SECTION 18.5

Comparative Review of Oral Hypoglycemic Agents in Adults

Harinder Chahal For WHO Secretariat

Table of Contents Acronyms: ...... 3 I. Background and Rationale for the review: ...... 4 II. Medications under comparative review: ...... 4 Table 1 - New oral hypoglycemic agents for comparison with current EML agents ...... 5 III. Literature searches and methodology: ...... 5 1. Title Search Results: ...... 6 2. Statement about quality of evidence: ...... 6 IV. Clinical efficacy and safety evaluation: ...... 6 1. DPP-4 Inhibitors (Sitagliptin, Saxagliptin) and Metformin: ...... 6 2. Glitazones (Rosiglitazone, Pioglitazone) and Metformin: ...... 7 3. Alpha-glucosidase inhibitors (AGIs – Acarbose, Miglitol) and Metformin: ...... 8 4. Meglitinides (Repaglinide, Nateglinide) and Metformin: ...... 8 5. DPP-4 Inhibitors (Sitagliptin, Saxagliptin) and Sulfonylureas:...... 9 6. Glitazones (Rosiglitazone, Pioglitazone) and Sulfonylureas: ...... 9 7. Alpha-glucosidase inhibitors (AGIs – Acarbose, Miglitol) and Sulfonylureas: ...... 10 8. Meglitinides (Repaglinide, Nateglinide) and Sulfonylureas: ...... 10 9. Statement on Amylin Analogues – Pramlintide: ...... 11 V. Cost, Regulatory and Current NEML Availability Evaluation: ...... 11 Table 2: Comparative Cost Chart and Drug Approval by US and Australian Regulatory Agencies ...... 12 Table 3: Oral hypoglycemics listed on selected NEMLs ...... 12 VI. Summary: ...... 12 Appendix: ...... 14 Table 4: Summary: Comparative efficacy and safety of oral hypoglycemics ...... 14 Table 5: Chart of systematic reviews used ...... 15 Table 6: Question: Should Metformin vs DPP-4 Inhibitors be used for Diabetes Mellitus, Type 2? ...... 16 Table 7: Question: Should Metformin vs Glitazones be used for Diabetes Mellitus Type 2? ...... 19 Table 8: Question: Should Acarbose vs Metformin be used for Diabetes Mellitus, Type 2? ...... 22 Table 9: Question: Should Metformin vs meglitinides be used for Diabetes Mellitus, Type 2? ...... 25 Table 10: Question: Should Glitazones vs SFU be used for Diabetes Mellitus, Type 2? ...... 28 Table 11: Question: Should Acarbose vs be used in SFU? ...... 31 Table 12: Question: Should SFU vs meglitinides be used for Diabetes Mellitus, Type 2? ...... 34 References: ...... 36

Page 2 of 37

Acronyms:

AGI - Alpha-glucosidase inhibitor

AHRQ – Agency for Healthcare Research and Quality

CHF – Congestive heart failure

CI – Confidence interval

CV – Cardiovascular

DM – Diabetes Mellitus

DPP-4 inhibitors – dipeptidylpeptidase-4 inhibitors

EC – Expert Committee

EML – Essential Medicines List

FDA – Food and Drug Administration

GRADE – Grading of Recommendations Assessment, Development and Evaluation

HbA1c – Glycosylated hemoglobin

HDL – High density lipoprotein-cholesterol

LDL – Low density lipoprotein-cholesterol

LMICs - Low- and Middle-Income Countries

MSH – Management Sciences for Health

NEML – National Essential Medicines List

RCT – Randomized controlled trial

SFU – Sulfonylureas

TG – Triglycerides

TGA – Therapeutics Goods Administration

US – United States of America

USD – United States dollar

WHO – World Health Organization

Page 3 of 37

I. Background and Rationale for the review:

Diabetes mellitus is a chronic disease that requires life-long pharmacological and non-pharmacological management to prevent complications such as cardiovascular disease, retinopathy, nephropathy, and neuropathy.[1, 2] While type 2 diabetes mellitus is the most common form of diabetes comprising of 90% to 95% of all diabetes cases.[2] An estimated 346 million people worldwide live with diabetes, resulting in 3.4 million deaths in 2004, with more than 80% of these deaths occurring in low- and middle income countries.[3] It is projected that the death burden from diabetes will double by the year 2030.[3] According to the 2010 WHO report on NCDs, the estimated prevalence of diabetes in 2008 was about 8% for men and women in low-income countries and 10% for both sexes in upper-middle-income countries with the highest global prevalence of diabetes in Eastern Mediterranean Region and Region of the Americas.[4] The high prevalence rate is of concern since diabetes in the leading cause of renal failure, visual impairment and blindness and increases the risk of lower limb amputation by at least 10 times.[4] Additionally, patients living with diabetes may need 2 to 3 three times the health-care resources compared to people without diabetes and diabetes care may require allocation of up to 15% of national health care budgets.[4] Furthermore, given the close link between poverty and NCDs, the NCDs impose a disproportionate burden on low and middle income countries.[4]

In the United States, 11 classes of medications are approved for management of DM; these include 8 oral agents such as – biguanides, sulfonylureas, meglitinides, thiazolidinediones (glitazones), alpha- glucosidase inhibitors, DPP-4 inhibitors, bile acid sequestrants, dopamine-2 agonists, and 3 injectable agents such as – GLP-1 receptor agonists (incretins), amylin analogues and insulin.[1, 5] The 18th WHO expert committee on the selection and use of essential medicines in 2011 requested a review of the current oral hypoglycemic medicines for use in adult to determine if updates to the EML are needed. [6] Currently, the EML contains two oral hypoglycemics, glibenclamide (sulfonylurea) and metformin. This document will conduct comparative analysis of four oral hypoglycemic agents – glitazones (thiazolidinediones), DPP-4 inhibitors, alpha-glucosidase inhibitors and meglitinides versus sulfonylureas (SFU) and metformin to determine their efficacy and safety, as well as conduct a cost- comparison. This review will also provide an overview of the current availability of the four agents in questions in LMICs by surveying NEMLs of 15 nations at random; as well as provide information on regulatory status of these agents in the US and Australia. The regulatory status in US and Australia was selected as an initial reference point given the stringent review and approval process required for therapeutic approval by these agencies and due to the availability of the databases in English.

II. Medications under comparative review:

Table 1 lists the medications reviewed by this document and the comparisons made. The 18th EC on the Selection and Use of Essential Medicines had also requested a review on pramlintide – this medication was not reviewed; a statement regarding this therapeutic peptide is made in section IV-9.

Page 4 of 37

Table 1 - New oral hypoglycemic agents for comparison with current EML agents Comparison # EML Medication Comparison Medication GRADE Table Comparison 1 Metformin DPP-4 Inhibitors (Sitagliptin) Table 6 Comparison 2 Glitazones (Pioglitazone, Rosiglitazone) Table 7 Comparison 3 Alpha-glucosidase inhibitors (Acarbose) Table 8 Comparison 4 Meglitinides (Repaglinide, Nateglinide) Table 9

Comparison 5 Sulfonylureas DPP-4 Inhibitors (Sitagliptin) None Comparison 6 Glitazones (Pioglitazone, Rosiglitazone) Table 10 Comparison 7 Alpha-glucosidase inhibitors (Acarbose) Table 11 Comparison 8 Meglitinides (Repaglinide, Nateglinide) Table 12 Comparison 9 Pramlintide acetate – Not reviewed None

III. Literature searches and methodology:

The purpose of this review is to present evidence for safety, efficacy and cost for DPP-4 inhibitors, glitazones, alpha-glucosidase inhibitors and meglitinides as compared to the current EML oral hypoglycemics, metformin (biguanide) and glibenclamide (sulfonylurea). Literature search was focused to answer this question.

The Cochrane library and PubMed databases were searched for existing systematic reviews on hypoglycemic medications up to July 2012 using the following terms: sitagliptin, saxagliptin, DPP-4 inhibitors, dipeptidylpeptidase-4 inhibitors; alpha-glucosidase inhibitors, acarbose; sulfonylureas, glibenclamide, glyburide, glimepiride, gliclazide; thiazolidinediones, glitazones, pioglitazone, rosiglitazone; biguanides, metformin; meglitinides, nateglinide, repaglinide, mitiglinide.

