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.

Some IBs have an additional complication compared to biospecimen biomarkers as they often incorporate contrast agents to derive the biomarker, the choice of which can be limited due to the market size required for commercial viability for a different purpose (i.e. as a clinical diagnostic) . This can complicate translation of the IB: for example some PET agents based on short-lived radionuclides such as 11 C will depend on the availability of a cyclotron to provide the tracer (8). In MRI, gadolinium-based contrast agents are designed and marketed to improve qualitative diagnosis. Low molecular weight agents such as gadopentetate (MW548) have regulatory approval for this purpose, but some animal studies show advantages for high molecular weight agents such as gadomelitol (MW6474) for deriving IBs of perfusion (9). Although gadomelitol was taken into clinical development, it did not prove sufficiently attractive as a commercial diagnostic product despite its appeal in providing a biomarker (10).
Most IBs require post-acquisition analysis. Many commonly studied IBs (e.g. ADC, the family of K trans and related perfusion biomarkers derived from either CT or MRI and 18 F-fluorothymidine PET) continue to be measured and analysed in disparate ways by different laboratories using a wide range of software systems, despite attempts to standardise and harmonise these methods (2, 3, 11-14). Lack of standardisation has impaired multicentre reproducibility for many IB measurements, which has made IB translation and meta-analysis problematic (5). For example, identical data can have order of magnitude differences in numeric values for K trans depending on choice of tracer kinetic model (with differing underlying assumptions) and the software used (15). This is the major reason why absolute values of K trans have not been qualified as prognostic or predictive IBs (Box 6). Even if software packages are commercialised, they are subject to instability following upgrades that can alter biomarker values significantly (16).

Tumor sampling
Many biospecimen biomarkers represent a molecular entity, or "an analyte" (17) that can be measured (18), although some such as tumor grade do not. Some IBs are associated with a specific molecular entity or process ("molecular imaging") which they may purport to directly quantify. However, many IBs lack any associated analyte that could be isolated and assayed (e.g. TNM, ADC, K trans or tumor texture metrics).
Further, biospecimen analytes are measured using an in vitro diagnostic assay in a process quite separate from collection of the sample from the patient (19). Biospecimen samples can be split following collection, allowing multiple laboratories to compare reproducibility between instruments, between supposedly similar assays, and between the different laboratories (20). In distinction, the IB arises from a signal detected at the moment when the patient is coupled to the in vivo imaging device, quite unlike the relationship of patient to machine seen with in vitro biochemical assays (1). Although raw scanner data can be analyzed at different laboratories, acquisition-specific sources of bias and variance are already embedded irreversibly in this work-flow.
A significant advantage of IBs is their ability to provide serial non-invasive measurements (unlike most tissue based assays). In addition, they can distinguish, characterize and track whole and/or multiple lesions within a subject (inter-tumoral heterogeneity) and also quantify intra-tumoral heterogeneity within a lesion (21). These factors are key indications for using imaging in cancer research. Tumor heterogeneity can be a major confounding factor in non-imaging studies where biomarkers can be at extremes of either under-sampling a single lesion (e.g. where biospecimen biomarkers derived from biopsy based on gene mutation or receptor status may be discordant between primary and metastatic lesions in the same individual (22)) or integrating samples from the entire patient at the expense of losing lesion specific information (e.g. circulating free DNA, circulating tumor cells or serum proteome). Investigators have greater control over tumor sampling in IB studies, where sampling can be restricted to part of a tumor, the whole tumor, comparison between multiple tumors within one patient, or total disease burden (21). While an advantage for IBs, this means technical and biological validation studies must be tailored to suit the sampling method employed for each imaging study.

Resources, patient commitment and staff requirements
Some IB studies can proceed on standard of care images (23). Many IB validation studies require acquisition of new data because appropriate raw images do not exist already in an image bank or repository. Recruiting even moderate numbers of patients to imaging studies can be challenging, costly and may require substantial patient commitment without therapeutic benefit.
Some CT, MRI and PET research protocols require intravenous access for contrast or radiopharmaceutical administration. Studies may require multiple examinations and scan time may last up to several hours (5). Some modalities can be uncomfortable, noisy (MRI) and require invasive procedures (e.g. arterial lines to provide input functions and allow metabolite analysis in some early validation studies in PET). Other modalities involve ionizing radiation. However, not all IB studies have as many practical issues; for example DCE-US protocols can be more rapid, cheaper and require less patient commitment (14).
The requirements of IB studies must be compared to those found with biospecimen biomarker studies. For example, if only one biopsy is performed or if biofluid samples are collected along with samples taken for routine clinical care, then this may be seen as attractive relative to imaging-based studies. Biopsies do however generally carry greater clinical risk than most imaging methods. Limited recruitment may lead to highly selected patient groups for initial validation studies that may be unrepresentative of IB performance in large populations (2).
Developing a program of clinical studies to validate and qualify the most promising IBs takes considerable time and resource. This is quite different to developing a biospecimen biomarker from biobanked samples (2) or the specific case developing an IB from readily available clinical scan data such as radiotherapy planning (24) or diagnostic CT scans (25,26). Qualification of an IB for a particular use (e.g. the role of PET 18 F-FDG SUV max in patient selection in an adaptive phase II study) (27) may take many years and will be expensive.
Finally, IBs depend enormously on the performance of clinicians (radiologists, nuclear medicine physicians and others), technicians (including radiographers) and scientists. This is well appreciated for ultrasound (28), but is true for all imaging modalities. While biospecimen biomarkers are also vulnerable to "preanalytical" factors, the scope for human error and impact of human variability are likely to be high in imaging studies with multiple acquisition and analysis steps. Unfortunately, mistakes and human variability associated with imaging are often not documented and can seldom be rectified retrospectively.