Clinical trial sample size calculation

Eg, blood pressure reduction mmHg , weight loss kg. Anticipated Means Group 1 The anticipated mean and standard deviation of group 1 eg, mmHg, 75 kg Standard deviation is determined by examining previous literature of a similar patient population.

Anticipated Mean Known population The anticipated mean and standard deviation of a known population eg, mmHg, 75 kg The mean and standard deviation are determined by examining previous literature of a similar patient population. Most medical literature uses a value of 0.

Sample Size Group 1 Group 2 Total View Power Calculations. Rosner B. Fundamentals of Biostatistics. Follow Us! Get Email Updates. Woodward M. Formulae for sample size, power and minimum detectable relative risk in medical studies.

A simple approximation for calculating sample sizes for comparing independent proportions. A sample survey is planned to test, at the 0. Woodward M Epidemiology Study Design and Data Analysis. Suppose that equal sized samples will be taken in each year i. Our test is to have a power of 0. The standard deviation of serum cholesterol in humans is assumed to be 1. International Agency for Research on Cancer; A matched cohort study is to be conduct to quantify the association between exposure A and an outcome B.

Assume the prevalence of event in unexposed group is 0. In order to detect a relative risk of 0. Our approach is based on Chapters 5 and 6 in the 4th edition of Designing Clinical Research DCR-4 , but the material and calculators provided here go well beyond an introductory textbook on clinical research methods.

Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH. Kohn MA, Senyak J. For this reason, several different scenarios are often examined. Consider the example of the antihypertensive study and the Figure. It is clear that an increase in scatter leads to an increase in required sample size.

A reduction in the level of significance also leads to an increase in required sample size, as this reduces the probability of mistakenly demonstrating the effect. Nevertheless, the level of significance may not be varied for the sake of sample size planning. Other relationships of this type are demonstrated in Table 2 for the unpaired t-test. In addition, it is important to bear in mind that the difference to be detected should also be clinically relevant.

The clinical investigator regards the 5 mm Hg greater reduction in blood pressure with drug B as being clinically relevant. However, if the effect expected in the study is too low, then the benefit of the study may be doubted. In such a case, even statistically significant results may be irrelevant 7. One important point in sample size planning is to consider losses to follow-up or drop-outs If, for example, it must be assumed that adequate data cannot be collected for a proportion of the volunteers in a study—for whatever reason—, the sample size must be proportionately increased.

The necessary increase in the sample size depends on the estimated rate of participation and the study conditions. It must, nevertheless, be pointed out that these adjustments may influence the representative character of the data and generally lead to biased results. This must also be considered when planning the study. Explicit formulae are available for calculating sample sizes for the most common tests 12 — Machin et al. It is advisable to use a validated program—such as one of the above.

Planning the sample size of a clinical study requires prior information. The type of prior information depends on the statistical methods which are to be used. If the desired parameters cannot be estimated, it may be desirable to perform a pilot study in advance, in order to estimate the appropriate population parameters. In any case, the expected effect should be at least as large as the minimal clinically relevant effect.

The size of the study group s have to be determined even for exploratory or descriptive studies 1 , so that the precision of the parameter estimates will not be excessive. If there is no sample size planned, this indicates that the quality of the study is poor. Sample size planning for a clinical study is based on an estimate from prior information, which may be of different precision in different studies.

This should be considered when interpreting the results. If the treatment effect is overestimated during the planning phase, this usually leads to an excessively small sample size.

The observed treatment effect may then not be significant—but only because the sample size is too small. Sample size planning must also include the procedures for dealing with missing values and with patients who leave the study. We have only been able to consider a few aspects of sample size planning. There are additional aspects, which may be important with specific study designs. For example, the method of sample size planning may be different if a clinical study is to include a test for superiority, non-inferiority, or equivalence Non-inferiority studies may require really large sample sizes, as the mean difference to be detected is often specified as the smallest relevant clinical difference, which then acts as the non-inferiority limit.

This is usually much smaller than the actual mean difference. It often happens that a data set is used to test several hypotheses. The problems of multiple testing must be considered during sample size planning. For this reason, only a single main question to be answered is often specified. Moreover, the sample size is not always totally specified in modern studies.

For example, an adaptive design can be used. The sample size may then be influenced or controlled during the study, in accordance with a scheme which is strictly specified in the planning phase. However, this procedure necessitates careful and statistically demanding planning and should never be performed without the support of an experienced biometrician.

As sample size calculation is so complex and has such important consequences, collaboration is desirable between experienced biometricians and physicians. The quality and validity of studies can be greatly improved if all important details are planned together 2 , 3 , Sample size planning requires the expert knowledge of clinicians or physicians, who provide an estimate of the relevant effect. Sample size planning depends on the planned method of statistical evaluation and thus on the medical question to be answered.

The chances of success in a clinical study and the quality of the research results are highly dependent on sample size planning. Sample size planning should always be carried out in collaboration with an expert statistician or biometrician.