Eight (8) reviews were identified relevant to topic of this review (Table 4); 6 reviews from Cochrane and 2 from AHRQ.[7-14] The primary reviews used for this report were by Bennett et al and Bolen et al due to their most recent publication dates and review of medications of interest.[7, 9] However, other reviews as shown in Table 4, were used and referenced as needed to clarify and to add to the body of evidence. Bennett et al reviewed literature up to April 2010 on all anti-diabetic medications except alpha-glucosidase inhibitors.[7] Bolen et al reviewed literature up to January 2006 on all anti-diabetic medications of interest.[9]

New, English-language literature beyond the periods covered by the systematic reviews was searched using Cochrane Central Register for Controlled Trials for titles addressing comparative safety and efficacy of monotherapy with medications for whom a paucity of data was determined. For alpha- glucosidase inhibitors the databases for searched from February 2006 up to July 2012. For meglitinides and DPP-4 inhibitors the databases were searched from June 2010 up to July 2012. The following search terms were used: sitagliptin, saxagliptin, DPP-4 inhibitors, dipeptidylpeptidase-4 inhibitors; alpha- glucosidase inhibitors, acarbose; sulfonylureas, glibenclamide, glyburide, glimepiride, gliclazide, glipizide; biguanides, metformin; meglitinides, nateglinide, repaglinide, mitiglinide. No additional search was conducted on glitazones or DPP-4 inhibitors versus metformin because it was determined the evidence available in the systematic reviews used on these comparisons was sufficient for review of efficacy and safety. The results of these searches are provided under section III-1 – Title Search Results.

Page 5 of 37

Since our search focused on comparative literature for the classes of medications in question, this review does not include many of the placebo-controlled studies conducted for safety and efficacy of these agents.

The WHO Essential Medicines website was used to reference NEMLs of 15 nations at random to determine how many of the surveyed nations had four classes of drugs in question on their NEML.[15] (Table 3)

MSH 2010 International Drug Price Indicator Guide was referenced first to obtain median buyer price per unit.[16] However, for majority of the medications of interest, prices were not available in MSH 2010 guide, Lexi-Comp online database was used for price and maximum daily dose of all medications as a reference source for pricing.[17]

1. Title Search Results: a. DPP-4 Inhibitors versus SFU: 80 trials resulted in the search; 6 studies were identified as relevant to the question from title review. Two were duplicates from the Bennett et al review. None of the 4 new studies identified compared DPP-4 monotherapy with SFU monotherapy.[18-21] These studies were included in this review due to their relevance to efficacy and safety outcomes. b. Meglitinides versus Metformin: The search resulted in 3 trials. The trials did not address the question of comparative efficacy and safety of these agents. No new trials of interest were identified. c. Meglitinides versus SFU: The search resulted in 38 trials. 2 new trials of interest were identified. These trials did not address all outcomes of interest; however, they were included in this review due to their relevance to the safety outcome data.[22, 23] d. Alpha-glucosidase inhibitors versus metformin: The search resulted in 17 trials. The trials did not address the question of comparative efficacy and safety of these agents. No new trials of interest were identified. e. Alpha-glucosidase inhibitors versus SFU: The search resulted in 11 trials. The trials did not address the question of comparative efficacy and safety of these agents. No new trials of interest were identified.

2. Statement about quality of evidence: The quality evidence presented in the systematic reviews and other clinical trials used in this review were evaluated using the GRADE methodology. GRADE tables were prepared for efficacy and safety outcomes, whenever possible, based on the evidence presented in the referenced reviews; other GRADE assessments were at the judgment of the author of this review. When necessary, the primary publication was referenced to determine GRADE rating. The strength of evidence evaluations are presented in Tables 6 through 12 in the appendix.

IV. Clinical efficacy and safety evaluation:

1. DPP-4 Inhibitors (Sitagliptin, Saxagliptin) and Metformin:

Page 6 of 37

Efficacy: Bennett et al reviewed three RCTs that compared metformin with DPP-4 inhibitors, and found greater reductions in HbA1c with metformin.[7] The between-group difference of -0.4 percent (95 percent CI -0.5 percent to -0.2 percent) were observed, favoring metformin.[7] For weight loss, 3 short duration RCTs comparing these 2 agents found greater reduction in weight with metformin.[7] Although evidence favors a greater reduction in LDL and a greater increase in HDL with metformin compared to DPP-4 inhibitors, no statistical significance is seen.[7] While a greater reduction in triglycerides is seen with DPP-4 inhibitors, these results are also not statistically significant.[7] Bennett et al, found insufficient data to make a determination regarding all-cause mortality and cardiovascular mortality benefits between DPP-4 inhibitors and metformin.[7]

Safety: DPP-4 inhibitors have a better safety profile in terms of mild to moderate hypoglycemia symptoms and gastrointestinal side effects.[7] In one 24-week RCT mild to moderate hypoglycemic symptoms were observed at a rate of 3.3% for metformin and 1.7% with DPP-4 inhibitors, however, the results were not statistically significant (p=0.12).[7, 24] One RCT showed an occurrence of adverse GI events (nausea/vomiting/diarrhea/abdominal discomfort) in metformin group at a rate of 20.7% and 11.5% in DPP-4 inhibitor group, in which diarrhea accounted for majority of the difference at 10.9% with metformin and 3.6% for sitagliptin.[7] The high incidence of diarrhea with metformin is consistent with published literature as a common side-effect of therapy and usually subsides with continued therapy.[25]

GRADE evidence is summarized in Table 6.

2. Glitazones (Rosiglitazone, Pioglitazone) and Metformin:

Efficacy: From the 14 RCTs comparing glitazones and metformin reviewed by Bennett et al, no between-group differences in reduction of HbA1c was observed.[7] A review of 8 RCTs comparing weight loss between therapy with metformin and glitazones, found weight loss with metformin and mild increases in weight with glitazone treatment.[7] A four-year RCT observed a between-group reduction in weight of 6.9kg favoring metformin over rosiglitazone.[7, 26] A review of 6 RCTs favors metformin for reduction in LDL and TG over rosiglitazone, with pooled between-group difference of -12.8mg/dL for LDL and -26.9mg/dL for TG.[7] However, an evaluation of 6 RCTs found no HDL benefit with either metformin or rosiglitazone.[7] There was no all-cause mortality or cardiovascular mortality benefit with either treatment.[7, 26]

Safety: A large 4-year double-blind RCT (ADOPT) with over 1400 participants in each arm found no significant differences in the occurrence of self-reported hypoglycemic events in patient assigned to the rosiglitazone or the metformin group, with one serious hypoglycemic event in each group.[7, 26] Bennett et al notes conflicting evidence for rate of CHF with metformin and glitazones.[7] Three RCTs and four observational studies provide a low grade evidence for increased risk of CHF with glitazones.[7] However, the ADOPT study notes no difference in CHF adverse events between either treatment group.[7, 26] It is important note that the FDA has placed a boxed warning for all thiazolidinedione agents, including rosiglitazone and pioglitazone for risk of congestive heart failure.[27-29] Metformin has been consistently shown to have a greater occurrence of GI adverse events over glitazones.[7] Page 7 of 37

GRADE evidence is summarized in Table 7.

3. Alpha-glucosidase inhibitors (AGIs – Acarbose, Miglitol) and Metformin:

Efficacy: Van de Laar et al. and Bolen et al. reviewed 2 RCTs comparing submaximal dosed metformin and maximally dosed acarbose showing no significant differences in HbA1c reduction between the two treatment groups.[9, 14] No statistically significant differences were observed for weight reduction with either AGIs or metformin.[9, 14] Reviews by Van de Laar et al and Bolen et al, found no benefits to HDL or TG with either therapy. [9, 14] One study, using submaximal doses of metformin and maximum doses of acarbose showed a reduction in LDL favoring acarbose.[14] No evidence is available to determine all-cause or CV mortality benefits with either treatment.

Safety: One RCT reported a low incidence of hypoglycemia risk with both agents, however, provided no statistical analysis.[30] Van de Laar et al and Bolen et al reviews based on two trials, report higher rate of side effects for acarbose, favoring metformin.[9, 14] For total adverse events, one study reported an odds ratio of 15 in favor of metformin.[14] Van de Laar et al, reviewed one RCT comparing miglitol (AGI) and metformin, in which no statistically significant differences in GI adverse events were observed.[14] Another study reports the incidence of withdrawal from the study due to GI adverse effects was 58% for acarbose arm and 14.8% for metformin.[9, 30]

GRADE evidence is summarized in Table 8.

4. Meglitinides (Repaglinide, Nateglinide) and Metformin:

Efficacy: Bennett et al reviewed 3 RCTs comparing metformin with meglitinides, which found similar effects on HbA1c with both treatments.[7] Two studies compared metformin and repaglinide at comparable doses showing non-significant HbA1c differences between treatment groups.[7] Evidence regarding benefits of weight reduction with meglitinides or metformin is low, however, indicates generally non-significant weight differences, with a slight trend that may favor metformin.[7] Similarly, evidence suggests a reduction in LDL and TG that may favor metformin over meglitinides, however is non-significant.[7] For HDL, their maybe a benefit with repaglinide over metformin, however the results are non-significant.[7] Overall, the evidence for benefits to lipid profile with meglitinides versus metformin is low.[7] There is low level of evidence to determine all-cause mortality or CV mortality benefits, however, one 24-week trial found one death in the metformin group and no deaths in the nateglinide treatment group.[7, 31] The one death in the metformin group was judged by investigators to be unlikely to be associated with therapy.[31] A recent nationwide study of over 100,000 Danish residents >20years of age, determined no statistical difference in all-cause mortality between patients taking repaglinide versus metformin.[32]

Safety: In Bennett et al review, 5 RCTs determined a favorable side effect profile for mild or moderate hypoglycemic events for metformin over meglitinides with an OR of 3.01.[7] Comparatively, meglitinides present with a favorable GI adverse effect profile over metformin.[7, 33] In one double-

Page 8 of 37

blind, RCT combined GI side-effects were 70% and 47% for metformin and repaglinide, respectively.[33]

GRADE evidence is summarized in Table 9.

