Skip to main content

Table 1 Predictors of response from a stepwise logistic-regression model on combined ‘improver’ criteria and combined ‘non-decliner’ criteria

From: Responder analysis of a randomized comparison of the 13.3 mg/24 h and 9.5 mg/24 h rivastigmine patch

Effect Maximum likelihood estimates
Parameter Week 24 Week 48
Estimate a Pr > ChiSq Estimate a Pr > ChiSq
‘Improvers’
Treatmentb Treatment 9.5 mg/24 h patch −0.580 0.0001 −0.461 0.021
Gender Female 0.234 0.137 −0.040 0.014
ADAS-cogc ADAS-cogc 0.044 0.017 −0.438 0.059
ADCS-IADLc ADCS-IADLc −0.026 0.050 na na
MMSEc MMSE ≤12c −0.854 <0.0001 na na
‘Non-decliners’
Treatmentb Treatment 9.5 mg/24 h patch −0.337 0.002 −0.270 0.058
Gender Female 0.218 0.094 na na
ADAS-cogc ADAS-cogc 0.031 0.014 na na
ADCS-IADLc ADCS-IADLc na na −0.027 0.020
MMSEc MMSE ≤12c −0.497 0.002 −0.351 0.040
Weight, versus >80 kg Weight 50 to 80 kg −0.363 0.068d −0.405 0.119d
  Weight <50 kg 0.208 0.344
  1. aPositive estimate indicates increased odds of response when increasing the value of the covariate, or when in the given category. Negative estimated effect means reduced odds of response. bTreatment effect compares 9.5 mg/24 h to 13.3 mg/24 h rivastigmine patch. cAt DB-BL. dAll categories combined. Candidate covariates were: treatment, gender, ADAS-cog score at DB-BL, ADCS-IADL score at DB-BL, MMSE score of ≤12 at DB-BL and weight category (<50, 50 to 80 and >80 kg). ‘Improver’ = patient with an improvement from DB-BL in ADAS-cog score ≥4 points and a change from DB-BL in ADCS-IADL score ≥0 points; ‘non-decliner’ = patient with a change from DB-BL in ADAS-cog score ≤0 points and a change from DB-BL in ADCS-IADL score ≥0 points. ADAS-cog, Alzheimer’s Disease Assessment Scale–cognitive subscale; ADCS-IADL, Instrumental domain of the Alzheimer’s Disease Cooperation Study–Activities of Daily Living scale; DB-BL, double-blind baseline; MMSE, Mini-Mental State Examination; na, not applicable because effect not selected as relevant by the stepwise regression model – only variables (effects) with a P-value <0.15 were included in the final model. Treatment was always retained within the model; Pr > ChiSq, P-value of the Chi-Square test.