Eligibility criteria in clinical trials are designed to ensure the safety of research participants and tailored based on scientific objective of a trial. The eligibility criteria differ from trial to trial, but a clear diagnosis of disease characteristics and disease staging are often essential before patient enrollment. In oncology, the TNM staging system is widely used in trials in all solid tumors and remains a vital component of eligibility criteria. The TNM may be classified as an imaging biomarker, as it usually relies on imaging to define the size and extent of the primary tumor (T), any lymphatic involvement (N), and the presence of metastases (M). Whole-body PET imaging improves TNM assessment from morphologic imaging, as it reveals metastases not evident on conventional CT or MR imaging. Molecular imaging techniques are increasingly used to augment conventional imaging to confirm patient eligibility. For example, imaging with [¹⁸F]-FDG has been used in the EORTC LungTech trial, to ensure that enrolled patients are with inoperable early stage non-small cell lung cancer (NSCLC) [17].

Clinical trials can be enriched by selecting subpopulations that may be more responsive to treatments, so as to improve the chance of trial success. Molecular and functional imaging provides additional information on tumor microenvironment, which could help to “preselect” sub-patient populations. For example, identification of tumor hypoxia could facilitate the use of hypoxia-stimulated pro-drugs, which selectively kill hypoxic cells. Tirapazamine, a cytotoxic agent with high selective toxicity toward hypoxic cells, is a good example. The relatively limited benefit obtained in a trial of NSCLC reported by the CATAPULT I study group was likely due to poor patient stratification with inclusion of patients with better-oxygenated tumors [18]. Rischin and his colleagues compared the cisplatin/5-FU versus cisplatin/tirapazamine regimen in patients with head and neck squamous-cell cancer (HNSCC), where [¹⁸F]-fluoromisonidazole (FMISO)-PET hypoxia imaging was used to stratify the tumors into hypoxic and non-hypoxic ones. The study showed that tirapazamine improved local tumor control in hypoxic but not in non-hypoxic tumors [19]. Similar examples are imaging with ^99mTc-etarfolatide used to identify folate receptor-positive patients with advanced ovarian cancer, who are most likely to benefit from treatment with vintafolide, a folate receptor-targeted therapy [20], and [¹⁸F]-FDG used in patients with gefitinib-treated NSCLC, which showed a low baseline SUV of [¹⁸F]-FDG associated with a higher response rate (53 % versus 18 %) and a prolonged PFS (median, 33.1 weeks versus 8.6 weeks) [21]. The prognostic value of those imaging biomarkers needs to be further validated with multicentric prospective studies, to confirm that baseline imaging without any treatment effect can predict the degree of risk for disease occurrence, or progression and ultimately to implement this patient stratification strategy in therapeutic drug trails.

A very direct approach in using imaging to verify therapeutic targets is to image the distribution of the drug itself. Typically, the drug molecule would be synthesized with a positron-emitting isotope (e.g., ¹¹C or ¹⁸F) replacing the natural isotope, allowing detection by PET. This approach is common in neuroscience although rare in other diseases. Alternatively, for detection with single-photon emission computed tomography (SPECT), a gamma-emitting moiety may be attached to the drug molecule (albeit with the risk that the drug’s properties are thereby altered). An example is radiolabeled derivative of trastuzumab, an antihuman epidermal growth factor receptor 2 (HER2) monoclonal antibody used to treat breast cancer patients with HER2-expressing tumors [22, 23]. Radiolabeled trastuzumab helps to visualize the affinity of the drug in vivo and provides useful information about the PK properties of the drug, such as injected dose versus accumulated drug concentration in the organs and its regional bio-distribution. The noninvasive whole-body imaging overcomes problems associated with biopsies, including sampling errors and discordance of expression between primary tumors and metastases. More importantly, the drug uptake by the target tissue can be quantified at sequential imaging scans, and might provide insight into drug’s action at the target tissue and its association with tumor response. Indeed, for the radiotherapeutic drug [¹³¹I]-tositumomab, FDA requires that PK be verified by imaging before a therapeutic dose be given [24].

