11 The objective of this chapter is to introduce interested but inexperienced researchers to a number of practical concepts that are important to the planning, design, and conduct of a research study. On the surface, institutional approval for conducting a study, data management, and budget formulation may not appear to be directly linked. Closer examination, however, reveals that they are strong determinants of the day-to-day operations of both large and small-scale research studies. These important pieces of a study are intended to ensure that it adheres to institute guidelines (application for institutional research approval), the data is accurate and credible (data management), and the appropriate resources are available to conduct the study as planned (budget formulation). Each of these topics is covered in this chapter, largely presented conceptually with practical advice for what to do or avoid. Where possible, issues of particular relevance to radiology research are highlighted. As is mentioned throughout the chapter, readers are encouraged to investigate the regulations and restrictions that affect the research activities at their local institutions and from relevant sponsors and granting agencies. A summary of key points to remember for each of the three sections is included at the end of this chapter. The objective of clinical research is to advance the knowledge and understanding of human health and biology.1 Individuals who consent to be research participants usually do not benefit directly from the research despite the potential for risk.2 Guidelines for the ethical conduct of clinical research exist to minimize the chance of exploitation of research participants by assuring that their rights are respected. This section provides an introduction to the role these guidelines play in clinical research and highlights important aspects of the independent ethics review and informed consent. Since the specific details of these processes vary by institution, it is recommended that readers familiarize themselves with the rules that govern the approval and conduct of research at their local institutions. Since the Second World War, numerous guidelines have been produced to address ethical issues in human research participants. The purpose of the majority of these guidelines is to ensure the protection of the rights of research participants and to guarantee data integrity. Detailed review of these important guidelines is beyond the scope of this chapter, however, and interested readers are encouraged to look up details related to the following: the Nuremberg Code,3 Declaration of Helsinki,4 the Belmont Report,5 United States’ Code of Federal Regulation (CFR), regulated by the U.S. Food and Drug Administration,6,7,8,9 the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH), and Guideline for Good Clinical Practice (GCP).10 These guidelines have played an integral role in the creation of rules that govern research conduct at institutions throughout the world, and should be followed when generating clinical trial data that are intended to be submitted to regulatory authorities. The principles established in these guidelines may also be applied to other clinical investigations that have an impact on the safety and well-being of human participants. It is strongly recommended that researchers become familiar with the rules and regulations that govern research at their local institutions. GCP guidelines require that all research involving human participants be reviewed and approved by an Institutional Review Board (IRB), also known as the Research Ethics Board (REB) or Independent Ethics Committee (IEC).10 The IRB consists of scientific and nonscientific members responsible for ensuring the protection of the rights and safety of human research participants.8 Regulatory authorities have given IRBs the power to approve studies, request for changes in planned research before approval, or reject research proposals.8 The IRB is also responsible for paying extra attention to vulnerable research participants, such as pregnant women, children, the elderly, mentally disabled persons, and members of groups with a hierarchical structure, such as medical and nursing trainees.10 It is imperative that the appropriate IRB has reviewed and approved a research study before any intervention or interaction with human research participants.11 The IRB will require specific documents to be submitted for review. These documents will vary by study design and the specific IRB from which approval is being requested. This is an important reason why investigators must familiarize themselves with their IRB guidelines prior to the submission of documents for approval. A majority of IRBs will request a specific application form to be completed and a research protocol to serve as the backbone of the application. The protocol should address the project summary, study rational, background information, study design, methodology, hypothesis, statistical details, and participant safety considerations.10 The ICH guidelines provide excellent recommendations regarding research protocol contents.10 For prospective studies that require the recruitment of patients, an Informed Consent Form (ICF) is essential. The basic elements of an ICF are generally standard and similar in all the ICH regions around the world and are described in this chapter. Many IRBs will have a suggested template available for modification by researchers. Most clinical trials also require the submission of other documents such as an Investigator’s Brochure (IB), study budget, data collection forms, and