What this article covers
- Yield accumulates over days. In a 26,751-patient registry, 29.9% of patients with a detected arrhythmia had their first event after the initial 48 hours.
- The trade runs both ways. Inside a matched 24-hour window a three-lead Holter outperformed a single-lead patch. Duration wins on paroxysmal events; leads win on morphology.
- Logistics decide the design. A patient who travels three hours for one application does not come back for a second attempt.
A patient walks into a district hospital in eastern Uttar Pradesh describing palpitations that come two or three times a month. She gets a 24-hour Holter, which is the correct first-line investigation and the one the department can deliver. The recording is clean. It shows sinus rhythm throughout.
She is told her heart is fine.
Her episodes occur two or three times a month. The recording covered one day. The probability that a monthly event landed inside that window was never high, and the normal result carries almost none of the reassurance she took home with it.
Where the yield sits
Turakhia and colleagues published the largest look at this in 2013, covering 26,751 consecutive patients wearing an extended adhesive patch, with a mean wear time of 7.6 days.1 Among the 60.3% of patients in whom an arrhythmia was found, 29.9% had their first arrhythmia after the initial 48 hours. For symptom-triggered arrhythmias the figure was 51.1%.
Swipe the chart sideways to see all of it.
Barrett's team ran the direct comparison in 2014, fitting 146 patients with a 24-hour Holter and a 14-day patch at the same time.2 Over total wear the patch recorded 96 arrhythmia events against the Holter's 61.
The detail worth sitting with is the one that goes the other way. Inside the matched 24-hour window, the Holter found more events than the patch, 61 against 50, and detected 11 events the patch missed. Three leads see things one lead cannot. The patch caught all 11 of those events later in its wear period, but the point stands on its own: extended single-lead monitoring is a trade, and if the question is a matched-window comparison, a proper Holter wins it.
Duration wins on paroxysmal events. Leads win on morphology and localisation. Anyone selling you a patch as a strict upgrade over a Holter is selling.
The gradient that matters clinically
The Turakhia data broke time-to-detection down by atrial fibrillation burden, and the relationship runs the wrong way for anyone relying on a short study.1
| Total AF burden | First detection after 48 h |
|---|---|
| 51% to 75% of monitored time | 11.2% |
| 26% to 50% | 10.5% |
| 1% to 25% | 20.8% |
| Below 1% | 38.0% |
Swipe the table sideways to see all columns.
Low-burden paroxysmal AF is the AF a short recording misses. It is also AF that carries stroke risk, and the patient in front of you has no way of knowing which group she belongs to before the recording is made. The population most likely to receive a falsely reassuring 24-hour study is the population with the least detectable and still dangerous disease.
Why we designed for 21 days rather than 14
If 90% of arrhythmias appear by day five, three weeks looks like over-engineering. Two arguments took us there, and only one of them is clinical.
The clinical argument is the tail described above, plus symptom correlation. A patient with daily symptoms is well served by a short study. A patient with fortnightly symptoms needs a window that contains at least two expected events, and 90% of symptom-triggered detections took eight days to accumulate in the registry data. Fourteen days covers monthly symptoms poorly. Twenty-one days covers them adequately.
The second argument is logistics, and in our setting it dominates.
If a patient travels three hours to reach the district hospital that holds the equipment, you get one application. Not one appointment, one application. A 14-day study that comes back inconclusive means a second journey, a second slot, and a second attempt at persuading someone who has already lost two days of wages that this is worth repeating. Most of the time the repeat does not happen, and the study that would have found the answer is never made.
India has roughly five cardiologists per million people.4 That number describes the reading capacity, and it also describes the queue. A repeat study costs more than rupees. It is a second position in a queue that barely exists, held by a patient who has already been told once that everything looked normal.
Designing for 21 days is designing for one shot.
What longer recording makes worse
Three weeks of continuous single-lead data produces findings the referring clinician did not ask about. In the stroke and TIA cohort Tung and colleagues studied, supraventricular tachycardia of four beats or more appeared in 70.2% of recordings, and runs of eight beats or more in 51%.3 Nearly all of those patients had no atrial fibrillation at all.
Most of that is noise in the clinical sense. Some of it will be acted on anyway, and a device that surfaces more incidental findings without helping the reader sort them has moved work rather than removed it. This is the reason on-device classification matters to us as much as recording duration. Handing a cardiologist 21 days of unsorted trace is a different failure from handing her 24 hours, but it is still a failure.
Longer recordings also demand more of the acquisition chain, because the adhesive and the hydrogel degrade across the wear period while the analysis assumes they have not. We wrote about that side of the problem separately.5
The honest summary
A 24-hour Holter is a good test, correctly indicated, that answers a narrow question: what is happening today. Roughly half the arrhythmia yield available in an extended study sits outside that window, and for low-burden AF the fraction is larger.
The reason to extend is not that longer is better in general. It is that the patient described at the top of this article gets one recording, and the recording should be long enough to have contained her arrhythmia.