between 2002–2012, only 12 of 86 phase I/II studies of antivascular agents that incorporated DCE-MRI biomarkers, such as the volume transfer co-efficient Ktrans, measured test–retest performance53 (Box 6). Multisite reproducibility studies of IBs are very rare.

Bias. Technical bias describes the systematic difference between measurements of a parameter and its real value85. Few in-human IB studies report bias because real values might not be possible to ascertain in a clinical setting. For some IBs, bias can be estimated by comparison with reference phantoms, as is the case when validating biomarkers on the basis of CT Hounsfield units86, and MRI longitudinal (R1) and transverse (R2) relaxation rates87. Imaging phantoms, however, seldom fully represent technical performance in animals or humans. For other IBs, such as the effective transverse relaxation rate (R2*), diffusion anisotropy, or Ktrans (all measurable by MRI), appropriate calibration phantoms are not available or poorly represent living tissue.

Availability. IB assessments must be feasible, safe, and well-tolerated in the target population, and must have regulatory and ethical approval for use in humans. Differing regulatory approvals (for example, different gadolinium-based contrast agents have been approved by regulatory agencies in Europe, North America and Japan13) and commercial pressures (for example, ferucarbotran iron oxide nanoparticles are no longer commercially available in Europe or North America) can substantially affect availability. With regard to PET, some short half-life tracers (such as 15O-H2O) are highly informative, but require the presence of on-site cyclotron facilities, making worldwide translation impossible with currently available technologies, for economic reasons. Finally, specialist techniques in all modalities require advanced radiographer and technician support, which might not be available locally.

Biological/clinical validation — domain 2

The terms 'biological validation', 'clinical validation' and 'clinical utility' describe the stepwise linking of biomarkers to tumour biology, outcome variables and value in guiding decision-making, respectively. Clinical validation requires demonstration that the biomarker merely relates to a clinical variable and, therefore, is less difficult to establish than clinical utility (but also less meaningful)86. Clinical utility is achieved when the biomarker leads to net improvement of health outcomes or provides information useful for diagnosis, treatment, management, or prevention of a disease10,88.

The REMARK guidelines89 provide a framework for the assessment of clinical utility and validation. Typically, prospective testing of IBs in clinical populations is required in adequately powered studies with follow-up times of 3–5 years to provide outcome data associated with the IB. For example, in patients with colorectal cancer, parameters such as circumferential resection margin56, depth of tumour spread, and extramural vascular invasion90, are assessed by MRI before surgery, have been validated as preoperative prognostic and predictive IBs, and are currently used to stratify patients into treatment groups. Similarly, ongoing work is evaluating the role of tumour CT, MRI and PET 'radiomic signatures' (Ref. 91) (Supplementary information S7 (box)) or 18F-FDG SUVmax baseline values65 as prognostic IBs. In these examples, IB data are derived from routine clinical images, making the process of establishing clinical utility much faster than is typically possible for IBs derived from techniques not yet established in routine healthcare61.

IBs that have demonstrated clinical validity can have important roles in drug development. For example, RECIST 1.1-defined objective response is a widely used phase II clinical trial end point that has proved useful for preliminary screening of new therapeutic agents45. More-rigorous criteria must be met to validate an IB as a trial-level surrogate end point to completely replace a well-established clinical trial end point. The Prentice criteria92 describe idealized conditions for demonstration of surrogacy, which unfortunately are rarely achieved and would typically require extremely large datasets93. Approaches for surrogate validation more pragmatic than the Prentice criteria involve the application of meta-analysis methods to a collection of large or moderate-sized datasets that measure both the putative surrogate biomarker and the definitive clinical end point94. These meta-analytic approaches often require some degree of IB standardization in order to combine datasets in a meaningful manner. Some IBs have been well-standardized (such as the RECIST 1.1 objective response), whereas others (for example, DCE-MRI-derived Ktrans53; Box 6) have not. Large prospective multicentre studies relating IBs to an outcome can only be initiated when exhaustive technical validation has established multicentre IB precision and accuracy. Clinical utility (and sometimes, validation) necessarily happens late in the IB roadmap; therefore, alternative validation strategies must be sought for many novel IBs. Biological validation can be approached by accumulating a platform of evidence linking the IB to meaningful biological features, and response to therapies with well-studied mechanisms of action71. These principles — outlined by Bradford Hill93,95 (Supplementary information S9 (box)) — establish evidence for IB performance on the basis of scientific coherence, specificity, strength of association, effect gradient, temporality and consistency.

If the Bradford Hill criteria are adopted, early preclinical imaging–pathology correlation studies5,96 provide an important component of the platform of evidence because whole-tumour histopathology is rarely possible in patients. Studies conducted with animals enable clinically relevant IBs to be related to fundamental biological processes that can only be measured with invasive techniques. For example, the relationship between DCE-MRI-derived Ktrans and tumour blood flow rate (a fundamental measure of response to antivascular agents) was investigated using a terminal radiotracer-based technique in rat tumour models97. Preclinical studies also enable the examination of the therapeutic dose–response relationship at different time points in a range of tumour models. Of note, however, pathophysiological assays are often considered as a 'reference standard', but are also biomarkers and thus imperfect, and also subject to sampling bias. Thus, some 'reference standard' tests might not enable the prediction of survival, or relate to the intended biological process. Moreover, precise pathophysiological correlates of some IBs might not exist, or can be almost impossible to obtain13.

Evidence of imaging–biology correlations and early response to therapies can provide sufficient biological validation to establish IBs as useful for PD assessments or monitoring response in early phase clinical trials, even in the absence of compelling outcome data53. The IBs 18F-FDG-PET SUVmax and Ktrans (Boxes 5,6) are illustrative of how IBs can cross translational gap 1 to guide decision-making for subsequent studies. For example, the biologically active dose for the antivascular agent ZD6126 calculated according to Ktrans measurements98 was shown to be greater than its maximum tolerated dose, effectively halting further clinical development of this agent. In other studies, Ktrans data informed of the biologically active dose for cediranib99 and the optimum scheduling for brivanib100 (Box 6).

Cost-effectiveness — domains 2 and 3

To be translated into the clinic, IBs must provide good 'value for money' and compare favourably with biospecimen-derived biomarkers resulting from technologies, such as 'liquid biopsies' of isolated circulating tumour cells or cell-free DNA. In the research setting, the value added by testing a key hypothesis (with an IB) should be greater than the cost of performing the study. In healthcare, the economic test is harsher, because even well-validated IBs will not cross translational gap 2 unless they offer an advantage in terms of cost per quality adjusted life year (QALY) gained13. Initially, translating IBs into the healthcare setting is costly and time-consuming101.

Studies in early stages of IB discovery and validation can access conventional biomedical research funding streams, whereas evidence-gathering studies to support IB qualification can be difficult to fund. For example, large-scale multicentre reproducibility studies can be unattractive to research funders if seen as 'incremental' and, unless such studies create exploitable intellectual property, they are not attractive to commercial sponsors. One approach to solve this problem is to assemble consortia of commercial and not-for-profit stakeholders, such as the QuIC-ConCePT consortium13,31, QIBA28 and QIN102, to undertake multicentre validation steps to meet the collective needs of the community.

Initial IB translation requires large-scale funding by government, charity or industry, or commercialization of the IB by an imaging company (for example, a major scanner manufacturer in the case of hyperpolarized 13C-MRI/MRS), or by a start-up business focused on a specific technique or IB. This strategy requires careful assessment of the likely risk–benefit of the development process, and strong intellectual property positions to ensure likely financial return on the imaging device, agent or biomarker38. Complex economic considerations are related to IBs developed as companion diagnostics associated with a specific therapeutic. Such IBs might be cost-effective for the healthcare payer (reducing futile expenditure and avoiding adverse effects in patients who cannot benefit from the drug) as well as beneficial for the supplier of the therapeutic (leading to a reduction in the size of trial cohorts, enabling marketing authorization, and providing better ratios of costs per QALY gained), but are associated with increased risks for the IB developer because, if clinical translation of the therapy fails (such as in the case of vintafolide48; Box 4), the market for the companion diagnostic disappears.