participant recruitment procedures and tools, such as advertisements and telephone follow-up scripts. Depending on the complexity of the study and where it is to be conducted, additional approvals, certifications, and contracts may have to be obtained from different hospital departments, governmental authorities, and standardization associations. Industry-sponsored studies usually require full contracts executed by the institution’s contracts and legal department before approval is granted. For multicenter studies initiated by a principal investigator (PI), the IRB may request a data sharing agreement to be in place between the lead site and other participating centers.12 The main purpose of this agreement is to assure that patient confidentiality rules and regulations are followed and that all data and documents shared between sites are fully anonymized.12,13 Informed consent has been a cornerstone of ethically conducted research dating back to the Nuremberg Code, the first formal research guideline.3 Prior to any research-related procedure, consent must be obtained and an ICF must be signed by the participant and the consenter. It is vital to remember that the form is not a contract. By signing the form, the participant does not give up any of his/her rights and is able to withdraw from the study at any time.14 In certain study designs, especially those addressing inter- and intrareader agreement, it might be necessary to identify radiologists as research participants. In these situations, the radiologists themselves will need to provide consent and the form should detail assurances that study results will not affect the participant’s employment in any way. Obtaining informed consent from research participants is not just a simple act of obtaining their signature on a document. It is an opportunity for sharing information with potential participants and for addressing their questions and concerns.9,12 It is also very important to remember that informed consent is an ongoing process and participants enrolled in an ongoing study may need to be re-consented if major changes are made to the study.12 If documents are updated and a new version of the consent becomes available, the researchers must ensure the most recent version of the form is used. A majority of IRBs and regulatory bodies ask for consent forms to be written in a grade 6–8 reading level, which discourages the use of scientific terms that may be foreign to the lay audience.15 The consent forms must also contain a set of specific elements. The purpose of the study must be stated clearly and the relevant investigators identified. The drug, device, or intervention being studied must be mentioned. The ICF should also describe all study procedures to be performed, possible risks and benefits, assurance of confidentiality, and the rights of the research participant. Most importantly, it must be stated that participation is voluntary and that the participant can leave the study at any time without affecting their ongoing and future standard of care.11,12 The IRB may require the inclusion of additional details in the ICF for more complex studies. Both retrospective and prospective study designs require IRB approval prior to initiating the study. In retrospective studies, investigators attempt to answer study questions using data that have already been collected by the time the IRB application has been submitted. The data, which can come from the radiology Picture Archiving and Communication System (PACS) or patient medical records, may have been generated for reasons other than research. Depending on the design and objectives, retrospective studies may not require full board approval and their review will be expedited by delegates of the IRB chair. Prospective studies are designed before any data is collected. Specific study participant inclusion and exclusi on criteria are set and participants are followed during the course of a defined study period. A majority of prospective studies will go through a full board approval where all members of the IRB must review the study.10,12,14 The scope of the review is determined by the IRB and not the researcher. Personal health information (PHI) includes data that can be used to identify participants on their own or in combination. Examples include, but are not limited to, names, medical record numbers, and accession numbers. IRBs often place restrictions on the use of PHI, and it is the researchers’ responsibility to keep this data confidential if it is collected.10,12,14 The investigator must assure the IRB that information related to the participant will be respected and not divulged to others without permission,12 providing details that describe the responsible collection and storage of data. Such assurances should also be described in the ICF in more general terms, as this helps establish trust between researchers and participants.16 As good practice, it is often recommended by IRBs that each enrolled participant be assigned a unique study identifier (study ID). If it is necessary to reidentify participants, a file that links study ID to PHI should be maintained separately outside of the study database and described in the original IRB application. Documents that are embedded with PHI should be anonymized using appropriate methods and tools so that the de-identified information cannot be recreated.17 Digital Imaging and Communications in Medicine (DICOM) is a standard term used for handling, storing, printing, and transmitting information in medical imaging, thus allowing images to be easily readable by other imaging programs. This