Posted as supplied by author.

Table A Checklist used to assess the RCTs on donepezil, rivastigmine and galantamine

Methodological considerations
Participants		Were inclusion and exclusion criteria clearly defined and strictly followed?
Randomization		Which method was used to generate and implement the random allocation sequence? Who assigned participants to the groups?
Stratification		Was patient allocation stratified according to the prognostic factors?
Blinding		Were the participants, those administering the interventions, those assessing the outcomes, and those analyzing the data blinded to group assignment? Was blinding effective?
Outcome measures		Were the primary and secondary outcome measures, including the tests used, the duration of the study, and the statistical methods applied for analysis clearly defined?
Significance level		Which significance level was defined and did the authors adhere to it?
Power		What was the planned power? Did the planned sample size meet the intended sample size?
Treatment		Were patients of intervention and placebo groups treated equally, apart from the drug given? Were they observed with the same attention?
Results
Baseline data	Were baseline demographic and clinical characteristics comparable between groups?
Multiplicity	Was multiple testing avoided or corrected for? Were all primary outcome measures still statistically significant when corrected according to multiplicity?
Adverse events	Were adverse events registered and reported adequately?
Intention-to-treat-analysis	Was analysis performed using the intention-to-treat (ITT) method? How many patients were excluded from "ITT" analysis? What results would emerge in a worst-case-scenario? How many patients were on treatment at the end of the trial?
Numbers analyzed	Are the numbers of participants in each group included in each analysis stated?
Outcomes and estimation	Are the outcomes for each primary outcome measure described including its confidence interval?
Consistency	Are data shown in text, figures and tables consistent? Does the abstract correctly summarize the results of the study?
Overall interpretation of the study
Methodology	Which impact did methodological insufficiencies have on the results? Can they be ignored or do they jeopardize the conclusions?
Clinical relevance	Are the observed differences between intervention and the placebo group clinically relevant? Was the duration of the study adequate with regard to the natural course of the illness?
Conflict of interest	Was the sponsoring of the study clear? Could it have influenced the interpretation of data?

Posted as supplied by author.

Table B Results reported in the 22 randomized controlled trials analyzed

Author and year

Characteristics of trial

Primary endpoint(s) based on ITT-population whenever possible

Dose

Treat-ment in weeks

Baseline MMSE (mean ± SD #)

ADAS-cog �

CIBIC-plus �

Percentage of patients

with benefit on CIBIC-plus (ChE-Inhibitor : placebo)

Other endpoints �

DONEPEZIL

Rogers 1996 (5)

1 mg

19.6

1.6 (p=0.11)

CGIC (sc.£ 4) 82:80 (p=0.5)

3 mg

18.6

2.1 (p=0.036)�

CGIC (sc.£ 4) 83:80 (p=0.5)

5 mg

18.0

3.2 (p=0.002)

CGIC (sc.£ 4) 90:80 (p=0.23)

placebo

18.2

Rogers 1998a (6)

5 mg

157

19.4 ± 5.0#

2.5 (1.3..3.6)

0.3 (0.08..0.50)

(sc.£ 3) 32:18 (p=0.004)Y �

10 mg

158

19.4 ± 5.0#

3.1 (1.9..4.2)

0.4 (0.13..0.55)

(sc.£ 3) 38:18 (p=0.0001)Y

placebo

153

19.8 ± 5.0#

Rogers 1998b (7)

5 mg

154

19.0 ± 5.0#

2.49 (p<0.0001)

0.36 (p=0.0047)

(sc.£ 3) 26:11 (p=0.0007)Y *

10 mg

157

18.9 ± 5.0#

2.88 (p<0.0001)

0.44 (p<0.0001)

(sc.£ 3) 25:11 (p=0.002)Y *

placebo

162

19.2 ± 5.1#

Burns 1999 (8)

5 mg

271

20 ± 5#

1.5 (0.5..2.4)&

(sc.£ 3) 21:14 (p=0.03)Y �

10 mg

273

20 ± 4#

2.9 (1.9..3.9)&

(sc.£ 3) 25:14 (p=0.0008)Y

placebo

274

20 ± 5#

Greenberg 2000 (9)

5 mg

21.8 ± 3.7

2.17 (0.20..4.10)

comment: cross-over-design

Homma 2000 (10)

5 mg

136

not reported

2.54 (p<0.001)

J-CGIC (sc.£ 3) 48:19 (p<0.0001)Y

placebo

132

adequately

Feldman 2001 (11)

10 mg

144

11.7 ± 4.8#

0.54 (0.29..0.79)&

(sc.£ 4) 63:42 (p<0.001)

146

placebo

12 ± 3.6#

Mohs 2001 (12)

10 mg

214

17.1 ± 2.9#

median time to clinically evident decline in function in days:

placebo

217

17.1 ± 2.9#

median (CI) 357 (280..¥ ) : 208 (165..252)

Tariot 2001 (13)

10 mg

103

14.4 ± 5.4

Neuropsychiatric Inventory � Nursing Home Version (NPI-NH): 2.6 (p>0.05)

placebo

105

14.4 ± 5.8

Winblad 2001 (14)

