Webinar Q&A Effective Trial Design: Clinical Trials in Small Patient Populations

Designing and conducting clinical trials in indications with inherently small patient populations, such as rare disease, can be quite difficult. The limited number of patients available for the trials has a direct impact on the difference in clinical outcomes that can be detected between treatment groups. Another challenge is the extended time to recruit the necessary number of patients.

On November 17, we hosted a complimentary webinar titled “Effective Trial Design: Clinical Trials in Small Patient Populations” to present several alternatives for small population study design to help overcome challenges listed above.

The webinar was conduced by Thomas Zwingers, Statistical Consultant of CROS NT. He has been working in the clinical trial environment since 1980. He is specialised in statistical analysis and reporting with particular expertise in Adaptive Trial Design. Prior to joining us at CROS NT, Thomas ran his own biometrics consulting CRO in Germany for over 20 years.
Thomas holds a Master’s degree in Cybernetics from the Technical University of Munich. He is member of numerous international associations, including the Biometric Society, International Society for Clinical Biostatistics and the Society for Clinical Trials. He has written and collaborated on over 100 publications in the pharmaceutical and clinical trial statistics.

Continue reading to discover the questions from the webinar’s Q&A session answered by the speaker.

What method should be used when the trial intervention is a permanent treatment (e.g. transplantation)?

When the trial intervention is a permanent treatment, all designs that use the patients as their own control cannot be applied. The reason is that obviously the condition of the disease is not comparable between treatment periods. If historical controls are available, methods that include historical controls into the control group might be of advantage over design that use concurrent controls alone.

Can you say something about the use of enrichment designs for clinical trials in small populations?

Enrichment designs are used to find a patient population which is sensitive to the treatment. It assigns more patients to a promising subgroup within a trial including a broader population in order to improve the power of the study for this subgroup.
They do not in general solve the problem of low patient numbers in small populations.

What about Bayesian analysis?

Bayesian designs play an important role in trials for small populations, because in contrast to Frequentist statistics, they do not give “yes”/”no” answers, but calculate the probability distribution of the effect size. This may be more useful in a “go”/”no go” decision in a drug development program than a “significant”/”not significant” outcome of an analysis.
Bayesian analysis methods are often used in trials designs that include historical controls.

Can you please explain about the washout period?

The wash-out period is included into cross-over designs in order to prevent so called carry-over-effects. A carry-over-effect is a condition that prevents the state of disease the be comparable between treatment periods and thus introduces bias into the results.

Can you please explain a bit more the sample size calculation and the power of the trial?

In general, sample size calculations are based on the idea that there is an inverse relationship between the sample size and the effect size, i.e. the smaller the effect size, the larger the sample size. The effect size is defined as the proportion of the difference to the deviation of the measurement.

Thus, the smaller the difference between treatment and control group for the endpoint is and the larger the standard deviation is, the more difficult it is to detect a difference and the required sample size is increased.  All formulas for sample size calculation also include the probabilities of false-positive and false-negative conclusions. There is a common agreement that the probability of a false-positive conclusion from a trial should not exceed 5% and that the probability of a false-negative conclusion should not exceed 20%. The lower these probabilities are chosen, the larger will be the required sample size, because the uncertainty of the result will be decreased.

Can the carry-over effect be minimised using a double cross-over design (a cross-over followed by another cross-over)?

No, the carry-over effect is related to the treatment and is not affected by the study design. A double cross-over enables you to distinguish between carry-over and period effects, but does not minimise the effect.

Is a double cross-over design used or not in clinical trials for small populations?

From my experience, double cross-over designs are very rare. It generates repeated measures and can be used to estimate the intra-patient variability.

How much do I save with the presented designs, especially with the cross-over design?

Cross-over designs save about 50% of the patients compared to a classical 1-period randomised trial.

Could you please repeat on the concept of N-of-1 design?

The basic concept of N-of-1 trials is to have each patient as his own control like in cross-over designs. The original motivation was to determine the actual benefit of a treatment in an individual patient opposed to group derived measures of efficacy. Usually different treatment sequences are randomised within the individual patient.

How do I analyse N-of-1 trials?

Data from multiple N-of-1 trials are combined to produce estimates of treatment effect at a population level using Bayesian hierarchical models.

How to analyse and calculate sample size when we have only one arm, so we do not have a delta in the formula?

Single-arm trials are usually designed to estimate the treatment effect. Therefor only the confidence interval of the estimated effect can be used to determine a sample size. Most of the commercially available sample-size calculation programs include this option.

Could you please explain why ANCOVA is the best statistical method to be used in small populations?

ANCOVA is not only the preferred method in small populations compared to a simple “change from baseline” analysis, but for all analyses, because this statistical method can use patient characteristics which are important for the response to calculate adjusted estimates.

Why should I know the natural history of a disease?

In rare disease very often information on the natural history of the disease is limited. This information is necessary to determine the appropriate trial design, e.g. does the disease change over time and if yes, how fast does this happen.

Are the presented methods accepted by the regulatory authorities?

No general statement on the acceptance of the described methods is possible. Regulatory authorities are open for “unconventional” designs in small population studies, if they are justified. It is recommended to discuss such a design upfront with the authorities to align on the acceptance.

Do you have more questions about designing clinical trials in small populations? Contact us and we will get back to you as soon as we can!