5. DPP-4 Inhibitors (Sitagliptin, Saxagliptin) and Sulfonylureas:

Efficacy: Bennett et al reviewed one 12-week moderately-sized double-blind RCT compared high dose sitagliptin with maximum dose glipizide and found similar reductions in HbA1c, -0.77% versus -1.00%, for DPP-4 inhibitor and SFU, respectively.[7] Additional studies comparing DPP-4 inhibitor or SFU add-on therapy to metformin have shown similar results for reduction of HbA1c, not favoring either agent.[7, 19, 20, 34] Evidence indicates a benefit for weight reduction with a DPP-4 inhibitor over SFU, either as monotherapy and as combination therapy with metformin.[7, 19, 20, 34] However, due to lack of direct monotherapy comparative data, unable to determine true effect. Bennett et al review of lipid profile indicated an increase in LDL and HDL with sitagliptin over SFU, while the increase in TG with sitagliptin was less than the increase with SFU (3.6% versus 7.0%).[7] However, in all lipid measures reviewers found an overlapping CI after placebo-subtracted change from baseline in each group.[7] There is insufficient data to determine all-cause mortality benefits for this comparison.[7]

Safety: Sitagliptin consistently has a better hypoglycemic profile compared to SFUs as monotherapy and as combination therapy with metformin.[7, 18-20] Additionally, reduced incidence of hypoglycemia with sitagliptin versus glipizide or glimepiride was observed during Ramadan in a multi-center study.[18] This is a specialized patient population since observers of Ramadan abstain from food or water from dawn until dusk for the duration of the month of Ramadan.[18] No differences in GI side- effects have been observed with DPP-4 inhibitors and SFU as monotherapy or combination therapy.[7, 20, 35]

GRADE evidence: For all outcomes, the evidence strength for DPP-4 inhibitor comparison with SFUs is Low, with the exception of hypoglycemia and GI adverse events, for which the evidence strength is Moderate. One short term RCT evaluating direct comparison of DPP-4 inhibitor with SFU was identified [35]; this trial is reviewed in Bennett et al.[7] Low risk of bias is detected, the outcomes observed were direct, however, it is not possible to determine consistency (due to only one study) and there is concern for precision since the trial moderately sized and no statistical analysis was provided for some outcomes (such as GI side effects), and differing doses on sitagliptin and glipizide (based on titration protocol) were compared leading to uncertainty in results.

6. Glitazones (Rosiglitazone, Pioglitazone) and Sulfonylureas:

Efficacy: Bennett et al reviewed 13 RCTs comparing glitazones or thiazolidinediones (TZDs) (pioglitazone and rosiglitazone) and second-generation sulfonylureas (glibenclamide, glimepiride, and glyburide). The review found both treatments had similar effects on HbA1c.[7] Five RCTs with up to 1 year or less in duration, compared glitazones and a SFU, showing greater weight gain with glitazones, favoring SFUs.[7] Five RCTs compared rosiglitazone or pioglitazone with a SFU, indicating a greater increase in LDL with glitazones relative to a SFU.[7] Eight RCTs compared rosiglitazone or Page 9 of 37

pioglitazone with a SFU, indicating a favorable increase in HDL with glitazones relative to a SFU.[7] Pioglitazone is favored for a greater decrease in TG over SFUs in 6 RCTs.[7] However, when comparing rosiglitazone and SFUs, Bennett et al found conflicting evidence for benefits of TG lowering. In one RCT, while both rosiglitazone (at 8mg dose) and a SFU were associated with a decrease in TG, the differences were non-significant; in another RCT a lower dose (4mg) of rosiglitazone lowered TG relative to a SFU, however, at a dose of 8mg, rosiglitazone increased TG relative to SFU with no statistical significance reported.[7] The ADOPT study showed all-cause mortality and cardiovascular mortality to be similar for rosiglitazone and glyburide at 2.3% and 2.2%, respectively.[7, 26] As above, it should be noted that the FDA has placed a boxed warning for all thiazolidinedione agents, including rosiglitazone and pioglitazone for risk of congestive heart failure.[27-29]

Safety: Five RCTs determined a greater risk of mild to moderate hypoglycemia with SFUs over glitazones with an OR of 3.9.[7] Although the ADOPT study with over 1300 participants in each arm reported no statistical significance for outcome of hypoglycemia between rosiglitazone and glyburide.[7] Bennett et al reviewed 4 RCTs looking at outcome of CHF with glitazones versus SFU and found an increase of CHF incidence with glitazones over SFUs with an OR of 1.68.[7] While the review did not show statistical significance, clinical significance could not be ruled out.[7] Three RCTs did not show a consistent difference in the occurrence of diarrhea between groups treated with pioglitazone or rosiglitazone and glyburide.[7]

GRADE evidence is summarized in Table 10.

7. Alpha-glucosidase inhibitors (AGIs – Acarbose, Miglitol) and Sulfonylureas:

Efficacy: Van de Laar et al reviewed 8 RCTs comparing SFU and acarbose showing a non-significant trend for reduction in HbA1c in favor of SFU.[14] However, for most comparisons the review found that the SFU was dosed sub-maximally.[14] Van de Laar et al review found no statistically significant differences in weight between AGIs and SU.[14] Five RCTs show Reviews of RCTs by Van de Laar et al and Bolen et al, found no benefit for lipid profile (LDL, HDL or TG) when comparing acarbose versus a SFU.[9, 14] All-cause mortality and cardiovascular mortality evidence is limited to allow for determination of mortality benefit between SFU and acarbose.[9] One small RCT comparing acarbose and tolbutamide showed no statistical difference in mortality benefit between the two treatments.[9, 14]

Safety: SFUs are favored for their overall and GI side effect profile. One RCT favors SFU over acarbose for GI side effects with an OR of 7.70.[14] However, in contrast, in terms hypoglycemic risks acarbose is favored over SFU.[9, 36]

GRADE evidence is summarized in Table 11.

8. Meglitinides (Repaglinide, Nateglinide) and Sulfonylureas:

Efficacy: Bennett et al reviewed 7 RCTs comparing a second-generation sulfonylurea with repaglinide, showing a pooled between-group difference of 0.1 percent (95 percent CI -0.2 percent to 0.3 percent) slightly favoring meglitinides.[7] However, when only studies using comparable doses of the two agents Page 10 of 37

were evaluated (3 out 7 studies), no differences in HbA1c reduction were observed.[7] The review found that no single study significantly influenced the results.[7] A review of 6 RCTs comparing meglitinides with SFUs showed no benefit for reduction in weight.[7] No statistically significant differences have been observed for improvement of lipid profile (LDL, HDL, TG) when comparing SFUs and meglitinides.[7] Evidence is limited to for mortality benefits when comparing these two classes of drugs. Bennett et al reviewed one, 1-year long RCT that looked at the all-cause mortality between repaglinide and glyburide and observed 3 deaths out of 362 participants in the repaglinide group and 1 death out of 182 participants in the glyburide group.[7] Each treatment group had one CV related death.[7] To note, the reviewers identified the strength of the evidence for all-cause mortality and CV mortality outcomes as low.[7]

Safety: Six studies reviewed by Bennett et al showed a lower incidence of hypoglycemia with meglitinides when compared with a SFU, however, the pooled results were not statistically significant.[7] The lower incidence of hypoglycemia is supported by 2 RCTs since the Bennett review.[22, 32] Additionally, a high-quality trial comparing repaglinide versus glibenclamide in patients observing Ramadan, showed statistically significant lower incidence of hypoglycemia with repaglinide than with glibenclamide (p<0.001).[7, 37] As mentioned above, Ramadan is a period during which the practitioners do not drink or eat from sunrise to sunset, so this study applies to a specific subset of patient population.[18, 37] The same study and two others have found that apart from incidence of hypoglycemia, both treatments were equally well tolerated.[37-39] However, there is paucity of data for evaluation of comparative GI side-effects between these agents.

GRADE evidence is summarized in Table 12.