term includes a definition for the file format and a protocol for network communications. For radiology studies where DICOM images are to be shared between sites, it is recommended that all images be fully anonymized using pretested and approved methods. This may be mandatory at some institutions. Data management refers to study activities related to the collection and storage of data. Because of the direct link to study design, it is important that study objectives and design details be documented prior to planning data-related activities. This will increase the likelihood that consensus and clarity will drive subsequent study-related decisions. It is also important that data management activities, especially the flow of information and the personnel and equipment required to carry them out, be determined as early as possible. These will likely need to be authorized by an IRB prior to the commencement of the study. The importance of good data management practices cannot be overstated. High study integrity and data transparency are important to maintain. More practical reasons for good management practices include easy communication between study team members, less time redoing tasks or correcting errors, being able to hit submission targets for conferences or grants, and more efficient analysis. The contents of this section describe concepts and make suggestions that may help principal investigators and the study team achieve the intended study goals. On the collection end, data elements should be defined. These elements include units of measure, precision, and scales. It is recommended that commonly accepted terminology and measurements be weighted heavily as they improve generalizability of the data and the eventual conclusions that will be made. Useful sources for such terminology are relevant clinical consortiums and organizations. Since the publication of results and integration of findings into clinical practice are two frequent and important goals of research, the use of common terms and data elements will provide a better context to outside reviewers and readers as to what the study has accomplished and how it can be translated in the local or external institutions. Most study designs should involve data that are objective and do not require interpretation. This means that the variables can be clearly defined to study team members and an external audience and are easy to understand and replicate during the conduct of the study as well as in follow-up studies. A data dictionary, also known as a glossary, is a document that lists and describes all data elements. It is an important tool for a number of reasons; most obvious is keeping track of what data are being collected. It should include variable descriptors such as format (e.g., string, numeric, categorical, date, etc.), category definition, coding scheme, and the definition of the variables. The dictionary can be maintained in list or table format in a word processor or spreadsheet program, with each descriptor listed as a field or column heading. It should be stored as a stand-alone document in electronic form, as this allows for easy use and access by study team members. The document is especially useful if a statistician is responsible for the data analysis. The data collection process is of utmost importance to study integrity. While data sources will differ from primary (imaging studies, study participants) to secondary (patient records), similar data collection methods and instruments can be applied for all types of studies. The process and data should be unbiased and objective, have a low likelihood of error, and result in data that are easy to analyze. Data collection forms (DCF), also known as case report forms (CRFs), are commonly used tools in clinical research. They can be paper or electronic, and consist of a collection of clearly identified fields that guide the user to search for and record relevant data. DCFs are especially useful if multiple people are involved in the data collection or generation stage, including study coordinators, research assistants, and radiologists. The template format prompts readers to record data in a consistent fashion, and the series of completed forms provides an audit trail. DCFs should be designed so that they are self-explanatory and require minimal interpretation by users. Units of measure and levels of precision should be indicated on the forms, especially for fields in which this can be variable (e.g., dates, anthropometrics). The inclusion of features such as checkboxes and options items to circle will reduce the need for handwritten responses and likelihood of error. The sequence and relative spacing of items in the horizontal and vertical axes of the DCFs should be chosen to minimize the moving of the eyes across a form as the user searches for and fills in a field. This can be accomplished by matching the layout to the natural order with which observations are recorded when radiologists interpret images or review reports. Data collection and entry should be tested prior to full study rollout. The same individuals who will perform these activities for the actual study should be involved in the testing phase. This will help identify inefficiencies in the flow and guide changes to areas of concern. As radiology research often involves multiple readers, this is also a good opportunity to ensure readers are similarly interpreting data elements when reading images. Research readings often involve different or more intricate levels of detail than standard clinical readings and take the radiologists outside of the