The percentage of publications that do not use ‘per cent predicted’ for spirometric indices, is negligible. This is because this usage is copied unthinkingly; in spite of its widespread adoption per cent predicted is not scientifically founded (see ref. 1). In fact, ‘Nowhere else in medicine is such a naive view taken of the limit of normal’ (Sobol). Similarly it is inappropriate to express a change in spirometric indices, for example due to an intervention such as bronchodilatation, as a percentage of the initial value because this introduces bias due to regression to the mean. The indiscriminate use of 80% of predicted as the ‘lower limit of normal’ (LLN) cannot be justified.
When would æ% of predicted valueÆ be correct? Simply, when the scatter is proportional to the predicted value. In other words, if in a reference group a low predicted value is associated with a small scatter, and a high predicted value associated with a proportionally larger scatter. The figure illustrates the relationship between scatter in FEV1 and age in asymptomatic lifelong non-smoking males. The figure also shows the regression line (solid line) and the lower limit of normal (5th centile) for predicted values in these individuals. There is no sign whatsoever that the scatter is small in persons with a low FEV1, and proportionately higher in subjects with a high FEV1.
See also: Bias due to percent predicted in adultsScatter not proportional to levelInterpretation of results: Use of percent of predicted
Ref. 1 - Do not use per cent predicted