The role of clinical imaging in oncology drug development: progress and new challenges

Abstract

In 2008, the role of clinical imaging in oncology drug development was reviewed. The review outlined where imaging was being applied and considered the diverse demands across the phases of drug development. A limited set of imaging techniques was being used, largely based on structural measures of disease evaluated using established response criteria such as response evaluation criteria in solid tumours. Beyond structure, functional tissue imaging such as dynamic contrast-enhanced MRI and metabolic measures using [¹⁸F]flourodeoxyglucose positron emission tomography were being increasingly incorporated. Specific challenges related to the implementation of imaging were outlined including standardisation of scanning across study centres and consistency of analysis and reporting. More than a decade on the needs of modern drug development are reviewed, how imaging has evolved to support new drug development demands, the potential to translate state-of-the-art methods into routine tools and what is needed to enable the effective use of this broadening clinical trial toolset. In this review, we challenge the clinical and scientific imaging community to help refine existing clinical trial methods and innovate to deliver the next generation of techniques. Strong industry–academic partnerships and pre-competitive opportunities to co-ordinate efforts will ensure imaging technologies maintain a crucial role delivering innovative medicines to treat cancer.

Introduction

The role of clinical imaging in oncology drug development was reviewed in 2008. ¹ The drug development process was outlined and the role of imaging in the different phases of development was described. Specific challenges relating to the successful implementation of clinical imaging were highlighted, including reproducibility in data acquisition, imaging biomarker validation and multicentre standardisation. At the time, response evaluation criteria in solid tumours (RECIST) 1.0 was widely used in oncology clinical studies and positron emission tomography (PET) had a limited role. However, the review came at an inflection point for imaging in oncology drug development. For example, RECIST 1.1, response assessment in neuro-oncology (RANO) and many other criteria were released, application of machine learning to study large imaging data sets was emerging, a broadening MRI toolset became available and there was increased translation of radiotracers investigating different aspects of tumour biology. Now 15 years later, the current state is reassessed, and opportunities identified to improve drug development success through innovative imaging approaches.

The clinical trial process aims to deliver an evaluation of drug candidate efficacy and safety to support marketing approval. Traditionally, this has been done in a sequential manner starting with evaluation of safety, pharmacokinetics and dosing (Phase 1), followed by assessing efficacy and safety for a selected cancer type (Phase 2), before demonstrating clinical efficacy (usually relative to standard of care) for the purposes of regulatory submission (Phase 3). Compared to other therapeutic areas, oncology clinical trials typically do not include healthy volunteers due to safety profiles of oncology drug candidates and instead enrol patients beginning in Phase 1. Evaluation in patients at the earliest stages of drug development enables an initial assessment of efficacy and mechanism at Phase 1 including identifying molecular and genetic subsets of responders. Clinical trial designs have continued to evolve, promoting faster and more efficient routes to deliver new cancer medicines. ² For example, new paradigms are emerging away from the sequential three stage clinical trial process. ^3,4 In some instances, clinical development can be conducted in a seamless fashion, seeing Phase 1 studies expanded to study hundreds of subjects to accelerate programmes. However, interpretation of safety and efficacy is difficult in these patient populations who typically have advanced, metastatic, heterogenous, and progressive disease after treatment with multiple lines of therapies. This increases the importance of measuring molecular and cellular characteristics of disease alongside structural measurement of disease. Imaging-derived endpoints are a critical component of clinical drug development, informing on mechanism, efficacy, safety and differentiation to enable programme decision-making and support regulatory submissions. This review provides an overview of established and emerging imaging techniques and how imaging innovations can address evolving clinical trial needs.

Current status of the pharmaceutical industry

The productivity of pharmaceutical research and development (R&D) remains low despite advances in our understanding of human disease, the use of genetics to identify novel therapeutic targets and an ever-widening range of technology platforms able to support drug development. In the last 15 years, the main trend in oncology has been the increasing focus on precision medicine. For example, replacing biomarker-agnostic chemotherapy with drugs targeting specific tumour driver mutations or patients stratified by programmed cell death protein 1 (PD-1) expression levels. Medications leveraging biomarkers are more likely to succeed with a streamlined approach to access through a higher benefit to risk ratio. Analyses suggest promising productivity in oncology drug development, with 2018 resulting in a record number of new drugs approved by the U.S. Food and Drug Administration FDA ⁵ and recent analyses suggesting the productivity persisting. ⁶ However, most resource is still applied to drug candidates that will never address important unmet patient needs.

At a time when innovative therapeutic development has never had so much momentum in oncology, out of the hundreds drug candidates in development, how do we ensure only the most promising progress and how do we identify failures early? Taking immuno-oncology (IO) as an example, Tang et al reviewed oncology programmes ongoing in 2018 and highlighted the unprecedented level of activity with 940 agents in clinical development against 303 targets. ⁷ This trend has continued with thousands of IO-based trials evaluating a broadening range of IO therapy types. ⁸ Interest in combinations beyond programmed death-ligand 1 (PD-L1), PD-1, and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) inhibitors are seeing growth ⁹ but efficiently delivering successful medicines of the future will demand highly informed decision-making.

