CT indicates computed tomography; ER, estrogen receptor; FDG, fludeoxyglucose F 18; PD, pharmacodynamics; PET, positron emission tomography; and PK, pharmacokinetics.
PET imaging of estrogen receptor (ER) expression. Images from a patient with left locally advanced breast cancer imaged using fluoroestradiol F 18 (FES) to visualize ER expression (A) and fludeoxyglucose F 18 (FDG) to indicate areas of metabolically active tumor (B). The patient’s arms are raised for the FES scan and lowered for the FDG scan. Arrowheads indicate the tumor site. PET imaging of cellular proliferation (C). Fluorothymidine F 18 (FLT) PET and CT images of a patient with multifocal breast cancer (solid arrowheads). Normal uptake of FLT is also seen in marrow, a highly proliferative normal organ (blue arrowhead).
Factors in boldface type indicate technical (eg, probe chemistry) or logistical (eg, regulatory approval) barriers; items in lightface type indicate key oncology and imaging considerations that should guide testing. FDA indicates US Food and Drug Administration; IND, investigational new drug.
Mankoff DA, Farwell MD, Clark AS, Pryma DA. Making Molecular Imaging a Clinical Tool for Precision OncologyA Review. JAMA Oncol. Published online December 29, 2016. doi:10.1001/jamaoncol.2016.5084
Individualized cancer treatment, tailored to a particular patient and the tumor’s biological features (precision oncology), requires a detailed knowledge of tumor biology. Biological characterization is typically performed on biopsy material, but this approach can present challenges for widespread and/or heterogeneous disease and for performing serial assays to infer changes in response to therapy. Molecular imaging is a complementary approach that provides noninvasive and quantitative measures of the in vivo biology of the full disease burden and is well suited to serial assay.
Molecular imaging can provide unique information to guide precision oncology that includes measuring the regional expression of therapeutic targets, measuring drug pharmacokinetics, measuring therapy pharmacodynamics, and providing a marker of therapeutic efficacy that is highly indicative of outcome. Thus far, most trials of novel molecular imaging in oncology have been small, single-center trials. Only a few methods have progressed to multicenter trials and even fewer have become part of clinical practice.
Conclusions and Relevance
Molecular imaging holds great promise for precision oncology, complementing tissue-based markers to guide more effective, less toxic, and more cost-effective cancer treatments. Beyond logistical and technical challenges, moving new imaging tests from the laboratory to the clinic requires a compelling use case that will benefit patients and/or improve cost-effectiveness, and it requires the collaboration of imagers, oncologists, and industry to reach its true clinical potential.
An important part of current oncologic practice is the ability to assay cancers for biological features to guide therapeutic choices.1 Cancer biomarkers are traditionally based on in vitro tissue assays.2 Advances in molecular biology and laboratory-based methods have increased the ability to measure cancer biology, including genomics, gene expression, and protein expression.3- 6 Similarly, advances in both technology and cancer science now provide the ability to perform noninvasive cancer assays, including molecular imaging.7 The Society of Nuclear Medicine and Molecular Imaging defines molecular imaging as “the visualization, characterization, and measurement of biological processes at the molecular and cellular levels in humans and other living systems.”8[p18N] Thus far, clinical molecular imaging has been largely radionuclide based in the form of positron emission tomography (PET) or single-photon emission computed tomography. Ongoing work in optical, photoacoustic, ultrasonographic, and hyperpolarized magnetic resonance imaging and spectroscopy offers capabilities highly complementary to radioisotope imaging.9,10
Molecular imaging is distinct from more commonly used structural and anatomical imaging, such as computed tomography (CT). An intriguing development for anatomical imaging is the ability to extract imaging features that correlate with tumor biology, often termed radiomics, which has shown promise for characterizing cancers and indicating probable behavior.11,12 This approach is also applicable to molecular imaging.13 Advances in both molecular imaging and radiomics offer new opportunities for imaging as a biomarker for precision oncology.
Although there has been clinical acceptance of a few specific types of molecular imaging in oncology (eg, fludeoxyglucose F 18 [FDG] PET/CT for cancer detection and staging14), there has been only modest progress in translating other novel molecular imaging methods beyond early trials. This review highlights areas of progress toward this goal and is focused on PET molecular imaging and its application to systemic cancer therapy. We also provide a critical appraisal of the barriers to moving new molecular imaging tests into clinical practice.
