Sample size

Sample size

Determining an appropriate sample size is a critical step in designing a randomized controlled trial (RCT). A well-calculated sample size ensures that the study has sufficient statistical power to detect a clinically meaningful effect, while also considering ethical, financial, and logistical constraints. This page outlines the main factors influencing sample size decisions in RCTs.

Statistical Considerations

Sample size determination starts with defining key statistical parameters. One of the most important is statistical power, typically set at 80% or 90%. This means the trial has a high probability of detecting a true effect if one exists. The corresponding Type II error rate (β) is 0.20 or 0.10, respectively. Alongside power, the significance level (α) is usually set at 0.05 for two-sided tests, controlling the risk of a Type I error—falsely declaring an effect when none exists.

Another essential parameter is the effect size, which reflects the minimum difference between groups that is considered clinically meaningful. This can be expressed as a standardized metric (e.g., Cohen’s d) for continuous outcomes or as an absolute risk reduction for binary outcomes. For instance, a 5 mmHg reduction in systolic blood pressure might be considered meaningful in a hypertension trial.

Outcome variability also plays a key role. For continuous outcomes, higher standard deviations require larger sample sizes to detect the same effect. For example, if the standard deviation of systolic blood pressure is 15 mmHg, more participants are needed than if the standard deviation were 5 mmHg.

The allocation ratio—how participants are divided between groups—also influences power. A 1:1 ratio is most efficient statistically, but unequal ratios (e.g., 2:1) may be justified for ethical or logistical reasons, with a modest increase in sample size to maintain power.

Study Design Considerations

Different study designs affect sample size requirements. Parallel-group RCTs are the standard and often require the largest sample size. In contrast, Cross-over trials reduce variability by having participants act as their own controls, generally requiring smaller samples. Cluster RCTs, where groups (not individuals) are randomized, require a larger sample size due to intra-cluster correlation—participants within the same cluster may respond similarly.

To adjust for clustering, the sample size must be inflated using the intraclass correlation coefficient (ICC). A higher ICC indicates greater similarity within clusters and a greater need to increase the sample size to achieve adequate power.

Adjustments for Missing Data and Attrition

Dropout and non-compliance are common in clinical trials and must be accounted for during sample size calculation. A simple adjustment involves dividing the calculated sample size by the expected retention rate:

To account for expected attrition, the calculated sample size should be inflated using the following formula:

n_final = n_calculated / (1 - dropout rate)

For example, if a trial requires 100 participants and expects a 10% dropout rate, the adjusted sample size would be:

n_final = 100 / (1 - 0.10) = 111

For example, if a trial requires 100 participants and expects a 10% dropout, the adjusted sample size would be 111 participants.

Ethical and Practical Considerations

While statistical precision is important, ethical and practical concerns must also guide sample size decisions. An excessively large trial can expose more participants than necessary to potential harms and increase costs unnecessarily. On the other hand, a trial that is too small may fail to detect important effects, wasting resources and participant time. Sample size should reflect a balance between scientific rigor and feasibility, considering recruitment capacity, study budget, and operational complexity.

In [[Adaptive trials, interim analyses may allow early stopping for efficacy or futility, which can reduce the total sample size needed.

Software for Sample Size Calculation

Several tools are available to support sample size estimation, including:

• R: Functions such as pwr and power.t.test()
• Stata: The sampsi command
• G*Power: A free, user-friendly tool for various power analyses
• PASS: Commercial software offering advanced features and flexibility

Conclusion

Sample size planning is a foundational element of trial design. It should integrate statistical power, clinical relevance, study design characteristics, expected attrition, and real-world feasibility. Thoughtful calculation helps ensure trials are ethical, efficient, and scientifically robust.

---

Bibliography

1. Julious SA. Sample Sizes for Clinical Trials. CRC Press; 2009. A comprehensive guide on calculating sample size for various trial designs.
2. Chow S-C, Shao J, Wang H. Sample Size Calculations in Clinical Research. 2nd ed. Chapman & Hall/CRC; 2008.
3. Wittes J. Sample size calculations for randomized controlled trials. Epidemiologic Reviews. 2002;24(1):39–53.
4. Jones SR, Carley S, Harrison M. An introduction to power and sample size estimation. Emergency Medicine Journal. 2003;20(5):453–458.
5. Piantadosi S. Clinical Trials: A Methodologic Perspective. 3rd ed. Wiley; 2017. Chapter 9: Sample size and power calculations.

---

Adapted for educational use. Please cite relevant trial methodology sources when using this material in research or teaching.