Across drug development, maximising success through an informed drug development is an ongoing enterprise. Towards this goal, there has been an emphasis on characterising specific aspects of each drug, such as the “three-pillar” framework ¹⁰ : (1) drug tissue levels, e.g. leveraging the sensitivity of PET to measure tumour uptake of a microdose radiopharmaceutical, (2) target engagement, where a successful study might reliably estimate the relationship between plasma pharmacokinetics and target occupancy, (3) proximal or distal measures of pharmacology such as early signals of efficacy measured with CT assessing tumour volume over time. These approaches will require complex, early stage clinical studies, defining an information-rich modern drug development paradigm. In addition to early drug candidate characterisation, better matching of the drug to the patient population is translating into successes towards broadly adopted precision medicine. ^11–14

Furthermore, alternative drug development paradigms have emerged such as the adoption of master protocols, including umbrella (same cancer types but different biomarker), basket (different cancer types but same biomarker) and platform studies (usually an adaptive design with several interventional arms against a common control group) in multiple diseases that share similar molecular alterations or common risk factors. ¹⁵ Whilst these studies aim to streamline clinical drug development, they present a challenge to optimise imaging-derived endpoints with multiple tumour types.

Specific benefits of imaging include whole-body assessment of disease with the ability to provide quantitative, localised and relevant imaging-derived biomarkers. However, imaging remains underutilised. Even serially acquired CT scans in oncology trials comprise of much more information besides the unidimensional measurements of RECIST target lesions. In the advent of innovative therapeutic regimens such as immunotherapies, all imaging methods, data and analytic approaches should be considered to adequately capture complex patterns of response. ¹⁶

As clinical trials evolve and imaging technology advances, it has never been more important to keep the fields synchronised, ensuring measurement gaps are clearly articulated by drug developers. This will enable the imaging community to innovate towards specific trial needs and ensure modern drug development can benefit from the whole-body, non-invasive toolset that imaging can deliver. This review identifies the current challenges across the stages of clinical drug development and highlights how imaging is adapting to evaluate the current wave of therapeutics.

Precision oncology toolset

Precision medicine for oncology is established, with a growing list of molecular pathology defined drug treatments approved. ¹⁷ Imaging remains a vital platform to support cancer drug development but must be considered alongside existing and emerging platforms based on blood (termed liquid biopsy including tumour cells and circulating tumour DNA, (ctDNA)) and tissue analyses including genetics and immunohistochemistry. ¹⁸ As these tissue and blood-based platforms expand, the integration of imaging data must be optimised. Imaging can be used to support validation of say blood-based markers or combined for improved prognostic purposes. Hypothesis testing using large-scale, multimodal data sets will likely improve disease understanding and deliver novel predictive tools. ¹⁹

As the toolset broadens and advanced analytical approaches enable us to work across modalities, clinical trials provide an opportunity to incorporate these methods as exploratory techniques to build evidence for broader use.

Established vs innovative imaging methods

Oncology is the only therapeutic area within drug development where broadly, a common imaging methodology is both practical and relevant from Phase 1 through to Phase 3 clinical trials (cp. other disease areas, such as rheumatoid arthritis where different techniques are applied to different clinical trial phases). Most clinical imaging methods applied to oncology drug development are based on established, routine radiological approaches adapted for trial use (e.g. response evaluation frameworks such as RECIST and variants thereof). ²⁰ Routine imaging is typically adapted for drug development by attempting to improve consistency and harmonisation across participating centres via implementation of training and guidance documents. Particular attention is given to standardisation of radiological review, often achieved through centralised image collection for the purposes of blinded independent central review (BICR).

In early drug development, unidimensional change in tumour measurements can provide initial evidence of efficacy in small patient cohorts, giving some confidence to proceed to larger clinical studies. However, such assessment has limitations as it: provides no information related to mechanism of action; provides limited information on tumour biology; is not sufficiently sensitive for early response prediction; and is subject to reader variability. Early clinical development studies have potential to incorporate complex methodologies, resulting in information rich studies. An array of clinical imaging methods can be considered for a given tumour type and therapeutic modality. Options include advanced structural measurements such as tumour volume and regression/growth kinetics including historic scans. This may lead to earlier markers of response and progression, and better intra- and interobserver reproducibility, but limitations such as time-consuming manual segmentation and small patient numbers so far have hampered clinical trial implementation. Advanced methods such as functional or microstructural tissue measurements, such as diffusion-weighted imaging (DWI) or dynamic contrast-enhanced MRI (DCE-MRI) or molecular methods (bespoke or generic PET radiotracers) offer opportunities, for certain mechanistic approaches and may provide an adjunct to measuring structural change. However, these novel techniques have been seldom applied to clinical trial decision-making, reasons for which will be considered further.