Molecular imaging has capabilities distinct from those of tissue-based molecular assays.15 Imaging can measure biological properties across the entire burden of disease when comprehensive tissue sampling is neither feasible nor well tolerated, and imaging can measure the regional heterogeneity of cancer biological properties, especially for widespread and/or heavily treated disease.16,17 Molecular imaging is noninvasive and quantitative, making it ideal for serial studies to infer drug action and predict therapeutic response.
Limitations of molecular imaging as a cancer biomarker include the ability to measure, at most, a few processes simultaneously. Imaging generally provides millimeter resolution compared with tissue-based methods, which can query many different biologic processes at once at much higher spatial resolution. In addition, although it is possible to batch tissue assays at a central processing facility, imaging must be performed in 1 individual at a time at a site local to the patient. The need for sophisticated equipment and imaging probes at each site of clinical practice contributes to the cost of imaging biomarkers and can lead to methodologic variability. These differences between molecular imaging and tissue-based assays must be considered in directing molecular imaging to the applications most likely to have a clinical impact.
The use of molecular imaging biomarkers to guide precision oncology can be broadly categorized as follows (Figure 1 and Table 1)17,28,29:
identifying therapeutic targets and selecting patients, with the goal of identifying patients likely to benefit from targeted therapy;
measuring drug pharmacokinetics and guiding dosing, with the goal of measuring drug transport, delivery, and clearance within the tumor and/or healthy organs to guide drug dosing and minimize toxic effects;
measuring drug pharmacodynamics, with the goal of measuring drug effect on the tumor to determine the probable therapeutic response and early indication of therapeutic futility; and
predicting patient outcomes, including survival, and possibly serving as a surrogate outcome for clinical trials and clinical practice.
In addition to the applications listed above, there has been considerable research in novel molecular imaging methods for cancer detection and staging, especially for cancers such as prostate cancer that are not well served by FDG PET/CT.30 Although accurate staging and restaging of cancer are important for guiding cancer therapy, we focus our discussion on molecular imaging as a cancer biomarker. Reviews of molecular imaging cancer detection and staging are available.31
Tissue assays for estrogen receptor (ER) and HER2 are routinely used to direct therapy.32,33 The ER- and HER2-targeted therapies have had a significant effect on the survival of patients with breast cancer, but only for those whose tumors express the targets.34 PET imaging probes for ER (eg, fluoroestradiol F 18 [FES]) (Figure 2A and B) and HER2 (eg, trastuzumab Zr 89) have been validated against in vitro assays of biopsy material.20,35,36 Preliminary data support quantitative measures of probe uptake as predictive assays for response to endocrine therapy37,38 and HER2-targeted therapy.21 HER2 PET has demonstrated predictive value in multicenter trials,21 and a multicenter trial of FES PET is under way (F-18 Fluoroestradiol PET as a Predictive Measure for Endocrine Therapy in Patients with Newly Diagnosed Metastatic Breast Cancer) based on promising single-center trials18,19,38 and an investigational new drug–enabling trial.39 These probes could help to guide treatment for ER- and HER2-targeted therapy in the clinic, especially for advanced disease where tissue sampling poses a challenge.
Molecular imaging can also be used to guide locoregional therapy, especially for radiotherapy treatment planning and response evaluation.40 The ability to measure the spatial distribution of biological features helps in determining the boundaries of the tumor and enables biologically based treatment planning.41
Labeled versions of cancer drugs have been developed and tested in humans, providing insights into drug kinetics and delivery.42 Alternatively, probes measuring drug targets can test the adequacy of dosing for drugs designed to engage or antagonize the target. For example, based on studies demonstrating FES PET’s ability to measure ER blockade for drugs such as tamoxifen citrate and fulvestrant,19,32,43 FES PET successfully guided early phase 2 testing of a novel ERα antagonist, ARN-810, demonstrating greater than 90% suppression of estradiol binding to ER in 90% of the patients.44 The ability of PET imaging to measure target engagement can be especially informative if the drug lacks efficacy. If the lack of efficacy is due to subtherapeutic drug levels in the tumor, then higher doses or use of a more potent drug should be explored. However, if the drug demonstrates a lack of efficacy despite significant target engagement, then alternative drugs for a different target or combination therapies should be considered.
