## GOLD v LLN Page 2

Secondly the use of a single cut off for FEV1/FVC of 0.7 to define the limit of normality for this index is also not supported by the ATS or ERS [4, 5]. The fact that the arbitrary definition may lead to an age bias in the diagnosis of COPD has already been presented [6] and we will now explore the reason behind this in more detail.

The figure to the right shows a plot of the FEV1 values from a sample population of normal men of the same age and height with the average value found being 3.5 L and the standard deviation of the spread of results is 0.5 L. For a subject with FEV1 of 2.678 L this is 1.645 SD below the predicted mean, that is 3.5 - (1.645 x 0.5), and means that 5% of the normal population have FEV1 values lower than this. The grey shaded area in the graph is 5% of the total area under the graph.

If we define the Lower Limit of Normal (LLN) as 1.645 SD below the predicted in effect we are accepting a false positive rate of 5% in identifying subjects as abnormal. Most subjects referred for spirometry have an a priori increased probability of an abnormality because of either symptoms, signs or exposures so the lower 95% confidence limit is used as the LLN. This increases the sensitivity of the test with some loss of specificity. If spirometry is being undertaken on people where the a priori probability of an abnormality is not increased, such as screening the general population, we should take the LLN as 1.96 SD below predicted and only accept a false positive rate of 2.5%. The ATS and ERS endorse using the LLN as 1.645 SD below predicted [4, 5].