As clinical trials are scaled to access larger patient cohorts (including expanded Phase 1 studies and into Phase 2 and 3), the technology must adapt to enable consistency and harmonisation across tens to hundreds of investigator sites, shifting towards simplicity, image quality, and operational efficiency. In later clinical drug development phases, RECIST-derived endpoints are often used to provide evidence to support regulatory approval, typically by defining disease progression events. Early time-to-event endpoints, such as progression-free survival (PFS), are used as surrogates for overall survival (OS) to accelerate approval of novel oncology agents. ²¹

Here we consider progress in developing and applying the available range of methods against the knowledge gaps that need to be addressed in early drug development, followed by how methods for later stages of development are applied.

Measuring drug tumour uptake

Lack of drug tumour uptake is a significant factor in limiting responsiveness to therapies. ^22–24 Tumour microenvironmental factors, host tissue barriers (e.g. blood–brain barrier (BBB)), abnormal vascularisation, tumour stromal factors, drug efflux mechanisms and properties of the drug itself can limit drug uptake and increase variability of uptake between lesions within and across patients. ^25–29 Given that plasma drug levels are poor predictors of tumour levels, direct measurements of tumour drug levels are required. ²⁷ Nerini et al review tumour heterogeneity and the influence on drug distribution and highlight the importance of considering complementary methodologies such as mass spectrometry imaging to understand these factors. ²⁹ Whilst analyses of biopsy-acquired tissue or microdialysis are important methodologies, they are not feasible in all settings and do not readily allow for measurement between lesions in the same subject or monitoring the same lesion over time. Importantly, they typically inform on a single lesion and do not provide a clear picture of overall tumour heterogeneity.

Molecular imaging, particularly with PET has a broad role in cancer drug development. ^30,31 PET with radiolabelled drug enables the evaluation of tumour drug levels quantitatively across the whole body. ^30,32 The broad use of these approaches is exemplified by the multiple studies of tyrosine kinase inhibitors (TKIs) where ¹¹C- and ¹⁸F-labelled analogues have been manufactured for clinical PET use. In the example by Saleem et al, the ¹¹C-labelled dual TKI [¹¹C]lapatinib was used to evaluate uptake within brain metastases in patients with human epidermal growth factor receptor (HER-2) positive metastatic breast cancer. ³³ In the study of only six subjects, it was concluded that lapatinib uptake was only observed in metastases through a disrupted BBB and not normal brain tissue. In a further example, [¹¹C]osimertinib was administered intravenously to eight subjects with an intact BBB. ³⁴ The study showed a rapid and widespread uptake in the brain and that [¹¹C]osimertinib penetrates an intact BBB. This provided additional validation of the clinical data to support data previously shown in non-human primates that [¹¹C]osimertinib has superior brain uptake compared to other epidermal growth factor receptor (EGFR) TKIs. ³⁵ The two studies are compared in Figure 1. It has been suggested that the findings support improved efficacy of osimertinib in non-small cell lung cancer (NCSLC) brain metastases compared to second generation EGFR-TKIs erlotinib and gefitinib. ³⁶

Figure 1.

1 (a) Representative PET images for [¹¹C]osimertinib overlaid on an MRI in a human subject. TACs for [¹¹C]osimertinib in a human subject. Total radioactivity concentration in blood and radioactivity concentration for parent radioligand in plasma (b). Inset shows blood radioactivity concentration during the initial 5 min after injection. (c) Shows regional brain radioactivity. Reproduced from JCBFM, SageJournals. Varrone et al., 2020. ³⁴ 2 (a) The image data show radioactivity distribution in normal brain and cerebral metastases (enclosed in blue circle) (bottom panel) and corresponding contrast-enhanced MRI images (top panel) of a patient with Her-2+ metastatic breast cancer brain metastases following [¹¹C]lapatinib-PET. TACs for individual metastases for all the subjects from the Day 8 PET scan (b) show variability in uptake between and within patients. In (c), mean TACs for normal brain (green) is plotted for comparison with mean whole blood (red) and plasma (plasma) TACs. PET, positron emission tomography; TAC, time activity curve. Reproduced from EJNMMI Research, SpringerOpen. Saleem A et al., 2015. ³³

These examples demonstrate that bespoke radiosyntheses combined with clinical PET scanning can answer specific and highly relevant questions to guide clinical drug development. Although the methodology is complex, PET centres with established capabilities have demonstrated that this can be efficiently completed in small studies. However, there remain multiple barriers to systematic adoption including cost and method development lead times. ³⁷ The imaging community has an opportunity to increase the access to these methods: improving radiosynthesis strategies to increase tractability and production reliability, sharing further clinical examples where the value to drug development programmes are clearly articulated and establishing early engagement between imaging scientists and drug developers to ensure sufficient lead time to guide clinical implementation. The recent growth in radioligand therapy across the industry will likely catalyse further integration of molecular imaging into programs, increasing the awareness of how labelled therapeutics can be incorporated to inform clinical development. ³⁸ In addition, there are potential opportunities for novel molecular imaging modalities to study drug distribution including optical methods like near-infrared fluorescence that are demonstrating early clinical potential. ³⁹