Qualification — domain 3

The term 'qualification' has different scientific and regulatory meanings. Generally, qualification is an evidentiary process of linking a biomarker with biological processes and clinical end points intended to establish that the biomarker is fit for a specific purpose96,103,104. For example, the use of IBs qualified as prognostic would enable the enrichment of a clinical trial population with patients at the highest risk of experiencing a clinical event, thus increasing the efficiency of the trial105. Such IBs have to be clinically validated to show their association with an outcome, but might not necessarily be associated with clinical utility for decision-making in healthcare. In distinction, the same IB could potentially be qualified for prediction of patient response (or lack of response) to a particular therapy based on evidence that the treatment effects in biomarker-positive patients differ from those in biomarker-negative patients (for example, superior outcome with experimental therapy in biomarker-positive patients versus biomarker-negative patients)106. The IB might be used for enrichment or stratification of patient populations in clinical trials and, if the trial is successful, such IB could be further developed into a companion diagnostic with established clinical utility for identifying patients likely to benefit from the new therapy.

Qualification requires demonstration of fitness for a particular purpose and, therefore, an IB might need to be evaluated in several scenarios to justify a broad claim of qualification. Hence, the IB might become qualified for one particular use (such as for a cancer type or a certain drug class), but not for others. Nevertheless, when a biomarker is qualified in multiple settings (for example, different tumour types, different therapies or different research questions), then the process of qualification for a new application can be expedited because most of the necessary validation requirements are likely to have been fulfilled107.

'Regulatory qualification' is a more-specific term that describes a framework for the evaluation and acceptance of a biomarker for specific use in regulatory decision-making108 (Supplementary information S1 (table)). Examples include new drug approvals or safety monitoring. This framework for regulatory qualification is not a formal requirement for a drug developer who merely wishes to 'qualify' an IB to support a proof-of-mechanism or a dose-selection decision.

Roadmap application

With this Consensus Statement we aim to accelerate IB translation. For this purpose, we have produced 14 key recommendations (Box 7), accompanied by a detailed imaging biomarker roadmap for cancer studies (Fig. 2). Together, the roadmap and recommendations provide a framework for understanding how to examine qualitative and quantitative IBs at all stages of their validation and qualification. Herein, we identify the key steps required to achieve these goals, recognizing the important differences between IBs and biospecimen-derived biomarkers. No biomarker is expected to be perfect; instead, this roadmap provides a tool to assess the current evidence for technical and biological and/or clinical performance of any given IB at any stage of development. The limitations of each IB can be identified, quantified, documented and made publicly available. In the process of creating this roadmap, we have identified several potential obstacles in the clinical translation of putative IBs.

Key recommendations

Recommendations 1–2 — grant submissions and study publications. Proposals for funding to support IB-related studies should state clearly how these will advance IB validation or qualification. Resulting journal publications should state explicitly how these aims have been achieved (recommendation 1). Study design, protocols, quality assurance processes and standard operating procedures should be reported exhaustively by making full use of supplementary materials for research publications. The software used for image processing should also be reported and be made available once intellectual property rights have been addressed (recommendation 2). These two recommendations will result in the greatest possible confidence in the reported IB-related data, will facilitate the conduct of statistically valid meta-analysis studies of imaging data, and will help investigators, reviewers, target audiences, funders and governments evaluate the risks associated with each IB for any given research or healthcare application.

Recommendations 3–7 — technical (assay) validation. A compelling rationale supports the accreditation of clinical imaging laboratories as being competent for measuring a given IB (recommendation 3), in line with the standards set by the biospecimen-derived biomarker community4. This approach has been adopted by the PET community when performing quantitative 18F-FDG-PET in clinical trials both in Europe and North America, focusing on performing regular equipment calibration and quality assurance. Moreover, data acquisition, analysis and reporting standards have been adopted32. In Europe, this initiative has been driven by the EANM and endorsed by the EORTC to ensure that institutions involved in multicentre clinical trials adopt best-practice procedures, and that quantitative reporting is harmonized across sites to improve reproducibility. Participants receive certification to distinguish them from nonparticipating institutions. Similarly, standardization of DCE-ultrasonography109 and LVEF measurements47 have been addressed by expert consensus. In the USA, the NCI evaluates performance of dynamic and static PET, volumetric CT and DCE-MRI in partnership with the ACRIN29. All NCI-designated cancer centres have been certified for measurement value equivalence, with ongoing annual inspection to provide regulation39. This certification involves the incorporation of IBs in clinical trial design and requires evaluation of the performance of imaging sites by the NCI Clinical Trials Network26.

Site accreditation is an important step towards improving technical performance in multicentre studies, but must reflect widely sought academic consensus, become adopted by international societies, and receive the backing of funders, industry and regulators for such accreditation to have value. Accreditation must simultaneously promote standardization and harmonization of IBs for multicentre use, while accommodating studies led by investigators who have scientific freedom to further develop and optimize IBs. To achieve this optimization, best-practice guidelines for each widely used IB (or related family of IBs) must be updated and reviewed regularly (recommendation 4), because biomarker drift is inevitable owing to technological advances in scanner performance (for example, clinical trials with ongoing data collection can be affected by hardware and software upgrades). Suitable statistical methods, such as multivariate linear regression analysis and other more complex statistical approaches, must be used to adjust for changes in IBs during ongoing studies (for example, defining pre-change and post-change data).

IB precision must be demonstrated early in IB development through single-centre repeatability studies, or few-site reproducibility studies (recommendation 5). This assessment is particularly important when testing the ability of an IB to measure the effect of therapeutic intervention. The choice of performance metric is important; for example, the coefficient of variation assumes that the standard deviation is approximately proportional to the mean. If studies of repeatability are performed under multiple conditions (such as imaging patients across different groups), then a plot of the standard deviation versus mean should be examined to determine whether this proportionality is valid. In most imaging studies of repeatability, only two replicates are acquired for each patient and thus, a Bland Altman plot will enable the evaluation of whether or not the mean of the measurements influence on variance, providing overall limits of agreement110.

Once the IB is shown to have sufficient technical and biological validity to cross translational gap 2, multicentre reproducibility must be evaluated (recommendation 6). Some studies have measured IB multicentre reproducibility (for example, DCE-CT evaluation of ovarian cancer111), but these are rare exceptions. The recruitment of patients with cancer to attend for scanning on devices at multiple centres is seldom possible and thus, studies usually require the use of data from different patients, scanned on different machines. Considerable centre-specific differences might exist regarding devices, contrast agents and tracers, and software. This variability must be accounted for when considering multicentre reproducibility. Mixed-effects modelling provides a statistically robust approach to maximising data inclusion, while acknowledging inevitable slight inconsistencies in the data112.

Data-analysis strategies must be developed for multicentre studies (recommendation 7). Analysis led by one central site can reduce data variation for studies with a moderate number of participating sites113,114. As IBs transition towards crossing gap 2, however, using one central site is inappropriate for IBs that will eventually be analysed at many cancer centres once the IB has been adopted in healthcare. To facilitate this transition, sites should compare their own technical performance against a central analysis, similarly to the assessment of objective responses in oncology trials.

Recommendations 8–11 — biological validation and clinical validation. Clinical validation occurs relatively late in the development process for most IBs compared with biospecimen-derived biomarkers71. Extensive preclinical studies can provide well-controlled data to examine the relationship of the IB to pathology, and the IB to the effects of interventions, and therefore are strongly encouraged (recommendation 8). The choice of experimental model is an important consideration. Tumour xenograft models in immunodeficient mice are well-studied and have reproducible growth characteristics115, and can be ideal for initial IB development, but tumour models that better portray the relevant biological characteristics found in human cancers are also needed116,117. Firstly, in situ tumours, or those implanted into their orthotopic site often better recapitulate the local microenvironment of human tumours117,118. Secondly, syngeneic models, with intact immune systems, are essential for some studies (for example, for the evaluation of IBs for immunotherapies). Thirdly, appropriately genetically-engineered mouse models can have lesions that accurately mimic human tumours119, an approach that can facilitate co-clinical trials in which preclinical studies are run in parallel with clinical trials in order to identify likely responders to targeted therapies118,120. Finally, models derived from patient tissue (PDX models)121 or circulating tumour cells (CDX models)122 offer potential insights into developing personalized therapeutic regimens123. The biological validation of IBs must incorporate the use of these models once proof-of-concept has been demonstrated in xenograft models. When possible, IBs should then be validated in clinical studies124, in order to confirm imaging–biology relationships in humans. Adaptive trial designs can be useful for early stage IB studies125. Sample-size re-estimation can be performed in those studies in which limited relevant data inform on sample size126. Similarly, group- sequential design127 can provide flexibility in the number of animals or patients entered into a study. Such approaches can ensure adequate power with small sample sizes — if effect sizes are suitably large — and, therefore, can make imaging studies more affordable128.