9. Statement on Amylin Analogues – Pramlintide:

The 18th EC on Essential Medicines had also requested a comparative review of pramlintide, an amylin analogue.[6] However, a detailed review on this medication was not prepared because it was determined to be not appropriate for a comparison with oral hypoglycemics, the primary focus of this review. Pramlintide is a subcutaneous injectable synthetic analog of the human amylin peptide, a hormone produced by the pancreas for glycemic control during postprandial period.[40] Pramlintide works by delaying gastric emptying to reduce the initial postprandial increase in glucose, preventing a rise in plasma glucagon during postprandial period and causes satiety to decrease caloric intake.[40] Pramlintide is indicated for use prior to meals as an adjunct to insulin with or without SFU or metformin.[40]

V. Cost, Regulatory and Current NEML Availability Evaluation:

Table 2 provides an overview of the cost per unit, per 30 units and estimated monthly cost of treatment with medications under review in US dollars. Metformin and glibenclamide prices are from MSH and Lexi-Comp online; however, the cost of other agents was not available from MSH, therefore, Lexi- Comp online was used to evaluation – this provides costs of medications as they pertain to US markets.[16, 17] Regulatory status of medications in the US (FDA) and Australia (TGA) is also

Page 11 of 37

shown.[27, 41] As mentioned above, the regulatory status in US and Australia was selected as an initial reference point given the stringent review and approval process required for therapeutic approval by these agencies and due to the availability of the databases in English. Table 3 evaluates the availability of DPP-4 inhibitors, glitazones, acarbose and meglitinides across 15 low and middle-income countries based on the NEML for each nation. The countries for this review were selected at random from the WHO website hosting NEMLs.[15]

Table 2: Comparative Cost Chart and Drug Approval by US and Australian Regulatory Agencies Medication (Name and Cost per unit Cost/30 Daily Maximum Monthly cost based on FDA TGA Strength) (USD) tablets (USD) Dose[17] maximum dosing (USD) Approved Approved Metformin 500mg* 0.0087[16] 0.26 2550mg/day 1.30 Yes Yes Glibenclamide 5mg* 0.0042[16] 0.13 20mg/day 0.52 Yes Yes Prices from Lexi-Comp Online (US based prices) [17] Metformin 500mg* 0.23 6.50 2550mg/day 26.00 Yes Yes Metformin 1000mg* 0.60 17.99 2550mg/day 44.98 Yes Yes Glibenclamide 5mg* 0.53 15.99 20mg/day 15.99 Yes Yes Sitagliptin 25mg 7.18 215.40 100mg daily 861.60 Yes Yes Sitagliptin 100mg 7.83 234.90 100mg daily 234.90 Yes Yes Rosiglitazone 2mg 3.03 90.90 8mg once or BID 363.60 Yes Yes Rosiglitazone 8mg 8.33 249.99 8mg once or BID 249.99 Yes Yes Pioglitazone 15mg 6.45 193.48 45mg/day 580.44 Yes Yes Pioglitazone 45mg 10.33 310.00 45mg/day 310.00 Yes Yes Acarbose 25mg* 0.84 25.20 100mg TID 302.40 Yes Yes Acarbose 100mg* 0.90 27.00 100mg TID 81.00 Yes Yes Nateglinide 60mg* 1.60 47.99 120mg TID 287.94 Yes No Repaglinide 0.5mg 3.04 91.21 16mg daily 2918.40 Yes Yes Repaglinide 2mg 2.93 87.92 16mg daily 703.20 Yes Yes *Denotes generic price

Table 3: Oral hypoglycemics listed on selected NEMLs # Country DPP-4 Inhibitors Glitazones Alpha-glucosidase Meglitinides (Sitagliptin) (Pioglitazone and inhibitors (Repaglinide and Rosiglitazone) (Acarbose) Nateglinide) 1 Bangladesh No No No No 2 China No No No No 3 Dominican Republic No No No No 4 Ecuador No No No No 5 Fiji No No No No 6 Ghana No No No No 7 India No No No No 8 Iran No Yes (pioglitazone) Yes (acarbose) Yes (repaglinide) 9 Kyrgyzstan No Yes (Rosiglitazone) No No 10 Malta No No Yes (acarbose) Yes (repaglinide) 11 Morocco No No Yes (acarbose) No 12 Malaysia No No No No 13 Namibia No No No No 14 Nigeria No No No No 15 Oman No Yes (rosiglitazone) No No Total # of surveyed countries with identified 0 3 3 2 medications on the NEML

VI. Summary:

Page 12 of 37

This document provides a comprehensive comparative efficacy, safety and cost profile of four classes of oral hypoglycemic agents – glitazones, DPP-4 inhibitors, alpha-glucosidase inhibitors and meglitinides versus sulfonylureas and metformin using GRADE methodology in Section IV and in Tables 6 through 12. Table 4 provides a summary of the key efficacy and safety outcomes alongside the strength of evidence. The summary table also provides a relative comparison of cost for the agents in review compared to metformin and glibenclamide as baseline agents. The cost comparison is based on US market as referenced from Lexi-Comp online database for the recommended maximum daily dose. Costs for metformin 500mg strength and glibenclamide 5mg strength was used as these dosage strengths are on the EML and compared to maximum available strength for the comparative agents. Additionally this review has shown the limited availability of these agents on NEMLs from a survey of 15 LMICs in Table 3 and provided information on regulatory status of these agents in the US and Australia in Table 2.

Evidence on efficacy, safety, cost and availability on selected NEMLs does not support the addition of any agent from the four classes of oral hypoglycemics reviewed – glitazones, DPP-4 inhibitors, alpha- glucosidase inhibitors and meglitinides – to the EML at this time.

Page 13 of 37

Appendix:

Table 4: Summary: Comparative efficacy and safety of oral hypoglycemics Comparison HbA1c Weight LDL HDL TG Hypoglycemia Adverse events Relative Cost (GI) (cost for maximum monthly dose) Metformin versus (outcome and strength of evidence) Relative Cost: Metformin Baseline [US$26.00/month for 500mg tablets (strength on EML)] 1 DPP-4 inhibitors Favors Favors Neither favored Neither favored Neither favored Neither favored Favors DPP-4-I Sitagliptin 100mg: (Sitagliptin, Metformin Metformin (Low) (Very Low) (Low) (High) (Very Low) 9x greater than Saxagliptin) (Moderate) (Moderate) Metformin 2 Glitazones Neither favored Favors Favors Neither favored Favors Neither favored Favors Rosiglitazone 8mg: (Rosiglitazone, (Moderate) Metformin Metformin (Moderate) Metformin (Low) (High) Glitazones 9.6x greater Pioglitazone) (High) (Moderate) (High) Pioglitazone 45mg: 11.9x greater 3 AGIs (Acarbose, Neither favored Neither favored Favors Acarbose Neither favored Neither favored Unclear Favors Acarbose 100mg: Miglitol) (Moderate) (Low) (Moderate) (Moderate) (Moderate) (Low) Metformin 3.11x greater (Moderate) 4 Meglitinides Neither favored Neither favored Neither favored Neither favored Neither favored Favors Favors Nateglinide 60mg: (Nateglinide, (Moderate) (Moderate) (Moderate) (Moderate) (Moderate) Metformin Meglitinides 11x greater Repaglinide) (Low) (High) Repaglinide 2mg: 27x greater Sulfonylureas versus (outcome and strength of evidence) Relative Cost: Glibenclamide Baseline [US$ 15.99/month for 5mg tablets (strength on EML)] 5 DPP-4 inhibitors Neither favored Unclear Neither favored Neither favored Neither favored Favors DPP-4-I Neither favored Sitagliptin 100mg: (Sitagliptin, (Low) (Low) (Low) (Low) (Low) (Moderate) (Moderate) 14.6x greater than Saxagliptin) Glibenclamide 6 Glitazones Neither favored Favors SFU Favors SFU Favors Glitazones Unclear Neither favored Neither favored Rosiglitazone 8mg: (Rosiglitazone, (Moderate) (Low) (Low) (Low) (Low) (High) (High) 15.6x greater Pioglitazone) Pioglitazone 45mg: 19x greater 7 AGIs (Acarbose, Neither favored Neither favored Neither favored Neither favored Neither favored Favors AGI Favors SFU Acarbose 100mg: Miglitol) (Moderate) (Moderate) (High) (High) (High) (High) (High) 5x greater 8 Meglitinides Neither favored Neither favored Neither favored Neither favored Neither favored Favors Unknown Nateglinide 60mg: (Nateglinide, (High) (High) (Low) (Moderate) (Low) Meglitinides (n/a) 18x greater Repaglinide) (Moderate) Repaglinide 2mg: 43x greater

Page 14 of 37

Table 5: Chart of systematic reviews used Drug Class Drugs Review Period Reviewed Alpha-glucosidase 1. Acarbose Alpha-glucosidase inhibitors for type 2 diabetes mellitus (Van Up to: 29 April 2003 inhibitors 2. Miglitol de Laar FA)[14] 3. Voglibose DPP-4 Inhibitors 1. Sitagliptin Dipeptidyl peptidase-4 (DPP-4) inhibitors for type 2 diabetes Up to: 30 January 2008 2. Saxagliptin Mellitus (Richter B)[12] Meglitinide 1. Repaglinide Meglitinide analogues for type 2 diabetes mellitus (Black C)[8] Up to : 30 October 2006 Analogues 2. Nateglinide Biguanides 1. Metformin Metformin monotherapy for type 2 diabetes mellitus (Saenz Up to: 29 September 2003. A)[13] Glitazones 1. Rosiglitazone Rosiglitazone for type 2 diabetes mellitus (Richter B)[11] Up to: 29 April 2007

1. Pioglitazone Pioglitazone for type 2 diabetes mellitus (Richter B)[10] Up to: 30 August 2006. All above All above Comparative Effectiveness and Safety of Oral Diabetes Up to: January 2006 Medications for Adults with Type 2 Diabetes. Comparative Effectiveness (Bolen et al.)[9] All above except All above except Oral Diabetes Medications for Adults With Type 2 Diabetes: An Up to: April 2010 Alpha-glucosidase Alpha-glucosidase Update (Bennett et al.)[7] inhibitors inhibitors

Page 15 of 37

Table 6: Question: Should Metformin vs DPP-4 Inhibitors be used for Diabetes Mellitus, Type 2? Bibliography: Oral diabetes medications for adults with type 2 diabetes: An update. Bennett W. et al.