The ability to image molecular events downstream of targeted drug action underlies interest in molecular imaging as a pharmacodynamic marker.45,46 A prime example is the use of FDG PET as an early indicator of response to imatinib mesylate, demonstrating a decline in FDG uptake within hours after dosing.47 Of other emerging methods, PET cellular proliferation imaging is particularly well suited as a pharmacodynamic marker and has been shown to determine the probability of response to both cytotoxic chemotherapy and targeted agents. Fluorothymidine F 18 (FLT) (Figure 2C), a tracer of thymidine flux into DNA, indicated responsiveness to epidermal growth factor receptor–targeted therapy in lung cancer early in the course of treatment.24 A recent multicenter trial carried out in more than 20 centers (including several community practice centers) demonstrated the ability of FLT PET to determine the probability of breast cancer pathologic complete response to neoadjuvant chemotherapy after a single treatment.25 Other novel probes, for example, ISO-1 F 18,48 are well suited to cell-cycle–targeted therapy and could guide clinical trials and clinical use.
The ability to measure the biological activity of cancer after therapy, as opposed to the presence or absence of a residual mass, provides a powerful basis for molecular imaging to determine probable therapeutic outcomes. This ability has been best demonstrated for FDG PET/CT and lymphoma, where posttherapy PET is often used as an integral biomarker to direct therapy in clinical trials and clinical practice.26 Molecular imaging can also assess response for challenging disease sites, such as bone. For example, studies have shown that FDG PET and FDG PET/CT estimate the probability of progression-free survival and time to skeletal events for patients with breast cancer who have bone metastases.27,49 Recent studies suggest that the combination of FDG PET/CT and more-specific molecular probes offers even better predictive value, for example, combining PET ER or HER2 imaging with FDG PET/CT.21,50
These examples indicate the potential of molecular imaging for precision oncology, highlighting methods that have undergone multicenter testing and, in rare cases, moved into clinical practice. There are, however, many more examples of early trials that have not progressed toward clinical translation. This issue is the subject of the remainder of this review.
The development and clinical translation of new molecular imaging methods requires many steps from conceptual development to clinical use (Figure 3). There are several logistical and technical barriers to translation that include (boldface type in lower boxes in Figure 3) probe chemistry, standardized probe production methods suitable for human use, standardized imaging acquisition and analysis for multicenter trials, regulatory approval for clinical trials (investigational new drug) and clinical practice (new drug application), a commercial supply for imaging probes, and payer reimbursement. Some hurdles may be overcome by technical innovation. For example, the long half-life of zirconium-89, a label for macromolecules, facilitates central distribution; however, the multitude of technical and logistical barriers challenges clinical translation.
In addition to technical and logistical challenges, a less well recognized and more poorly addressed challenge relates to the need for the imaging and oncology communities to work together to guide imaging test development and translation (Table 2). How do we, as partners, decide which methods and applications make sense to move out into the clinic? How do we test promising new methods? These questions are highlighted in lightface type in the lower boxes in Figure 3, and below we suggest a framework for addressing these issues. We also refer the reader to a recent review of related questions posed to imaging probes developers.51
A clinically useful imaging tool should fill a need not addressed by existing methods. Although this requirement seems intuitively obvious, it is often not considered in efforts to move new methods to the clinic. An effective test should help to determine the most effective, least toxic treatment and should steer clinicians away from options unlikely to improve patient outcome. The effect of imaging on patients should be measured in outcomes, such as progression-free survival and quality of life, and also the economic benefit of avoiding drugs unlikely to be helpful. Molecular imaging use should be justified in terms of its ability to direct effective—and cost-effective—treatment. These considerations suggest application to patients in whom biopsy is difficult or misleading (eg, bone metastases) and to treatments that are toxic and/or expensive.