Imaging drug mechanism and the tumour microenvironment

If the drug candidate can reach the site or sites of disease, bind to the intended therapeutic target, the downstream consequences will need to be interrogated. Size changes may not occur for months following commencement of treatment, may be complex and challenging to interpret. For example, in cases of pseudoprogression in immunotherapy, ⁴⁰ size-dependent criteria underestimate response in certain settings and importantly offer little to elucidate drug mechanism. Methods sensitive to proximal pharmacology will provide valuable input into a drug development programme. ⁴¹ Furthermore, understanding characteristics of the tumour microenvironment, such as levels of T cells and inflammatory mediators, that may influence response and resistance could help identify which patients will benefit from therapies. ⁴²

Molecular imaging

PET is providing a growing array of tools able to characterise disease quantitatively, non-invasively and at multiple disease sites. The molecular imaging toolset, although dominated by the use of [¹⁸F]fluorodeoxyglucose ([¹⁸F]FDG) PET for routine response assessment and integrated into several response criteria, now extends to many probe molecules interrogating metabolic processes, ⁴³ receptor expression, ^44,45 tumour microenvironmental factors such as hypoxia, ^46,47 angiogenesis ^48–50 and immune cell populations. ^51–54 Several broad reviews are available. ^31,55,56 The use of immunoPET (long-lived radioisotope radiolabelling of antibodies) has become an established tool to study antibodies in clinical research. Recent reviews have summarised the extensive progress in the application of immunoPET methods to evaluate therapeutic radiometals, drug pharmacokinetics, receptor occupancy and target expression across tissues. ^31,57,58

To date, specific molecular imaging agents have seen limited adoption in clinical trials to support programme decision-making. Reasons for this include lack of validation of these tracers in the context of treatment response; limited evaluation of the performance of these agents especially in multicentre clinical studies; cost; timelines and complexity to establish a new agent at a required trial centre. Consequently, only a small proportion of promising molecular imaging methods have translated into clinical trial tools. It is imperative that the drug development community communicates the measurement gaps and what level of evidence is needed to incorporate these methods into future clinical trials to provide key trial endpoints.

Although PET provides key advantages in terms of quantification, opportunities from other modalities should be considered. These include both established modalities such as single photon emission computed tomography (SPECT) with existing and new radiotracers and emerging clinical approaches including optical molecular imaging with novel molecular probes. ⁵⁹

Functional and microstructural imaging

In addition to molecular imaging, magnetic resonance-based functional tissue imaging techniques have an established role in the evaluation of treatment response. MR methods can study the tumour microenvironment through a range of scanning techniques and derive semi-quantitative indices. ⁶⁰ For example, DCE-MRI offers a methodologically straightforward way to probe the vasculature. Across many studies it has been demonstrated that DCE-MRI enables a mechanistic evaluation of treatment response associated with anti-vascular or anti-angiogenic therapies. ⁶¹ Other techniques such as DWI-MRI to study tissue microstructure provides an early response marker with changes measured assumed to be associated with cell death. Successful implementation of complex MR techniques across centres is still challenged by differences in scanning hardware, acquisition and analysis. ⁶² To address this, performance of methods through test/re-test studies have been evaluated and standardisation efforts promote the utility of such methods. ^62–64

Beyond these methods, emerging MR techniques offer promise to provide molecular specific information including: chemical exchange saturation transfer to study endogenous metabolites ⁶⁵ ; dynamic nuclear polarisation to study metabolic processes using injected ¹³C metabolic substrates ^66–69 ; and imaging of deuterium-labelled metabolic substrates. ⁷⁰ It is anticipated that a flow of clinically relevant MR methods will continue to emerge to inform on tissue structure, function and metabolism based on the unique adaptability of MR scanning techniques and advancements in hardware.

In addition to the array of MRI techniques available, the toolset available to the drug developer will broaden across modalities. For example, photoacoustic methods are emerging as a potential clinical tool in addition to a nascent set of optical tools to study tissue function. ^71,72

Despite the breadth of existing and emerging techniques available and the demand for sensitive and mechanistically relevant measures of drug effect, very few of these specialised methods are incorporated into Phase 1 clinical trials. It is critical to understand the limited uptake of these methods even to generate exploratory endpoints in trials.

Established functional techniques like DCE-MRI and DWI-MRI demonstrate the challenge to progress a promising quantitative (or semi-quantitative), functional tissue biomarker from concept to clinical trial adoption. ⁷³ Significant centre to centre collaboration is required to harmonise methodology, reaching agreement on scanning details, analysis and derived endpoints. Test/re-test studies are then required to demonstrate the robustness of implementation in order to confidently interpret treatment response. ⁷⁴ In parallel to addressing standardisation, understanding the clinical relevance and biological underpinning of the measurement is required, necessitating studies in different tumour types, optimisation of methods in different anatomical sites, correlation of in vivo imaging with tissue and studies of treatment response.

