The 11-item ADAS-Cog assessed a range of cognitive abilities including memory, comprehension, orientation in time and place and spontaneous speech. Items were evaluated by tests and clinician ratings on a 5-point scale. Scores ranged from 0-70 with higher scores indicates more dysfunction. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline is defined as value at W0. Estimated value was calculated by Active treatment minus Placebo. The adjusted means were presented. A hierarchical testing procedure was used to control for the two rosiglitazone dose groups and the three genetic subgroups. There was no adjustment for the donepezil versus placebo comparisons, which were included to assess the sensitivity of the trial to detect a treatment effect.

ITT Population. Only those par. available at the indicated time points were analyzed. These comparisons were conducted for APOE e4 negative (neg) par. (e2/e2, e2/e3, and e3/e3) cohort

The 11-item ADAS-Cog assessed a range of cognitive abilities including memory, comprehension, orientation in time and place and spontaneous speech. Items were evaluated by tests and clinician ratings on a 5-point scale. Scores ranged from 0-70 with higher scores indicates more dysfunction. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline is defined as value at W0. Estimated value was calculated by Active treatment minus Placebo. A hierarchical testing procedure was used to control for the two rosiglitazone dose groups and the three genetic subgroups. There was no adjustment for the donepezil versus placebo comparisons, which were included to assess the sensitivity of the trial to detect a treatment effect.

ITT Population. Only those par. available at the indicated time points were analyzed. These comparisons were conducted in all except e4/e4's: comprised of APOE e4 neg par. (e2/e2, e2/e3, and e3/e3) and APOE e4 heterozygote par. (e2/e4, e3/e4)

The 11-item ADAS-Cog assessed a range of cognitive abilities including memory, comprehension, orientation in time and place and spontaneous speech. Items were evaluated by tests and clinician ratings on a 5-point scale. Scores ranged from 0-70 with higher scores indicates more dysfunction. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline is defined as value at W0. Estimated value was calculated by Active treatment minus Placebo. A hierarchical testing procedure was used to control for the two rosiglitazone dose groups and the three genetic subgroups. There was no adjustment for the donepezil versus placebo comparisons, which were included to assess the sensitivity of the trial to detect a treatment effect.

ITT Population

The CIBIC+ assessment comprised of a 7-point rating of severity (at baseline) and change (at indicated time points). It was rated on a scale of 1 to 7 as 1: markedly improved, 2.: moderately improved, 3: minimally improved, 4: no change, 5: minimally worse, 6: moderately worse and 7: markedly worse; higher score means greater dysfunction. It was based on interviews with the par. and caregiver by an independent rater. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived. Estimated value was calculated by Active treatment minus Placebo. A hierarchical testing procedure was used to control for the two rosiglitazone dose groups and the three genetic subgroups. There was no adjustment for the donepezil versus placebo comparisons, which were included to assess the sensitivity of the trial to detect a treatment effect.

ITT Population. Only those par. available at the indicated time points were analyzed. These comparisons were conducted in APOE e4 negative (neg) par. (e2/e2, e2/e3, and e3/e3).

The CIBIC+ assessment comprised of a 7-point rating of severity (at baseline) and change (at indicated time points). It was rated on a scale of 1 to 7 as 1: markedly improved, 2.: moderately improved, 3: minimally improved, 4: no change, 5: minimally worse, 6: moderately worse and 7: markedly worse; higher score means greater dysfunction. It was based on interviews with the par. and caregiver by an independent rater. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived. Estimated value was calculated by Active treatment minus Placebo. A hierarchical testing procedure was used to control for the two rosiglitazone dose groups and the three genetic subgroups. There was no adjustment for the donepezil versus placebo comparisons, which were included to assess the sensitivity of the trial to detect a treatment effect.

ITT Population. Only those par. available at the indicated time points were analyzed. These comparisons were conducted in all except e4/e4's: comprised of APOE e4 neg par. (e2/e2, e2/e3, and e3/e3) and APOE e4 heterozygote par. (e2/e4, e3/e4)

The CIBIC+ assessment comprised of a 7-point rating of severity (at baseline) and change (at indicated time points). It was rated on a scale of 1 to 7 as 1: markedly improved, 2.: moderately improved, 3: minimally improved, 4: no change, 5: minimally worse, 6: moderately worse and 7: markedly worse; higher score means greater dysfunction. It was based on interviews with the par. and caregiver by an independent rater. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived. Estimated value was calculated by Active treatment minus Placebo. A hierarchical testing procedure was used to control for the two rosiglitazone dose groups and the three genetic subgroups. There was no adjustment for the donepezil versus placebo comparisons, which were included to assess the sensitivity of the trial to detect a treatment effect.

ITT Population. Only those par. available at the indicated time points were analyzed. These comparisons were conducted in full population.

The 11-item ADAS-Cog assessed a range of cognitive abilities including memory, comprehension, orientation in time and place and spontaneous speech. Most items were evaluated by tests, but some were dependent on clinician ratings on a 5-point scale. Scores ranged from 0-70 with higher scores indicating greater dysfunction. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline is defined as value at W0. Endpoint treatment differences were adjusted to take account of missing data. It was evaluated at Baseline, W8, W16 and W24.

