All clinical trials share core requirements such as the need to exclude bias and provide evidence of effectiveness. However, diagnostics trials differ from drug trials in many important ways. As a result, diagnostics development requires substantial differences in planning, trial design and trial execution.
It will come as no surprise to diagnostics professionals that diagnostics development is different from drug development. In fact, when they hear vocabulary from drug trials applied to diagnostics studies, diagnostics professionals tend to be vocal. However, a quick summary of key differences in drug development and diagnostics development may be useful for diagnostics startups, clinical research professionals new to diagnostics trials and ex-US diagnostics companies that wish to enter the US market.
Key ways in which diagnostics development differs from drug development include the following:
- For diagnostics developers, many development decisions depend on identifying the intended use and user (lab assay performed by trained technician, point-of-care diagnostic used in physician’s office, use for screening vs. treatment decisions, home diagnostic to aid in managing a chronic health condition and so on).
- Diagnostics developers must choose among a wider range of regulatory pathways for approval depending on factors such as intended use and user, test complexity and role in patient-care decisions [e.g. 510(k) clearance vs. PMA (Premarket Approval) vs. CLIA waiver].
- Drug trials usually deal with one office within one FDA center, but diagnostics trials must sometimes consider policies of multiple regulatory agencies and offices [FDA’s Center for Devices and Radiological Health (CDRH), Center for Medicaid Services (CMS) and Centers for Disease Control and Prevention (CDC)].
- EU and US regulatory and approval processes for diagnostics differ far more than for drugs, increasing complexity of global development strategy.
- Collecting complete data on primary endpoints can be more challenging in diagnostics trials. Diagnostics trials often require more layers of data, with data needed on both the gold-standard diagnostic and the test diagnostic. Samples may be of different types (e.g. blood sample for investigational molecular diagnostic vs. tissue sample for gold-standard assay). Often the gold standard diagnostic itself consists of a hierarchy of multiple tests; whether the subject has the next test performed depends on the result of the prior test. Thus, the available data set can vary for individual subjects. Since subjects with incomplete data will likely be excluded from analysis, trial efficiency depends on accurate tracking of primary endpoint data for each subject. Furthermore, unlike processes for standard-of-care diagnostics, processes for test diagnostics are unfamiliar to sites, increasing risk of mistakes in sample collection, handling and shipment.
- Statistical analysis in diagnostics trials focuses not on efficacy and safety endpoints, but on sensitivity (agreement with true positive results of the gold-standard test), specificity (agreement with true negative results), positive predictive value (percentage of time gold standard is positive when the experimental test is positive) and negative predictive value (percentage of time the gold standard is negative when the experimental test is negative).
- Diagnostics developers must decide whether their goal is a quantitative result or qualitative result, such as a positive or negative determination of the presence of a health condition, perhaps derived from a quantitative result. If developing a qualitative test, the basis for drawing the line between positive and negative result is important.
- While drug developers must wait for availability of investigational product to begin trials, diagnostics developers can collect samples before assay validation; optimal timing of sample collection depends on risk of sample degradation, storage requirements to preserve valid samples, cost of sample storage and actual vs. predicted time of assay validation. Waiting to collect samples until after validation entails delays. Collecting samples before assay validation increases risk.
- Diagnostics developers must manage enrollment efficiently when there is a requirement to test a diagnostic in stratified populations, such as different age groups. There can be substantial delays because of failure to enroll enough subjects from one stratum or substantial waste from enrolling too many subjects from another.
- When the gold-standard comparator is a pathology outcome, diagnostics trials must monitor pathology reports. In addition, when there is a central reader, slides from local pathologists must be shipped to a central pathologist and this requires authorizations for shipment at the source and checks on the central reader’s receipt of samples.
Importance of Operational Planning in Diagnostics Studies
It is essential to begin a diagnostics study with a clear plan to address operational considerations listed above. Such considerations include closely tracking primary endpoint data, managing stratified enrollment and monitoring pathology reports. The clinical affairs, data management and statistical teams assigned to a diagnostics trial should all agree on the plan at the outset as should the sponsor and CRO. Throughout each diagnostics trial, the study team must be mindful of the importance of precise execution to ensure availability of the data required to evaluate the test diagnostic.
Clint Dart, MS, Director of Biometrics at Health Decisions, has provided statistical services in trials evaluating a variety of devices and diagnostics as well as phase 1 – 3 drug and biologics studies. Clint has provided statistical support for a variety of successful 510(k) and PMA submissions and led statistical efforts for multiple pivotal studies with successful regulatory filings.