Improved methods for imaging–biology correlation are needed for IB biological validation (recommendation 9), because currently available methods often fail to account for extensive spatial heterogeneity in the IB129 and in the tissue pathology118, and for the difference in scale of the two measurements. Moderate-level concordance between imaging data (slice or volume) and pathology (single section) provide only limited biological validation. Instead, 3D comparisons are strongly encouraged, to better co-localise imaging and pathology data and reduce sampling bias. In some studies, genomic, transcriptomic and proteomic biomarkers, or biofluid-derived biomarkers, including circulating tumour cells, might be more-appropriate validation tools than tissue-derived measurements91,130. The use of appropriate statistical methods is required to define associations between imaging and pathology measurements. Spearman's rho or Kendall's tau rank correlation coefficients enable the comparison of IB values to reference standard pathology readouts within a cohort; however, if these two measurements (or transformations of these variables, such as logarithms) relate linearly to one another, regression models provide estimates of biases and the scale of the relationship.

IB studies often generate rich datasets that can be collected and banked for future re-use. These data include raw images, processed images and parameter maps, ancillary files (such as regions of interest, mask files, and stored header information detailing scan parameter settings), and essential patient metadata covering demographics, treatment history and clinical outcome (such as response, progression-free survival and/or overall survival). Useful data-archiving systems require financial support; collaboration between multiple academic, industry and funding partners; and ongoing curation and active management — as exemplified by the NCI Informatics Technology for Cancer Research platform131. Technological advances (such as cloud-based solutions) must be accompanied by suitable information governance arrangements, potentially crossing international legal and regulatory frameworks102, and by commitment from funders to resource these initiatives.

Creation of animal and human cancer image repositories, following the lead of organizations including the NCI and the ESR132, is strongly encouraged (recommendation 10) to enable rapid testing of new analyses by researchers from different institutions. In some cases, this approach can even lead to a reduction in the number of animal experiments (in line with the 3Rs of animal welfare in cancer research116) or in the number of new patients recruited. Standardized data collection, analysis and archiving are required to support multisite clinical trials102.

Complete and transparent study reporting is essential to avoid selective reporting bias and publication bias (recommendation 11). Results of all prespecified analyses (secondary analyses as well as primary analyses) should be reported, regardless of whether they are consistent with expectations, or whether they achieve statistical significance133,134. Highlighted exploratory analyses (either prespecified or post hoc) should be accompanied by a description of all other exploratory analyses performed. This strategy will more-accurately reflect the potential for false-positive and false-negative findings. Adherence to these principles of reporting will help to eliminate the distortion of research findings resulting from selective reporting and failure to publish negative studies. These biases can also be reduced by registering studies and detailing their hypotheses on publically available websites (such as before the study is initiated or before outcome data are unblinded89,135.

Recommendation 12 — qualification. Robust study design and predefined statistical analysis plans are vital to ensuring that the highest quality evidence is available to qualify IBs. Late-stage multicentre clinical trials should be powered adequately to demonstrate clinically useful effects. At present, few IBs are rigorously validated or qualified as prognostic for quality of life, progression-free survival or overall survival136. Appropriate statistical methods (for example, control of the false-discovery rate and cross-validation techniques137) are needed to avoid spurious findings and overfitting caused by measurement of large numbers of image parameters relative to numbers of patients — a common problem in several published biospecimen-based biomarker studies138. Efficient study designs should be pursued, for example, in studies of diagnostic accuracy135 or test impact, in which each patient is regarded as his or her own control; examples include adding functional imaging to standard anatomical radiology in order to improve diagnosis139,140. In later-stage clinical trials, several treatments can be tested concurrently in the same trial using enrichment or predictive IBs, as is the case of current molecular profiling-based studies141. This measure can increase recruitment efficiency, reduce costs, and accelerate achievement of primary end points142.

Qualification requires rigorous and detailed statistical reporting standards. Screening and diagnostic accuracy IB studies should report sensitivity and specificity, results of receiver operating characteristic (ROC) curve analysis, and negative and/or positive predictive values135, whereas prognostic and predictive IB studies should report estimated effect sizes, such as hazard ratios89. Estimates should be accompanied by measures of uncertainty (for example, 95% confidence intervals) as well as statistical significance. In studies in which predictive biomarkers are evaluated using randomized controlled trials, CONSORT guidelines should be followed143.

Recommendations 13–14 — cost-effectiveness. New models for funding and regulation should be developed for investigational devices, tracers and contrast agents, and software that have value as IBs in the research setting, but lack commercial viability as a diagnostic product in healthcare systems (recommendation 13). Imaging studies are perceived to be expensive and thus, integration of investigational IBs should be linked to existing radiological tests (addition of the IB to a clinically approved ultrasonography, CT or MRI examination, for example) whenever possible. Large-scale studies should include health-economic considerations, including measuring the cost-effectiveness of IBs versus competitor tests (other IBs or biospecimen-derived biomarkers)144 (recommendation 14).


Clinical imaging has transformed contemporary medicine145. IBs have enormous potential to facilitate further advances in cancer research and oncology practice by accurately informing clinical decision-making, but must undergo rigorous scrutiny through validation and qualification to achieve this end13. This process of evaluation enables investigators and consumers to make informed decisions about IB translation for each research and healthcare application10. The roadmap and recommendations that we present herein will, if adopted, mark a change in the development and use of IBs in cancer research and patient management.


  1. 1

    Biomarkers Definitions Working Group. Biomarkers and surrogate endpoints: preferred definitions and conceptual framework. Clin. Pharmacol. Ther. 69, 89–95 (2001).

  2. 2

    Food and Drug Administration & National Institutes of Health. BEST (Biomarkers, Endpoints, and other tools) resource. NCBI (accessed 4 February 2016).

  3. 3

    Lesko, L. J. & Atkinson, A. J. Jr. Use of biomarkers and surrogate endpoints in drug development and regulatory decision making: criteria, validation, strategies. Annu. Rev. Pharmacol. Toxicol. 41, 347–366 (2001).

    CAS  Article  PubMed  Google Scholar 

  4. 4

    Micheel, C., Nass, S. J. & Omenn, G. S. & Institute of Medicine Committee on the Review of Omics-Based Tests for Predicting Patient Outcomes in Clinical Trials. Evolution of translational omics: lessons learned and the path forward (National Academy of Sciences, 2012).

    Book  Google Scholar 

  5. 5

    Workman, P. et al. Minimally invasive pharmacokinetic and pharmacodynamic technologies in hypothesis-testing clinical trials of innovative therapies. J. Natl Cancer Inst. 98, 580–598 (2006).

    CAS  Article  PubMed  Google Scholar 

  6. 6

    Yap, T. A., Sandhu, S. K., Workman, P. & de Bono, J. S. Envisioning the future of early anticancer drug development. Nat. Rev. Cancer 10, 514–523 (2010).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  7. 7

    Hait, W. N. Forty years of translational cancer research. Cancer Discov. 1, 383–390 (2011).

    Article  PubMed  Google Scholar 

  8. 8

    La Thangue, N. B. & Kerr, D. J. Predictive biomarkers: a paradigm shift towards personalized cancer medicine. Nat. Rev. Clin. Oncol. 8, 587–596 (2011).

    CAS  Article  PubMed  Google Scholar 

  9. 9

    Poste, G. Bring on the biomarkers. Nature 469, 156–157 (2011).

    CAS  Article  PubMed  Google Scholar 

  10. 10

    Hayes, D. F. et al. Breaking a vicious cycle. Sci. Transl. Med. 5, 196cm6 (2013).

    Article  PubMed  Google Scholar 

  11. 11

    Macleod, M. R. et al. Biomedical research: increasing value, reducing waste. Lancet 383, 101–104 (2014).

    Article  PubMed  Google Scholar 

  12. 12

    Lee, J. W. et al. Method validation and measurement of biomarkers in nonclinical and clinical samples in drug development: a conference report. Pharm. Res. 22, 499–511 (2005).

