Visual analogue scales (VAS) and numerical rating scales are commonly used to assess breathlessness in patients with cancer.1-3 Their repeatability in this situation has not, however, been assessed in a way that allows calculation of the sample size required to design studies with sufficient power to detect a change in breathlessness. We asked 31 patients with breathlessness due to cancer to complete a 100 mm VAS and a numerical rating scale (from 0 to 10) on three occasions, twice on one day one hour apart and then again after a mean (range) of 2 (1–8) days. Patients were asked to rate their breathlessness “right now” and the worst and average severity of breathlessness and the degree of bother it had caused in the preceding 24 hours. Identical words were used to anchor the ends of both scales—for example, “not breathless at all” and “breathlessness as bad as you can imagine”.

Sample sizes were calculated using the following equation4:

where ς=the standard deviation of the difference between tests, δ=a specified reduction in the mean, Z=the critical values from the Normal distribution to achieve the selected levels of significance and power, α=level of significance, and β=the power of the test. Table 1 illustrates the sample sizes required to have a reasonable chance of detecting the effect of an intervention that reduced breathlessness scores over a few days in patients with cancer by 25% (p = 0.05; power = 90%).

The sample size required, reflecting the repeatability, varies with the particular aspect of breathlessness being measured. The numerical rating scale was a more repeatable measure than the VAS. We conclude that many studies that have examined the effect of an intervention on breathlessness in patients with cancer have not had sufficient power to detect a 25% change following the intervention.