daily reading routines with which they are most familiar. A suggestion for performing a test run is to have study team members independently review the same small set of images (radiologists) or reports while filling out the DCFs (non-radiologists), followed by a discussion and comparison of the results. Principal investigators should be open to feedback provided during this stage and be willing to make changes to study materials or the process itself. If changes are made following IRB approval, an amendment may need to be submitted to the IRB to notify them about changes to the data collection process. If at all possible, those involved in data collection or reading images should be kept unaware of information that identifies important status or is strongly correlated with outcome. This reduces the potential for bias as the data collection or generation is unlikely to be affected by important characteristics of the study participants. Examples of information to be considered for withholding include diagnosis, prognosis, predisposing risk factors, and results of other tests or procedures. Blinding can often be accomplished by anonymizing images provided to readers so that radiologists do not have access to patient history or clinical reports, sorting images or records to be reviewed in random order for studies involving competing tests or re-reads, or using separate DCFs for different portions of the data collection process. Data will eventually be stored into a database or spreadsheet that is likely to follow a row-column format. Rows are intended to contain data for the units of measure (e.g., participants, lesions) and columns represent the variables themselves. It is advised to follow to the row-column format, especially if working with a statistician, as this is the format that is recognized by analysis software packages. Data format should be consistent within columns. Common formats include dates, text (also known as string), and numeric. Only one data element or observation should be included per column or cell. Deviations lead to confusion when reviewing the data and will likely involve laborious data cleaning to allow for the appropriate data analysis (Fig. 11.1). Practical tips to aid with this include formatting the column so that it adds the appropriate number of decimal places or configures dates into the desired format. Many database or spreadsheet programs also allow data to be entered by selecting an option from a drop-down list or checkbox, which minimizes keystrokes and reduces error; these features are useful for entry of categorical data. The preferable mode of data storage is as a numerical code. A simple example is gender, where female and male are two options. Arbitrarily, if possible, this variable could be entered as 1 for female and 2 for male. This will make it easier to perform data analysis, where numeric codes are often required for more advanced analytical techniques. Although this will require entry of numeric codes for basic spreadsheets, the use of database programs that include various data entry options might allow for built-in numeric coding. Continuing with the above example, such a program could allow for female to be selected from a dropdown list and be stored as a 1 in the actual data table that will be used for analysis or exported for use by a statistician. These and other more advanced database options are discussed below. It is to be expected that a certain number of participants will have missing data. It will be useful to distinguish between items that are evaluated but intentionally left blank because the data could not be identified versus those that were left blank because they were unintentionally passed over by study personnel. Two important considerations when determining a code for missing data are to maintain the format of the variable in question (i.e., if numeric, the code should be numeric and not string) and to use values that are impossible for a given variable to indicate it is missing (e.g., 999 for age or 99 for body mass index). Shading portions of the spreadsheet with different colors, while helpful to visually highlight important variables or subgroups, should not serve as a replacement for variables or proper coding as colors cannot be integrated into analysis. The responsible use of personal health information (PHI) was discussed in detail in the ethics section, and much of what was described there applies to this section. On a more practical level, assigning a unique study identifier (study ID) for each participant on the DCF and in the database will go a long way toward reducing the need for entering some PHI parameters into the study database. If it is necessary to re-identify participants, a file that links study ID to PHI should be maintained separately outside of the database, and described in the original IRB application. Readers are encouraged to verify the regulations that govern the collection and use of PHI at their local institute.
Obtaining Ethics Approval, Data Management, and Budget Formulation
Learning Objective
Obtaining Ethics Approval and Informed Consent
Important Ethics Guidelines
Institutional Review Board
Required Documentation
Informed Consent
Elements of an Informed Consent Form
Institutional Research Board Review of Retrospective vs. Prospective Studies
Data Confidentiality
Data Management
Defining Data Elements
Data Dictionary
Data Collection
Data Collection Forms
Test Runs
Blinding
Electronic Data Considerations
Standardizing Data Elements
Data Coding
Handling of Personal Health Information
Radiology Key
Fastest Radiology Insight Engine
