High correlation between salivary cortisol and free serum cortisol measurements does not mean good agreement

摘要

To the Editor:I read with great interest the study by Estrada-Y-Martin and Orlander published recently in CHEST (November 2011). According to the study, salivary cortisol can replace free serum cor-tisol in patients with septic shock because of a good correlation between these variables. However, the correlation coefficient is of no practical use in the statistical analysis of comparison data. Indeed, the high correlation between salivary cortisol and free cortisol measurements does not mean that the two methods agree. The correlation coefficient in this case depends on both the variation between individuals and the variation within individuals (measurement error). Therefore, the correlation coefficient will partly depend on the choice of patients; if the variation between patients is high compared with the measurement error, the correlation will be high, whereas if the variation between patients is low, the correlation will be low no matter how good the agreement between the two methods. Consequently, when measurements are taken by two methods on a group of patients, the variation between patients will introduce a strong correlation between the two sets of measurements. This correlation is, of course, irrelevant for the method comparison. Thus, it is quite wrong to infer from high correlation that the salivary cortisol and free serum cortisol measurements may be used interchangeably.