When proceeding to implement a method within a clinical drug study, different challenges emerge: do the centres required to conduct the trial have the right imaging methodology and expertise?; how robust is the technique likely to be within a study such as a Phase 1 study with heterogenous disease presenting in multiple anatomical locations?; cost and complexity of implementation often limit integration of exploratory imaging methods without well-defined value; complex clinical trials can be burdensome on the patient and the addition of complex scanning can be considered prohibitive.

To ensure the opportunities from new functional and molecular methods are realised, close co-ordination between drug development and imaging is required. The specific challenges of implementing new techniques into early clinical trials need to be addressed. Cancer Research UK (CRUK) and the European Organisation for Research and Treatment of Cancer (EORTC) assembled experts to identify the challenges of imaging biomarker validation and qualification and describe a conceptual framework to promote a better process to prepare imaging-based biomarkers for clinical trials. ⁷⁵ This includes parallel tracks of technical validation, biological/clinical validation and assessment of cost-effectiveness; the need for standardisation and accreditation systems; the need to continually revisit measurement precision; and the essential requirements for multicentre studies to qualify methods for clinical use. Also, recommendations include comprehensive publication of technical detail associated with the imaging tests. If followed these recommendations will ensure novel methods can be developed and applied robustly with confidence in the data generated. Organisations such as the RSNA Quantitative Imaging Biomarkers Alliance (QIBA), the UK National Cancer Imaging Translational Accelerator (NCITA), and Pharma Imaging Network for Therapeutics and Diagnostics (PINTAD) aim to bring together academic, pharmaceutical imaging scientists and other experts to work towards improved standardisation, method adoption and knowledge sharing.

Measuring drug efficacy using established methods

Response criteria today

Response criteria such as RECIST are used to report on patient responses in early studies and define progression events later in clinical development. For registration studies, these approaches can provide surrogate endpoints of survival to facilitate accelerated drug approval. The criteria were introduced to harmonise the interpretation of results within and between patients, across trial centres and across studies. Data from over 4000 patients supported the initial RECIST recommendations ⁷⁶ published in 2000 which were subsequently widely adopted. In 2009, the criteria were updated based on data analysis of over 6500 patients and 18,000 lesions to include clarifications and modifications reflecting clinical practice, e.g. integration of [¹⁸F]FDG-PET in some contexts ^77,78 and reducing the number of target lesions to five per patient. In 2013, Liu et al, reported on a survey of response criteria to evaluate adoption, understand gaps and guide future use of these methodologies. ⁷⁹ Issues persisting were outlined including relevance of the criteria when applied to specific tumour types and with novel targeted agents. Although these approaches have been broadly adopted, many users have applied modified criteria to address specific needs applied to certain tumour types (e.g. use of [¹⁸F]FDG-PET to support lymphoma response assessment), incorporate methods used in clinical practice and react to unique response patterns seen in novel, targeted therapies in development (e.g. immunotherapies). Table 1 summarises the growing list of response criteria developed for specific indication and treatment settings. Changes in clinical practice, including new imaging technologies and novel drugs in development will inevitably lead to further evolution in criteria to ensure relevance.

Table 1.

A summary of the most common imaging-based response criteria used in oncology drug development