    CAS  Article  PubMed  Google Scholar 

  13. 13

    Waterton, J. C. & Pylkkanen, L. Qualification of imaging biomarkers for oncology drug development. Eur. J. Cancer 48, 409–415 (2012).

    CAS  Article  PubMed  Google Scholar 

  14. 14

    Cooksey, D. A review of UK health research funding (TSO, 2006).

  15. 15

    Sung, N. S. et al. Central challenges facing the national clinical research enterprise. JAMA 289, 1278–1287 (2003).

    Article  PubMed  Google Scholar 

  16. 16

    Lee, J. W. et al. Fit-for-purpose method development and validation for successful biomarker measurement. Pharm. Res. 23, 312–328 (2006).

    CAS  Article  PubMed  Google Scholar 

  17. 17

    Goodsaid, F. M., Frueh, F. W. & Mattes, W. Strategic paths for biomarker qualification. Toxicology 245, 219–223 (2008).

    CAS  Article  PubMed  Google Scholar 

  18. 18

    Srivastava, S. et al. Translational Research Working Group developmental pathway for biospecimen-based assessment modalities. Clin. Cancer Res. 14, 5672–5677 (2008).

    PubMed  PubMed Central  Article  Google Scholar 

  19. 19

    Chetty, R. K. et al. A systematic approach to preclinical and clinical safety biomarker qualification incorporating Bradford Hill's principles of causality association. Clin. Pharmacol. Ther. 88, 260–262 (2010).

    CAS  Article  PubMed  Google Scholar 

  20. 20

    Khleif, S. N., Doroshow, J. H. & Hait, W. N. AACR-FDA-NCI Cancer Biomarkers Collaborative consensus report: advancing the use of biomarkers in cancer drug development. Clin. Cancer Res. 16, 3299–3318 (2010).

    CAS  Article  PubMed  Google Scholar 

  21. 21

    Cancer Research UK. Pharmacological biomarker roadmap. Cancer Research UK (accessed 30 November 2015).

  22. 22

    NIH. (2004).

  23. 23

    Food and Drug Administration. Clinical trial imaging endpoint process standards guidance for industry[draft guidance]. (2015).

  24. 24

    Woodcock, J. & Woosley, R. The FDA critical path initiative and its influence on new drug development. Annu. Rev. Med. 59, 1–12 (2008).

    CAS  Article  PubMed  Google Scholar 

  25. 25

    Clarke, L. P. et al. The Quantitative Imaging Network: NCI's historical perspective and planned Goals. Transl. Oncol. 7, 1–4 (2014).

    PubMed  PubMed Central  Article  Google Scholar 

  26. 26

    Shankar, L. K. The clinical evaluation of novel imaging methods for cancer management. Nat. Rev. Clin. Oncol. 9, 738–744 (2012).

    CAS  Article  PubMed  Google Scholar 

  27. 27

    Huang, E. P. et al. Meta-analysis of the technical performance of an imaging procedure: guidelines and statistical methodology. Stat. Methods Med. Res. 24, 141–174 (2015).

    Article  PubMed  Google Scholar 

  28. 28

    Sullivan, D. C. et al. Metrology standards for quantitative imaging biomarkers. Radiology 277, 813–825 (2015).

    PubMed  PubMed Central  Article  Google Scholar 

  29. 29

    National Cancer Institute. Cancer Imaging Program. (accessed 30 November 2015).

  30. 30

    European Society of Radiology. ESR statement on the stepwise development of imaging biomarkers. Insights Imaging 4, 147–152 (2013).

  31. 31

    Hunter, A. J. The Innovative Medicines Initiative: a pre-competitive initiative to enhance the biomedical science base of Europe to expedite the development of new medicines for patients. Drug Discov. Today 13, 371–373 (2008).

    Article  PubMed  Google Scholar 

  32. 32

    Boellaard, R. et al. FDG PET/CT: EANM procedure guidelines for tumour imaging: version 2.0. Eur. J. Nucl. Med. Mol. Imaging 42, 328–354 (2015).

    CAS  Article  PubMed  Google Scholar 

  33. 33

    Tofts, P. S. et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J. Magn. Reson. Imaging 10, 223–232 (1999).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  34. 34

    Leach, M. O. et al. The assessment of antiangiogenic and antivascular therapies in early-stage clinical trials using magnetic resonance imaging: issues and recommendations. Br. J. Cancer 92, 1599–1610 (2005).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  35. 35

    Shankar, L. K. et al. Consensus recommendations for the use of 18F-FDG PET as an indicator of therapeutic response in patients in National Cancer Institute Trials. J. Nucl. Med. 47, 1059–1066 (2006).

    CAS  PubMed  Google Scholar 

  36. 36

    Padhani, A. R. et al. Diffusion-weighted magnetic resonance imaging as a cancer biomarker: consensus and recommendations. Neoplasia 11, 102–125 (2009).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  37. 37

    Leen, E. et al. Dynamic contrast enhanced ultrasound assessment of the vascular effects of novel therapeutics in early stage trials. Eur. Radiol. 22, 1442–1450 (2012).

    Article  PubMed  Google Scholar 

  38. 38

    Dorfman, G. S., Sullivan, D. C., Schnall, M. D. & Matrisian, L. M. The Translational Research Working Group developmental pathway for image-based assessment modalities. Clin. Cancer Res. 14, 5678–5684 (2008).

    Article  PubMed  Google Scholar 

  39. 39

    Buckler, A. J. et al. Quantitative imaging test approval and biomarker qualification: interrelated but distinct activities. Radiology 259, 875–884 (2011).

    PubMed  PubMed Central  Article  Google Scholar 

  40. 40

    Barker, A., Kettle, J. G., Nowak, T. & Pease, J. E. Expanding medicinal chemistry space. Drug Discov. Today 18, 298–304 (2013).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  41. 41

    Blagg, J. & Workman, P. Chemical biology approaches to target validation in cancer. Curr. Opin. Pharmacol. 17, 87–100 (2014).

    CAS  Article  PubMed  Google Scholar 

  42. 42

    American College of Radiology. Breast imaging reporting and data system atlas (BI-RADS atlas) (American College of Radiology, 2013).

  43. 43

    Edge, S. B. et al. AJCC Cancer Staging Handbook 7th edn (Springer, 2010).

    Google Scholar 

  44. 44

    Therasse, P. et al. New guidelines to evaluate the response to treatment in solid tumors. J. Natl Cancer Inst. 92, 205–216 (2000).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  45. 45

    Eisenhauer, E. A. et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur. J. Cancer 45, 228–247 (2009).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  46. 46

    Reid, D. M. et al. Guidance for the management of breast cancer treatment-induced bone loss: a consensus position statement from a UK Expert Group. Cancer Treat. Rev. 34 (Suppl. 1), S3–S18 (2008).

    CAS  Article  PubMed  Google Scholar 

  47. 47

    Plana, J. C. et al. Expert consensus for multimodality imaging evaluation of adult patients during and after cancer therapy: a report from the American Society of Echocardiography and the European Association of Cardiovascular Imaging. J. Am. Soc. Echocardiogr. 27, 911–939 (2014).

    Article  PubMed  Google Scholar 

  48. 48

    European Medicines Agency Committee for Medicinal Products for Human Use. Summary of opinion. EMA (accessed 30 November 2015).

  49. 49

    Food and Drug Administration. Novel drugs approved using surrogate endpoints. FDA (accessed 30 November 2015).

  50. 50

    Bergstrom, M. et al. Human positron emission tomography studies of brain neurokinin 1 receptor occupancy by aprepitant. Biol. Psychiatry 55, 1007–1012 (2004).

    Article  CAS  PubMed  Google Scholar 

  51. 51

    Willett, C. G. et al. Direct evidence that the VEGF-specific antibody bevacizumab has antivascular effects in human rectal cancer. Nat. Med. 10, 145–147 (2004).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  52. 52

    Avril, N. & Propper, D. Functional PET imaging in cancer drug development. Future Oncol. 3, 215–228 (2007).