ITT Population. Only those participants available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The CIBIC+ assessment comprised of a 7-point rating of severity (at baseline) and change (at indicated time points). It was rated on a scale of 1 to 7 as 1: markedly improved, 2.: moderately improved, 3: minimally improved, 4: no change, 5: minimally worse, 6: moderately worse and 7: markedly worse; higher score means more dysfunction. The scale was based on interviews with the par. and caregiver and was completed by an independent rater. It required separate structured 15-20 minute interviews with the par. and caregiver. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived. It was evaluated at Baseline, W8, W16 and W24.

ITT Population. Only those participants available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The NPI assessed behavioral disturbances comprises 10 dimensions: delusions, hallucinations, dysphoria, apathy, euphoria, disinhibition, aggressiveness and agitation, irritability, anxiety and aberrant motor activity. The par. caregiver asked about behavior in the par. If "Yes", the informant then rates both the severity on a 3-point scale, 1: mild to 3: severe (total range: 0-36) and the frequency using a 4-point scale, 1: occasionally to 4: very frequently. The total domain score was frequency × severity. The distress was scored on 5-point scale, 0: no distress to 5 - very severe or extreme. A total NPI score can be calculated by adding all domain scores together; NPI total score: from 0-144 and NPI distress score: from 0-60, all with higher scores indicating more severe behavioral disturbance. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0.

ITT Population. Only those participants available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles). Endpoint treatment differences which were adjusted to take account of missing data are derived.

The DAD, assessed the ability of a par. to execute basic and instrumental activities of daily living (ADL) and leisure activities. The scale consists of 40 questions assessing basic and instrumental ADLs. This scale assesses a participants' ability to initiate, plan, and perform activities related to hygiene, dressing, continence, eating, meal preparation, telephoning, going on an outing, finance and correspondence, medications, leisure, and housework. Each item was scored as yes: 1, no: 0 and N/A: not applicable. Higher scores indicate less disability with a score of 100 indicating no disability and 0 indicating no functional ability. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. The percentage score was calculated as (DAD total score/total number of applicable items) multiplied by 100. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Change from Baseline in short term memory assessment score was assessed from a combined analysis of items 1 (word recall task) and 7 (word recognition task) of ADAS-Cog scale. Word recall task consist of the participants score was the mean number of words not recalled on three trials (maximum score 10) and word recognition task, to score this item the number of incorrect responses was counted (maximum error score was 12). Higher score indicating greater dysfunction. Total score is sum of individual score. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The EQ-5D Proxy was a 2 part scale used to assess the quality of life and utility benefit. The data for Part 1 is presented. It is a 5 dimensional Health State Classification. Caregivers were asked to respond as they feel the par. would on dimensions of mobility, self-care, usual activities, pain/discomfort and anxiety/depression. Answers to each question were recorded on a 3-point scale which indicates the level of impairment (level 1= no problem; level 2=some or moderate problem(s) and level 3=unable, or extreme problem with higher scores indicating greater dysfunction. EQ-5D Proxy assessments was performed at Baseline, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The EQ-5D Proxy is a 2 part scale used to assess the quality of life and utility benefit. The data for Part 2 is presented. It is a the visual analogue scale Thermometer which assessed caregiver's impression of par. overall health. The Thermometer has endpoints of 100 (best imaginable health state) and 0 (worst imaginable health state). EQ-5D Proxy assessments was performed at Baseline, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The RUD instrument was developed as a comprehensive tool to assess the amount of resource use among demented par. RUD assess both formal and informal resource use of the par. and the primary caregiver, making it possible to calculate costs from a societal perspective. Q1 relates to assisting par. with basic activities of daily living and Q2 relates to instrumental activities of daily living. The assessments was performed at Baseline, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The ACQLI was an assessment of caregiver quality of life. This instrument consists of 30 questions exploring various aspects of carer's quality of life. Each of the questions had a two point response and the 30 questions were summed to provide a total score. Items are assumed to be unidimensional (i.e., represent a single variable) and are scored 0/1 (false/true) before summation into a total score with a 0-30 range. To ease comparisons between scales, ACQLI scores were transformed to range between 0-100 (100: worse). The assessments was performed at Baseline, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The MMSE consists of 11 tests of orientation, memory (recent and immediate), concentration, language and praxis. Scores range from 0 to 30, with lower scores indicating greater cognitive impairment. The scale is completed by the investigator, based on the performance of the par. and takes approximately 5 to 10 minutes to administer. The assessments was performed at Baseline and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

The blood sample was collected for assessments of HbA1c levels at Baseline and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0. Endpoint treatment differences which were adjusted to take account of missing data are derived.