Indication	Criteria	Imaging modalities	Reference
Solid tumours
Oncology	RECIST 1.1	CT, MRI, [18F]FDG-PET, X-ray, bone scintigraphy	Eisenhauer et al., 2009 77
ImmunoOncology	irRC (superseded) irRECIST iRECIST	CT, MRI, PET, X-ray, bone scintigraphy	Seymour et al., 2017 80
Oncology	PET Response Criteria in Solid Tumours (PERCIST)	FDG PET, CT	Wahl et al., 2009 81
Intratumoral (IT)	itRECIST	CT, MRI, [18F]FDG-PET, X-ray, bone scintigraphy, photography/caliper.	Goldmacher et al., 2020 82
Gastrointestinal stromal tumour (GIST)	Choi	CT, [18F]FDG-PET	Choi et al., 2007 83
Prostate cancer	Prostate cancer working group guidelines	CT, MRI, [18F]FDG-PET, X-ray, bone scintigraphy	Scher et al., 2016 84
Hepatocellular carcinoma (HCC)	Modified RECIST (mRECIST) for HCC	Triphasic (pre-contrast, arterial, venous) CT or MRI	Llovet and Lencioni, 2020 85
Malignant plueral mesothelioma	mRECIST v. 1.1	CT	Armato and Nowak, 2018 86
Central nervous system malignancies
High-grade glioma (HGG) and low-grade glioma (LGG)	RANO-HGG/LGG	Brain MRI - T1 pre-/post-contrast, T2/fluid attenuated inversion recovery (FLAIR) and optional DWI)	Chuckwueke and Wen, 2019 87
Brain metastases (BM)	RANO-BM	Brain MRI - T1 pre-/post-contrast, T2/fluid attenuated inversion recovery (FLAIR) and optional DWI) CT	Lin et al., 2015 88 Camidge et al., 2018 89
Leptomeningeal (LM) disease	RANO-LM	Brain MRI - T1 pre-/post-contrast, T2/fluid attenuated inversion recovery (FLAIR) and optional DWI) Spine MRI	Chamberlain et al., 2017 90 Le Ruhn et al., 2019 91 Le Ruhn et al., 2022 92
Neurofibromatosis type 1 (NF1)	Response evaluation in neurofibromatosis and schwannomatosis (REINS)	MRI-short tau inversion recovery (STIR). Volumetric measurment	Dombi et al., 2013 93
Paediatric neuro-oncology	International neuroblastoma response criteria (INRC) Response assesement in paediatric neuro-oncology (RAPNO)	MRI—T1, axial T2 FLAIR, DWI. Apparent diffusion coefficient (ADC) measurement	Erker et al., 2020 94 Park et al., 2017 95
Hematological malignancies
Acute myeloid leukaemia (AML)	Cheson	X-ray, CT, MRI, PET	Cheson et al., 2003 96
Chronic lymphocytic leukaemia (CLL)	International workshop on CLL (iwCLL)	CT	Hallek et al., 2018 97
Hodgkin & non-Hodgkin’s Lymphoma	Lugano International Working Group consensus response evaluation criteria in lymphoma RECIL 2017 PINTaD Response Criteria in Lymphoma Working Group (PRoLoG)	[18F]FDG-PET, CT	Cheson et al., 2014 98 Younes et al., 2017 99 Ricard et al., 2023 100,101
Multiple myeloma	International myeloma working group (IMWG)	[18F]FDG-PET, CT, MRI	Kumar et al., 2016 102

PET, positron emission tomography.

Included are the specific imaging methods incorporated.

Evolving response criteria

Gerwing et al, expand on the current limitations of RECIST making the case that these methodologies have not kept pace with modern therapeutic development since they are unable to define all response types. ⁴¹ Specifically, they propose that response assessments must benefit from advanced methods such as immunoPET for studies of immune cells, MRI evaluation of the tumour microenvironment, advanced image analysis and integration of imaging with other biomarkers.

Through consortia, criteria can be systemically updated via data analysis across studies and user surveys. Incorporation of new methods (e.g. volumetrics, PET and advanced MR) must provide clear benefits in terms of objective and reproducible assessments over current criteria, require analyses of large data sets and simulations until both validity and multicentre implementation are addressed. ⁷⁹ In addition, current methods benefit from simplicity of processing and for the majority rely on human expert input to derive the measurements. Volumetric measurement of total tumour burden, e.g. will only be feasible for Phase 3 trials if software-based approaches can reduce manual workload. Advancements have been made to develop automated approaches and reduce processing time compared to manual annotation. ¹⁰³ However, the challenge is to ensure the software is validated for all relevant organs and can deal with complex clinical cases, e.g. in case of NSCLC lung nodules within atelectasis, confluent metastases without clear borders, or malignant pleural effusions. Integration of software-based tools to support clinical trials is likely to increase, enabling more complex analyses with reduced workload.

Immune-checkpoint inhibitors represent one of the most important therapy advancements in modern oncology. The assessment of treatment responses following immunotherapy has led to unusual response patterns in RECIST 1.1 and in some cases pseudoprogression, which can lead to a shortened exposure to treatment in clinical studies. To address this limitation, Wolchok et al developed a modified immune-related Response Criteria (irRC) based on the World Health Organisation (WHO) criteria. ¹⁰⁴ In a further development, the bidimensional irRC measurements were adapted to the unidimensional immune-related RECIST (irRECIST) criteria. Subsequent use of irRC and irRECIST led to other modifications being recommended, leading to inconsistent application. To address this issue, the official RECIST Working Group (http://www.eortc.org/recist) published the new iRECIST guideline in 2017 ²⁵ for assessing response to immunotherapy in clinical trials. Many late phase clinical trials now include iRECIST endpoints to improve assessment of iPFS. However, the added complexity of continuing randomised treatment and collecting scans post RECIST 1.1 progression means that there will be a subset of patients who continued to benefit from treatment but had no scans to build this evidence. Given the relatively low incidence of pseudoprogression (5–10%), it should be noted that the added information from iRECIST may only lead to a modest improvement in the correlation of iPFS with OS. ¹⁰⁵ With increased utility of intratumoral (IT) delivery of immunotherapy itRECIST was designed to address the unique needs of IT immunotherapy trials but, where possible, aligns with RECIST 1.1 and iRECIST. ⁸²

The imaging community plays a vital role to ensure current response criteria are optimally applied and supporting the development of new approaches through generation of evidence.

