Many continuous variables should be analyzed using the relative scale: a case study of β2-agonists for preventing exercise-induced bronchoconstriction

Table 2 Comparison of the intercept and slope to explain the effect of β₂-agonists by IPD

	Intercept	Slope	χ²(3 df) *	P *	AIC **
Uniform effect (absolute scale)	27.7	–			1638.3
Intercept and slope	7.9	−0.69	82.3	10⁻¹⁷
Slope (relative scale)	–	−0.90	13.9	0.003	1576.8

The intercept and the slope were calculated with a linear mixed-effects model using the type of β₂-agonist and study as grouping variables. The intercept alone indicates that β₂-agonists decrease the post-exercise FEV₁ decline by an absolute effect of 27.7 pp. (95% CI: 22.1 to 33.4 pp). The slope alone indicates that β₂-agonists decrease the post-exercise FEV₁ decline by a relative effect of 90% (95% CI: 72% to 109%). Abbreviations: pp., percentage points
* Anova test comparing the model on the row with the model on the row above
** AIC, Akaike information criterion. AIC estimates the relative quality of statistical models, lower AIC value is preferable

ISSN: 2046-4053