Bias due to the use of % predicted in adults
Let us consider the following two males:
Height (m) |
Age (yr) |
Predicted FEV1 (L) |
80%predicted (L) |
Observed FEV1 (L) |
SDS (Z-score) |
1.70 m |
70 |
3.01 L |
2.41 L |
2.35 |
-1.29 |
1.85 m |
25 |
5.12 L |
4.10 L |
4.10 |
-2.00 |
In the youngest of the two, according to the rule of thumb ‘over 80% of predicted is within normal limits’ an FEV1 of 4.10 liter would be regarded as satisfactory; it should be noted, however, that such a value occurs in only about 2% of healthy subjects. In the elderly man an FEV1 of 2.35 L would have been regarded as abnormally low, whereas this would be frequently observed in a population of healthy subjects. An analogous situation occurs in women.
This problem may also affect scientific projects. Just consider this example:
Problem |
assess the effect of drug X on FEV1 in patients with airway obstruction |
Method |
double blind intervention trial of drug X and placebo |
Subjects |
subjects with airway obstruction defined as FEV1< 80% predicted |
Results |
on average the effect of drug X is smaller in the elderly subjects than in young patients |
Conclusion |
the ‘reversibility’ of airflow limitation is less in elderly subjects, which appears plausible to you; the drug is less effective in elderly subjects |
Comments:
due to the ’80% predicted criterion’ each group comprises
elderly and short subjects who in effect do not have airway obstruction,
whilst younger and taller subjects with mild airway obstruction
have been eliminated by the selection strategy. Who knows the conclusion
might have been the same had a more correct selection criterion
been used, but the present selection criterion introduced a bias
towards overrepresentation of elderly and short subjects and underrepresentation
of young subjects into the material which may (at least in part)
explain the findings.
One can graphically depict the ‘lower limit of normal’ based on either a fixed percentage (e.g. 80%) of the predicted value, or the 5th percentile. Each criterion may lead to different conclusions for the same subject, and this varies with age and standing height, but also with gender and ethnic group.
In keeping with this Hardie et al. [1] reported how the application of the GOLD (Global
Initiative for Chronic Obstructive Lung Disease) criteria for defining
stage I of COPD (FEV1%FVC < 70%
and FEV1 % predicted > 80% leads
to 35% of healthy, elderly nonsmoking subjects being wrongly classified
as having COPD stage I. In those aged > 80 yrs this increased to
50%, and approximately one-third would be classified as having
stage II COPD.
Such
results, as explained above, are perfectly predictable and arise
from the arbitrary and unscientific definition of cut-off criteria.
Such criteria do the patient no good at all and potentially lead
to over-diagnosis and over-treatment. The latter is only of benefit
to the pharmaceutical industry.
1. Hardie JA, Buist AS, Vollmer WM, Ellingsen I, Bakke PS, Mokve O. Risk of over-diagnosis of COPD in asymptomatic elderly never-smokers. Eur Respir J 2002; 20: 1117-1122.