Point-of-Care Fingerstick PT/INR System

Against the point-of-care comparator (n = about 350), the all-sites Passing-Bablok slope was slightly below 1 (about 0.9), the intercept near zero and r² about 0.9. Site-level slopes ranged from about 0.9 to 1.0. Against the laboratory reference on venous plasma (n = about 350), the all-sites slope was about 1.0, the intercept near zero and r² about 0.9, with site-level slopes from about 0.9 to 1.1.

Combined-site whole-blood repeatability was about 5% CV for INR <2.0 (n = about 150), about 5% CV for INR 2.0 to 3.0 (n = about 400) and about 4% CV for INR 3.1 to 4.5 (n = about 100). The highest single-site value (about 9% CV in the <2.0 range at one site) included an outlier, as the decision summary notes. In the control reproducibility study (n = about 300 per level), total CV across all sites was about 4% for the normal control (mean about 1.3 INR) and about 5% for the therapeutic control (mean about 3.2 INR). The decision summary reports that both precision studies passed their acceptance criteria.

Hematocrit between 22% and 52% did not significantly affect results, and the normal reference range was 0.9 to 1.1 INR. FDA found the system substantially equivalent and cleared it, with its PT controls, for professional point-of-care use.

Read the two regressions together. Against the laboratory method, an all-sites slope of about 1.0 and an intercept near zero indicate no systematic proportional or constant bias across the range. Against both comparators, the slope estimates suggest that, if anything, fingerstick INRs ran slightly lower than the comparators at higher values. The r² against venous plasma is a little lower than against the capillary comparator. Part of that gap is expected: the venous comparison adds matrix and sampling differences that a capillary-to-capillary comparison does not.

The decision summary does not publish numeric acceptance criteria for the method comparison, and 21 CFR 864.7750 does not codify performance limits. Readers should treat the regression and precision figures as the evidence FDA found sufficient for this device, not as a universal threshold. Site-level results deserve as much attention as pooled ones. One site's therapeutic control showed about 8% total CV, compared with about 2% to 4% at the other sites, a reminder that pooled estimates can hide a single site's operational variability.

The device is intended for professional healthcare providers at the point of care. It reports an INR, based on a prothrombin time response, for monitoring oral anticoagulation with warfarin. The intended population is patients 18 years and older who have been stabilized on warfarin for more than six weeks; the system is not intended for patients transitioning from heparin to warfarin. The only specimen is fresh capillary (fingerstick) whole blood applied directly to a single-use test strip, and the system is labeled for prescription use.

To support substantial equivalence, fingerstick INR results had to agree with a legally marketed point-of-care PT/INR device and with a laboratory reference method across the claimed 0.8 to 4.5 INR reportable range. The reference method was INR on venous citrated plasma, measured on a laboratory coagulation analyzer using a recombinant human tissue thromboplastin reagent. Because no linearity study applies to a calculated INR, FDA's decision summary notes that the reportable range itself was established through these method comparison studies.

The enrolled population therefore had to include both people not on anticoagulants (INR <2.0) and warfarin patients (INR 2.0 to 4.5). Testing had to happen at intended-use sites with multiple operators and strip lots, so that lot-to-lot and operator variability were part of the result rather than tested only in a development lab.

Prothrombin time tests are classified under 21 CFR 864.7750 as Class II devices. The regulation covers PT tests used as a general screen for extrinsic-pathway clotting factor deficiencies and to monitor patients on coumarin anticoagulants, including fingerstick PT/INR meters. The accompanying liquid controls are reviewed as a coagulation plasma control under 21 CFR 864.5425. FDA does not codify device-specific special controls for this regulation in the CFR, so the evidence expectations visible in decision summaries come largely from FDA-recognized CLSI consensus standards for precision (EP05), method comparison and bias estimation (EP09), interference (EP07) and commutability of processed control material (EP14).

An INR is not a direct measurement. The meter measures a clot time and converts it to an INR using an International Sensitivity Index (ISI) and mean normal prothrombin time assigned to each strip lot by calibration against a reference thromboplastin traceable to the World Health Organization International Reference Preparation. Method comparison therefore tests the whole chain at once: strip reagent, lot calibration, meter algorithm and sample handling.

The category is hard for practical reasons. Capillary whole blood must be compared against venous citrated plasma, which is a different matrix. Hematocrit, interfering drugs and heparin exposure can shift clot times. Dosing decisions are made inside a narrow therapeutic interval, so small biases matter clinically. Every claimed segment of the reportable range also has to be populated with real patients, including warfarin patients at the high end.

• Using two comparators separated two questions: equivalence to a device already used at the point of care (capillary blood versus capillary blood) and accuracy against the laboratory INR that drives clinical practice (capillary blood versus venous plasma).
• Enrolling both non-anticoagulated subjects and stabilized warfarin patients spread results across the reportable range. That matters because the decision summary states the 0.8 to 4.5 INR range was established through method comparison, not linearity.
• Running three strip lots at every site, with at least three operators each, put lot calibration and operator technique into the comparison data instead of leaving them as untested assumptions.
• Duplicate fingersticks gave a direct measure of capillary repeatability in real patients. Control-material reproducibility over 20 days isolated instrument and strip variability from sampling variability.
• Collecting venous EDTA hematocrit alongside each capillary test let the hematocrit claim come from the same clinical population, rather than only from spiked or manipulated samples.