10 mg

142

19.4 ± 4.4

Gottfries-Bråne-Steen-Skala (GBS): 3.4 (p=0.054)

placebo

144

19.3 ± 4.5

AD 2000 2004 (15)

5 or 10 mg

12, 48 or more

283

19 (10-27)F

Relative risk of entry to institutional care 0.97 (0.72..1.30)

placebo

283

19 (10-26)F

Relative risk of reaching disability endpoint (BADLS): 1.02 (0.72..1.45)

Holmes 2004 (16)

10 mg

20.8± 3.8#

Neuropsychiatric Inventory, change in the total score: 6.2 (p=0.02)

placebo

21.1± 6.7#

comment: drug withdrawal study

RIVASTIGMINE

Agid 1998 (17)

4 mg

136

not reported

(CGIC sc.£ 2) 32:30 (p=0.5)

6 mg

133

(CGIC sc.£ 2) 43:30 (p=0.05)�*

placebo

133

Corey-Bloom 1998 (18)

1-4 mg

233

19.5

1.73 (0.64..2.82)Y

0.26 (0.03..0.49)�

PDS - 0.29 (- 2.16..1.58)

6-12 mg

231

19.6

3.78 (2.69..4.87)Y

0.29 (0.07..0.51)

PDS 3.38 (1.51..5.25)

placebo

235

Forette 1999 (19)

6-12 mg b.i.d.

not reported

adequately

1.8 (- 1.7..7.1)&

(sc.£ 3) 57:16 (p=0.007)

6-12 mg t.i.d.

4.8 (- 0.7..10.2)&

(sc.£ 3) 35:16 (p=0.13)

placebo

Rösler 1999 (20)

1-4 mg

243

not reported

- 0.03 (p>0.05)

0.14 (p>0.05)

(sc.£ 3) 30:20 (p<0.05)�*

PDS - 1.19 (p>0.05)

6-12 mg

243

1.60 (p<0.1)

0.47 (p<0.001)

(sc.£ 3) 37:20 (p<0.001)�

PDS 2.23 (p<0.1)

placebo

239

Potkin 2001 (21)

3-9 mg

not reported adequately

(sc.£ 4) 75:29 (p<0.03)

placebo

GALANTAMINE

Raskind 2000 (22)

24 mg

212

19.5 � 4.4#

3.9 (p<0.001)

(sc.£ 4) 73:57 (p<0.05)*

32 mg

211

19.1 � 4.4#

3.4 (p<0.001)

(sc.£ 4) 69:57 (p<0.05)�*

placebo

213

19.2 � 4.4#

Rockwood 2001 (23)

24-32 mg

261

19.7 � 3.2#

1.6 (p<0.01)

(sc.£ 4) 79:63 (p<0.01)

placebo

125

19.6 � 3.4#

Tariot 2000 (24)

8 mg

140

18.0 � 3.5#

1.3 (p>0.05)

(sc.£ 4) 53:49 (p>0.05)

16 mg

279

17.8 � 3.3#

3.1 (p<0.001)

(sc.£ 4) 66:49 (p<0.001)*

24 mg

273

17.7 � 3.3#

3.1 (p<0.001)

(sc.£ 4) 64:49 (p<0.001)*

placebo

286

17.7 � 3.4#

Wilcock 2000 (25)

24 mg

220

19.5 � 3.4

2.9 (p<0.001)

(sc.£ 4) 62:50 (p<0.05)�*

32 mg

218

19.0 � 3.8

3.1 (p<0.001)

(sc.£ 4) 65:50 (p<0.001)*

placebo

215

19.3 � 3.5

Wilkinson 2001 (26)

18 mg

18.8 � 2.8#

1.7 (p>0.05)

24 mg

18.2 � 3.0#

3.0 (p<0.01)

36 mg

18.8 � 3.7#

2.3 (p=0.08)

placebo

18.7 � 2.8#

Legend to table 2

N patients randomized

MMSE Mini-Mental State Examination, SD = standard deviation

ADAS-cog Alzheimer�s Disease Assessment Scale-cognitive Subscale

CIBIC-plus Clinician�s Interview-Based Impression of Change-Scale plus Caregiver supplied information

CGIC Clinical Global Impression of Change Scale, J-CGIC = Japanese version

PDS Progressive Deterioration Scale

b.i.d.; t.i.d. two treatments per day; three treatments per day

sc. score

* p>0.05 in worst-case scenario, i.e. all missing patients in the placebo group are assumed as improved and all missing patients in the intervention groups are assumed as worsened

� after consideration of worst-case scenario and correction for multiplicity p>0.05

� p>0.05 after correction for multiplicity

� mean differences between treatment and placebo (95% confidence interval or p value).

# Following the suggestion of the reviewer original baseline standard errors were transformed into standard deviations. This might have led to rounding errors. Transformed data are marked.

F Median and range

& Confidence intervals were not available in the original publication and were extracted from figures.

p values or confidence intervals were not directly available in the original publication and had to be calculated from data given in text or tables.