    CAS  Article  PubMed  Google Scholar 

  53. 53

    O'Connor, J. P., Jackson, A., Parker, G. J., Roberts, C. & Jayson, G. C. Dynamic contrast-enhanced MRI in clinical trials of antivascular therapies. Nat. Rev. Clin. Oncol. 9, 167–177 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  54. 54

    Lassau, N. et al. Standardization of dynamic contrast-enhanced ultrasound for the evaluation of antiangiogenic therapies: the French multicenter Support for Innovative and Expensive Techniques Study. Invest. Radiol. 47, 711–716 (2012).

    CAS  Article  PubMed  Google Scholar 

  55. 55

    van Elmpt, W. et al. The PET-boost randomised phase II dose-escalation trial in non-small cell lung cancer. Radiother. Oncol. 104, 67–71 (2012).

    Article  PubMed  Google Scholar 

  56. 56

    Taylor, F. G. et al. Preoperative magnetic resonance imaging assessment of circumferential resection margin predicts disease-free survival and local recurrence: 5-year follow-up results of the MERCURY study. J. Clin. Oncol. 32, 34–43 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

  57. 57

    Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: the next generation. Cell 144, 646–674 (2011).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  58. 58

    Ueda, S. et al. Baseline tumor oxygen saturation correlates with a pathologic complete response in breast cancer patients undergoing neoadjuvant chemotherapy. Cancer Res. 72, 4318–4328 (2012).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  59. 59

    Gallagher, F. A., Kettunen, M. I. & Brindle, K. M. Imaging pH with hyperpolarized 13C. NMR Biomed. 24, 1006–1015 (2011).

    CAS  Article  PubMed  Google Scholar 

  60. 60

    Lambin, P. et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur. J. Cancer 48, 441–446 (2012).

    PubMed  PubMed Central  Article  Google Scholar 

  61. 61

    Gillies, R. J., Kinahan, P. E. & Hricak, H. Radiomics: images are more than pictures, they are data. Radiology 278, 563–577 (2016).

    Article  PubMed  Google Scholar 

  62. 62

    Fletcher, J. W. et al. Recommendations on the use of 18F-FDG PET in oncology. J. Nucl. Med. 49, 480–508 (2008).

    Article  PubMed  Google Scholar 

  63. 63

    Wahl, R. L., Jacene, H., Kasamon, Y. & Lodge, M. A. From RECIST to PERCIST: evolving considerations for PET response criteria in solid tumors. J. Nucl. Med. 50 (Suppl. 1), 122S–150S (2009).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  64. 64

    Barrington, S. F. et al. Role of imaging in the staging and response assessment of lymphoma: consensus of the International Conference on Malignant Lymphomas Imaging Working Group. J. Clin. Oncol. 32, 3048–3058 (2014).

    PubMed  PubMed Central  Article  Google Scholar 

  65. 65

    Na, F. et al. Primary tumor standardized uptake value measured on F18-Fluorodeoxyglucose positron emission tomography is of prediction value for survival and local control in non-small-cell lung cancer receiving radiotherapy: meta-analysis. J. Thorac. Oncol. 9, 834–842 (2014).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  66. 66

    Kayani, I. et al. Sequential FDG-PET/CT as a biomarker of response to sunitinib in metastatic clear cell renal cancer. Clin. Cancer Res. 17, 6021–6028 (2011).

    CAS  Article  PubMed  Google Scholar 

  67. 67

    Sarker, D. et al. First-in-human phase I study of pictilisib (GDC-0941), a potent pan-class I phosphatidylinositol-3-kinase (PI3K) inhibitor, in patients with advanced solid tumors. Clin. Cancer Res. 21, 77–86 (2015).

    CAS  Article  PubMed  Google Scholar 

  68. 68

    Michaelis, L. C. & Ratain, M. J. Measuring response in a post-RECIST world: from black and white to shades of grey. Nat. Rev. Cancer 6, 409–414 (2006).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  69. 69

    Lassau, N. et al. Advanced hepatocellular carcinoma: early evaluation of response to bevacizumab therapy at dynamic contrast-enhanced US with quantification—preliminary results. Radiology 258, 291–300 (2011).

    Article  PubMed  Google Scholar 

  70. 70

    Patterson, D. M., Padhani, A. R. & Collins, D. J. Technology Insight: water diffusion MRI—a potential new biomarker of response to cancer therapy. Nat. Clin. Pract. Oncol. 5, 220–233 (2008).

    Article  PubMed  PubMed Central  Google Scholar 

  71. 71

    Waterton, J. C. in New Applications of NMR in Drug Discovery and Development (eds Garrido, L. & Beckmann, N.) 341–368 (Royal Society of Chemistry, 2013).

    Google Scholar 

  72. 72

    Parkinson, D. R., Johnson, B. E. & Sledge, G. W. Making personalized cancer medicine a reality: challenges and opportunities in the development of biomarkers and companion diagnostics. Clin. Cancer Res. 18, 619–624 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

  73. 73

    Gonzalez de Castro, D., Clarke, P. A., Al-Lazikani, B. & Workman, P. Personalized cancer medicine: molecular diagnostics, predictive biomarkers, and drug resistance. Clin. Pharmacol. Ther. 93, 252–259 (2013).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  74. 74

    Hayes, D. F. Biomarker validation and testing. Mol. Oncol. 9, 960–966 (2015).

    Article  PubMed  Google Scholar 

  75. 75

    Schilsky, R. L. Personalized medicine in oncology: the future is now. Nat. Rev. Drug Discov. 9, 363–366 (2010).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  76. 76

    Cai, W. & Chen, X. Multimodality molecular imaging of tumor angiogenesis. J. Nucl. Med. 49 (Suppl. 2), 113S–128S (2008).

    CAS  Article  PubMed  Google Scholar 

  77. 77

    International Organization for Standardization. Statistics–vocabulary and symbols–part 2: applied statistics. ISO (accessed 30 November 2015).

  78. 78

    Barnes, S. L., Whisenant, J. G., Loveless, M. E. & Yankeelov, T. E. Practical dynamic contrast enhanced MRI in small animal models of cancer: data acquisition, data analysis, and interpretation. Pharmaceutics 4, 442–478 (2012).

    PubMed  PubMed Central  Article  Google Scholar 

  79. 79

    Jackson, A. et al. Reproducibility of quantitative dynamic contrast-enhanced MRI in newly presenting glioma. Br. J. Radiol. 76, 153–162 (2003).

    CAS  Article  PubMed  Google Scholar 

  80. 80

    Galbraith, S. M. et al. Reproducibility of dynamic contrast-enhanced MRI in human muscle and tumours: comparison of quantitative and semi-quantitative analysis. NMR Biomed. 15, 132–142 (2002).

    Article  PubMed  Google Scholar 

  81. 81

    Koh, D. M. et al. Reproducibility and changes in the apparent diffusion coefficients of solid tumours treated with combretastatin A4 phosphate and bevacizumab in a two-centre phase I clinical trial. Eur. Radiol. 19, 2728–2738 (2009).

    Article  PubMed  Google Scholar 

  82. 82

    O'Connor, J. P. et al. Quantifying antivascular effects of monoclonal antibodies to vascular endothelial growth factor: insights from imaging. Clin. Cancer Res. 15, 6674–6682 (2009).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  83. 83

    Haacke, E. M. et al. New algorithm for quantifying vascular changes in dynamic contrast-enhanced MRI independent of absolute T1 values. Magn. Reson. Med. 58, 463–472 (2007).

    Article  PubMed  Google Scholar 

  84. 84

    Leach, M. O. et al. Screening with magnetic resonance imaging and mammography of a UK population at high familial risk of breast cancer: a prospective multicentre cohort study (MARIBS). Lancet 365, 1769–1778 (2005).

    CAS  Article  PubMed  Google Scholar 

  85. 85

    Kessler, L. G. et al. The emerging science of quantitative imaging biomarkers terminology and definitions for scientific studies and regulatory submissions. Stat. Methods Med. Res. 24, 9–26 (2015).

    Article  PubMed  Google Scholar 

  86. 86

    Lewis, M. et al. Quality control within the multicentre perfusion CT study of primary colorectal cancer (PROSPeCT): results of an iodine density phantom study. Eur. Radiol. 24, 2309–2318 (2014).

    Article  PubMed  Google Scholar 

  87. 87

    Henoumont, C., Laurent, S. & Vander Elst, L. How to perform accurate and reliable measurements of longitudinal and transverse relaxation times of MRI contrast media in aqueous solutions. Contrast Media Mol. Imaging 4, 312–321 (2009).

