Data Integrity in Decentralized Clinical Trials (DCTs)
A major challenge around recruitment and retention of subjects in clinical trials has been the requirement for them to have easy access to medical centers so they can travel to the trial sites for screening and enrolment, as well as for follow-up visits for tests and procedures required as per the protocol, and medication refill. Technological advances have helped Decentralized Clinical Trials (DCTs) emerge as a viable solution to this problem, and the pandemic has fast-tracked its adoption.
Benefits of DCTs: Enhancing Patient Experience and Efficiencies for the Sponsor
DCTs benefit both sponsors and patients. The sponsor benefits because DCTs improve speed of recruitment, increase patient retention & diversity & result in overall reduction of time and cost. At the same time, the enrollment and participation are patient friendly, thereby resulting in better patient experience. Improved compliance and retention, and increase in safety lead to better patient outcomes. Trial outcomes tend to reflect the real-world environment closely since the trial conduct resemble home-based routine care.
Digital Enablers of DCTs: Revolutionizing Data Capture and Analysis
Digital enablers of DCTs include eConsent tools, telehealth visits, tools for electronic patient reported outcomes (ePROs) and electronic clinical outcome assessment (eCOA), wearables, sensors, remote monitoring tools and patient engagement platforms. With a large majority of trials projected to use digital health technologies in the next few years to capture both objective (e.g. wearables/sensors) and subjective (e.g. ePRO) data, monitoring and data management methods outside the traditional data cleaning and reconciliation activities would be required for the breadth and depth of data that will come in from all these sources. Of course, there’s also tremendous opportunity along with the inherent challenges, for e.g., applying advanced analytics to the digital data, detecting patterns, and exploring novel endpoints.
Data Management Challenges in DCTs
Every disruptive innovation must be complemented by adapted procedures to address new challenges it brings up, and this applies to DCTs as well. Traditionally, sites entered clinical trial data in an EDC system and these source data were verified at the site to confirm accuracy. Risk based monitoring focused on site level metrics such as screen failure rates, query rates, reported SAEs, missed/late visits, and others. With DCTs, the data volume is high, and the data are heterogenous, from widely disparate sources which are dynamic and may not follow data standards. Other means of ensuring quality and integrity of data are required to address the security risks and data monitoring challenges.
As a rule, a comprehensive understanding of all sources for data capture in a clinical trial and the process for centralization is essential. Also, it is pertinent to evaluate data collected in real time to allow early interventions that will ensure data integrity for regulatory submission. It is important to outline data flow to a centralized location during the clinical trial, including data from mobile technologies. This understanding is essential to ensure all data are accounted for, in a manner that is transparent to all members of the study team. In addition to a high-level mapping, the details of transfers must also be documented. This provides clarity regarding which individuals are responsible for sending and receiving the data and how data will be securely transferred and stored at each stage.
Let’s examine a few facets of ensuring data integrity in DCTs. We focus on:
- Application of Centralized Statistical Monitoring (CSM) and role of technology platforms
- Role of mobile Health Care Providers (HCPs), and
- Role of Artificial Intelligence (AI) and Machine Learning (ML) in identification of corrupted data from wearables.
Application of CSM and role of technology platforms
Centralized Statistical Monitoring takes a holistic data surveillance approach, using advanced statistical and analytical tools to identify data gaps & anomalies to identify study issues more serious than those identified by transactional data reviews. Centralized monitoring strategies built on unified data platforms help in real-time aggregation and analysis of data, systemic risk monitoring and early issue detection, as well as provide insights into a variety of functional teams across the trial continuum. Optimized approach to centralized monitoring needs aggregation of data from siloed and disparate systems to build a unified picture of the patient journey, site performance and overall trial health.
Ensuring Data Integrity with Centralized Statistical Monitoring (CSM)
Monitoring and data management are integrated in the CSM approach. Risk-based data management strategies focus on core processes and critical data points most likely to impact data integrity and interpretability. There may be a higher risk of fraudulent data or misconduct in DCTs with data from ePROs and wearables. System audits and geo-tagging can be used to detect misconduct. Non-compliance is not easy to define when data is being streamed. Is lag or variability in the data indicative of a signal? Missing data may arise because patients may not comply with completion of questionnaires/diaries. It’s important to ensure consistency between patient-reported data and data collected during procedures or physician’s assessment, and data during on-site visits has to be compared with data collected remotely during home-health visits. Statistical monitoring methods like the following are well-suited for this:
- Descriptive statistics to identify outliers or influential observations
- Comparison of study-wise and subject-specific confidence intervals with multiplicity-adjusted alpha
- SD analysis – repeated measures model for SD estimates
- Correlations – subject, site & study level correlations between variables; compared using CIs
- Mahalanobis distance (MD): multidimensional risk assessment method based on multidimensional risk score; flexible method to combine dimensions and identify risk factors; inliers, outliers
There are platforms that transform data from various sources to a common data model so that medical data can be queried. The ontology system enables standardization of medication and diagnosis codes and interoperability of all these medical data as well as data from procedures and lab tests. The architecture of the platform is such that while querying the data from the hospital information system and running aggregated queries, the data keeps its full resolution and integrity.