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality of Study event rates Relative Anticipated absolute effects (studies) bias bias evidence (%) effect Follow up (95% CI) With DPP-4 With Risk with DPP-4 Inhibitors Risk difference with Metformin Inhibitors Metformin (95% CI)

Mean difference in HbA1c for Metformin vs DPP-4 Inhibitors (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

1921 serious1 no serious no serious no serious undetected ⊕⊕⊕⊝ 972 949 - The mean mean difference The mean mean difference (3 studies) inconsistency indirectness imprecision MODERATE1 in hba1c for metformin vs in hba1c for metformin vs 24 weeks due to risk of bias dpp-4 inhibitors in the dpp-4 inhibitors in the control groups was intervention groups was -0.7 % 0.37 lower (0.54 to 0.2 lower)

Mean difference in weight for Metformin vs DPP-4 Inhibitors (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

1918 serious1 no serious no serious no serious undetected ⊕⊕⊕⊝ 969 949 - The mean mean difference The mean mean difference (3 studies) inconsistency indirectness imprecision MODERATE1 in weight for metformin vs in weight for metformin vs 24 to 54 due to risk of bias dpp-4 inhibitors in the dpp-4 inhibitors in the weeks control groups was intervention groups was -0.6 Kg 1.40 lower (1.8 to 1 lower)

Mean difference in LDL for Metformin vs DPP-4 Inhibitors (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

1918 serious2 no serious no serious serious2 undetected ⊕⊕⊝⊝ 969 949 - The mean mean difference The mean mean difference (3 studies) inconsistency indirectness LOW2 in ldl for metformin vs dpp- in ldl for metformin vs dpp-4 24 to 54 due to risk of bias, 4 inhibitors in the control inhibitors in the intervention weeks imprecision groups was groups was -0.5 mg/dL 5.85 lower (9.65 to 2.05 lower)

Mean difference in HDL for Metformin vs DPP-4 Inhibitors (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by higher values)

1918 serious3 serious3 no serious serious3 undetected ⊕⊝⊝⊝ 969 949 - The mean mean difference The mean mean difference (3 studies) indirectness VERY LOW3 in hdl for metformin vs dpp- in hdl for metformin vs dpp- 24 to 54 due to risk of bias, 4 inhibitors in the control 4 inhibitors in the

Page 16 of 37 weeks inconsistency, groups was intervention groups was imprecision 3.9 mg/dL 2.30 higher (0.28 lower to 4.88 higher)

Mean difference in triglycerides for Metformin vs DPP-4 Inhibitors (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

1918 serious4 no serious no serious serious4 undetected ⊕⊕⊝⊝ 969 949 - The mean mean difference The mean mean difference (3 studies) inconsistency indirectness LOW4 in triglycerides for in triglycerides for metformin 24 to 54 due to risk of bias, metformin vs dpp-4 vs dpp-4 inhibitors in the weeks imprecision inhibitors in the control intervention groups was groups was 3.4 higher -3 mg/dL (0.39 lower to 7.19 higher)

All-cause mortality for Metformin vs DPP-4 Inhibitor5 (CRITICAL OUTCOME; measured with: Number of events; Better indicated by lower values)

5 0 See comment - 0 - See comment See comment (0)

Cardiovascular mortality for Metformin vs DPP-4 Inhibitor5 (CRITICAL OUTCOME; assessed with: Number of events)

0 See comment - - not See comment See comment (0) pooled5

Hypoglycemia for Metformin vs DPP-4 Inhibitor (CRITICAL OUTCOME; assessed with: Number of events)

1050 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 9/528 17/522 RR 1.88 17 per 1000 15 more per 1000 (1 study) risk of inconsistency indirectness imprecision HIGH6 (1.7%) (3.3%) (0 to 0)7 (from 17 fewer to 17 fewer) 24 weeks bias6

Combined GI adverse effects for Metformin vs DPP-4 Inhibitor (Nausea/Vomiting/Diarrhea/Abdominal discomfort) (IMPORTANT OUTCOME; assessed with: Number of events)

1155 very no serious no serious serious8 undetected ⊕⊝⊝⊝ 88/534 146/621 RR 1.42 165 per 1000 69 more per 1000 (2 studies) serious8 inconsistency8 indirectness VERY LOW8 (16.5%) (23.5%) (0 to 0) (from 165 fewer to 165 24 weeks due to risk of bias, fewer) imprecision

1 Bennett et al reviewers rated 3 RCTs as Moderate due to medium risk of bias. No identification for the source of bias was provided. 2 Bennett et al reviewers rated 3 RCTs as Moderate due to medium risk of bias and imprecision. No identification for the source of bias was provided. Possible reason for the rating of imprecision may be the MD of -5.85mg/dL with a somewhat wide CI [-9.65, -2,.05]. However, using the GRADE criteria, the rating for this evidence has been decreased to Low. 3 Bennett et al reviewers rated 3 RCTs as Low due to medium risk of bias, inconsistency and imprecision. No identification for the source of bias was provided. Possible reason for the rating of imprecision may be the MD of 2.30mg/dL with a CI [-0.28, 4.88] that cross the line of no difference. A possible reason for inconsistency may be due to the I-squared value of 49%. However, using the GRADE criteria, the rating for this evidence has been decreased to Very Low. Page 17 of 37

4 Bennett et al reviewers rated 3 RCTs as Low due to medium risk of bias and imprecision. No identification for the source of bias was provided. Possible reason for the rating of imprecision may be the MD of 3.40mg/dL with a relatively wide CI [-0.39, 7.19] that crosses the line of no difference. 5 Insufficient data 6 Bennett et al reviewers rated 3 RCTs as High with medium risk of bias. No identification for the source of bias was provided. This table is based on 1 double-blind, multi-center RCT with over 500 participants in each treatmentment group. No points were deducted for bias. The overall rating of the evidence remains consistent with that of the reveiwers as High. 7 No statistically significant (p=0.12) 8 Bennett et al reviewers rated 2 RCTs as Low due to high risk of bias, unknown inconsistency and imprecision. No source of bias was identified. The review did not provide a meta-analysis of these trial, therefore, the reveiwers determination of bias and imprecision is accepted, however, the no points will be deducted for inconsistency. Further, using the GRADE criteria, the rating for this evidence has been decreased to Very Low.

Page 18 of 37

Table 7: Question: Should Metformin vs Glitazones be used for Diabetes Mellitus Type 2? Bibliography: Oral Diabetes Medications for Adults With Type 2 Diabetes: An Update (Bennett et al)

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality of Study event rates Relative Anticipated absolute effects (studies) bias bias evidence (%) effect Follow up (95% CI) With With Risk with Glitazones Risk difference with Glitazones Metformin Metformin (95% CI)

Mean difference in HbA1c for Metformin vs Glitazones (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

2662 serious1 no serious no serious no serious undetected ⊕⊕⊕⊝ 1334 1328 - The mean mean difference The mean mean (14 studies) inconsistency2 indirectness imprecision MODERATE1,2 in hba1c for metformin vs difference in hba1c for 12 to 52 due to risk of bias glitazones ranged across metformin vs glitazones in weeks control groups from the intervention groups -2.6 to -0.3 % was 0.07 lower (0.18 lower to 0.04 higher)

Mean weight difference for Metformin vs Glitazones (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

1914 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 972 942 - The mean mean weight The mean mean weight (8 studies) risk of inconsistency indirectness imprecision HIGH difference for metformin vs difference for metformin 16 to 52 bias glitazones in the control vs glitazones in the weeks groups was intervention groups was -0.3 Kg 2.61 lower (4.06 to 1.16 lower)

Mean difference in LDL for Metformin vs Rosiglitazone (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

482 no serious no serious no serious serious3 undetected ⊕⊕⊕⊝ 246 236 - The mean mean difference The mean mean (6 studies) risk of inconsistency indirectness1 MODERATE1,3 in ldl for metformin vs difference in ldl for 16 to 52 bias due to imprecision rosiglitazone in the control metformin vs rosiglitazone weeks groups was in the intervention groups 5.1 mg/dL was 12.76 lower (23.96 to 1.56 lower)

