QIBA Newsletter

In My Opinion

The Value of QI in Early and Late-phase Clinical Trials

By Gudrun Zahlmann, PhD

Biomarkers are used in clinical trials to guide the rationale of therapeutic intervention and to enable decisions on continuation/discontinuation of development processes for medical devices and especially for the development of novel therapeutics and drugs. Imaging techniques provide specific in vivo biomarkers that are additionally used for patient stratification, to inform trial protocol and trial design, to provide information for modeling of pharmacokinetics/pharmacodynamics, and to predict response as well as to monitor short-, mid- and long-term therapeutic outcomes.  Which imaging biomarker to use depends on which question needs to be answered.

In early drug development phases, questions are, e.g., whether the new drug candidate reaches the target and, if so, in sufficient quantity or whether the expected mechanism of action can be observed. These activities often start pre-clinically. Here, a full tool box of potential biomarkers, including quantitative imaging biomarkers, is available. The biomarker(s) that are most informative are used. Often we see a variety of imaging techniques for pre-clinical in vivo imaging that provide valuable insights into biological processes and have the potential to guide decision in early drug development ^(1).

It becomes challenging when pre-clinical biomarkers need translation into clinical procedures. From an imaging perspective, it is easier to translate, e.g., MR sequences from animal into human use. One successful example was the use of DCE-MRI for the development of anti-angiogenic drugs ^(2). This example also illustrates how good communication and collaboration between drug companies and regulators can speed up drug development by introducing, validating and accepting a new biomarker (K^trans based on DCE-MRI) for efficacy of, at that time, a novel drug class.

Such a seamless translation process is not always possible. Novel tracers for PET/SPECT need almost the same approval process as a new drug and are not always available for large scale, late-phase clinical trials. Therefore, we typically have a larger tool box for early development compared to late development phases ^(1).

Nevertheless, imaging biomarkers play an important role in late-phase clinical trials. The importance of imaging endpoints in clinical trials is reflected by specific FDA guidance to industry published April 2018 after thorough discussion ^(3).

QIBA´s mission is to support quantitative imaging biomarker applications in clinical trials and clinical care by standardizing all imaging procedure-related activities to enable reproducible and reliable measures from imaging techniques. This is in full agreement with the need of the researchers and regulators to build decisions on solid data. QIBA´s profiles are dedicated to this task. As a volunteer organization, QIBA has its own timelines in developing profiles, and needs synchronization with the needs of Profile users in clinical research, supported by industry or public funding. The next phase of QIBA activities should reflect this necessity.

References:

1. Mayer AT, Gambhir SS; The Immunoimaging Toolbox; the Journal of Nuclear Medicine, 2018; 59;8;1174-1182.O´Connor JPB, Parker GJM, Jackson A, Jayson GC; Dynamic contrast-enhanced MRI in clinical trials of antivascular therapies; Nature Reviews Clinical Oncology 9(2012)3, 167-177.

    

2. Clinical Trial Imaging Endpoint Process Standards, Guidance for Industry, FDA Center for Drug Evaluation and Research (CDER), Center for Biologics Evaluation and Research (CBER); April 2018.

PubMed

QI / IMAGING BIOMARKERS IN THE LITERATURE

PubMed Search on: The Value of QI in Early and Late-phase Clinical Trials.

Each issue of QIBA Newsletter features a link to a dynamic search in PubMed, the National Library of Medicine's interface to its MEDLINE database. Link to articles on: The Value of QI in Early and Late-phase Clinical Trials.

Analysis Tools and Techniques

The QIBA CT Volumetry Profile Conformance Assessment

By Nicholas Petrick, PhD, and Kevin O’Donnell, MASc

The goal of a QIBA Profile is to achieve the measurement performance defined in the Profile Claims.  The measurement process is modeled as a chain of "Activities" and, for each Activity, requirements are placed on the participating people and systems ("Actors") as needed to achieve the Claims.  Below we discuss conformance testing for the CT Tumor Volume Change for Advanced Disease (CTV-AD) Profile ^[1] and our initial experiences with testing. To simplify conformance, requirements were grouped into checklists, including both simple observations and detailed assessment procedures. 

Scanner and Reconstruction Software conformance requirements focus on achieving image quality sufficient for tumor segmentation by image analysis software.  Based on expert opinion and our groundwork studies, most CT acquisition parameters were determined to manifest as an impact on resolution or noise, or as not having a strong impact on segmentation.  Thus, conformance came down to meeting specific targets for spatial resolution (f50-value for the modulation transfer function [MTF]) and noise (voxel noise standard deviation) with only a few additional scanning requirements.  Similar simplification may be effective in other QIBA Profiles. Scanner assessment of MTF and voxel noise was based on AAPM recommendations and the American College of Radiology (ACR) CT Accreditation Phantom such that the procedures are familiar and available at most sites. 

Image Analysis Software conformance testing includes demonstrating tumor segmentation performance.  A challenge is that performance on clinical datasets is most relevant, but "ground truth" is only known for phantom datasets.  The current compromise assesses repeatability using clinical images and segmentation bias and linearity using images of artificial lesions in an anthropomorphic phantom (see Fig. 1).  Test data is provided via the QIDW (Quantitative Imaging Data Warehouse) website, which means sites do not need to collect clinical cases or have access to an anthropomorphic phantom. 

These assessments are non-trivial exercises due to the need to segment many lesions, but once a software package is qualified, testing does not need to be repeated.  The phantom dataset has multiple scans of seven different nodules for a total of 39 volume estimates.  The clinical dataset has 20 patients with two repeated scans each.  These numbers are small and impact the robustness of our assessment procedure, but a balance between conformance testing effort and statistical power is required.  Ideally larger datasets, especially for repeatability analysis, would be available to improve the statistical power of the analyses.  In the future, digitally inserting realistic lesions into clinical images, such that truth is known, may provide a better solution for assessing both repeatability and bias/linearity within a single dataset ^[2].

While Technologist and Physicist conformance mostly involves training to highlight specific details of familiar procedures, the radiologist plays a critical QA role in the Profile by flagging problematic studies.  For semi-automated algorithms, the radiologist affects segmentation performance and thus must perform the Image Analysis Software conformance testing for each software package used.  This is a substantial burden, which may hinder the utilization of the Profile, and is an incentive to fully automate segmentation.

Focus On

QIBA® Profile Conformance Testing Now Available