    CAS  Article  PubMed  Google Scholar 

  88. 88

    McShane, L. M. & Hayes, D. F. Publication of tumor marker research results: the necessity for complete and transparent reporting. J. Clin. Oncol. 30, 4223–4232 (2012).

    PubMed  PubMed Central  Article  Google Scholar 

  89. 89

    McShane, L. M. et al. Reporting recommendations for tumor marker prognostic studies (REMARK). J. Natl Cancer Inst. 97, 1180–1184 (2005).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  90. 90

    Chand, M. et al. The prognostic significance of postchemoradiotherapy high-resolution MRI and histopathology detected extramural venous invasion in rectal cancer. Ann. Surg. 261, 473–479 (2015).

    Article  PubMed  Google Scholar 

  91. 91

    Aerts, H. J. et al. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nat. Commun. 5, 4006 (2014).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  92. 92

    Prentice, R. L. Surrogate endpoints in clinical trials: definition and operational criteria. Stat. Med. 8, 431–440 (1989).

    CAS  Article  PubMed  Google Scholar 

  93. 93

    Freedman, L. S., Graubard, B. I. & Schatzkin, A. Statistical validation of intermediate endpoints for chronic diseases. Stat. Med. 11, 167–178 (1992).

    CAS  Article  PubMed  Google Scholar 

  94. 94

    Buyse, M., Sargent, D. J., Grothey, A., Matheson, A. & de Gramont, A. Biomarkers and surrogate end points—the challenge of statistical validation. Nat. Rev. Clin. Oncol. 7, 309–317 (2010).

    Article  PubMed  Google Scholar 

  95. 95

    Hill, A. B. The environment and disease: association or causation? Proc. R. Soc. Med. 58, 295–300 (1965).

    PubMed  PubMed Central  CAS  Google Scholar 

  96. 96

    Wagner, J. A. Strategic approach to fit-for-purpose biomarkers in drug development. Annu. Rev. Pharmacol. Toxicol. 48, 631–651 (2008).

    CAS  Article  PubMed  Google Scholar 

  97. 97

    Maxwell, R. J. et al. Evaluation of the anti-vascular effects of combretastatin in rodent tumours by dynamic contrast enhanced MRI. NMR Biomed. 15, 89–98 (2002).

    CAS  Article  PubMed  Google Scholar 

  98. 98

    LoRusso, P. M. et al. Phase I clinical evaluation of ZD6126, a novel vascular-targeting agent, in patients with solid tumors. Invest. New Drugs 26, 159–167 (2008).

    CAS  Article  PubMed  Google Scholar 

  99. 99

    Drevs, J. et al. Phase I clinical study of AZD2171, an oral vascular endothelial growth factor signaling inhibitor, in patients with advanced solid tumors. J. Clin. Oncol. 25, 3045–3054 (2007).

    CAS  Article  PubMed  Google Scholar 

  100. 100

    Jonker, D. J. et al. A phase I study to determine the safety, pharmacokinetics and pharmacodynamics of a dual VEGFR and FGFR inhibitor, brivanib, in patients with advanced or metastatic solid tumors. Ann. Oncol. 22, 1413–1419 (2011).

    CAS  Article  PubMed  Google Scholar 

  101. 101

    Gazelle, G. S., McMahon, P. M., Siebert, U. & Beinfeld, M. T. Cost-effectiveness analysis in the assessment of diagnostic imaging technologies. Radiology 235, 361–370 (2005).

    Article  PubMed  Google Scholar 

  102. 102

    NIH. (accessed 30 November 2015).

  103. 103

    O'Connor, J. P. et al. Quantitative imaging biomarkers in the clinical development of targeted therapeutics: current and future perspectives. Lancet Oncol. 9, 766–776 (2008).

    CAS  Article  PubMed  Google Scholar 

  104. 104

    Micheel, C. M. & Ball, J. R. Evaluation of Biomarkers and Surrogate Endpoints in Chronic Disease (National Academy of Sciences, 2010).

    Google Scholar 

  105. 105

    European Medicines Agency. Qualification opinion. (2015).

  106. 106

    Polley, M. Y. et al. Statistical and practical considerations for clinical evaluation of predictive biomarkers. J. Natl Cancer Inst. 105, 1677–1683 (2013).

    PubMed  PubMed Central  Article  Google Scholar 

  107. 107

    Sargent, D. J. et al. Validation of novel imaging methodologies for use as cancer clinical trial end-points. Eur. J. Cancer 45, 290–299 (2009).

    CAS  Article  PubMed  Google Scholar 

  108. 108

    Food and Drug Administration. Biomarker qualification program. FDA (accessed 30 November 2015).

  109. 109

    Dietrich, C. F. et al. An EFSUMB introduction into Dynamic Contrast-Enhanced Ultrasound (DCE-US) for quantification of tumour perfusion. Ultraschall Med. 33, 344–351 (2012).

    CAS  Article  PubMed  Google Scholar 

  110. 110

    Bland, J. M. & Altman, D. G. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 327, 307–310 (1986).

    Article  Google Scholar 

  111. 111

    Lee, T. Y. et al. CT perfusion as an early biomarker of treatment efficacy in advanced ovarian cancer: An ECOG-ACRIN and NRG GOG study 9 [abstract]. J. Clin. Oncol. 33, 5522 (2015).

    Article  Google Scholar 

  112. 112

    Chu, R. et al. Comparing methods to estimate treatment effects on a continuous outcome in multicentre randomized controlled trials: a simulation study. BMC Med. Res. Methodol. 11, 21 (2011).

    PubMed  PubMed Central  Article  Google Scholar 

  113. 113

    Nuyts, J. et al. Cardiac phantom measurement validating the methodology for a cardiac multi-centre trial with positron emission tomography. Eur. J. Nucl. Med. Mol. Imaging 29, 1588–1593 (2002).

    Article  PubMed  Google Scholar 

  114. 114

    Yankeelov, T. E. et al. Quantitative imaging in cancer clinical trials. Clin. Cancer Res. 22, 284–290 (2016).

    PubMed  PubMed Central  Article  Google Scholar 

  115. 115

    O'Connor, J. P. B. et al. Oxygen-enhanced MRI accurately identifies, quantifies, and maps tumor hypoxia in preclinical cancer models. Cancer Res. 76, 787–795 (2016).

    CAS  Article  PubMed  Google Scholar 

  116. 116

    Workman, P. et al. Guidelines for the welfare and use of animals in cancer research. Br. J. Cancer 102, 1555–1577 (2010).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  117. 117

    Frese, K. K. & Tuveson, D. A. Maximizing mouse cancer models. Nat. Rev. Cancer 7, 645–658 (2007).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  118. 118

    Nardella, C., Lunardi, A., Patnaik, A., Cantley, L. C. & Pandolfi, P. P. The APL paradigm and the “co-clinical trial” project. Cancer Discov. 1, 108–116 (2011).

    PubMed  PubMed Central  Article  Google Scholar 

  119. 119

    Sharpless, N. E. & Depinho, R. A. The mighty mouse: genetically engineered mouse models in cancer drug development. Nat. Rev. Drug Discov. 5, 741–754 (2006).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  120. 120

    Graham, T. J. et al. Preclinical evaluation of imaging biomarkers for prostate cancer bone metastasis and response to cabozantinib. J. Natl Cancer Inst. 106, dju033 (2014).

    Article  CAS  PubMed  Google Scholar 

  121. 121

    Hidalgo, M. et al. Patient-derived xenograft models: an emerging platform for translational cancer research. Cancer Discov. 4, 998–1013 (2014).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  122. 122

    Hodgkinson, C. L. et al. Tumorigenicity and genetic profiling of circulating tumor cells in small-cell lung cancer. Nat. Med. 20, 897–903 (2014).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  123. 123

    Malaney, P., Nicosia, S. V. & Dave, V. One mouse, one patient paradigm: new avatars of personalized cancer therapy. Cancer Lett. 344, 1–12 (2014).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  124. 124

    National Institute of Health. Oncology co-clinical imaging research resources to encourage consensus on quantitative imaging methods and precision medicine (U24). (accessed 19 February 2016).

