Sample size calculation using Bland and Altman

[Mads Lumholdt]

Hello stata community,

This is my first post here. I have searched forum posts for an answer to this problem, but could not find anything useful.

I want to calculate required sample size for a method comparison study. A new blood gas analysis method has beed developed and pilot studies have compared agreement between the normal blood gas analysis method and the new method using bland and altman's limits of agreement. I found the following article during a search on pubmed which describes sample size calculation for agreement studies: Sample Size for Assessing Agreement between Two Methods of Measurement by Bland−Altman Method

Pubmed id: 27838682

Is it possible to calculate sample size using this method in State?


Sincerely
Mads Lumholdt

[Marcos Almeida]

Welcome to the Stata Forum | Statalist.

Your query is interesting. However, it demands an extra level of effort of the reader, since we still need to browse on the Web in order to know the parameters used to do the estimates.

I gather the parameters are the crucial issue here.

[Mads Lumholdt]

I apologize. Parameters for comparison are pH, carbondioxide, and oxygen in venous samples and arterial samples, respectively. All floating variables. I am most interested in carbondioxide and plan to calculate sample size using this parameter. As I understand a maximum limit of agreement should be predefined. I plan to use +/- 1 kilopascal. Does this clarify my request?

[Nick Cox]

I think Marcos' query was not about your data (note that measured variables are not parameters in the statistical sense in any case) but about what that paper recommends. Most such calculations seem to involve plugging numbers into about one line of algebra, although I have not read the paper.

Please see FAQ Advice https://www.statalist.org/forums/help#references for our standards on references. A fuller reference is

Lu, M., Zhong, W., Liu, Y., et al. (2016). Sample Size for Assessing Agreement between Two Methods of Measurement by Bland−Altman Method. The International Journal of Biostatistics, 12(2), doi:10.1515/ijb-2015-0039

and I find that the journal wants USD 42 from me for access, so no go there.

[Marcos Almeida]

You are fully right, Nick. That's what I meant by demanding information about the parameters. I gather the "sample size for assessing agreement" eventually boils down to a process where one enters the parameters and gets the estimated sample size.

After all, means, difference between continuous variables and SDs are part and parcel of Bland-Altman plots.

That said, I'm not in the spenthrift mood as well, hence I'll wait and see if Mads wishes to clarify the setting, as recommended in the FAQ.