Executive summary of AAPM Report Task Group 113: Guidance for the physics aspects of clinical trials.

The charge of AAPM Task Group 113 is to provide guidance for the physics aspects of clinical trials to minimize variability in planning and dose delivery for external beam trials involving photons and electrons. Several studies have demonstrated the importance of protocol compliance on patient outcome. Minimizing variability for treatments at different centers improves the quality and efficiency of clinical trials. Attention is focused on areas where variability can be minimized through standardization of protocols and processes through all aspects of clinical trials. Recommendations are presented for clinical trial designers, physicists supporting clinical trials at their individual clinics, quality assurance centers, and manufacturers.

ABOUT THIS EXECUTIVE SUMMARY

The full report of AAPM Task Group 113 on Guidance for the Physics Aspects of Clinical Trials is available at the AAPM Reports website. This executive summary provides an overview of the major headings of the full report. In addition, details were retained in this report to highlight a few areas where there has been an evolution in clinical trials. Appendices A–D include all of the TG113 recommendations with the reference information contained in the full report.

INTRODUCTION AND CHARGE OF THE REPORT

There is growing evidence1, 2, 3, 4, 5 on the need for standardization of treatment planning and delivery methods to ensure quality in clinical trials to help support the investigation of new safe and effective treatments and/or assessment methods in multi‐institutional settings. Such standardization will improve the consistency of the radiotherapy received by patients and the radiotherapy data submitted for a given clinical trial. These data are required to validate that all patients in each arm of a given study received the therapy as intended. Violating this assumption can jeopardize the validity of the outcomes reported by the trial group.

A related consideration that affects overall quality is the ability of those participating in clinical trials to create plans as part of their standard clinical flow that are both compliant with protocol specifications and optimal. The importance of compliance in trials and the impact on detecting changes in outcome have been demonstrated in a number of trials,1, 2, 3, 4, 6 such as TROG 02.02 on advanced head and neck cancer (Fig. 1), and in meta‐analyses of other trials. When designing a trial, the planning guidelines are set to be able to answer the clinical trial questions. However, there may be variation in planning methods, and a planner may not know when a better (such as improved target coverage with reduced dose to normal tissues) plan is reasonably achievable without real‐time feedback during the planning process. Knowledge‐based planning, where the achievable dose‐volume metrics from previous patients can used to predict each new patient's DVH, was shown to retrospectively identify plans which were clinically acceptable but suboptimal in the context of the clinical trial.7 For example, plan quality was analyzed for patients treated on RTOG 0126 exploring the relationship between plan quality and rectal toxicity. Suboptimal plans were identified by comparing predictions for target and organ‐at‐risk doses to those that were submitted as part of a trial for 219 IMRT patients. The library was created from plans which were defined as the best from the protocol based on a risk evaluation. This work highlights the challenge of using a series of DVH points alone as the primary guidance to create a treatment plan. There is a richness of information available when comparing a new plan against a library of plans that have been previously determined to be optimal and protocol compliant. Improved planning tools such as those with knowledge‐based planning have been needed for some time to provide detailed feedback to institutions on whether or not their treatment plans not only meet the dose‐volume histogram requirements but are also optimal for use in clinical trials. With respect to quality assurance requirements, there are important ongoing efforts toward global harmonization of quality assurance6 (such as structure nomenclature addressed by AAPM Task Group (TG) 2638) for radiation therapy clinical trials.

Figure 1

Peters et al. assessed the impact of protocol compliance for TROG 02.02 on advanced head and neck cancer and demonstrated an impact on a) overall survival and b) time to locoregional failure as a function of the deviation status. (Figures 2 and 3 reprinted with permission from Peters et al., JCO, 28: p. 2999.)

Figure 2

This diagram shows the flow of the AAPM TG 113 report with respect to the patient treatment process (top row). Example credentialing activities related to quality in clinical trials are shown in the middle row. The bottom row shows AAPM and other reports that are related to the areas in the top row.

Figure 3

Moore et al. retrospectively evaluated the impact of knowledge‐based methods, such as calculation of a predicted DVH (pDVH), on the overall IMRT plan quality for RTOG 0126 for prostate cancer and the resulting predicted grade 2 rectal normal tissue complication probability (NTCP). (Figure courtesy of Moore et al.7).

The charge of AAPM TG 113 is to:

recommend physics practices for clinical trials involving external photon and electron beam radiation therapy that ensure minimum standards for data quality in clinical trials.

identify opportunities to improve consistency in each part of the planning and delivery process.

provide guidance to QA organizations on how best to support the spectrum of radiotherapy clinical trials, from those with basic to advanced technology.

provide suggestions regarding the credentialing requirements to reduce potential inconsistencies in the radiotherapy process.

