3
min read
From ECG Collection to Signal Interpretation: Building a Smarter Cardiac Safety Strategy
In clinical trials, ECG data moves from site collection through transmission to centralized interpretation. Each step affects quality. Standardized acquisition, expert over-read and consistent QT/QTc measurement turn raw traces into reliable cardiac safety signals, reducing the variability that can obscure or exaggerate proarrhythmic risk.
A twelve-lead ECG takes ten seconds to record. Everything that determines whether those ten seconds produce usable evidence happens before and after: how the electrodes were placed, whether the patient had rested, how the trace travelled, and who measured the interval and how.
This is the part of cardiac safety that sponsors rarely see, because it happens between the site visit and the summary table. It is also where most of the avoidable variability lives.
The ECG Data Journey in a Clinical Trial
Four stages separate a patient lying on an examination table from a number in a safety dataset. Each one has its own owner and its own characteristic failure.
Stage | What happens | Owner | Typical failure |
|---|---|---|---|
Collection | Electrodes are placed, the patient rests, the recording is triggered at the protocol-defined timepoint | Site staff, trained by the core lab | Lead misplacement, insufficient rest, timing drift against the PK sample |
Transmission | The digital trace is transferred to the core lab, or a paper ECG is scanned and digitized | Site and core lab | Missing recordings, unreadable scans, incorrect subject or visit identifiers |
Over-read | Trained analysts measure intervals under a fixed convention; a cardiologist reviews abnormal findings | Core lab | Reader drift, inconsistent lead selection, unblinded review |
Interpretation | Corrected intervals are calculated, alert values escalated, datasets reconciled and exported | Core lab and sponsor | Correction method changed after seeing data, reconciliation gaps at lock |
Where ECG Data Quality Breaks Down
Almost none of the failures above are equipment failures. They are procedural, which is good news, because procedure can be controlled.
Electrode placement. The precordial leads are the usual culprit. V1 and V2 placed one intercostal space too high change the morphology of the trace and can produce patterns that look like a genuine conduction abnormality. Vertical variation across the same patient over successive visits introduces movement in the data that has nothing to do with the drug.
Patient state. Heart rate responds to posture, recent activity, anxiety and caffeine. Since the correction of the QT interval depends directly on heart rate, an inconsistent resting period before recording propagates straight into the corrected value.
Site-to-site variation. Different machines, different filter settings, different sampling rates, different local habits about which lead to print. A study running in twenty countries collects data through twenty slightly different lenses.
Timing. In any analysis that pairs ECG with drug exposure, the timestamp is part of the measurement. An ECG recorded twenty minutes off schedule is not a slightly imperfect data point. It may be an unusable one.
The pattern is consistent: these are small errors, each defensible in isolation, that accumulate into a variance the statistical analysis then has to absorb.
From Raw Trace to Signal: The Interpretation Layer
Once traces reach the core lab, the question changes. It is no longer about collecting data. It is about measuring the same thing, the same way, several thousand times.
Manual expert over-read vs algorithmic reads
Every clinical ECG machine embeds an automated interval measurement algorithm. Those algorithms are useful in clinical care and were designed for it. They differ by manufacturer, they were not built for regulatory endpoint analysis, and their behavior on noisy or unusual morphologies is not standardized across devices.
An ECG over-read is the review of the trace by a trained human analyst who applies a single documented measurement convention across the whole study, with cardiologist review for abnormal or clinically significant findings. The value is not that a person is more accurate than an algorithm on any given beat. It is that one convention applied by a trained and monitored team produces a consistent bias rather than a variable one, and consistent bias is something statistics can handle.
Hybrid workflows are now common: algorithmic pre-measurement to flag and prioritize, human over-read to decide. What matters is that the escalation rules are written down before the first trace arrives.
Measuring QT/QTc consistently (Fridericia and Bazett)
The QT interval shortens as heart rate rises, so it has to be corrected before values can be compared. Two formulas dominate. Bazett's correction (QTcB) divides QT by the square root of the RR interval. Fridericia's correction (QTcF) divides QT by the cube root of the RR interval.
They do not behave the same way. Bazett over-corrects at high heart rates and under-corrects at low ones, which means it can manufacture apparent QT prolongation in a patient who is simply tachycardic. Fridericia is less sensitive to heart rate and is generally preferred in drug development. Study-specific and individual corrections are also used, particularly where the compound affects heart rate directly.
The methodological point matters more than the choice itself: the correction method belongs in the protocol, decided before anyone has seen the data. Selecting a formula after the fact, because it produces a more comfortable number, is the kind of decision that becomes very difficult to explain during review.
Building a Smarter Workflow
A well-built ECG workflow is mostly about removing decisions from the moment of collection and putting them into documents written in advance.
One acquisition standard across every site, covering electrode placement, rest period, posture, recording conditions and timing windows
Site training and qualification before enrollment, not troubleshooting after the first queries
Centralized transmission with automated identifier checks, so a mismatched subject number is caught in hours rather than at lock
A single measurement convention, applied by a trained analyst team whose performance is monitored over the life of the study
Cardiologist review with predefined escalation thresholds and a documented turnaround commitment
The QTc correction method specified in the protocol
Continuous quality control rather than a review at database lock
This is the workflow Banook has been refining since 1999, when the group started as Cardiabase, a core lab built specifically for cardiac safety. Digital ECGs, paper ECGs, Holter recordings and extracted ECGs are centralized on a platform developed in house and aligned with 21 CFR Part 11, with QT and QTc analysis covering QTcB, QTcF and QTcL. Board-certified cardiologists and chief medical officers are involved from protocol review onward, not only at the reading stage, and they support sponsors in responding to investigator and regulatory agency questions. On the operational side, Banook states that an ECG is analyzed in 30 minutes rather than days, that a study starts in less than a month, and that 99% of its commitments are met on time. Submission support extends to the FDA ECG warehouse, alongside statistical analysis and medical writing.
The wider picture of why cardiac safety has moved up the agenda is covered in Why Cardiac Safety Matters More Than Ever in Modern Clinical Trials
👉 Contact our team today to discover how we can support your next study
Frequently Asked Questions
How are ECGs collected in clinical trials?
Through twelve-lead recordings taken at protocol-defined timepoints, using equipment provisioned and configured for the study, following a standardized acquisition procedure that specifies electrode placement, rest period, posture and timing. Recordings are then transmitted to a central facility rather than interpreted locally. Holter and event monitoring are added where continuous data is required.
What is an ECG over-read?
An ECG over-read is the independent review and measurement of a trace by trained analysts at a core lab, applying one documented convention across the whole study, with cardiologist review of abnormal findings. It replaces site-level and algorithmic interpretation as the source of the endpoint data.
What is QT/QTc and why does it matter?
The QT interval measures the time from the start of ventricular depolarization to the end of repolarization. Because it varies with heart rate, it is reported as a corrected value, QTc. A prolonged QTc is the accepted surrogate marker for proarrhythmic risk, including torsade de pointes, which is why it is measured so carefully in drug development.
What is a cardiac safety core lab?
A cardiac safety core lab centralizes the collection, measurement and interpretation of cardiac data across all sites in a trial: equipment provisioning and site training, standardized acquisition, over-read with cardiologist review, consistent QTc correction, quality control and submission-ready datasets.
Manual or automated ECG interpretation?
Both, in sequence. Automated algorithms are efficient for pre-measurement and triage but vary by manufacturer and were not designed for regulatory endpoint analysis. Manual expert over-read provides the single consistent convention that comparison across sites and timepoints requires. Most core lab workflows combine the two with predefined escalation rules.