In the field of medical research, BYOD, in its broadest definition, means allowing participants in a clinical trial to use their own computer devices (e.g., smartphone, tablet, laptop, desktop PC) to access and respond to study-related PRO questionnaires. As discussed, this is a departure from traditional studies utilizing ePRO where patients are provided with the hardware they need to enter study data, typically a tablet or desktop computer for site-based questionnaires and a PDA or smartphone for remote questionnaire completion. There are a number of factors that make BYOD a compelling model for clinical trials (Box 1); however, there are key outstanding issues that need to be adequately addressed before BYOD becomes a viable model for clinical trials. Although the technology is available to do it today, enthusiasm for doing it should not trump the need to approach it in a measured and thoughtful manner.
Downloading Software to Devices
One of the key issues standing in the way of widespread adoption of BYOD in clinical trials is how one gives the study subject access to the PRO instruments they are required to complete during the study. With Web-enabled devices such as smartphones, subjects could access the instruments on their Web browser. However, this approach requires active Internet access in order to view the PRO instruments and enter data. The process of entering data on a mobile device via the Internet can vary based on the subject’s network connection; subjects using their mobile device to enter data on a website may have to wait for the browser screens to load and respond. In addition, not all browsers are created equally and functionality can vary between different browsers, and between different versions of the same browser. Security is a heightened concern when transmitting data over an open Internet connection.
An alternative to entering data on a website is the use of apps, where the subject downloads a small piece of software onto their smartphone that would display the PRO instruments(s) in a relatively consistent manner across handsets, barring differences in screen size. This has the important advantage of allowing subjects to respond to questionnaires and provide PRO data when they do not have Internet access. While native apps can address some of the issues above, they present unique considerations. New functionality for a Web browser can be built and deployed quickly; a native app will take longer, since it must go through an entire software development lifecycle.
However with both of these approaches, Web and app, there can be an issue with ensuring the wide range of devices, operating systems, and Web browsers available are compatible with any app and Web system that is developed. This can be overcome by creating the app and Web system on the widely used operating systems such as Android and iOS, and the widely used Web browsers such as Internet Explorer, Firefox, Safari, and Chrome, to ensure greater access and compatibility. However, this does risk creating issues in countries and populations where other operating systems and browsers are more predominant.
Equivalence Across Data Collection Modes
Another key issue is that of equivalence of PRO instruments across different data collection platforms or modes, particularly when they are being used to support primary or key secondary endpoints. The FDA expects the sponsor to demonstrate that a PRO instrument that has been ‘modified’ is capturing equivalent data regardless of data collection mode [5]. Based on current recommendations [20, 21], using mixed modes in a clinical trial would require a quantitative measurement equivalence study if there are more than minimal differences in the presentation of the PRO instrument on the different modes (e.g., smartphone screen vs the large monitor of a desktop PC). This approach becomes impractical in the BYOD model where the number of potential devices, operating systems, and browsers available could be extensive. However, with appropriate programming across the various operating systems and browsers, the only material difference becomes the size of the screen (and font) on the device the subject is using to enter data. In this case, an assessment of the measurement equivalence of the smallest screen size to the largest screen that would realistically be available for each operating system should be sufficient. As mentioned previously in regard to ePRO more generally, an accumulation of evidence from BYOD-based studies may, at some point in the future, support the assumption of measurement equivalence across screen-based devices in most circumstances where a faithful migration has been rendered.
Paying for Data
During traditional ePRO studies where clinical trial subjects are provided hardware by the trial sponsor, data transmission costs (e.g., submitting data from a daily diary completed on a smartphone) are automatically covered by the sponsor with the included SIM card on the device. However, in the BYOD model, the expectation is that the subject is using his or her own device and, as a result, will incur data transmission costs as part of, or in addition to, the subjects’ cellular plan. The subject should have a reasonable expectation of being reimbursed for these costs in addition to any other compensation received for participating in the study.
