Recruitment for clinical trials is a knowledge- and data-intensive task that often presents significant challenges for clinical researchers. The efficacy and safety of new therapeutic agents depend on these trials, but failure to recruit a sufficient number of participants within defined timeframes can jeopardize their successful completion, leading to delays or even premature termination. Existing methods of identifying eligible patients, which involve examining medical records and matching characteristics to complex eligibility criteria, are time-consuming, dynamic, and prone to overlooking potentially eligible individuals. This is particularly true in large academic clinics where dozens of trials may be active simultaneously.
The article, “A Methodology for Determining Patients’ Eligibility for Clinical Trials,” by Tu et al., introduces a groundbreaking computer-assisted methodology designed to address these critical recruitment challenges. This innovative approach aims to increase the accrual rate of eligible patients into clinical trials by automating parts of the eligibility-determination process, thereby reducing the information load on clinicians and enabling a more systematic identification of suitable candidates.
Key Components of the Methodology:
- Computer-Interpretable Language for Eligibility Criteria: The methodology begins with a simple yet powerful language to express the complex eligibility criteria of clinical-trial protocols. This language allows for the definition of time-stamped parameter values, interval-based events, simple comparisons, arithmetic combinations, prior conditions, and logical conjunctions and disjunctions of criteria. These criteria are then translated into database queries to check a patient’s condition against protocol requirements.
- Classification of Eligibility Criteria: To make the program sensitive to the variability and controllability of patient conditions, eligibility criteria are organized into five distinct groups:
- Stable Requisite: Immutable preconditions, such as disease history or drug intolerance.
- Variable Routine: Criteria based on routinely collected, relatively stable data like laboratory test results (e.g., CD4 count, platelet count).
- Controllable: Patient circumstances that a physician can modify (e.g., medication restrictions).
- Subjective: Criteria relying on physician judgment (e.g., patient compliance, estimated survival, Karnofsky score).
- Special: Criteria requiring unusual, often costly or invasive, laboratory tests not typically part of routine care (e.g., lumbar-puncture studies). This classification allows the system to identify patients who may become eligible later or through specific clinician actions.
- Dual Approaches for Computing and Summarizing Eligibility Scores: The article describes two primary approaches to calculate and summarize eligibility status, particularly when data are insufficient:
- Qualitative Approach: This method defines discrete, nonnumeric eligibility scores (P: meets criterion, PP: probably meets, N: data not available, FP: probably fails, F: fails). It uses heuristic missing-value assumptions (e.g., Default, Immutable, Stable satisfied, Assume satisfied) to assign scores when current data are absent. Eligibility status is then summarized using a multivalued propositional logic with AND/OR truth tables. This approach allows for sensitivity-versus-specificity tradeoffs, meaning the program can be made more inclusive (more sensitive, but more false positives) or less inclusive (less sensitive, but more specific) in its search. The THERAPY-HELPER system implemented this qualitative approach.
- Probabilistic Approach: This approach defines eligibility as the subjective probability (between 0 and 1) that a patient satisfies a criterion. It employs a Bayesian belief network to represent criteria and the dependencies among data values. This network allows for the computation of posterior probabilities, even for nodes with no direct observation, providing a stronger, quantitative statement about a patient’s overall probability of satisfying all eligibility criteria simultaneously. However, this method requires modeling complex dependencies and assessing prior distributions, which can be computationally intensive and rely on expert elicitation.
- The Combined Approach (A Synthesis): Recognizing the complementary strengths of the two methods, the article proposes a combined approach. This synthesis leverages the Bayesian belief network for a uniform and consistent representation of individual eligibility criteria and missing-value assumptions, allowing for a principled calculation of the probability that a patient satisfies each criterion. Subsequently, these probabilities are mapped into qualitative scores, which are then combined using the simple AND and OR truth tables from the qualitative approach to derive a summary eligibility status. This combined strategy avoids the complexity of modeling all dependencies for summary scores while still providing a clear, semantically defined outcome.
Tangible Results and Broader Impact:
The application of this methodology to a database of HIV-positive patient cases demonstrated its significant potential. Retrospective analysis using the THERAPY-HELPER system revealed that clinicians inadvertently missed a substantial number of opportunities for enrolling patients. For example, 20% of sampled patients were provisionally eligible for at least one protocol but were not enrolled, and 17% satisfied all but controllable criteria. Furthermore, key data for eligibility evaluation were frequently missing, highlighting the need for automated alerts. The system’s ability to identify such patients and to indicate missing information or modifiable characteristics suggests it can significantly increase patient accrual rates by prompting clinic staff to collect necessary data or adjust treatments.
This methodology is adaptable to both mass screening and referral recruitment strategies. Its principles extend beyond clinical research, offering a valuable tool for determining the applicability of medical practice guidelines in specific clinical situations.
By providing automated assistance, this methodology promises to enhance the efficiency and effectiveness of clinical trial recruitment, ultimately accelerating the evaluation of new therapies and improving patient care.
Reference for the Article:
Tu, S. W., Kemper, C. A., Lane, N. M., Carlson, R. W., & Musen, M. A. (1993, September). A methodology for determining patients’ eligibility for clinical trials. Methods of Information in Medicine. doi:10.1055/s-0038-1634933