Evaluate the Efficacy and Safety of Saxagliptin Added to Insulin Monotherapy or to Insulin Combined With Metformin in Chinese Subjects With Type 2 Diabetes Who Have Inadequate Glycaemic Control

Study Design: Randomized, prospective, double-blind, two-arm, parallel group, multi-center trial.

Target Subject Population: Subjects aged ≥18 who have type 2 diabetes (HbA1c of ≥7.5% and ≤11.0% and FPG<270 mg/dL (15 mmol/L)) on stable baseline therapy (insulin alone or insulin combined with metformin, with insulin at doses of ≥20 and ≤150 units per day total) for at least eight weeks at the time of screening. Insulin may be long-acting, intermediate-acting, or pre-mixed. 444 patients are planned to be randomized.

Investigational Product, Dosage and Mode of administration: Active treatment will comprise Saxagliptin 5 mg tablets once daily.

Comparator, Dosage and Mode of administration: Matching placebo tablets will be used as comparator.

Duration of Treatment: The study is divided to a single blind placebo lead in period of 8 weeks and a double-blind treatment phase of 24 weeks. Patients will be rescued based on high FPG values.

Statistical Methods: The analysis of the primary endpoint of change from baseline to week 24 of treatment in HbA1c will consist of an analysis of covariance (ANCOVA) model with treatment group and metformin use at enrolment as fixed effects and baseline HbA1c value as a covariate. The analysis will be performed on the Full analysis Set (FAS) consisting of randomised subjects who received at least 1 randomised investigational product dose and had at least 1 non-missing baseline and 1 post-baseline efficacy assessment. Within the framework of the ANCOVA model, point estimates and two-sided 95% confidence intervals (CI) for the mean change within each treatment group as well as for the difference in mean change between treatment groups will be calculated.

The Per Protocol (PP) analysis set is a subset of the full analysis set and will consist of subjects who do not deviate from the terms of the protocol which may affect the study outcome significantly as specified in the pre-defined protocol deviation list prior to unblinding the study. All decisions to exclude subjects from the primary data set will be made prior to the unblinding of the study. The primary efficacy endpoint of change from baseline in HbA1c, demographics, and baseline diabetes related characteristics and all secondary efficacy endpoints are to be analyzed using the PP Data Set. The analyses of PPG AUC, 120 minute PPG, FPG and mean total daily dose of insulin will also be done on the FAS and use a similar ANCOVA model as described above. Subjects achieving a therapeutic glycaemic response (A1C <7%) will be analyzed using a Fisher's exact test and will include exact 95% confidence intervals. The FPG analyses will utilize the mean of the latest two FPG values prior to randomization as the baseline value. The endpoint for the FPG analysis will be the mean of the week 20 and week 24 FPG values. All analyses (except the analysis of mean total daily dose of insulin) will utilize only observations at visits prior to rescue or where the subject's mean total daily dose of insulin has not increased by >10% from baseline. If an observation at week 24 is missing or does not meet these criteria, the latest post-baseline value that does will be carried forward (LOCF). The analysis of mean total daily dose of insulin will utilize the latest, non-missing, post-baseline value regardless of rescue. Multiplicity for the primary and secondary endpoints will be controlled via a hierarchical testing procedure that utilizes the full alpha (0.05) for each test. The sequence of testing will be: 1. Change from baseline in HbA1c at Week 24. 2. Change from baseline in PPG AUC at Week 24.3. Change from baseline in 120 minute PPG at Week 24. 4. Proportion of subjects achieving HbA1c <7.0% at Week 24. 5. Change from baseline in FPG to the mean at Week 20 and Week 24. 6. Change from baseline in MTDDI at Week 24.