People Without Access to a Suitable Device
The underlying assumption of the BYOD model is that study subjects have their own device, which they can, in fact, bring. In FDA parlance, the subjects are required to provide a device that is ‘fit for purpose.’ However, even high global Internet and mobile phone penetration does not guarantee that the patient population for a particular study will all have a suitable device that can access the study measures, and there is little reason to believe that smartphone penetration will ever reach 100 %. For example, in the US ~45 % of all adults own smartphones, but ownership varies widely based on age, income, education, and geographic location. However, Blacks (47 %) and Hispanics (49 %), who are traditionally underrepresented in clinical trials in the US, had higher rates of smartphone ownership than non-Hispanic whites (42 %) [26].
There are two potential approaches to addressing the device access issue:
-
Require ownership of an appropriate device in the inclusion criteria. This would help to ensure that participants in the study have an appropriate device on which to enter data. However, this approach raises serious concerns about selection bias within the sample. Potentially, this could lead to exclusion of important segments of the target population from the clinical trial.
-
Provide participants who do not have appropriate devices with stand-alone hardware, while allowing those with an appropriate device to follow the BYOD model. This approach would avoid issues around excluding certain segments of the target population, but attenuates one of the key perceived strengths of the BYOD model (i.e., avoiding costs associated with providing devices).
Whether selection bias is a problem will depend on the size of the target patient population. If the target population is hard to find, then the idea of providing devices would be attractive as it will allow those that have the right device to use their own, but it would not exclude those that do not. If the target population is easy to attain, then the ownership of the correct device can be used as an inclusion criterion.
Security, Both Physical and Electronic
When evaluating BYOD solutions, ensuring protection of the subject’s private information and the data being collected should be a paramount concern.
Box 2 provides examples of questions that should be considered when evaluating any BYOD system. The answers to these questions and others like them will impact the level of confidence with BYOD in general and specific solutions in particular.
Smartphones and tablets present problems when lost or stolen. The physical security can be breached with a brute force attack, so extra design precautions need to be taken to ensure that the patient’s privacy and anonymity are protected. There are also additional security considerations that need to be addressed to make sure that authentication controls are not bypassed. Such privacy and authentication controls are part of a ‘normal’ ePRO build, no matter what modality is being used, and apply to all aspects of a person’s digital life, not just his or her health-related data.
Ownership
In BYOD studies, the sponsor no longer owns and controls the device in the same way as if the device were deployed to the subjects. It should be recognized that these are multi-purpose devices that cannot be locked down in a way that disables any of the other device’s functions during data collection. In addition to completing the study protocol activities (i.e., responding to the PRO instrument at the appropriate time) on their devices, study subjects will be making calls, sending text messages, playing games, surfing the Internet, and interacting with friends on social media sites on this same device. These are ‘environmental risks’ of collecting data in an unsupervised setting and can cause data collection to be interrupted or delayed. However, this issue is not unique to BYOD studies; clinical trial sponsors have never been able to control a subject’s behavior when they are completing any PRO assessment remotely, be it on paper or any other data collection mode.
Storage, or the amount of space available on the device, also becomes an issue with devices that are not under the sponsor’s control. This is especially true when considering all the tasks for which the subject uses his or her mobile device (e.g., watching videos, taking pictures, playing music, and downloading other apps). In addition, the subject can delete the app at any time, even with captured and un-submitted data. However, as part of the inclusion criteria, potential subjects could be required to agree to not delete the study app. Some checks can be conducted by site staff and via online monitoring; however, these options are limited.
Compliance questions are also raised since the sponsor does not own the device and can’t force the subject to have notifications turned on. If the subject chooses to turn off notifications, these automated alerts to complete required data entry task do not work. Like the agreement to not delete the app as part of the inclusion criteria, subjects could be required not to turn off notifications, which are associated with high compliance [27, 28]. If the subject does turn off notifications during the trial and compliance with data entry is negatively impacted, the study staff’s review of online reports should alert them to a compliance problem.
Technology Changes
The set of mobile operating systems that support the mobile market are changing constantly and quickly. For sponsors of clinical trials, the pace with which change happens will become a major challenge. Agile software developers iterate through releases quickly, with some deploying new features on a monthly basis. Support for the previous version of the software is dropped as soon as the new app is made available because app stores only allow one version of the software to be sold at any time. Increases in the amount of infrastructure necessary to support the complex functionality of existing solutions and increased maintenance cycle costs for vendors may drive prices upward. The result may be that the cost-benefit ratio of switching to a BYOD solution is not as large as anticipated.