Commentary: Diagnostic devices in clinical trials have high stakes for patient careBMJ 2016; 354 doi: https://doi.org/10.1136/bmj.i5197 (Published 28 September 2016) Cite this as: BMJ 2016;354:i5197
- Vinay K Rathi, postgraduate trainee1,
- Harlan M Krumholz, Harold H Hines Jr professor of medicine and epidemiology and public health2,
- Joseph S Ross, associate professor of medicine and public health2
- 1Department of Otolaryngology, Harvard Medical School, Boston, Massachusetts
- 2Yale School of Medicine PO Box 208093 New Haven, CT 06520-8093, USA
- Correspondence to: J S Ross
The widely used anticoagulant rivaroxaban (Xarelto) has become the centre of considerable scientific and regulatory dispute. A key clinical trial supporting US Food and Drug Administration (FDA) approval was recently found to have used faulty point-of-care devices for monitoring international normalised ratio (INR).1 2 This ongoing episode provides important insights into regulation challenges for diagnostic devices and the far reaching effects these devices have on clinical research and patient care.
The FDA defines in vitro devices as devices using human specimens to guide the treatment or prevention of disease, and clears the large majority for use through the 510(k) process for moderate risk devices, which requires only that a device be “substantially equivalent” to another previously cleared device.3
In 2002, the FDA cleared the INRatio model used in the ROCKET trial4 based on substantial equivalence to the CoaguChek S point-of-care monitor. This approval shows several critical challenges that the FDA faces with clearing diagnostic devices:
Predicates—Analytical and clinical validity of predicate devices may be inherently limited. The CoaguChek S was found to underestimate INR compared with laboratory reference methods, with >40% of measurements deviating by ≥15%,5 and in 2006 was subject to a class II (moderate risk) FDA recall because defective testing strips produced falsely raised results. Currently, no restrictions prohibit using voluntarily recalled devices as predicates, and devices with malfunctioning predicates cannot be readily identified for further testing after clearance.
Clinical testing—Diagnostic devices rarely require prospective clinical studies for clearance, and these studies may not be generalisable when …