ECG assessment of drug-induced QT prolongation in clinical trials
Drug-induced QT prolongation is a common electrocardiographic finding during exposure to drugs that affect ventricular repolarisation. Its clinical importance ranges from an isolated laboratory observation to a precursor of torsades de pointes. Interpretation becomes difficult when heart rate, rhythm, electrolytes, structural heart disease, or the measurement method also changes.
Clinical trials must translate this variable signal into reproducible evidence. A rise in a corrected QT interval may reflect genuine pharmacodynamic risk, a mathematical artefact of heart-rate correction, or differences in ECG acquisition and reader technique. Trial protocols therefore need methods precise enough to distinguish these explanations before they interpret safety data.
This article examines the methodological challenges of electrocardiographic assessment of drug-induced QT prolongation in clinical trials. It covers collection, correction, adjudication, risk stratification, endpoint definition, pharmacovigilance and analysis, with attention to Australian clinical environments and consumer medicine use.
Why QT correction remains consequential
The QT interval represents total ventricular depolarisation and repolarisation. Many QT-prolonging medicines reduce outward potassium current through cardiac hERG/IKr channels, delaying phase 3 repolarisation. The increase can be concentration-dependent and may become more prominent at slow heart rates or with hypokalaemia.
QT prolongation is a surrogate for malignant ventricular arrhythmia, not a direct measure of torsades de pointes. A larger corrected QT interval raises concern, but no universal millisecond threshold predicts events across every population. Risk also reflects baseline vulnerability, drug accumulation, heart rate, electrolytes and the arrhythmia mechanism.
Trial inference depends on whether a signal is reproducible within participants, related to exposure and distinguishable from baseline disease. A generic statement that one arm had a “longer QT” is not interpretable unless readers know the correction, heart-rate range, analysis set and adjudication process.
Designing trial-ready ECG specifications
A trial-ready schedule should anchor ECGs to dose timing, not simply to study visits. Baseline recordings should precede treatment; subsequent ECGs should cover first-dose effects, loading, expected steady state, accumulation, dose interruptions and washout where applicable. Record the time since the last dose, fed state, vital signs and any acute illness that could alter repolarisation.
Correction-formula choice can change the apparent drug effect. Bazett overcorrects at high heart rates and undercorrects at low rates, while Fridericia often performs more consistently across wider ranges; Framingham may provide a useful sensitivity analysis. Protocols should prespecify one primary formula, apply it consistently and examine plausible alternatives in blinded analyses.
ECG acquisition specifications should fix the validated device family, calibration, sampling speed, filter settings, lead set and software version. Central acquisition can reduce site variation, yet a fixed protocol remains essential for multicentre studies, hybrid workflows and repeat visits. The same rules should apply to screening, on-treatment and post-treatment records.
Automated measurement and manual adjudication
Modern ECG machines estimate QT automatically using waveform-detection and correction algorithms, but outputs can vary with T-wave morphology, noise and proprietary thresholds. Automated values support rapid processing, while a prespecified manual review provides a check on outliers, implausible values and readings affected by ectopy, artefact or unusual repolarisation.
For manual measurement, define the onset of the QRS complex and the end of the T wave, then specify the tangent or threshold method and lead-selection rule. U waves, low-amplitude T waves and borderline termination create reader variability. Images should be reviewed at adequate magnification with the calibration visible, and every manual override should be attributable to a named reader.
A quality system should include reader training, duplicate readings and prespecified disagreement resolution. Readers should remain blinded to treatment assignment, dose sequence and suspected event status. Atrial fibrillation, frequent premature beats, bundle-branch block, paced rhythms and fusion beats need explicit handling rules rather than silent exclusion from the analysis set.
Defining risk strata and cardiac endpoints
Risk stratification should combine baseline QTc with age, cardiovascular disease, bradycardia, electrolyte abnormalities, renal and hepatic impairment, interacting medicines and the investigational dose. Monitoring should intensify when several factors coincide. Eligibility criteria based on a single QTc threshold may miss vulnerable participants if those other determinants are ignored.
Structural heart disease can add arrhythmia risk without producing a markedly prolonged baseline QTc. The use of cardiac sarcoidosis imaging during ablation planning for ventricular arrhythmias illustrates why baseline substrate evidence may complement QT surveillance in a high-risk trial population.