Mean difference in HDL for Metformin vs. Rosiglitazone (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by higher values)

Page 19 of 37

482 serious4 no serious no serious no serious undetected ⊕⊕⊕⊝ 246 236 - The mean mean difference The mean mean (6 studies) inconsistency indirectness imprecision MODERATE4 in hdl for metformin vs. difference in hdl for 16 to 52 due to risk of bias rosiglitazone in the control metformin vs. weeks groups was rosiglitazone in the 3.5 mg/dL intervention groups was 0.45 lower (2.34 lower to 1.43 higher)

Mean difference in TG for Metformin vs. Rosiglitazone (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

482 serious5 serious5 no serious serious5 undetected ⊕⊕⊝⊝ 246 236 - The mean mean difference The mean mean (6 studies) indirectness LOW5 in tg for metformin vs. difference in tg for 16 to 52 due to risk of bias, rosiglitazone in the control metformin vs. weeks inconsistency, groups was rosiglitazone in the imprecision, large -4.2 mg/dL intervention groups was effect 26.86 lower (49.26 to 4.47 lower)

All-cause mortality for Metformin vs Rosiglitazone (CRITICAL OUTCOME6; assessed with: Number of events)

2910 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 34/1456 31/1454 RR 0.91 23 per 1000 2 fewer per 1000 (1 study6) risk of inconsistency indirectness imprecision6 HIGH6 (2.3%) (2.1%) (0 to 0) (from 23 fewer to 23 4 years bias fewer)

Cardiovascular mortality for Metformin vs Rosiglitazone (CRITICAL OUTCOME; assessed with: Number of events)

2910 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 2/1456 2/1454 RR 1 1 per 1000 0 fewer per 1000 (1 study7) risk of inconsistency indirectness imprecision7 HIGH7 (0.14%) (0.14%) (0 to 0) (from 1 fewer to 1 fewer) 4 years bias7

Hypoglycemia for Metformin vs Glitazones (CRITICAL OUTCOME; assessed with: Number of events)

2910 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 141/1456 167/1454 RR 0.9 97 per 1000 10 fewer per 1000 (1 study) risk of inconsistency indirectness imprecision HIGH1,8 (9.7%) (11.5%) (0.8 to 1) (from 19 fewer to 0 more) 4 years bias1,8

Incidence of Heart Failure for Metformin vs Rosiglitazone (IMPORTANT OUTCOME; assessed with: Number of events)

2910 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 22/1456 19/1454 RR 0.86 15 per 1000 2 fewer per 1000 (1 study9) risk of inconsistency indirectness imprecision HIGH (1.5%) (1.3%) (0 to 0) (from 15 fewer to 15 4 years bias fewer)

Page 20 of 37

Combined GI adverse effects for Metformin vs Rosiglitazone (Nausea/Vomiting/Diarrhea/Abdominal discomfort) (IMPORTANT OUTCOME; assessed with: Number of events)

2910 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 335/1456 557/1454 RR 1.66 230 per 1000 152 more per 1000 (1 study10) risk of inconsistency indirectness imprecision HIGH (23%) (38.3%) (0 to 0) (from 230 fewer to 230 4 years bias fewer)1

1 Rating based on documentation by Bennett et al reviewers for 16 studies. No reason for down-grading of evidence was provided. 2 Rating based on documentation by Bennett et al reviewers for 16 studies. Consistent for short-duration studies. One long-term study inconsistent. No points were deducted. 3 Rating based on documentation by Bennett et al reviewers for 7 studies. No reason for down-grading of evidence was provided. However, a MD of -12.76 is reported with a wide CI [-23.96, -1.56], which may account for imprecision of evidence. 4 Rating based on documentation by Bennett et al reviewers for 6 studies. No reason for down-grading of evidence was provided. However, a modest MD of -0.45 is reported with a CI crossing the line of no difference [-2.34, 1.43], which may account for imprecision of evidence. 5 Rating based on documentation by Bennett et al reviewers for 7 studies. No reason for down-grading of evidence was provided. An I-squared value of 70% is reported, which could account for the imprecision rating. A wide CI [-49.26, -4.47] with a mean difference of -26.86mg/dL favoring metformin is reported, which could account for the inconsistency rating. The median change for rosiglitazone was -4.2mg/dL versus -26.86mg/dL for metformin, which could account for the large effect reported by the authors. 6 Bennett et al reviewers report Low strength of evidence based of 4 RCTs for all-cause mortality due to imprecision, however, no further explanation was provided. The all-cause mortality outcome on this table is based on the ADOPT study, a large double-blind, RCT; the strength of this evidence is ranked as High. 7 Bennett et al reviewers report Low strength of evidence based of 2 RCTs for CV mortality due to imprecision and moderate level of bias, however, no further explanation was provided. The CV mortality outcome on this table is based on the ADOPT study, a large double-blind RCT; the strength of this evidence is ranked as High. 8 Bennett et al reviewers report Moderate strength of evidence based on ADOPT study for hypoglycemia due to moderate level of bias and unknown consistency, however, no further explanation was provided. Given only 1 RCT, the consistency of this evidence can be classified as unknown. The hypoglycemia outcome on this table is based on the ADOPT study, a large double-blind RCT; we are classifying the strength of this evidence as High. 9 Bennett et al reviewers report Moderate strength of evidence based of 3 RCTs and 4 observational studies for CHF due to moderate level of bias, inconsistency and imprecision, however, no further explanation was provided. The reviewers note low-grade evidence showing increased risk of HF with glitazones, which could explain the Moderate strength of evidence. The CHF outcome on this table is based on the ADOPT study, a large double-blind, RCT; the strength of this evidence is ranked as High. 10 Bennett et al reviewers report High strength of evidence based of 5 RCTs for GI side-effects. The GI side-effects outcome on this table is based on the ADOPT study, a large double-blind RCT; the strength of this evidence is ranked as High.

Page 21 of 37

Table 8: Question: Should Acarbose vs Metformin be used for Diabetes Mellitus, Type 2? Bibliography: 1. Alpha-glucosidase inhibitors for type 2 diabetes mellitus. Van de Laar FA, et al. 2. Comparative Effectiveness and Safety of Oral Diabetes Medications for Adults with Type 2 Diabetes. Bolen, et al. 3. Comparison of acarbose and metformin in patients with Type 2 diabetes mellitus insufficiently controlled with diet and sulphonylureas: a randomized, placebo-controlled study. Willms B. et al.

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality Study event rates Relative Anticipated absolute effects (studies) bias bias of evidence (%) effect Follow up (95% CI) With With Risk with Metformin Risk difference with Acarbose Metformin Acarbose (95% CI)

Mean difference in HbA1c for Acarbose vs. Placebo (IMPORTANT OUTCOME; measured with: %; Better indicated by lower values)

2831 no no serious no serious no serious undetected ⊕⊕⊕⊕ 1442 1389 - The mean mean difference The mean mean difference in (28 studies1) serious inconsistency indirectness imprecision HIGH in hba1c for acarbose vs. hba1c for acarbose vs. 16 to 52 risk of placebo ranged across placebo in the intervention weeks bias control groups from groups was -1.61 to 1.6 % 0.77 lower (0.9 to 0.64 lower)

Mean difference in HbA1c for Acarbose vs Metformin (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

62 no serious2 no serious no serious undetected ⊕⊕⊕⊝ 31 31 - The mean mean difference The mean mean difference in (1 study3) serious indirectness imprecision MODERATE2 in hba1c for acarbose vs hba1c for acarbose vs 24 weeks risk of due to metformin in the control metformin in the intervention bias inconsistency groups was groups was -0.86 0.25 lower (0.61 lower to 0.11 higher)

Mean difference in LDL for Acarbose vs Metformin (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

62 no serious2 no serious no serious undetected ⊕⊕⊕⊝ 31 31 - The mean mean difference The mean mean difference in (1 study) serious indirectness imprecision MODERATE2 in ldl for acarbose vs ldl for acarbose vs metformin 24 weeks risk of due to metformin in the control in the intervention groups bias inconsistency groups was was 0.05 mg/dL 0.94 lower (1.52 to 0.36 lower)

Mean difference in HDL for Acarbose vs Metformin (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by higher values)

Page 22 of 37

62 no serious2 no serious no serious undetected ⊕⊕⊕⊝ 31 31 - The mean mean difference The mean mean difference in (1 study) serious indirectness imprecision MODERATE2 in hdl for acarbose vs hdl for acarbose vs 24 weeks risk of due to metformin in the control metformin in the intervention bias inconsistency groups was groups was -0.01 mg/dL 0.24 higher (0.02 lower to 0.5 higher)

Mean difference in triglycerides for Acarbose vs Metformin (measured with: mg/dL; Better indicated by lower values)

62 no serious2 no serious no serious undetected ⊕⊕⊕⊝ 31 31 - The mean mean difference The mean mean difference in (1 study) serious indirectness imprecision MODERATE2 in triglycerides for acarbose triglycerides for acarbose vs 24 weeks risk of due to vs metformin in the control metformin in the intervention bias inconsistency groups was groups was -0.12 mg/dL 0.28 lower (0.8 lower to 0.24 higher)