The use of protons or brachytherapy in clinical trials is outside of the scope of this document. Throughout the report, recommendations are presented in each section for major areas of the process from simulation through treatment delivery in the context of clinical trials. The recommendations are organized by the categories of clinical trial designers, physicists (at the local institution), quality assurance (QA) centers, manufacturers, and advanced technology trials and are also presented by category in Appendices A–D. The full report includes information on restructuring of the clinical trials network and associated QA centers funded by the NCI.

THE ROLE OF THE PHYSICIST IN CLINICAL TRIALS

Physicists play different roles with respect to clinical trials. At institutional, national, and international levels, physicists may be lead or co‐investigators representing clinical and technical components. In the context of clinical trial groups, physicists may lead or co‐design a clinical trial. For national trials supported at individual institutions, physicists play a key role with physicians in ensuring protocol compliance. Other perspectives include physicist roles in QA centers and as employees of a manufacturer whose products are being used to support clinical trials.

TG 113 considers the entire process designing a trial and its QA through the activities of the local team from simulation to planning and treatment delivery to improve the consistency for clinical trials, whether trials are funded by NCI, industry, or other entities. Many AAPM task group reports are relevant to the work of TG 113. Figure 2 shows an overview of the major areas involved once a patient is enrolled in a clinical trial. For each area, both sample relevant task group reports and credentialing types are noted. Many of the referenced task group reports are ones that are already relevant to the practice of clinical medical physics in radiation therapy which then have an impact on the treatment of patients enrolled in clinical trials. Therefore, minimal additional references are made to task group reports throughout this report.

IMAGING

Image quality is paramount to many clinical trials for both target definition and treatment assessment. This section makes recommendations to facilitate consistent and accurate volume definition for clinical trials. Numerous collaborative efforts are focused on standardization of imaging, including quantitative applications. Formed in 2008, the Quantitative Imaging Biomarkers Alliance (QIBA) involves drug and equipment companies and imaging societies and has a charge to develop and advance standards for the use of volumetric computerized tomography (CT), positron emission tomography (PET), and magnetic resonance imaging (MRI) in clinical trials. QIBA has created validated datasets including ones that can be used for evaluating lung nodules9 and phantom datasets that are used to validate analytical tools such as dynamic contrast enhanced MRI.10 The Uniform Protocols for Imaging in Clinical Trials (UPICT) initiative has created a protocol for trials involving imaging with FDG‐PET/CT.11 Several groups within the AAPM are actively advancing the use of quantitative imaging information, and guidance will continue to evolve in this area.

Some clinical trials require credentialing or a central imaging review by QA centers that have expertise in quantitative imaging, such as IROC Ohio, IROC Philadelphia (DI), and IROC Rhode Island. Credentialing may evaluate characteristics, such as image quality, spatial integrity, and contrast; the requested characteristics depending on the role of imaging within a given trial. For example, considerations with respect to understanding uncertainties in molecular imaging have been described.12 More details regarding quantitative imaging in clinical trials are presented in the full report.

SEGMENTATION

Accurate segmentation is a critical task in clinical trials. Important technical sources of variation in segmentation include variable window and level settings, the use and sensitivity of auto‐segmentation algorithms to input parameters, and inappropriate margin expansion algorithms. For example, inappropriate window and level parameters can lead to significant bias and errors in volume definition with one study identifying factors leading to variations up to 42% by clinician which were reduced by using a standard protocol.13 Improvements in the consistency of contours are seen when pretreatment reviews of contoured structures are performed by protocol principal investigators. Training, such as via workshops or webinars, should be provided to physicians and other personnel for a given trial if there could be significant variability in the delineation of structures.

For organs which will be evaluated with dose‐volume histograms (DVHs), the protocol should specify how much of the organ must be contoured. For example, it may be appropriate to specify a region of spinal cord to be contoured with respect to the superior and inferior borders of the PTV. Structures with mean dose objectives should be contoured in their entirety. For structures where the entirety may not be included within the planning scan, the protocol should specify dose limits in absolute (cc) instead of relative (%)volume.

