Nomogram for sample size calculation in assessing validity of a new method based on a regression line

Abstract

The validity of a newly developed diagnostic method is usually proven by comparing with a well-grounded reference method. When measurements from a new method are continuous but in different units from a reference standard and having a linear relationship, validity can be usually assessed by Pearson correlation coefficient, but it does not provide clinical guidance for judging validity. We defined a limits-of-agreement derived from regression models for assessing validity of new method, and developed a sample size formula. The sample size formula to achieve a certain probability that the limits-of-agreement is within a pre-defined, clinically acceptable range [-delta, delta] was derived and the result is presented as a nomogram. When a ratio of upper bound of a limits-of-agreement to delta is expected to be 0.95, a sample size of approximately 300 achieves a 90% probability that the limits-of-agreement lies within +/- delta. The simulation showed that the suggested sample size formula had the targeted coverages. The sample size determination based on a limits-of-agreement is practical for showing validity of new methods, measuring the same attribute but in different units, and the presented nomogram is useful.