Continuous and bimonthly publication
ISSN (on-line): 1806-3756

Licença Creative Commons
7951
Views
Back to summary
Open Access Peer-Reviewed
Educação Continuada: Metodologia Científica

What is the importance of calculating sample size?

Qual a importância do cálculo do tamanho amostral?

Cecilia Maria Patino1,2, Juliana Carvalho Ferreira1,3

DOI: http://dx.doi.org/10.1590/S1806-37562016000000114

PRACTICAL SCENARIO



In a controlled, randomized clinical trial on the management of asthma in pregnant women, researchers evaluated the effect of implementing a program in which a portable device was used for asthma control, as assessed by the Asthma Control Questionnaire (ACQ). In clinical research, our goal is to make an inference regarding something about a population by studying a sample of that population. This sample has to be representative of the target popu-lation, and the number of participants must be appropriate. It should be large enough that the probability of finding differences between groups by mere chance is low and that of detecting true, clinically significant differences is high. However, the number of participants should not be so large that resources are wasted or participants are exposed to unnecessary risk. Therefore, in the study design phase, it is essential to perform sample size calculation. To perform this calculation, one must define the key characteristics of the study, such as the study design, the primary endpoint, the expected variability, the degree of certainty desired, and the predicted number of participants who will drop out of the study. To define these parameters and calculate the ideal sample size, we need to obtain deep knowledge of the field of research in question by reviewing the literature and biostatistics.



In our example, the researchers tested the effect that using a new device had on asthma control (the primary out-come) compared with the usual treatment. They estimated that the difference between the groups would be 0.55 points on the ACQ, with a standard deviation of 0.66 points, a power of 80%, and a level of significance of 5%. In addition, they estimated that 25% of the participants could be lost to follow-up. Using those data, the authors calculat-ed that they needed to include 72 participants. At the end of the study, the researchers analyzed the results of 69 participants and showed that the new intervention improved asthma control in pregnant women.



BASIC CONCEPTS



Power



In biostatistics, power is defined as the probability of obtaining a statistically significant result when there is a real difference between treatments. In general, a power of at least 80% is needed in order to ensure a high probability of observing the effect, if any, of the intervention. To increase the power to detect differences, it is necessary to increase the sample size (Figure 1).

 






Critical level of significance



The critical level of significance is usually ≤ 5%. If we want greater certainty that a difference observed in the study population is not coincidental, we need to increase the sample size.



Effect size and variability



The greater the effect of the new intervention on the outcome is, the smaller is the sample size needed in order to prove it. Conversely, to show smaller effects, it is necessary to increase the sample size. If there is great variability of the effect in the population, we will also need a larger sample size (Figure 1).



It should be borne in mind that the sample size calculation is based on estimates and assumptions that can be inac-curate and is therefore subject to error. It is also important to be realistic when choosing the estimates employed in calculating the sample size. Highly optimistic choices about the effect size increase the risk of calculating an insuffi-cient number of participants for the sample, whereas highly pessimistic choices can make the study unviable by resulting in a sample size that is too large to be practical.



RECOMMENDED READING



1. Zairina E, Abramson MJ, McDonald CF, Li J, Dharmasiri T, Stewart K, et al. Telehealth to improve asthma control in pregnancy: a randomized controlled trial. Respirology. 2016 Mar 31. doi: 10.1111/resp.12773 [Epub ahead of print]

2. Hulley SB, Cummings SR; Browner WS; Grady DG. Delineando a Pesquisa Clínica. 3rd ed. Porto Alegre: Artmed; 2008.

3. Ferreira JC, Patino CM. What does the p value really mean? J Bras Pneumol. 2015;41(5):485. http://dx.doi.org/10.1590/S1806-37132015000000215

Indexes

Development by:

© All rights reserved 2024 - Jornal Brasileiro de Pneumologia