It is crucial that protocol designers provide explicit guidance in how structures are defined, especially when multiple structures are involved. Significant differences have been shown in dosimetric parameters for lung cancer for different definitions of normal lung, the gross tumor volume, clinical target volume, internal target, or planning target volume.14 Variability of such definitions in a clinical trial would have a significantly detrimental impact on the ability of the trial to resolve the study question. It may also lead to inconsistency in the application of dose goals if the same dose goals are used but with different definitions from one trial to another. Therefore, definitions and dose goals across trials to the same body site should be standardized as much as possible with the expectation of evolution of care over time. In addition, the protocol should specify any additional limits to doses to organs outside the treatment field.15 A final critical concern is that some systems ignore the volume of an organ outside the dose calculation grid when reporting dose‐volume parameters. For such systems, the dose‐grid should cover the entire organ of interest so that derived dose‐volume parameters used for treatment planning represent the entire organ. Additional details and recommendations regarding segmentation are found in the full report.

IMAGE REGISTRATION

Clinical studies that require multiple image datasets need to use image registration software. When multiple image modalities are used for treatment planning, the protocol designers should consider providing specific recommendations for internal or external landmarks that can validate the adequacy of the registration for treatment planning.

If the accuracy of the image registration for each patient affects the quality of the trial (such as in defining the target volume), the protocol designers and QA centers should require credentialing of the image registration software by using phantoms of known geometry and should follow the guidance of AAPM TG 132.16 The physician directive should specify the goals of the image registration, the method and what anatomical region should be emphasized in the registration.16

With respect to how image registration is used at the treatment unit, the trials designers should determine if it is necessary to distinguish between applications for target and normal tissue definition compared with daily online treatment guidance. Image registration considerations, which are described in the full report, may also differ if there is a midcourse plan adaptation and dose accumulation methods are utilized.17

PATIENT AND TARGET POSITIONING

Patient and target positioning is affected by immobilization and the frequency and type of image guidance used at the treatment unit. The margins for treatment planning are affected, as well as the achievable accuracy of image registration using multimodality imaging scans which are used to design and assess patient treatments, especially dose–response studies for clinical trials.

In the context of clinical trials, the type of recommended immobilization described and/or required in a particular trial depends on (a) the available and acceptable equipment in potentially accruing clinics, (b) the accuracy required by the protocol; and (c) the frequency and accuracy of the treatment guidance methods that may be recommended during patient treatment. Trial designers should determine if a given trial requires specific immobilization, such as for stereotactic radiosurgery or stereotactic body radiation therapy. More details regarding immobilization considerations are available in the full report.

Protocols should be specific with respect to the type and frequency of image guidance. The relationship between localization methods and the appropriate PTV margin18 should be considered in the design of all clinical trials. For example, a trial involving treatment of breast cancer may involve weekly portal imaging, whereas a trial involving SBRT may require daily volumetric imaging. As described in the full report, the designers of clinical trials should be specific with respect to the recommendations for intra‐ and inter‐treatment margins in a given trial for consistency and reproducibility.

MOTION ASSESSMENT AND MANAGEMENT

For many treatment sites, physiological motion must be assessed to determine if management of that motion is necessary for segmentation and treatment delivery. The AAPM Task Group 76 report, published in 2006, provides guidance for considerations at simulation and for treatment planning.19 Efforts are under way to update that report with guidance needed today for clinic care and clinical trials. In 2017, several members of the Medical Physics Committee of NRG Oncology reviewed guidance in the context of stereotactic body radiation therapy for thoracic and upper abdominal tumors and made recommendations in the context of clinical trials.20 They describe considerations regarding both motion assessment and motion management.20 The full report of TG113 contains further discussion of these considerations.

TREATMENT PLANNING CONSIDERATIONS

With respect to treatment planning, there are considerations related to the treatment planning system itself as well as the creation of treatment plans for a given clinical trial. For example, more accurate model‐based algorithms rather than pencil beam algorithms should be used for planning for patients in clinical trials. Recommendations are also provided in the full report for clinical trial designers and physicists at local institutions emphasizing tools that support improved quality for clinical trials and that may improve efficiency as well.

Protocol designers and manufacturers may be able to provide templates and tools that can be used to support the uniform implementation of clinical trial guidelines. These tools may include structure templates that work on multiple vendor platforms such as following the nomenclature recommendations of AAPM TG 263 and advanced planning tools that aid in meeting the dosimetric requirements of a protocol. For example, a dosimetric model could be developed for knowledge‐based planning or a script could be created with standard input such as the beam energy, beam arrangement, and modality to best meet a given protocol.

Advances are being made in the use of automated tools for planning and for assessing the consistency of a treatment plan with respect to previous clinical trials. This development has important implications for clinical trials both for secondary analyses and for more robustly assessing plan quality during the accrual phase of a trial. The ability to improve plan quality using knowledge‐based methods was evaluated for RTOG 0126 where predictive DVHs showed that further sparing of normal tissues was achievable with a group of plans (Figure 3).7 Figure 3(e) demonstrates that plans which were defined as “low quality” had significant improvements with respect to the predicted rectal toxicity based on the calculated normal tissue complicated probability values for each plan. Such tools will be valuable both for the teams at the institution performing treatment planning for protocol patients as well as for the analysis of plan quality at the QA centers (https://www.nrgoncology.org/Scientific-Program/Center-for-Innovation-in-Radiation-Oncology).