Such platforms increasingly incorporate application of AI and ML to identify patterns and anomalies, which is the future trend for ensuring data integrity.
Evolving Role of Mobile Healthcare Providers(HCPs) in Improving Trial Access and Ensuring Data Integrity
Improvement of trial access, one of the major advantages of DCTs, is often the result of the contribution of mobile health care providers who transform clinic visits into home visits, real or virtual. By reaching out to the participant at home they decrease the burden and remove obstacles to participation. As a result, they make it possible to have longer trials, with more frequent visits, with participants who are more compromised and have difficulties with travelling to their clinic.
Also, as a network of health care providers, they can even follow participants who travel to a milder climate on a seasonal basis.
Completeness, consistency, accuracy, attributability and authenticity are among the key elements of data integrity from a regulatory perspective. We can appreciate how the mobile HCPs can contribute towards data integrity when we look at how their role is rapidly evolving, especially in the context of DCTs.
Traditionally, the role of mobile healthcare providers (in the context of a clinical trial or routine care) included blood draws, clinical assessments, treatment administration which included administration of the investigative product in a clinical trial, participant education, or re-enforcement of education received in the clinic and assessment of adherence to treatment such as pill counts.
But the role of mobile HCPs is changing. Some of the traditional tasks such as blood draws and clinical assessments have remained unchanged but new tasks have been added.
It is not rare for a mobile healthcare provider to have to re-assess the participant’s understanding of an ePRO instrument. This needs to be done whenever the data flow is found by the central monitoring staff to be interrupted or corrupted. There may often be a need to provide basic IT support and, when required, coordinate IT support with specialists at the site or the CRO.
In the absence of clinic visits, the mobile HCP inherits some components of the role of research coordinator. Mobile HCPs are in a good position to evaluate the social environment of the participant, a key variable that influences treatment outcomes.
Recent advances in the field of portable imaging could further add to the role of mobile HCPs. With the rapid progress in imaging technology, portable, handheld devices are now available, especially for ultrasound imaging. It’s possible that mobile HCPs may soon use such devices for imaging of organs, assuming that secure transmission can be established, and interpretation can be done by a licensed radiologist. There could be applications in many therapeutic areas, from cancer (tumor response) to chronic inflammatory disease.
Expansion of the role of mobile healthcare providers will have to be accompanied by proper training and certification. This requires considerable effort and expense on the part of the sponsor and the CRO. A visiting nurse or nursing assistant will need quite a bit of training to become a competent mobile healthcare provider.
Data Integrity through AI-based Detection Methods for Wearable Technologies in DCTs
Remote health monitoring based on non-invasive and wearable technologies is an integral part of DCTs. These systems allow healthcare personnel to monitor important psychological signs in real time, assess health conditions and provide feedback from distant facilities. Detecting the integrity of data obtained from wearable technologies at the network level enables us to eliminate corrupted or unusable data. This reduces the storage and power requirements of wearable devices without compromising on functionality.
Several AI-based binary classifiers are available for detecting integrity of data obtained from Internet of Things (IoT) enabled wearable sensors. Let’s consider the example of ECG signals from wearable devices.
In a recent study1, the authors tested 13 ML algorithms for predicting ECG signal integrity after training and testing them on the Physionet/Cinc 2017 challenge training dataset. The dataset consists of single-lead ECG records collected using the AliveCor mobile measurement device, which were annotated by clinical experts.
The neural network algorithms that were tested were based on decision trees (learning methods in which the important features that can be used to represent data are selected from regression models), boosting (an ensemble technique in which a sequence of variables is constructed additively to find the best prediction and the iteration that minimizes false predictions, which is then adopted for decision trees) and bootstrap aggregation/bagging (an ensemble method whereby a group of neural network models is generated by sampling with replacement at random of a training dataset to find the best fitting model).
The neural network ensemble using bagging and gradient boosting exhibited the best integrity classification across all parameters in this example.
Computer chips with energy-efficient deep neural networks have the potential to be deployed on IoT wearable devices for reliable ECG signal quality
Centralized Statistical Monitoring and automation of the methods it entails are crucial to protect integrity of data captured in DCTs. Technology platforms used for data standardization, aggregation and analysis need to have in-built mechanisms for data security, data privacy and data integrity. AI and ML methods are being increasingly used to detect and ensure integrity of data from wearables. In the conduct of DCTs, mobile HCPs are increasingly playing bigger roles and the right training will place them in a good position to ensure integrity of data.
1 – [Reference: John, Arlene & Panicker, Rajesh & Cardiff, Barry & Lian, Yong & John, Deepu. (2020). Binary Classifiers for Data Integrity Detection in Wearable IoT Edge Devices. IEEE Open Journal of Circuits and Systems. 1. 88-99. 10.1109/OJCAS.2020.3009520]
President and Founder
30+ years in industry and academia; 24 years in Pharma & CRO in Clinical research, post-marketing and safety; cancer epidemiology; Entrepreneurial experience; Statistics and Actuarial qualification.