Comparison of cognitive tests and CSF biomarkers (logistic regression analysis).

*<p>p<0.05 compared with AUC for cognitive tests and AUC for CSF. Note that some variable are continuous and others dichotomous, which greatly affects the OR.</p><p>The Hosmer and Lemeshow goodness-of-fit test was >0.05 for all models, indicating a good fit of the model to the data.</p><p>Demographic variables entered in all models: age and sex; CSF variables entered: Tau, Aβ₄₂ or Aβ₄₂ dichotomised at <208 and P-tau; Cognitive test variables entered: MMSE (orientation & recall) and clock drawing dichotomised at <4. All were entered with the backward LR method.</p><p>AUC = Area under the curve; CI = Confidence interval; MMSE (O & R) = the orientation and delayed word recall parts of the MMSE; MCI-AD = MCI patients who later convert to AD; MCI-other dementias = MCI patients who later convert to a dementia other than AD; OR = Odds ratio.