Mean difference in weight for Acarbose vs Metformin (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

62 no serious2 no serious serious2,4 undetected ⊕⊕⊝⊝ 31 31 - The mean mean difference The mean mean difference in (1 study) serious indirectness LOW2,4 in weight for acarbose vs weight for acarbose vs 24 weeks risk of due to metformin in the control metformin in the intervention bias inconsistency, groups was groups was imprecision -0.5 mg/dL 0.30 lower (5.45 lower to 4.85 higher)

Adverse effects for Acarbose vs Metformin (Total) (IMPORTANT OUTCOME; assessed with: Number of events)

64 no serious2 no serious serious5 undetected ⊕⊕⊝⊝ 2/32 16/32 OR 15 62 per 1000 438 more per 1000 (1 study) serious indirectness LOW2,5 (6.3%) (50%) (3.06 to (from 107 more to 768 more) 24 weeks risk of due to 73.58) bias inconsistency, imprecision

Hypoglycemia Acarbose vs Metformin (CRITICAL OUTCOME; assessed with: Number of events)

60 no serious6 no serious serious7 undetected ⊕⊕⊝⊝ 5/29 3/31 RR 0.5 172 per 1000 86 fewer per 1000 (1 study8) serious indirectness LOW6,7 (17.2%) (9.7%) (0 to 0) (from 172 fewer to 172 12 weeks risk of due to fewer) bias inconsistency, imprecision

1 Bases on the review by Van De Laar, et al. Review by Bolen et al. found 4 additional trials comparing alpha-glucosidase inhibitors with placebo that showed similar results. 2 The trial compared maximal doses of acarbose with submaximal doses of metformin. 3 Bases on the review by Van De Laar, et al. Review by Bolen et al. found 1 additional review that "compared submaximal doses of metformin to maximal doses of acarbose and showed no meaningful or consistent effects on HbA1c."

Page 23 of 37

4 Mean difference of -0,30Kg is reported with a wide CI [-5.45, 4.85]. May include benefit for treatment with either treatment group. 5 An OR of 15.00 is reported in favor of metformin, however, with a wide CI [3.06, 73.58]. Given one study with small N and wide CI, unable to determine true effect. 6 Maximum dose of acarbose was compared with sub-maximal doses of metformin. A point for inconsistency was deducted. 7 No definition of hypoglycemia or criteria of evaluation for hypoglycemia was provided. Absolute number of events for hypoglycemia were reported with no stastical significance. A point was deducted for imprecision. 8 Data from: Comparison of acarbose and metformin in patients with Type 2 diabetes mellitus insufficiently controlled with diet and sulphonylureas: a randomized, placebo-controlled study. Willms B. et al.

Page 24 of 37

Table 9: Question: Should Metformin vs meglitinides be used for Diabetes Mellitus, Type 2? Bibliography: 1. Metformin monotherapy for type 2 diabetes mellitus (Review). (Saenz A. et al) 2. Oral diabetes medications for adults with type 2 diabetes: An update. (Bennett W. et al) 3. Comparative Effectiveness and Safety of Oral Diabetes Medications for Adults with Type 2 Diabetes. Comparative Effectiveness Review No. 8. 2007. (Bolen S. et al)

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality Study event rates Relative Anticipated absolute effects (studies) bias bias of evidence (%) effect Follow up (95% CI) With With Risk with Meglitinides Risk difference with Metformin Meglitinides Metformin (95% CI)

Mean difference in HbA1c for Metformin vs Meglitinides (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

413 serious1 no serious no serious no serious undetected ⊕⊕⊕⊝ 208 205 - The mean mean difference The mean mean difference (2 studies2) inconsistency indirectness imprecision MODERATE1 in hba1c for metformin vs in hba1c for metformin vs 12 to 24 due to risk of bias meglitinides ranged across meglitinides in the weeks control groups from intervention groups was -0.38 to -0.3 % 0.16 lower (0.36 lower to 0.03 higher)2

Mean difference in weight for Metformin vs Meglitinides (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

56 serious3 no serious no serious no serious undetected ⊕⊕⊕⊝ 29 27 - The mean mean difference The mean mean difference (1 study) inconsistency4 indirectness imprecision MODERATE3,4 in weight for metformin vs in weight for metformin vs 12 weeks due to risk of bias meglitinides in the control meglitinides in the groups was intervention groups was -2.98 Kg 0.41 higher (0.12 lower to 0.94 higher)5

Mean difference in LDL for Metformin vs Meglitinides (NOT IMPORTANT OUTCOME; measured with: SD units; Better indicated by lower values)

56 serious3 no serious no serious no serious undetected ⊕⊕⊕⊝ 29 27 - The mean mean difference The mean mean difference (1 study) inconsistency4 indirectness imprecision MODERATE3,4 in ldl for metformin vs in ldl for metformin vs 12 weeks due to risk of bias meglitinides in the control meglitinides in the groups was intervention groups was 0.41 SD units 0.43 lower (0.96 lower to 0.1 higher)6,7

Mean difference in HDL for Metformin vs Meglitinides (NOT IMPORTANT OUTCOME; measured with: SD units; Better indicated by higher values)

56 serious3 no serious no serious no serious undetected ⊕⊕⊕⊝ 29 27 - The mean mean difference The mean mean difference (1 study) MODERATE3,4 in hdl for metformin vs in hdl for metformin vs

Page 25 of 37

12 weeks inconsistency4 indirectness imprecision due to risk of bias meglitinides in the control meglitinides in the groups was intervention groups was 0.21 SD units 0.45 lower (0.95 lower to 0.06 higher)6,7

Mean difference in TG for Metformin vs Meglitinides (NOT IMPORTANT OUTCOME; measured with: SD units; Better indicated by lower values)

56 serious3 no serious no serious no serious undetected ⊕⊕⊕⊝ 29 27 - The mean mean difference The mean mean difference (1 study) inconsistency4 indirectness imprecision MODERATE3,4 in tg for metformin vs in tg for metformin vs 12 weeks due to risk of bias meglitinides in the control meglitinides in the groups was intervention groups was 1.1 SD units 0.24 lower (0.76 lower to 0.29 higher)6,7

All-cause mortality for Metformin vs. Meglitinides (CRITICAL OUTCOME; assessed with: Number of events)

8 357 no serious no serious no serious serious undetected ⊕⊕⊕⊝ 0/179 1/178 - - (1 study) risk of bias inconsistency4 indirectness MODERATE4,8 (0%) (0.56%)9 24 weeks due to imprecision

Hypoglycemia for Metformin vs Meglitinides (CRITICAL OUTCOME10; assessed with: Number of events)

915 serious11 no serious no serious serious8 undetected ⊕⊕⊝⊝ 59/457 25/458 OR 3.01 129 per 1000 179 more per 1000 (5 studies) inconsistency indirectness LOW8,11 (12.9%) (5.5%) (1.76 to (from 78 more to 304 16 to 52 due to risk of 5.15) more) weeks bias, imprecision

Combined GI adverse effects for Metformin vs Meglitinides (Nausea/Vomiting/Diarrhea/Abdominal discomfort) (IMPORTANT OUTCOME12; assessed with: Number of events)

165 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 42/82 65/83 RR 1.53 512 per 1000 271 more per 1000 (1 study) risk of bias inconsistency4 indirectness imprecision HIGH4 (51.2%) (78.3%) (0 to 0) (from 512 fewer to 512 8 months fewer)

1 For one of the studies in the analysis (Moses 1999), the allocation concealment is unclear; the weight of this study in the analysis is high (44.7%), resulting in deduction of a point. 2 Based on Saenz A. et al Cochrane review. Bennett et al identified another study that did not show meaningful between-group differences. (Derosa G, Mugellini A, Ciccarelli L, et al. Comparison of glycaemic control and cardiovascular risk profile in patients with type 2 diabetes during treatment with either repaglinide or metformin. Diabetes Res Clin Pract 2003;60(3):161-169. ) 3 The allocation concealment in Moses 1999 study unclear, resulting in deduction of a point. 4 Given only one study, inconsistency is unknown. No points are deducted. 5 Based on Saenz A. et al Cochrane review. Bennett et al identified a study indicating weight differences as non-significant. (Derosa G, Mugellini A, Ciccarelli L, et al. Comparison of glycaemic control and cardiovascular risk profile in patients with type 2 diabetes during treatment with either repaglinide or metformin. Diabetes Res Clin Pract 2003;60(3):161-169.)