Imaging implementation and central review

In the previous review, it was noted that standardisation of scanning and analysis was hindering the value of imaging within clinical trials. This remains problematic across different trial sizes, scales and across all imaging methods. The operational implementation of clinical trial imaging, often supported through Contract Research Organisations (CROs), intends to harmonise how imaging is conducted to maximise confidence in the imaging data generated, which in the majority of clinical trials is providing primary or secondary endpoints. This aims to achieve standardised scanning procedures, consistent local analysis and/or radiological review and processes to centralise scanning for blinded central review if required. In addition, for more complex imaging procedures (e.g. advanced MRI) the CRO may also conduct a review of scanning hardware and quality assurance steps including phantom measurements and quality control to ensure adherence to a given scanning protocol.

Harmonising scanning will only be successful with the engagement of imaging experts at clinical trial centres. In addition, a level of pragmatism is required to ensure robust control of standardisation when required (e.g., strictly defined scanning protocols and quality control steps) though with less stringent demands when standardisation can be more forgiving (CT scanning protocols for RECIST-based trials). That said, even extensions of basic response assessments, e.g. PERCIST for incorporating [¹⁸F]FDG-PET response assessments can be problematic if not applied uniformly. ¹⁰⁶ Ultimately, these methods will be more impactful if there is stringent consistency applied within trials, between subjects and across centres.

In many trials, CROs facilitate the transfer of scans electronically to a core lab, perform quality control and then present data to a single radiologist or multiple radiologists depending upon the study needs (Figure 2). The purpose is twofold: (1) to promote a consistent application of image analysis or radiological review, e.g. with specific oncology response criteria (2) To blind the radiological assessment from clinical bias, which is considered particularly important in randomised controlled trials. ¹⁰⁷ Although the purpose of this process has been established for many years, it has been argued that site-based assessments are reliable and mechanisms can be employed to monitor for systematic bias. ¹⁰⁸ Following decades of analysis, there is acceptance that there will be disagreement rates of 25–40% for independent radiological reviews. These rates should be monitored to identify any performance issues and the analysing the underlying causes of disagreement may improve future approaches. ^109,110

Figure 2.

A typical data flow of images undergoing a blinded independent central review by an external imaging CRO as part of an oncology clinical trial. CRO, contract research organisation.

The evolution of response criteria challenges the reliance upon site-based assessments. Can different clinical centres involved in multiple trials simultaneously, with different response approaches prescribed, be expected to apply reviews with the level of consistency required? In the future, it seems unlikely we will necessitate the operational complexity of transferring thousands of scans associated with a single study to a central location for a radiological review. Opportunities to enable reliable and consistent site-based assessments are likely to be explored perhaps using algorithms able to support the radiological assessment of volumetric total tumour burden. Until such methods can be reliably deployed, regulatory engagement is required to ascertain when blinded independent central review (BICR) is required and when approval can be achieved without. With extensive data collected across industry, meta-analyses on different tumour types and end points can begin to support decisions around central reading needs.

Imaging drug-induced toxicity

The review has focused on the utility of imaging to measure drug efficacy responses. In addition, imaging modalities such as high-resolution CT (HRCT) can be used for the diagnosis and prediction of treatment toxicity and adverse events. For example, pneumonitis and interstitial lung disease are observed in some non-small cell lung cancer (NSCLC) patients treated with immune checkpoint inhibitors (ICIs), EGFR-TKIs, and antibody drug conjugates (ADCs), especially following radiotherapy to the lungs. In the PACIFIC trial, which studied the PD-1 blocker durvalumab vs placebo sequential to chemoradiotherapy in Stage 3 NSCLC, pneumonitis occurred in 34% patients in the durvalumab-treated group and 25% in the placebo group. ¹¹¹ On lung HRCT scans, the characteristic ‘ground-glass opacities’ usually appear early on in treatment. Novel image analysis software interpreted by expert clinicians allows accurate segmentation of lung structures and extraction of features such as volume, emphysema, honeycombing, and the blood vessel structure. Quantification of these parameters over time may improve the expedited detection of pneumonitis, and potentially predict patients at risk based on pre-treatment scans. This could create the opportunity to develop imaging software as a companion diagnostic, allowing for improved patient selection to screen out patients at risk of pulmonary manifestations of drug toxicity.

Emerging methods

There is a persistent lag between emerging clinical technologies being demonstrated in small methodology studies to being integrated as decision-making endpoints in clinical trials. However, it is likely that with robust validation activities, a wider range of methods will become part of a systematic approach to enable highly informed drug development. Early phase clinical studies can be challenging due to heterogenous disease, yet are likely to offer the ideal opportunity to integrate advanced methods using a single or a small number of centres. In such settings, advanced scanning methods can be most successfully applied.