Dose escalation and schedule, the main purpose of phase I trials, is usually undertaken to define the maximum tolerated dose (MTD), with the assumption that the most pronounced changes are likely to be detected at the highest dose. But, target saturation may already be reached at lower dose levels, and imaging changes are likely to be apparent and help to define an optimal biological dose. The effects of antiangiogenic therapies on dynamic contrast enhanced (DCE)-MRI have been documented in 39 phase I and II trials with a significant reduction in the transfer constant (K ^trans ) and/or initial area under the gadolinium curve (IAUGC) with multiple agents [25]. In a study of brivanib, a dual tyrosine kinase inhibitor of both vascular endothelial growth factor receptor and fibroblast growth factor receptor, DCE-MRI has been used in several dose schedules and then selected the optimal schedule for a phase II trial [26]. A similar approach has been used with vatalanib [27] and cediranib [28] in advanced cancer, and with sorafenib in renal cancer [29].

Using imaging biomarkers as surrogate endpoint to assess new drug efficacy has been addressed in the above content. Clinical outcome such as mortality often takes years of follow-up to establish, and the determination of morbidity could be subjective. The use of imaging biomarkers to assess response or progression can reduce sample size, trial duration, and cost and accelerate the introduction of new drugs. A notable example is the approval of etanercept, a tumor necrosis factor inhibitor for the treatment of rheumatoid arthritis. A phase III trial [30] was designed to evaluate the efficacy and safety of etanercept and methotrexate (the standard of care) in patients with early rheumatoid arthritis, based on two sets of criteria: (1) the American College of Rheumatology scores, which used patient outcome report such as pain and function, in addition to serum C-reactive protein level and (2) imaging evidence of progression, such as joint-space narrowing and erosion. The clinical scoring only showed etanercept a more rapid treatment effect within the first 6 months but was approximately the same thereafter between two groups. Imaging-based erosion score showed significant differences of both immediate and long-term effect. The FDA granted etanercept a conditional marketing authorization. Subsequent studies demonstrated that etanercept achieved sustained improvement compared to methotrexate on both clinical and imaging scores [31]. A conditional approval is a continuum approach that drug developers should continue conducting studies to collect data on the effectiveness of drugs in use after initial approval and eventually provide evidence to keep the drug on the market and obtain a full approval within a predefined timeline.

Many tyrosine kinase inhibitors and monoclonal antibodies targeting signal transduction inhibitor of tumor growth without tumor regression. Imatinib mesylate is such an example, a tyrosine kinase inhibitor for the treatment of chronic myelogenous leukemia and gastrointestinal stromal tumors (GIST). As opposed to the classical cytotoxic treatment, response rates are low, despite a high percentage of patients having prolonged stable disease and sometimes improvements in survival compared to standard therapies. Acute changes in tumor size are seldom significant in GIST patients treated by imatinib. However, when using [¹⁸F]-FDG-PET as a biomarker of tumor metabolism, response could be detected as early as 8 days following the start of treatment and was also associated with a longer PFS [32]. Other examples of early imaging assessment by MRI biomarkers have been reported [33, 34].

The use of functional and molecular imaging to predict the likelihood of response to a particular treatment at early stage is very attractive for drug development. This strategy has been used in early phase trials for proof of concept, although they still await validation to prove their relationships with clinical benefit. In these proof-of-concept trials, drug safety and efficacy will require the confirmation in subsequent trials.

Noninvasive imaging biomarkers have a huge potential in monitoring drug safety during clinical trials. In addition to LVEF and BMD, which have achieved surrogate endpoint status, emerging imaging biomarkers can detect drug-induced changes of vital organs and provide region-specific information about tissue abnormality, while serum and urine biomarkers can still be normal due to the functional reserve of the affected organs. The reserve of kidney function can compensate for up to 75 % of the loss, so serum and urine biomarkers are insensitive for early renal damage. A wide range of drugs can cause renal papillary necrosis, and the diagnosis tends to be made when irreversible destructive changes have occurred. Contrast-enhanced multiphasic CT can identify early signal changes of renal papillary necrosis and medullary necrosis and at the same time monitor lesion progression and regression. The liver is also a metabolic organ, often suffers from drug toxicity. In some cases, drug-induced hepatic steatosis can lead to a rapid evaluation of severe hepatic failure and ultimately death [35]. Both morphologic and functional liver imaging can display manifestations of hepatotoxicity, easier and earlier than histology techniques.

Although many molecular and functional imaging biomarkers have been explored in clinical trials, few have yet proved adequate to support decision-making in drug development, in regulatory approval, or in clinical practice. Table 21.3 summarizes some common pitfalls of imaging biomarker implementation in trials, and each pitfall is accompanied by proposed solutions.