Additional considerations include considerations specific to adaptive therapy and re‐irradiation. Emerging new technologies in radiation treatment planning and image guidance will place additional requirements on the capabilities of the TPS. Investigators and manufacturers are developing tools to better support adaptive therapy such as deformable image registration and the creation of a model based on the accumulated dose to a patient.21 Many of these considerations are beneficial for patients who are retreated which may also be a component of a clinical trial. Deformable registration and fusion algorithms are currently being investigated and should ultimately be included in the software tool set available at individual institutions and at QA centers. These algorithms are an integral part of accurately assessing and reporting the dose given to the patient throughout the course of therapy. To fully appreciate the impact of anatomical changes for case review in a clinical trial, the composite delivered dose would be best, but if not available, multiple imaging studies, their time sequences, and all treatment plans should be submitted to the QA center.

TREATMENT DELIVERY DOCUMENTATION

Treatment management systems permit verification that the correct energy, beam modifiers, monitor units, treatment dates, and number of fractions were used for individual patient treatments. A summary of this information should be exportable in a standard format for a clinical trial. This information is crucial because it has been shown that some patients may have poorer outcomes as a result of missed radiation therapy treatments.22 Missed treatments may also impact the interpretation of the effectiveness of a clinical trial if not documented and considered. Clinical trial groups should consider the implications of missed treatments and how best to collect the information.

QA CORE FUNCTIONS AND INSTITUTIONAL PREPARATION

Credentialing for clinical trials is the performance and documentation of specific processes by an institution and its team to demonstrate their ability to accurately plan and treat patients for a particular protocol or treatment modality. In addition, a part of credentialing verifies that the institution is capable of submitting the required datasets to the QA center. The credentialing process is designed to ensure that all participating institutions can faithfully apply the protocol guidelines and deliver comparable doses in a clinical trial. This improves the ability to detect outcome differences within a given trial.

Clinical trial groups face a challenge in determining the safest way to adopt and incorporate new technologies in both existing and newly developed clinical trials. When incorporating new or less uniformly applied technologies in clinical trials, the results of credentialing tests aid in discovering and correcting variable, outlier, or noncompliant performance by participating institutions, and this helps to lessen the variability in protocol performance across all institutions. The test consist of a combination of questionnaires, benchmark plans, dry‐run digital data submissions, and phantom irradiations. If the institution passes the test, then it is approved for enrollment of patients for the pertinent protocol and the specified treatment modality. The full report has details regarding the purpose and types of benchmarks, credentialing techniques, phantom considerations, pretreatment, and on‐treatment review.

A kick‐off meeting is recommended with the appropriate research staff, clinical trials coordinator, principal investigator, physicist, dosimetrist, and a therapist before patients are enrolled on the protocol. Examples of the types of things to discuss at a kick‐off meeting are included in the full report.

SUMMARY

It has been shown that the quality and consistency of the trial impacts patient outcomes.1, 2, 3, 4, 5 This report identifies physics and other team member practices that specifically improve the treatment planning and delivery data for clinical trials. It provides benchmark and other quality assurance recommendations for groups which design and conduct clinical trials to minimize inconsistencies in the radiotherapy processes and treatment. The details for each major section along with recommendations are provided in the full report. The recommendations for the full report are presented in the appendices for the clinical trial designers (Appendix A), physicists at individual institutions (Appendix B), QA centers (Appendix C), and manufacturers (Appendix D).

There are unique challenges posed by advanced technology trials in a multi‐institutional setting. To achieve the desired level of statistical power in a clinical trial, the QA center must verify that the technology is implemented uniformly in multiple settings. The QA centers have had to adapt quickly as new technology becomes available and is implemented into clinical practice. Other guidance will need to be developed as current advanced technologies mature and other technologies develop.

With technological advancements, manufacturers play a role in the development of improved technology and in providing updates to software tools to enhance the conduct of clinical trials. Important work has been ongoing in harmonization of credentialing for clinical trials which the NCI has advocated along with other changes.23 Quality for NCI‐funded clinical trials continues to be supported by the IROC infrastructure. Finally, successful clinical trials involve a partnership relationship among all of those involved.24 Improved consistency in the design and performance of the physics aspects of clinical trials will help ensure that the data are of high integrity and can be used to answer the clinical trial questions and ultimately affect clinical practice.