New technologies with potential to support drug development will likely emerge broadly from three technology domains: (1) molecular probes, (2) scanning hardware, (3) advanced analysis.

1. Molecular probes—integrating new molecular probes in early drug development will demand an increasing range of probes able to study multiple processes pertinent to the drug mechanism and the tumour microenvironment. The chemistry toolkit has never been broader, using radioisotopes to label small molecules, peptides, antibodies and nanoparticles. A range of isotopes will be required including ¹¹C and ¹⁸F to long-lived isotopes such as ⁸⁹Zr or ⁶⁴Cu where biological processes need to be studied over longer timescales. Interchangeable metals will also offer the opportunity for a broad set of theragnostic methods. The creativity of the radiochemist combined with robust approaches to biological and clinical validation provides the potential to translate many novel radiolabelled drug candidate molecules, highly relevant trial markers and therapeutic radioligands.

2. Scanning hardware—new scanning hardware will increase sensitivity of measurement and deliver increasingly rich data sets. For example, total body PET methods will enable high sensitivity measurements with lower radiation dose to interrogate total disease burden. ¹¹² In addition, these techniques can extend the measurement window out to days post-injection and will enable new experimental paradigms such as multitracer imaging. New magnetic resonance methods (very high field and high gradient strength) will increase the range of MR biomarkers to study the tumour microenvironment. Beyond the domains of radiology and nuclear medicine, techniques such as optical (with and without optical probes) bring in vivo measurements closer to information only currently available by studying ex vivo tissue. Extensions of optical methods such as photoacoustic imaging will overcome some of the measurement limitations of existing methods such as measurement depth to deliver new functional and molecular tissue measurements.

3. Advanced analysis—deriving quantitative image-based biomarkers reproducibly from increasingly complex data sets will require a step change from current methods. A new era of precision medicine is driving a change from single marker solutions (e.g. PD-L1) to multimarker approaches. Emerging examples include the use of ctDNA and image-derived tumour volumes to predict response and recurrence risk. ¹¹³

The promise of artificial intelligence/machine learning (AI/ML) will evolve from development for diagnostic application to extraction of response markers and novel phenotyping strategies, deriving more than structural indices. The field of radiomics is beginning to show some promise in the identification of substructural image features with potential to predict biology. ¹¹⁴ Extensive validation activities will be required to ensure confidence in these methods. In the meantime, AI/ML methods will likely provide tools to support efficiency of the imaging clinical trial process (e.g. automated quality control assessments). To enable these new techniques large, high-quality, annotated datasets will be required to develop and validate methods before deployment. Industry and academic collaboration will be required to ensure clinical trial data can be amalgamated to support the development of these methods. To this end, the Image Biomarker Standardisation Initiative standardised 169 quantitative radiomic features for high throughput image-based phenotyping and has set standardised definitions and validated reference values for clinical use. ¹¹⁵ Of these, 169 candidate radiomics features, good to excellent reproducibility was achieved for 167 radiomics features using MRI, [¹⁸F]FDG-PET and CT images obtained in 51 patients with soft-tissue sarcom.

Combined, new hardware, probes and analysis methods will deliver whole body, longitudinal evaluations across tumour types. Specific areas for further development include validation activities, particularly the ability to use tissue to compare across measurement scales; increased study of temporal change to ensure the dynamics of response can be characterised; ongoing commitment to standardise promising methods to ensure validation can be supported through multicentre studies; access to existing clinical trial data to develop next generation analysis algorithms.

Conclusions

Response assessment in drug development is still based largely on morphological change and response patterns may not be relevant in all tumour types and with new targeted therapies. The last 15 years has seen limited progress in the introduction of new imaging technologies within clinical trials. The field of imaging science will continue to extend the information content of existing techniques and deliver new clinical modalities. However, few of these imaging innovations are translating to clinical trial tools. At most, some novel techniques (methods and criteria) are incorporated as exploratory endpoints in addition to established techniques such as RECIST. This highlights the continued importance of ensuring the value of imaging is fully considered with imaging experts contributing to drug development planning.

Response criteria are evolving in tandem with drug development and technology innovations and will require data-driven validation. Emerging functional, microstructural, and molecular methods have more potential to be integrated into drug development to study drug biodistribution and distal and proximal pharmacology. Collaborations between expert clinical imaging centres and drug development organisations will ensure the right methods are guided towards supporting novel therapies to address important unmet clinical need.

The range of new therapeutic candidates in early drug development provides tremendous excitement. The imaging community will play a significant role to develop and deploy techniques that will ensure ineffective medicines are stopped early and only promising candidates progress towards patient benefit. The next 10 years holds great promise for imaging to underpin a new paradigm of precision drug development, providing (1) established structural, molecular and functional methods are robustly implemented across trial centres and (2) state-of-the-art hardware, probes and analysis techniques are developed and integrated to provide the next generation of decision-making imaging biomarkers adapting to study emerging treatment mechanisms.