In this context, molecular imaging to augment biopsy-based target assessment and measure early response to treatment seems particularly apropos. A recent example is combined HER2 and FDG imaging in the ZEPHIR T-DM1 (Phase II Prospective Imaging Study Evaluating the Utility of Pre-treatment Zr-89 Labeled Trastuzumab PET/CT and an Early FDG-PET/CT Response to Identify Patients With Advanced HER2+ BC Unlikely to Benefit From a Novel antiHER2 Therapy: TDM1) trial.21 Combined imaging before and after 1 cycle of T-DM1 had 100% accuracy in correlating with therapeutic response. In this trial, only 55% of the patients responded, toxic effects were modest, and there was considerable cost associated with drug treatment. Molecular imaging could have spared nearly 50% of the patients the toxic effects and cost of ineffective therapy. Although the imaging in this example carries cost, the imaging costs would be offset by the savings of discontinuing ineffective and costly drugs.52
Recent advances in immunotherapy, including immune checkpoint inhibitors, vaccines, and cell-based therapies, have created challenges in patient selection and assessment of response to therapy.53,54 Many new imaging tools are in preclinical development, such as radiolabeled anti-programmed death ligand 1 (PD-L1) and anti-PD-1 antibodies,55- 57 a radiolabeled anti-CD8 cys-diabody,58 and others.59 Although it is not yet clear whether these will be useful imaging biomarkers, incorporating the most promising candidates as integrated markers in therapeutic trials will be the best way to validate or reject them. With relatively low response rates but high impact in responders at significant toxic effects and cost, immunotherapies provide a clinically compelling need for an imaging biomarker.
Traditional oncologic practice relegates imaging to cancer staging and late-phase response evaluation (restaging). If the tumor shrinks after a few months of treatment, the therapy is working; if the tumor grows, the treatment should be discarded. Both imagers and oncologists perceive cancer imaging as largely qualitative, imprecise, and unable to measure response early. Cancer physicians perceive a biomarker to be a tissue assay, thinking that only tissue-based markers should guide the choice of therapy. Imagers contribute to this perception by emphasizing cancer imaging for detection and staging and by providing largely qualitative interpretations.
To move beyond preconceived notions, we need to support uniform methodology and rigorous testing of quantitative imaging biomarkers compared against tissue-based biomarkers in compelling clinical applications. We should consider only applications where imaging can affect clinical decision making and augment practice beyond current imaging tests and tissue-based assays.1,28,45 Imagers and oncologists need to work together to determine how best to validate imaging biomarkers for clinical decision making. Guidelines for tissue-based biomarkers60 and evolving standards for quantitative imaging61 provide a basis for these discussions.
In moving toward these goals, it will be helpful to start with examples in which using molecular imaging to make therapeutic decisions is only a minor departure from current clinical practice. An example lies in FES PET to guide endocrine therapy in metastatic breast cancer. Current guidelines call for biopsy of a metastatic site at diagnosis and possibly at progression to determine ER expression; however, this procedure is not uniformly done in practice, especially for bone-dominant disease.62 Molecular imaging can help. Studies have shown that the FES uptake correlates with in vitro assay of ER expression performed on biopsy material.35,36 Single-center trials also showed that, in the endocrine salvage setting, imaging identified up to 40% of patients with low or absent ER expression whose cancers responded to subsequent endocrine therapy.18 In this case, using FES PET to guide metastatic breast cancer treatment target to the ER, once validated in a prospective trial, would provide a compelling case for imaging that is both congruent with current practice and effective in directing patients away from ineffective therapy.
Clinical use of biomarkers should be based on high-quality clinical trial data supporting the marker’s accuracy and utility. For tissue-based biomarkers, data can be collected by central assay of archived tissue and then compared with therapeutic outcomes in a retrospective and observational (integrated) fashion.2 Testing new assays has minimal effect on the trial and its patients since tissue collection is already part of standard clinical practice. These considerations do not hold for investigational imaging biomarkers, which require prospective data collection at each trial site and add additional patient burden at no benefit.63 Sites are often unprepared to carry out and analyze novel imaging procedures. The inclusion of experimental imaging biomarkers adds cost and complexity to therapeutic trials, limiting enthusiasm for their inclusion. These factors challenge testing molecular imaging methods as integrated biomarkers, the key to validating their use as integral markers in larger clinical trials and clinical practice.