  125. 125

    Wason, J., Marshall, A., Dunn, J., Stein, R. C. & Stallard, N. Adaptive designs for clinical trials assessing biomarker-guided treatment strategies. Br. J. Cancer 110, 1950–1957 (2014).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  126. 126

    Gould, A. L. Sample size re-estimation: recent developments and practical considerations. Stat. Med. 20, 2625–2643 (2001).

    CAS  Article  PubMed  Google Scholar 

  127. 127

    Jennison, C. & Turnbull, B. W. Group Sequential Methods with Applications to Clinical Trials (Chapman & Hall/CRC Press, 1999).

    Google Scholar 

  128. 128

    Gerety, E. L. et al. Prospective study evaluating the relative sensitivity of 18F-NaF PET/CT for detecting skeletal metastases from renal cell carcinoma in comparison to multidetector CT and 99mTc-MDP bone scintigraphy, using an adaptive trial design. Ann. Oncol. 26, 2113–2118 (2015).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  129. 129

    O'Connor, J. P. et al. Imaging intratumor heterogeneity: role in therapy response, resistance, and clinical outcome. Clin. Cancer Res. 21, 249–257 (2015).

    CAS  Article  PubMed  Google Scholar 

  130. 130

    Halle, C. et al. Hypoxia-induced gene expression in chemoradioresistant cervical cancer revealed by dynamic contrast-enhanced MRI. Cancer Res. 72, 5285–5295 (2012).

    CAS  Article  PubMed  Google Scholar 

  131. 131

    National Cancer Institute. Informatics technology for cancer research. NCI (accessed 30 November 2015).

  132. 132

    European Society of Radiology. ESR position paper on imaging biobanks. Insights Imaging 6, 403–410 (2015).

  133. 133

    Bernsen, M. R., Kooiman, K., Segbers, M., van Leeuwen, F. W. & de Jong, M. Biomarkers in preclinical cancer imaging. Eur. J. Nucl. Med. Mol. Imaging 42, 579–596 (2015).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  134. 134

    Boult, J. K. et al. False-negative MRI biomarkers of tumour response to targeted cancer therapeutics. Br. J. Cancer 106, 1960–1966 (2012).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  135. 135

    Bossuyt, P. M. et al. STARD 2015: an updated list of essential items for reporting diagnostic accuracy studies. BMJ 351, h5527 (2015).

    PubMed  PubMed Central  Article  Google Scholar 

  136. 136

    O'Connor, J. P. & Jayson, G. C. Do imaging biomarkers relate to outcome in patients treated with VEGF inhibitors? Clin. Cancer Res. 18, 6588–6598 (2012).

    CAS  Article  PubMed  Google Scholar 

  137. 137

    Stone, M. Cross-validatory choice and assessment of statistical predictions. J. R. Stat. Soc. Series B Stat. Methodol. 36, 111–147 (1974).

    Google Scholar 

  138. 138

    Ein-Dor, L., Zuk, O. & Domany, E. Thousands of samples are needed to generate a robust gene list for predicting outcome in cancer. Proc. Natl Acad. Sci. USA 103, 5923–5928 (2006).

    CAS  Article  PubMed  Google Scholar 

  139. 139

    Meads, C. et al. Positron emission tomography/computerised tomography imaging in detecting and managing recurrent cervical cancer: systematic review of evidence, elicitation of subjective probabilities and economic modelling. Health Technol. Assess. 17, 1–323 (2013).

    PubMed  PubMed Central  Google Scholar 

  140. 140

    Mowatt, G. et al. The diagnostic accuracy and cost-effectiveness of magnetic resonance spectroscopy and enhanced magnetic resonance imaging techniques in aiding the localisation of prostate abnormalities for biopsy: a systematic review and economic evaluation. Health Technol. Assess. 17, 1–281 (2013).

    PubMed  PubMed Central  Article  Google Scholar 

  141. 141

    Lawler, M., Kaplan, R., Wilson, R. H. & Maughan, T. Changing the paradigm-multistage multiarm randomized trials and stratified cancer medicine. Oncologist 20, 849–851 (2015).

    PubMed  PubMed Central  Article  Google Scholar 

  142. 142

    Wason, J. M. et al. A Bayesian adaptive design for biomarker trials with linked treatments. Br. J. Cancer 113, 699–705 (2015).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  143. 143

    Moher, D. et al. CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomised trials. BMJ 340, c869 (2010).

    PubMed  PubMed Central  Article  Google Scholar 

  144. 144

    Griebsch, I. et al. Cost-effectiveness of screening with contrast enhanced magnetic resonance imaging versus X-ray mammography of women at a high familial risk of breast cancer. Br. J. Cancer 95, 801–810 (2006).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

  145. 145

    Looking back on the millennium in medicine [Editorial]. N. Engl. J. Med. 342, 42–49 (2000).

  146. 146

    McLeod, D. G. et al. Bicalutamide 150 mg plus standard care versus standard care alone for early prostate cancer. BJU Int. 97, 247–254 (2005).

    Article  CAS  Google Scholar 

  147. 147

    World Health Organization. WHO Handbook for Reporting Results of Cancer Treatment (World Health Organization, 1979).

  148. 148

    Choi, H. et al. Correlation of computed tomography and positron emission tomography in patients with metastatic gastrointestinal stromal tumor treated at a single institution with imatinib mesylate: proposal of new computed tomography response criteria. J. Clin. Oncol. 25, 1753–1759 (2007).

    Article  PubMed  Google Scholar 

  149. 149

    Zagar, T. M., Cardinale, D. M. & Marks, L. B. Breast cancer therapy-associated cardiovascular disease. Nat. Rev. Clin. Oncol. 13, 172–184 (2016).

    CAS  Article  PubMed  Google Scholar 

  150. 150

    Food and Drug Administration. Herceptin. Highlights of prescribing information. Accessdata.FDA (accessed 8 October 2015).

  151. 151

    Avram, A. M. Radioiodine scintigraphy with SPECT/CT: an important diagnostic tool for thyroid cancer staging and risk stratification. J. Nucl. Med. 53, 754–764 (2012).

    Article  PubMed  Google Scholar 

  152. 152

    Kam, B. L. et al. Lutetium-labelled peptides for therapy of neuroendocrine tumours. Eur. J. Nucl. Med. Mol. Imaging 39 (Suppl. 1), S103–S112 (2012).

    Article  CAS  PubMed  Google Scholar 

  153. 153

    Ambrosini, V., Nanni, C. & Fanti, S. The use of gallium-68 labeled somatostatin receptors in PET/CT imaging. PET Clin. 9, 323–329 (2014).

    Article  PubMed  Google Scholar 

  154. 154

    National Institute for Clinical Excellence. Familial breast cancer: classification and care of people at risk of familial breast cancer and management of breast cancer and related risks in people with a family history of breast cancer. (2013).

  155. 155

    European Medicines Agency Committee for Medicinal Products for Human Use. Summary of opinion. EMA (accessed 30 November 2015).

  156. 156

    Quarles van Ufford, H. M. et al. Added value of baseline 18F-FDG uptake in serial 18F-FDG PET for evaluation of response of solid extracerebral tumors to systemic cytotoxic neoadjuvant treatment: a meta-analysis. J. Nucl. Med. 51, 1507–1516 (2010).

    Article  PubMed  Google Scholar 

  157. 157

    Lara, P. N. Jr. et al. SU5416 plus interferon alpha in advanced renal cell carcinoma: a phase II California Cancer Consortium Study with biological and imaging correlates of angiogenesis inhibition. Clin. Cancer Res. 9, 4772–4781 (2003).

    CAS  PubMed  Google Scholar 

  158. 158

    Bendell, J. C. et al. Phase I, dose-escalation study of BKM120, an oral pan-Class I PI3K inhibitor, in patients with advanced solid tumors. J. Clin. Oncol. 30, 282–290 (2012).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  159. 159

    Maynard, J., Ricketts, S. A., Gendrin, C., Dudley, P. & Davies, B. R. 2-Deoxy-2-[18F]fluoro-D-glucose positron emission tomography demonstrates target inhibition with the potential to predict anti-tumour activity following treatment with the AKT inhibitor AZD5363. Mol. Imaging Biol. 15, 476–485 (2013).