ITT Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

An AE is defined as any untoward medical occurrence in a participant, temporally associated with the use of a medicinal product, whether or not considered related to the medicinal product. An AE included significant or unexpected worsening or exacerbation of the condition/indication under study, exacerbation of a chronic or intermittent pre-existing condition, new conditions detected or diagnosed, signs, symptoms, or the clinical sequelae of a suspected overdose of either investigational product or a concurrent medication. Number of participants with any AE and as per severity were reported. Refer to the general AE/SAE module for a list of AEs and SAEs.

Safety population: This included all par. randomized to treatment who have taken at least one dose of study medication.

SBP, DBP and HR of par. were recorded in sitting posture as vital signs, while body weight was measured without shoes and wearing light clothing at each visit. The blood pressure (BP) and HR values were identified as of PCC if the vales were out of the reference range (for SBP, 90 to 140 millimeters of mercury (mmHg), DBP, 50 to 90 mmHg, and HR >100 or <50 beats per minute [bpm]) or meet a change from Baseline criterion. For SBP it was increase from Baseline (high) if increased by more than or equal to (>=) 40 mmHg; decrease from Baseline (low) if decreased by >=30 mmHg. For DBP, increase from Baseline (high) if increased by >=30 mmHg; decrease from Baseline (low) if decreased by >=20 mmHg. For HR, increase from Baseline (high) if increased by >=30 bpm; decrease from Baseline (low) if decreased by >=30 bpm. For weight, increase from Baseline (high) if increased by >=7%; decrease from Baseline (low) if decreased by >=7%. Baseline was defined as value at W0.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Triplicate 12-lead ECG measures was obtained digitally, approximately one minute apart after the par. had rested in the supine position in a quiet room (no TV, minimal talking) for at least 10 minutes. Conduction intervals from the 12-lead ECGs were manually read and confirmed by an external cardiologist/vendor. The ECG parameters includes PR interval, QRS duration, QT - uncorrected interval, QTc Bazett (QTcB), QTc Fridericia (QTcF) and RR interval of Central Cardiologist are reported. The assessments was performed at Baseline and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at Screening/Visit 1/W-6.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Triplicate 12-lead ECG measures was obtained digitally, approximately one minute apart after the par. had rested in the supine position in a quiet room (no TV, minimal talking) for at least 10 minutes. Conduction intervals from the 12-lead ECGs were manually read and confirmed by an external cardiologist/vendor. The ECG HR of Central Cardiologist reported. The assessments was performed at Baseline and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at Screening/Visit 1/W-6.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Body weight will be measured at all visits, without shoes and wearing light clothing. The assessments was performed at Baseline, W4, W8, W12, W16, and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Hematology parameters were assessed at Baseline, W4, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Hematology parameters were assessed at Baseline, W4, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

HbA1c assessment was performed par. with type 2 diabetes mellitus or HbA1c >=6.5% at Screening only. HbA1c levels were assessed at Baseline, W12 and W24. Change from Baseline was calculated as value at scheduled time point minus Baseline value. Baseline was defined as value at W0.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Haematology parameters were identified as of PCC (high [H], low [L]), if the values were out of the reference range (RR). The range for parameters was: platelet (100AV-500AV), red blood cell (RBC, 0.8-1.2), hemoglobin (L: female [F]:10, male [M]:11; H: F:16.5-AV, M:18), hematocrit (0.8-1.2), white blood cell (WBC, 3-15), neutrophils (0.75-1.5), lymphocytes (0.75-1.5), monocytes (0.75-2), eosinophils (none-2), basophils (none-2), mean corpuscle volume (MCV, 0.8-1.2), mean corpuscular hemoglobin (MCH, 0.8-1.2), mean corpuscular hemoglobin concentration (MCHC, 0.8-1.2), red cell distribution width (RDW, 0.8-1.2). Data for mean platelet volume (reference range not established) not reported

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).

Clinical chemistry parameters were identified as of PCC (H, L), if values were out of RR: Alanine aminotransferase (ALT, none-120 [250% upper limit of RR, ULRR]), Albumin (0.75-2), Aspartate aminotransferase (AST,none-105 (3-64y), 137.5 (65+y), >250%ULRR), Alkaline phosphatase (ALP,none-312.5 (20+y), >250%ULRR), blood urea nitrogen (BUN)/Creatinine ratio (none-1.25), BUN (none-11), Chloride (80-115), Calcium (0.75-1.25), Carbon dioxide (CO2, 15-40) content, Creatinine (22, <50% lower limit of RR [LLRR]-155, >125%ULRR), Creatine phosphokinase (CPK, none-1.25), Gamma glutamyl transferase (GGT,none-2.5), Glucose (3.6-7.8), High density lipoprotein (HDL,0.65-none), Lactate dehydrogenase (LDH,none-1.25), Low density lipoprotein (LDL,none-2), Magnesium (0.5-2), Potassium (3-5.5), Phosphorus inorganic (0.5-1.5), Sodium (130-150), Total protein (0.8-1.5), Total cholesterol (none-1.25), Total bilirubin (none-1.95), Triglycerides (none-9). Data for Creatinine clearance not reported.

Safety Population. Only those par. available at the specified time points were analyzed (represented as n=x,x,x,x in the category titles).