How do we overcome these challenges? First, and most important, we need to choose imaging methods and applications that provide a compelling use case. We too often study new imaging approaches based on excitement about the technology without fully considering clinical impact. Clinical trials of novel molecular imaging are compelling to patients and their referring oncologists only when they address clinical challenges not met by existing approaches. In addition, the complexity of testing novel imaging methods requires an infrastructure for imaging biomarker clinical trials that can distribute imaging probes, qualify and calibrate advanced imaging devices, train technologists at each site in image acquisition, and provide support for local interpretation and advanced central analysis. Recent US National Clinical Trials Network and European group studies involving novel imaging methods and end points indicate some progress.21,25,64,65 National Cancer Institute and commercial entities have supported the distribution of molecular imaging probes for multicenter trials through both commercial and academic suppliers.25,64 Networks of expertise now provide guidelines for uniform instrument calibration and image acquisition, as well as imaging processing and analysis laboratories capable of central analysis for novel molecular imaging studies.61,64 These developments all improve the landscape for molecular imaging clinical trials, but more work is needed. Finally, industry and public funding must share the cost of testing new imaging markers. While industry supports therapeutic clinical trials, neither pharmaceutical companies nor imaging manufacturers have fully embraced the need to support imaging biomarker trials. They should. Imaging device manufacturers should recognize an increasing need to justify the cost of their devices by demonstrating that imaging improves patient outcomes and reduces overall cost. Pharma should recognize that the ability to select the patients most likely to benefit from new but costly drugs improves clinical acceptance and reimbursement. All sides must recognize the need for embedded studies of patient impact and cost-effectiveness to justify third-party payment for novel imaging tests.
Surgeons and radiation oncologists have recognized the need for training for new procedures.66 This need also applies to imaging. Advances in imaging are heralded by the availability of new devices and imaging agents, but not always accompanied by advances in training and certification. Prior examples, such as breast magnetic resonance imaging and FDG PET/CT, have shown that new tests may perform poorly in their early clinical implementation, often at much lower accuracy than in trials at experienced centers.67,68 Translation of new molecular imaging tests into the clinic setting must include training for interpreting physicians and technologists in patient preparation, imaging acquisition, and interpretation for new imaging tests. The inclusion of community oncology sites in imaging clinical trials, as accomplished in a recent FLT PET study,25 can help to vet new imaging procedures for clinical practice. Recognition of the need for specialty training in new imaging tests has led to more tailored training and certification, including certifying organizations such as the American Board of Nuclear Medicine and American Board of Radiology,67,69,70 but a broader approach is needed.
Molecular imaging holds great promise for precision oncology, complementing tissue-based markers to guide more effective, less toxic, and more cost-effective cancer treatments. There are many logistical and technical factors that limit the translation of new methods for the laboratory in the clinic. Beyond such challenges, moving new imaging tests from the laboratory to the clinic requires a compelling use case that will benefit patients and/or improve cost-effectiveness. Both imagers and oncologists need to consider imaging biomarkers in a broader context than their current use for detection and staging. Both the imaging and oncology communities must support evolving frameworks for rigorous trials of molecular imaging biomarkers that can provide level 1 evidence for clinical practice. Both pharma and imaging manufacturers need to support clinical trials to promote more efficient and effective application of imaging and therapy. Only when these forces align will molecular imaging reach its true clinical potential.
Corresponding Author: David A. Mankoff, MD, PhD, Division of Nuclear Medicine and Molecular Imaging, Department of Radiology, Perelman School of Medicine at the University of Pennsylvania, 3400 Spruce St, Philadelphia, PA 19104 (firstname.lastname@example.org).
Accepted for Publication: September 17, 2016.
Published Online: December 29, 2016. doi:10.1001/jamaoncol.2016.5084
Author Contributions: Dr Mankoff had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: All authors.
Acquisition, analysis, or interpretation of data: Mankoff.
Drafting of the manuscript: Mankoff.
Critical revision of the manuscript for important intellectual content: All authors.
Obtained funding: Mankoff.
Administrative, technical, or material support: Mankoff, Clark.
Conflict of Interest Disclosures: None reported.
Funding/Support: Manuscript writing was supported in part by grant SAC140060 from the Susan G. Komen Foundation, grant DE-SC0012476 from the Department of Energy, funding from the University of Pennsylvania Health System Breast Cancer Translational Center of Excellence, and Cancer Center Support grant P30CA016520 from the National Institutes of Health.
Role of the Funder/Sponsor: The support provided relevant research funding for the authors but did not have any influence on the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.