QT and confidence in Phase I

E14 calls for a study design that will measure the drug’s effect on QTc (QT corrected to allow comparison of QT measured at different heart rates), reporting the mean delta-delta QTc (ddQTc) and 95% upper confidence interval at each analyzed time point. And the study population must have an expected QTc response to positive control drug exposure. If, upper confidence interval after study drug exceeds 10 msec, you have achieved a “positive” TQT study result. There may be a mean of only, say, 2 msec; but if the confidence interval is too wide, the study is “positive” for significant QT effect. And “positive” may prove fatal to the candidate drug.

Key question

What can be done to narrow the confidence interval around the mean QTc, at each analyzed time point?

The selected ECG core lab contributes heavily to TQT study design. That’s entirely appropriate. But if you asked above why the Phase I team needs to worry about your QT study (“Can’t they just read the protocol and run the study?”), please read on.

Designed N

Generally, a confidence interval narrows as N – the number of evaluable participants – increases. But N is expensive. The statistician’s calculation of N depends on your need for confidence in the study’s results – the specified level of study power. It also depends on the expected magnitude of the drug’s true ddQTc effect and on the expected intra-individual QTc variation:

• Recent advances in cardiac waveform analysis – additional to the established QT analytic approach – promise to increase confidence in studies of fewer participants.

Prior QT data

Were QT data obtained in the earliest single-ascending-dose studies? Such data can be inexpensively acquired and saved for possible later QT analysis. If MTD was exceeded in the study, the tested dose(s) higher than MTD will never be studied again but might be expected to suggest any real QT effect. It’s possible that QT analysis may suggest a mean effect on ddQTc, and the assumption of a non-zero mean effect would significantly affect the statistician’s calculations of N.

Similarly, a “pilot” QT study can be useful, preceding the TQT study:

• The pilot study can demonstrate to sponsor and FDA that a particular proposed supratherapeutic dose is or is not appropriate for TQT study. Is there a dose above which drug is not tolerable or interferes with study objectives – e.g., sedation causing participants to doze off during “extraction windows” of their continuous ECG data acquisition, or drug causes GI effects, with risk that autonomic changes may affect QT intervals?
• A mean drug effect may be suggested and may inform choice of N for the definitive TQT study.
  
• The investigative site’s study performance can be evaluated.

Evaluable N

At each QT analysis time point, the real N will be determined by the number of participants retained in study whose QT data can be evaluated. As N declines, the confidence interval widens.:

• Retain enrolled participants in the study. Follow up on good enrollment decisions by treating the participants well. With few procedural exceptions, their time in-house should be as close to “spa” or “summer camp” as possible.
• Establish representative (averaged) baseline safety ECG measures. Avoid basing a post-dose delta-QTc calculation (and related decisions about withdrawal from study) on a single, possibly invalid baseline QTc measurement.
  
• Ensure data quality in all extraction windows. Recognize that the standard continuous 12-lead Holter recording, while a major improvement over site-selected 10-second ECGs, can be compromised by events (e.g., lead malfunction or muscular artifact) that are not readily observed or require disturbance of the study participant to observe. Do not rely on using an alternate lead (e.g., V5 instead of II) for QT measurements; the apparent QT varies considerably between leads, and ddQTc calculations should ideally be done only with data obtained from the same ECG lead. 12-lead telemetry is the ideal method for real-time ECG data monitoring and quality control. It allows review of all 12 leads of each participant’s ECG just prior to and during each extraction window, so staff can intervene to ensure acquisition of clean data.

QT-RR variability

All methods of correcting QT for the effect of heart rate depend on a somewhat predictable relationship between QT and RR. Increased variation in the QT-RR relationship causes increased variation in ddQTc, and increased confidence intervals around the mean ddQTc at each time point.

This is the study site’s place to shine. Confidence intervals will narrowed by each of these measures:

• Select participants who display calm behavior at screening and agree that they will tolerate a quiet, controlled – boring – environment during ECG acquisition.
  
• Select participants with low potential for variation in the QT-RR relationship. Inclusion/exclusion criteria often give the Investigator discretion to judge significance of screening and post-admission ECG findings. Volunteers with normal ECGs may need to be excluded – e.g., with exaggerated sinus arrhythmia creating increased QT-RR hysteresis.
  
• Discuss with participants the importance of the study, the small magnitude of QT change that can be significant, and the sensitivity of the QT measurement to many controllable stimuli. They can be surprisingly motivated to support the study’s data integrity.
  
• Use telemetry monitoring between admission and enrollment. Frequent premature beats are often not seen on isolated 10-second ECG tracings but are seen on telemetry. At best, such dysrhythmias challenge the ECG core lab to extract ECGs showing only sinus rhythm. At worst, increases in QT-RR hysteresis are unavoidable, and confidence intervals widen.
  
• Adopt protocol or procedure modifications to minimize changes in QT-RR relationship during extraction windows. Most importantly, avoid stimuli that may affect autonomic/vagal activity. Control the study environment – including music, movies/videos, TV. (They were warned this could be boring!). Consider separating the dose times for IV investigational drug (or its placebo) and oral administration of positive control (or its placebo), to avoid a gastrointestinal stimulus coincident with IV study drug administration. Require position changes (e.g., sitting vs supine) between extraction windows, to avoid excessive response to postural change after prolonged supine position. Serve meals and prompt visits to toilet immediately after extraction windows, to maximize time for rest/equilibration before subsequent windows. Substitute 12-lead telemetry acquisition of safety ECGs, instead of using roll-up ECG machines, to minimize procedural disturbances (approaching, conversation, uncovering chest, attaching leads).

A summary plea:

Give your Phase I team time to review and suggest revisions of protocol. Consider approaches described above, and more. Worry less.

Royce Morrison, MD is the Clinical Strategist at Charles River Laboratories. In this role, he is charged with developing new product lines for Phase I Operations. Charles River Laboratories based in Wilmington, Massachusetts, partners with global pharmaceutical and biotechnology companies, government agencies and leading academic institutions to advance the drug discovery and development process, bringing drugs to market faster and more efficiently.

About Charles River Laboratories

Charles River Laboratories based in Wilmington, Massachusetts, partners with global pharmaceutical and biotechnology companies, government agencies and leading academic institutions to advance the drug discovery and development process, bringing drugs to market faster and more efficiently. Charles River’s 8,300 employees serve clients worldwide. For more information on Charles River, visit our website at www.criver.com.