ECG Sampling Design and Assay Sensitivity for QT Studies

Why Sampling Design Matters

A concentration-QTc analysis can only describe the exposure-response pattern that the study has had a chance to observe. If ECGs are collected only before dosing and near a single nominal peak, the data may miss the rising or falling exposure profile, a delayed QTc response, or a time-related change also present on placebo. The schedule is therefore part of the evidence: it determines whether exposure and response are observed with enough temporal resolution to interpret their relationship.

A useful design aligns three elements: concentration measurements that describe exposure over time, ECG measurements that describe QTc over time, and controls that separate drug-related changes from background time trends and demonstrate that the study could detect a QTc effect. No single element substitutes for the others.

Build the ECG and Concentration Schedule Around Time

Start with the expected concentration-time profile at each planned dose. The schedule should represent the rise, approximate peak, and decline in concentrations, and should include enough observations to distinguish these phases. The most informative ECG times are not necessarily identical to the times of the highest measured concentration: as established in the earlier material on delayed QTc responses and concentration-effect hysteresis, the response can lag behind concentration. Sampling should therefore cover plausible response delays as well as the concentration peak.

Whenever practical, collect ECGs and plasma samples at matched or closely adjacent times. Record actual collection times rather than relying only on nominal visit times. A concentration measured substantially before or after an ECG may poorly represent the exposure relevant to that ECG, particularly when concentrations are changing quickly. A concentration profile with few samples around absorption or elimination can also leave exposure assignment uncertain, even if the ECG schedule is dense.

The schedule should include predose observations and postdose observations over a period long enough to characterize the relevant exposure and response. For repeat dosing, consider whether accumulation, time to steady state, and the dosing interval affect the exposure range that needs to be covered. Choose times based on the drug's expected pharmacokinetics and plausible response delay; a fixed list of clock times is not automatically suitable for every compound or formulation.

The schedule is a compromise between temporal coverage, participant burden, and operational feasibility. More observations are not automatically better if they are poorly timed, inconsistently collected, or measured with low-quality ECGs. Conversely, a sparse schedule can obscure a short-lived effect or make a delayed pattern indistinguishable from a time trend. Prespecify the key time windows and the reason each is included.

Baseline Replicates and Placebo Have Different Roles

Baseline ECGs establish the participant's QTc before treatment. Replicate ECGs at a visit reduce the influence of short-term measurement variation; using a time-matched baseline also helps account for regular changes across the day. If a time-matched change is defined as the postdose QTc minus the baseline QTc at the corresponding clock time, the comparison is more informative than subtracting a single baseline value from every postdose measurement when circadian variation is relevant.

Replication improves precision when replicate measurements are collected under comparable conditions and their errors are not perfectly correlated. If each measurement has within-occasion standard deviation \(\sigma\), and \(m\) replicate measurements have independent errors, the standard error of their mean is \(\sigma/\sqrt{m}\). Real replicates may share sources of error, so the reduction can be smaller. Replication cannot correct systematic problems such as inconsistent ECG timing, a biased correction method, or poor acquisition quality.

Placebo addresses a different question. In a randomized placebo-controlled design, placebo participants undergo the same schedule and procedures, so their time-matched changes help describe changes not attributable to active treatment, including circadian patterns, study procedures, and background variation. A placebo group is not a substitute for baseline measurements: baseline anchors within-participant change, while placebo provides a concurrent comparison for temporal changes.

A crossover design can allow each participant to receive active treatment and placebo, potentially improving precision by controlling for stable between-participant differences. It requires suitable washout and attention to period, sequence, and carryover effects. A parallel design avoids within-person carryover between treatments but generally requires more participants to achieve similar precision. The design choice should reflect pharmacokinetics, treatment duration, feasibility, and the possibility that prior treatment affects later observations.

Positive Control and Assay Sensitivity

A positive control is a treatment expected to produce a measurable QTc effect, selected and administered under conditions that allow its effect to be assessed. Its purpose is to show that the study procedures and analysis can detect an effect of prespecified relevance. It is not merely another active arm, and it should not be confused with placebo. Placebo estimates background change; the positive control tests the study's ability to detect a known kind of change.

The assay-sensitivity criterion should be specified before examining results. In a commonly used thorough-QT design, one criterion is that, at a prespecified time point, the lower bound of the 90% confidence interval for the positive-control versus placebo mean QTc change exceeds 5 ms. The exact criterion and analysis must follow the applicable protocol and regulatory context. A positive-control result meeting its criterion supports assay sensitivity under the study conditions; a failure to meet it does not prove that the control has no QT effect.

Precision, Diagnostics, and Design Limits

The following simple calculation explains why allocation and measurement variability matter. For independent groups with equal within-group standard deviation \(\sigma\), the standard error of a difference in group means is:

The first equality follows because the variance of a difference between independent sample means is the sum of their variances; each mean has variance \(\sigma^2/n\). Thus, noisy measurements or small groups widen uncertainty. This formula is a planning approximation, not a complete sample-size calculation: repeated observations, participant-level correlation, missing data, and the analysis model also affect precision.

Review the data during conduct for failures that compromise the intended schedule: missed or mistimed ECGs, concentration samples not aligned with ECGs, inadequate coverage of the rising or falling exposure phase, and missing observations near the expected delayed response. Examine whether placebo changes vary systematically by clock time, whether baseline replicates are unusually inconsistent, and whether positive-control results are compatible with the prespecified assay-sensitivity criterion. Document acquisition and reading quality, exclusions, and departures from the planned time windows.

If observations cluster at only one concentration range, a slope or delayed pattern may be weakly identified regardless of the number of ECGs. If a control effect is imprecise, a confidence interval can fail the criterion even when its point estimate is in the expected direction. If the positive control was not administered or measured as planned, apparent failure of assay sensitivity may reflect implementation rather than absence of pharmacologic activity. These possibilities should be investigated rather than resolved by changing the criterion after seeing the data.

Alternative approaches include adding ECGs around expected rapid changes, using a crossover rather than parallel comparison when washout is adequate, or increasing participant numbers when between-participant variability dominates. Each has costs: dense schedules increase burden, crossover designs introduce period and carryover concerns, and larger samples do not repair systematic timing or measurement errors. Sampling design also cannot eliminate uncertainty in concentration assignment or establish that the observed exposure range covers every clinically relevant situation.

Common Mistake