    Article  PubMed  Google Scholar 

  160. 160

    Engelman, J. A. Targeting PI3K signalling in cancer: opportunities, challenges and limitations. Nat. Rev. Cancer 9, 550–562 (2009).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  161. 161

    Kierkels, R. G. et al. Comparison between perfusion computed tomography and dynamic contrast-enhanced magnetic resonance imaging in rectal cancer. Int. J. Radiat. Oncol. Biol. Phys. 77, 400–408 (2010).

    Article  PubMed  Google Scholar 

  162. 162

    Naish, J. H. et al. Comparison of dynamic contrast-enhanced MRI and dynamic contrast-enhanced CT biomarkers in bladder cancer. Magn. Reson. Med. 66, 219–226 (2011).

    CAS  Article  PubMed  Google Scholar 

  163. 163

    Ng, C. S. et al. Dependence of DCE-MRI biomarker values on analysis algorithm. PLoS ONE 10, e0130168 (2015).

    PubMed  PubMed Central  Article  CAS  Google Scholar 

  164. 164

    Sabbatini, P. et al. Prognostic significance of extent of disease in bone in patients with androgen-independent prostate cancer. J. Clin. Oncol. 17, 948–957 (1999).

    CAS  Article  Google Scholar 

  165. 165

    Food and Drug Administration. 510(k) summary. Accessdata.FDA (accessed 30 November 2015).

  166. 166

    Food and Drug Administration. FDA approves Lymphoseek to help locate lymph nodes in patients with certain cancers. FDA (accessed 30 November 2015).

  167. 167

    European Medicines Agency. Withdrawn applications, vynfinit. EMA (accessed 30 November 2015).

  168. 168

    Lassau, N. et al. Validation of dynamic contrast-enhanced ultrasound in predicting outcomes of antiangiogenic therapy for solid tumors: the French multicenter support for innovative and expensive techniques study. Invest. Radiol. 49, 794–800 (2014).

    PubMed  PubMed Central  CAS  Article  Google Scholar 

Download references


Development of this roadmap received support from Cancer Research UK and the Engineering and Physical Sciences Research Council (grant references A/15267, A/16463, A/16464, A/16465, A/16466 and A/18097), the EORTC Cancer Research Fund, and the Innovative Medicines Initiative Joint Undertaking (grant agreement number 115151), resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and European Federation of Pharmaceutical Industries and Associations (EFPIA) companies' in kind contribution.

Author information




J.P.B.O'C., K.M.B., L.M.MS., A.J. and J.C.W. planned the article. J.P.B.O'C. and J.C.W. managed article writing. All authors researched data for the article according to their area of expertise, wrote relevant sections and reviewed the manuscript before submission. J.P.B.O'C., E.O.A., H.J.W.L.A., F.A.G., D.M.K., N.L., E.L., P.M., M.P., S.P., R.A.S. and J.C.W. prepared the display items. All authors edited the article before submission.

Corresponding author

Correspondence to James P. B. O'Connor.

Ethics declarations

Competing interests

J.P.B.O'C. (CRUK, Eli Lilly), H.J.W.L.A. (NIH), S.F.B. (AstraZeneca), R.B. (Philips), S.B. (CRUK), M.B. (CRUK, EPSRC), K.M.B. (CRUK, EPSRC, GE Healthcare), T.L.C. (NIH), S.C. (European Union FP7), G.J.C. (Alliance Medical, Bayer, Blue Earth Diagnostics, CRUK, GE Healthcare, Siemens Healthcare), N.M.dS. (CRUK, European Union FP7, NIHR), C.D. (CRUK), C.F-F. (AstraZeneca, Merck), F.A.G. (CRUK, GE, GSK), F.J.G. (CRUK, EPSRC), R.J.G. (NIH), V.G. (GE Healthcare, Siemens Healthcare), A.M.G. (GSK), A.L.H. (CRUK), D.J.H. (CRUK, EPSRC, Elekta, Philips, Truebeam), B.F.H. (GE Healthcare, Mediso, Siemens Healthcare), A.J. (CRUK, EPSRC), G.C.J. (AstraZeneca, Novartis, Oxigene, Roche), N.L. (Bracco), M.L. (CRUK, EPSRC), E.L.L. (GE Healthcare, Pfizer, Philips), J.S.L. (Eli Lilly, MabVAx Inc, NIH, The Curadel Companies), Y.L. (European Union FP7), T.N. (CRUK, EPSRC), A.R.P. (Acuitas Medical, Siemens Healthcare), G.J.M.P. (CRUK, EPSRC), M.P. (CRUK), A.C.P. (NIHR), R.A.S. (CRUK, HEFCE, MRC, Sirtex Technology), S.A.T. (NIHR), K.J.W. (Eli Lilly), P.W. (CRUK) and T.E.Y. (NIH) receive grants and financial support from the indicated sources. E.A. (Lightpoint Medical), H.J.W.L.A. (Genospace LLC), R.J.G. (HealthMyne Inc), T-Y.L. (GE Healthcare), K.A.M. (Imaging Endpoints), G.J.M.P. (GSK), R.A.S. (Affidea, BTG plc, Cancer Research Technology, Sirtex Medical, Vertex), D.C.S. (RSNA), S.A.T. (Robarts), M.vH. (Aquilab), P.W. (Astex Pharmaceuticals, Nextechinvest, Nuevolution) have acted as consultants for the indicated organizations. H.J.W.L.A. (Genospace LLC), J.L.E. (Merck & Co), R.J.G. (HealthMyne Inc), D.J.H. (IXICO PLC, visionRT), J.S.L. (pHLIP Inc), K.A.M. (Feedback PLC), G.J.M.P. (Bioxydyn Ltd), J.C.W. (Alderley Imaging Ltd, AstraZeneca), P.W. (Chroma Therapeutics, Nextechinvest, Piramed Pharma) hold stock and/or directorship from the indicated institutions. All other authors declare no competing interests. No authors declare relevant honoraria for speaking or patents.

Supplementary information

Supplementary information S1 (table 1)

Definitions of key terminology (PDF 252 kb)

Supplementary information S1 (table 2)

Important factors to consider when validating and qualifying imaging biomarkers (IBs) (PDF 192 kb)

Supplementary information S1 (box 3)

Receptor occupancy: pharmacodynamic IB (PDF 179 kb)

Supplementary information S1 (box 4)

DCE-US AUC: pharmacodynamic and prognostic IB (PDF 155 kb)

Supplementary information S1 (box 5)

Quantitative ADC: putative pharmacodynamic IB (PDF 132 kb)

Supplementary information S1 (box 6)

MRI 13C-pyruvate kp: putative pharmacodynamic IB (PDF 150 kb)

Supplementary information S1 (box 7)

Radiomic signature of heterogeneity: putative prognostic IB (PDF 195 kb)

Supplementary information S1 (figure 8)

Generating biomarkers from an imaging technique (PDF 250 kb)

Supplementary information S1 (box 9)

Examples of Bradford Hill criteria for biological validity of IBs (PDF 68 kb)

PowerPoint slides



Maximum standardized uptake value is a parameter used routinely in clinical medicine to identify and quantify avidity of tracer uptake in PET studies.

K trans

Volume transfer co-efficient describes the transendothelial transport of low molecular weight contrast agent from blood vessels into the extravascular-extracellular space by diffusion. This imaging biomarker is commonly used in studies of antivascular agents, and measures a composite of blood flow, vessel permeability and vessel surface area.


Apparent diffusion co-efficient is a commonly used imaging biomarker in studies of various therapies, including cytotoxic chemotherapy, targeted agents and radiotherapy. The biomarker is sensitive to motion of water molecules.

Double baseline studies

Biomarker precision can be assessed by measuring on multiple occasions and calculating the repeatability or reproducibility of the parameter. Typically, this precision is achieved by measuring the biomarker twice at baseline in the absence of any treatment effects.


Integrated area under the curve at 60 seconds is a nonspecific imaging biomarker of the tumour vasculature used in studies of antivascular agents. The biomarker measures a composite of blood flow, blood volume, leakage space, vessel permeability and vessel surface area.

Rights and permissions

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

O'Connor, J., Aboagye, E., Adams, J. et al. Imaging biomarker roadmap for cancer studies. Nat Rev Clin Oncol 14, 169–186 (2017).

Download citation

Further reading


Quick links