Endpoints need operational definitions. Distinguish a central-reader QTc change, a protocol-defined QTc threshold, confirmed torsades de pointes, polymorphic ventricular tachycardia, nonsustained ventricular tachycardia and arrhythmia-related discontinuation. Exposure windows, adjudication criteria and the method for relating an ECG to the nearest dose should be specified before unblinding.
Australian trial settings and everyday influences
Multisite Australian studies may enrol participants in Melbourne, Sydney, Brisbane or Perth while centralising ECG reading at one laboratory. Differences in staffing, transport times, device maintenance and software updates can affect data quality. Using the same recording equipment where feasible, calibrating machines on schedule and tracking device failures helps separate protocol effects from site effects.
Hot summers, dehydration and electrolyte loss through sweating can influence heart rate and repolarisation. Energy drinks, caffeine intake and alcohol also vary around trial visits and may change heart rate or symptom reporting. Rather than imposing unrealistic lifestyle restrictions, protocols should record these exposures consistently and schedule blood sampling and ECGs under comparable fasting and hydration conditions.
Participants may obtain prescription, over-the-counter, complementary or internet-purchased medicines without reporting them unless asked directly. Medication reconciliation should therefore include community-pharmacy records where permitted. The coexistence of atrial fibrillation, PFO closure decisions and anticoagulation can alter eligibility and follow-up, as discussed in atrial fibrillation and PFO; irregular RR intervals also make QT correction less stable.
Pharmacovigilance and Australian regulation
A serious suspected adverse reaction requires a chronology covering symptoms, dose changes, ECG findings, relevant laboratory results, concomitant medicines and the time from exposure to onset. Original tracings should be retained because source verification can change the apparent QTc or reveal an alternative rhythm. Syncope, palpitations and unexpected ventricular ectopy require structured assessment rather than being treated as isolated events.
In Australia, medicinal-product trials conducted through the TGA’s Clinical Trials Notification scheme operate within the Therapeutic Goods Act 1989. Ethics review remains a separate requirement, while the Australian New Zealand Clinical Trials Registry provides public trial visibility. Data handling should also comply with applicable privacy, health-records and consent requirements, including the Privacy Act 1988 and relevant state and territory legislation.
Sponsors should integrate blinded trial findings with post-marketing data when assessing a safety signal. Interpretation requires exposure denominators, background event rates, dose–response patterns, duration of therapy and the completeness of ECG ascertainment. Reporting only spontaneous cases can exaggerate or obscure risk, particularly when severe events are more likely to reach regulators than unremarkable measurements.
Strengthening analysis and transparent reporting
Statistical analysis should use participant-level repeated-measures methods, with prespecified handling of missing ECGs, early discontinuation and protocol deviations. Concentration–QT modelling can test whether any ECG change is compatible with known pharmacology. Models should account for heart rate and relevant covariates while examining individual participants with the largest changes, because an average arm-level result can conceal a small high-risk group.
Reports should state the correction formula, accepted heart-rate range, lead-selection rule, device and software versions, averaging method, reader process and adjudication rules. Raw traces, anonymised ECGs and code supporting the analysis should be shared when consent and privacy permit. The open-access Journal of Arrhythmia provides peer-reviewed arrhythmia resources that can help investigators refine these protocols.
Australian datasets should also address under-representation of Aboriginal and Torres Strait Islander participants and other populations. Recruitment strategies developed with relevant communities and appropriate governance can improve generalisability. Automated QT algorithms require validation in the trial population, particularly when low-amplitude T waves, conduction abnormalities or irregular rhythms are common.
Before the first participant is dosed, trial teams should adopt a harmonised ECG protocol, appoint the central reading laboratory, lock the statistical analysis plan and establish pharmacovigilance escalation pathways. Investigators, statisticians, ECG readers and safety monitors should each understand their responsibilities before any treatment code is broken.
A trustworthy assessment preserves the raw trace, corrected interval, heart rate, concentration data and clinical context as one continuous evidence trail. Adopt the protocol, centralise the read and report the complete ECG record from acquisition through adjudicated endpoint.