Time in Range

The test underneath is the prothrombin time, or PT: add a reagent called thromboplastin to plasma and time how many seconds it takes to clot. The trouble is that different labs' reagents clot at different speeds, so the raw seconds don't travel. The INR fixes that with two corrections: divide by the lab's mean normal PT, then raise the ratio to the reagent's ISI, a calibration number found by testing it against a reference reagent (University of Iowa pathology handbook; Clinical Laboratory Science). The WHO adopted a reference thromboplastin in 1983, and makers calibrate against it (Roche Diagnostics).

Same blood, two reagents

Run it backwards. Pick one INR and see how many seconds two reagents would show for it, both with a mean normal PT of 12 s. With an ISI of 1.0 the INR is just the plain ratio. A less sensitive reagent (higher ISI) clots sooner for the same blood, and the exponent stretches its shorter time back to the same INR.

A home meter does the same timing on a fingerstick drop instead of lab plasma. The strip carries dried reagent and electrodes, and the meter watches for the moment clotting gets going. Device descriptions show two approaches: some strips hold a peptide that thrombin cuts, releasing a piece that carries a current the meter can see, and some measure the clot through a change in the sample's electrical impedance (NICE's review of the meters). Either way, the meter turns a time into an INR with its own built-in calibration.

1. A drop of blood from a fingerstick lands on the strip.
2. The dried thromboplastin dissolves and starts the clotting cascade.
3. Thrombin builds up. Depending on the strip, it cuts a peptide that gives off a current, or the forming clot changes impedance.
4. The meter notes when the signal crosses its threshold. That's the clotting time.
5. A stored calibration turns the time into an INR on the screen.

Below are two strips of chart paper. The red trace is a made-up "true" INR over six months: it drifts slowly, jitters, and now and then takes a bump (an illness or a diet change, say). Each schedule only sees the line on test days, the black dots. Between them, the usual way to estimate time in range is to draw a straight line from one reading to the next and count the days that fall inside the band, a method from Rosendaal and colleagues (Frontiers in Cardiovascular Medicine, 2022). That's the blue dashed line.

Both strips share the same bumps and jitter. With "act on readings" on, an out-of-range reading nudges the line back toward the middle over a few days, a stand-in for whatever a clinician would change. That's why the two red traces can drift apart.

Things to try

• Set A to 28 and B to 7, then press "New trace" a few times. Monthly testing often reports a tidier picture than the line really had, because short bumps fall between visits.
• Turn "act on readings" off. Now testing can't change the line, only what you see of it. Both strips have the same true time in range; only the estimates differ.
• Set B to 1. Daily tests make the guess match the truth, but every reading also carries a little meter noise, and acting on each one can chase jitter.

The sampling question is roughly what the self-testing trials asked, with real people and real outcomes instead of a toy line.

• THINRS, a Veterans Affairs trial published in the New England Journal of Medicine in October 2010, randomized 2,922 people on warfarin to weekly home testing or monthly high-quality clinic testing. It found no significant difference in the main outcome, the time to a first stroke, major bleed, or death. The home group had a small gain in time in range, more minor bleeds, and higher satisfaction (ACP Internist Weekly; MDedge).
• A Cochrane review updated in 2016 pooled 28 randomized trials with 8,950 people. Self-testing and self-management together came with about half the thromboembolic events and no drop in major bleeds, on evidence it rated low to moderate in quality (Cochrane).

So the lab's lesson, that more frequent looks catch more, shows up in the trials as a modest gain rather than a dramatic one. The people in them were trained and chosen as able to do it, which is part of the picture.

• MedlinePlus: Prothrombin time test and INR, the NIH's plain-language page on the test.
• The Warfarin SMART trial, a randomized trial of home self-management against usual care.
• A study of supervised self-testing with an online system, on PubMed Central.
• A registered trial of patient self-testing on ClinicalTrials.gov.

Those cover what the test is and what self-testing did in practice. This one tries something smaller: putting the formula and the sampling question on sliders, so you can watch them work.

• The true lines are a random walk plus jitter plus a few bumps. Real INR depends on diet, other medicines, illness, and much else that isn't modelled.
• "Act on readings" moves the line part way back to the middle over four days. That is a placeholder for a clinician's decision, not a dosing rule.
• Each reading gets a small random error. Real meter-versus-lab differences vary by device and by INR level.
• The band options are two common ones; some conditions use others (Journal of Pharmacy Practice). Rosendaal's method also skips gaps over 56 days, which is why the interval stops there.