We evaluated the impact of selection of reference values on the categorisation of measured maximal oxygen consumption (V'O2peak) as "normal" or "abnormal" in an ageing population. We compared measured V'O2peak with predicted values and the lower limit of normal (LLN) calculated with five equations. 99 (58 males and 41 females) disease-free subjects aged ≥70 years completed an incremental maximal exercise test on a cycle ergometer. Mean V'O2peak was 1.88 L·min-1 in men and 1.26 L·min-1 in women. V'O2peak ranged from 89% to 108% of predicted in men, and from 88% to 164% of predicted in women, depending on the reference equation used. The proportion of subjects below the LLN ranged from 5% to 14% in men and 0-22% in women, depending on the reference equation. The LLN was lacking in one study, and was unsuitable for women in another. Most LLNs ranged between 53% and 73% of predicted. Therefore, choosing an 80% cut-off leads to overestimation of the proportion of "abnormal" subjects. To conclude, the proportion of subjects aged ≥70 years with a "low" V'O2peak differs markedly according to the chosen reference equations. In clinical practice, it is still relevant to test a sample of healthy volunteers and select the reference equations that better characterise this sample.