Increased statistical power with combined independent randomization tests used with multiple-baseline design.

Area of Science:

• Clinical Research Methodology
• Biostatistics
• Behavioral Interventions

Background:

• Physicians frequently evaluate treatment efficacy using small patient cohorts.
• Standard multiple-baseline designs (MBDs) with few subjects, such as the Wampold-Worsham (WW) method with four subjects, often exhibit limited statistical power.
• Existing methods may require extensive time commitments when increasing the number of subjects.

Purpose of the Study:

• To propose a novel MBD approach with enhanced statistical power.
• To develop a method that mitigates the time demands associated with larger sample sizes in MBDs.
• To improve the feasibility and efficiency of treatment effectiveness assessments.

Main Methods:

• Computer simulations were employed to estimate the statistical power of a design combining two four-subject MBDs.
• The power of the combined design was compared against standard four- and eight-subject MBDs.
• The impact of delayed linear treatment response on statistical power was investigated.

Main Results:

• The combined two four-subject MBD design achieved adequate power (>80%) for detecting standardized mean differences (SMDs) as low as 0.8.
• This combined approach identified smaller effect sizes (SMD=0.8) compared to a single four-subject MBD (SMD=1.3).
• Power was comparable to an eight-subject MBD (SMD=0.6) but with reduced study duration. Delayed linear responses significantly reduced power by 20-35%.

Conclusions:

• Combining two four-subject MBDs provides a statistically powerful and practical method for evaluating treatment effectiveness.
• This strategy allows for the detection of smaller, clinically relevant effect sizes (SMD=0.8).
• The proposed method offers a more timely and feasible study design compared to traditional approaches with larger sample sizes.