Page 26 of 37

6 Based on review by Saenz A. et al. Bennett et al, identified an RCT, indicating similar findings that between-group differences in LDL, HDL and TG were not statistically significant. 7 Not statistically significant. 8 Bennett et al reviewers classified the evidence as Low due to unknown consistency and imprecision, with no source of imprecision was identified. However, imprecision may stem from this outcome being based on 1 small, short-term RCT with no clear conclusion on mortality benefit with either treatment. A point for imprecision was deducted. However, with GRADE assessment the the level of evidence is classified as Moderate. 9 The relationship of the death was judged to be unlikely to be due to therapy. 10 This outcome based on review by Bennett et al. 11 Bennett et al reviewers classified the evidence as Moderate due to medium level of bias and imprecision, with no source of bias or imprecision identified. However, imprecision may stem for this outcome based on wide 95% CI for 3 out of the 5 individual studies in this analysis. A point for imprecision and bias was deducted. However, with GRADE assessment the the level of evidence is classified as Low. 12 The RCT for this outcome is discussed in Bennett et al, however, GRADE evidence outcome and GI effect incidence rate determined directly from the study. Lund, S.S., et al., Targeting hyperglycaemia with either metformin or repaglinide in non-obese patients with type 2 diabetes: results from a randomized crossover trial. Diabetes Obes Metab, 2007. 9(3): p. 394-407.

Page 27 of 37

Table 10: Question: Should Glitazones vs SFU be used for Diabetes Mellitus, Type 2? Bibliography: Oral Diabetes Medications for Adults With Type 2 Diabetes: An Update (Bennett et al)

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality Study event rates Relative Anticipated absolute effects (studies) bias bias of evidence (%) effect Follow up (95% CI) With SFU With Risk with SFU Risk difference with Glitazones Glitazones (95% CI)

Mean difference in HbA1c for Glitazones vs SFU (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

2170 serious1 no serious no serious no serious undetected ⊕⊕⊕⊝ 1003 1167 - The mean mean difference The mean mean difference (13 studies) inconsistency indirectness imprecision MODERATE1 in hba1c for glitazones vs in hba1c for glitazones vs 12 to 52 due to risk of sfu in the control groups sfu in the intervention weeks bias was groups was -0.9 % 0.10 lower (0.22 lower to 0.01 higher)

Mean weight difference for Glitazones vs SFU (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

1533 very no serious no serious no serious undetected ⊕⊕⊝⊝ 680 853 - The mean mean weight The mean mean weight (5 studies) serious1,2 inconsistency indirectness imprecision LOW1,2 difference for glitazones vs difference for glitazones vs 14 to 52 due to risk of sfu in the control groups sfu in the intervention weeks bias was groups was 1.9 Kg 1.24 higher (0.63 to 1.85 higher)

Mean difference in LDL for Pioglitazone vs SFU (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

465 very no serious no serious no serious undetected ⊕⊕⊝⊝ 239 226 - The mean mean difference The mean mean difference (3 studies) serious3 inconsistency indirectness imprecision LOW3 in ldl for pioglitazone vs sfu in ldl for pioglitazone vs sfu 24 to 52 due to risk of in the control groups was in the intervention groups weeks bias -1.4 mg/dL was 7.12 higher (5.26 to 8.98 higher)

Mean difference in HDL for Pioglitazone vs SFU (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by higher values)

616 serious1 no serious no serious serious4 undetected ⊕⊕⊝⊝ 312 304 - The mean mean difference The mean mean difference (6 studies) inconsistency indirectness1 LOW1,4 in hdl for pioglitazone vs in hdl for pioglitazone vs sfu 14 to 52 due to risk of sfu in the control groups in the intervention groups

Page 28 of 37 weeks bias, imprecision was was 0.5 mg/dL 4.27 higher (1.93 to 6.61 higher)

Mean difference in TG for Pioglitazone vs SFU (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

616 very no serious no serious serious5 undetected ⊕⊝⊝⊝ 312 304 - The mean mean difference The mean mean difference (6 studies) serious5 inconsistency indirectness VERY LOW5 in tg for pioglitazone vs sfu in tg for pioglitazone vs sfu 14 to 52 due to risk of in the control groups was in the intervention groups weeks bias, imprecision -3.6 mg/dL was 31.62 lower (49.15 to 14.1 lower)

All-cause mortality for Rosilitazones vs SFU (CRITICAL OUTCOME; assessed with: Number of events)

2897 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 32/1441 34/1456 RR 1.04 22 per 1000 1 more per 1000 (1 study7) risk of bias inconsistency6 indirectness imprecision HIGH6 (2.2%) (2.3%) (0 to 0) (from 22 fewer to 22 fewer) 4 years

Cardiovascular mortality for SFU vs Rosiglitazone (CRITICAL OUTCOME; assessed with: Number of events)

2897 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 3/1441 2/1456 RR 0.66 2 per 1000 1 fewer per 1000 (1 study9) risk of bias inconsistency6 indirectness imprecision HIGH6 (0.21%) (0.14%) (0 to 0)8 (from 2 fewer to 2 fewer) 4 years

Hypoglycemia for Glitazones vs SFU (CRITICAL OUTCOME; assessed with: Number of events)

3281 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 238/1650 56/1631 RR 3.88 144 per 1000 415 more per 1000 (5 studies10) risk of inconsistency indirectness imprecision HIGH10 (14.4%) (3.4%) (3.05 to (from 296 more to 568 1-3 years bias10 4.94) more)

Incidence of Heart Failure for Glitazone vs SFU (CRITICAL OUTCOME; assessed with: Number of events)

5323 no serious no serious no serious serious1 undetected ⊕⊕⊕⊝ 22/2653 37/2670 OR 1.68 8 per 1000 6 more per 1000 (4 studies12) risk of bias1 inconsistency indirectness MODERATE1 (0.83%) (1.4%) (0.99 to (from 0 fewer to 15 more) 16-52 weeks due to 2.85)11 imprecision

Combined GI adverse effects for Rosiglitazone vs SFU (Nausea/Vomiting/Diarrhea/Abdominal discomfort) (IMPORTANT OUTCOME; assessed with: Number of events)

2897 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 316/1441 335/1456 RR 1.05 Study population Page 29 of 37

(1 study13) risk of bias inconsistency indirectness imprecision HIGH (21.9%) (23%) (0 to 0) 219 per 1000 11 more per 1000 4 years (from 219 fewer to 219 fewer) Moderate - 1 Rating based on documentation by Bennett et al reviewers for 14 studies. No reason for down-grading of evidence was provided. 2 Rating based on documentation by Bennett et al reviewers for 7 studies. No reason for down-grading of evidence was provided. 3 Rating based on documentation by Bennett et al reviewers for 3 studies. No reason for down-grading of evidence was provided. 4 Rating based on documentation by Bennett et al reviewers for 5 studies. No reason for down-grading of evidence was provided. However, the I-squared stastic 99% and the meand difference between the studies ranges from -1.17 to 8.0, which could account for the down-grading of the evidence due to imprecision. 5 Rating based on documentation by Bennett et al reviewers for 6 studies. No reason for down-grading of evidence was provided. However, down-grading due to imprecision maybe accounted by large mean difference variation between studies ranging from -65mg/dL to -6mg/dL, with overall MD of -31.62, CI [-49.15, -14.10]. 6 Given only 1 study, inconsistency is unknown. 7 Bennett et al reviewers report Low strength of evidence based of 3 RCTs for all-cause mortality due to imprecision, however, no further explanation was provided. The all-cause mortality outcome on this table is based on the ADOPT study, a large double-blind, RCT; the strength of this evidence is ranked as High. 8 No stastical significance tests were provided by the trial. 9 Bennett et al reviewers report Low strength of evidence based of 1 RCTs for all-cause mortality due to imprecision, however, no further explanation was provided. The all-cause mortality outcome on this table is based on the ADOPT study, a large double-blind, RCT; the strength of this evidence is ranked as High. 10 Bennett et al reviewers rated 8 RCTs and 1 observational study as High with medium risk of bias; no explanation for rating of bias was provided. However, the rating of medium bias may be due to the inclusion of the observation study. This table only evaluated 5 RCTs, therefore no points are deducted for bias, and the evidence rating remains consistent with the reviewers' as High. 11 Not statistically significant, however, clinical significance in unknown given CHF RR of 1.68 CI [0.99, 2.85] associated with glitazones. 12 Bennett et al reviewers rated 4 RCTs and 5 observational studies as Moderate with medium risk of bias and imprecision; no explanation for rating of bias was provided. However, the rating of medium bias and imprecision may be due to the inclusion of the observation studies. This table only evaluated 4 RCTs, therefore no points are deducted for bias; point for imprecision was deducted based variability of OR in 4 RCTs ranging from 1.0 to 67.06. and the evidence rating remains consistent with the reviewers' as Moderate. 13 Bennett et al reviewers rated 4 RCTs High. This table is based on 1 large, double-blind, RCT (ADOPT).

Page 30 of 37

Table 11: Question: Should Acarbose vs be used in SFU? Bibliography: 1. Alpha-glucosidase inhibitors for type 2 diabetes mellitus. Van de Laar FA, et al. 2. Comparative Effectiveness and Safety of Oral Diabetes Medications for Adults with Type 2 Diabetes. Bolen, et al. 3. Prospective multicentre trial comparing the efficacy of, and compliance with, glimepiride or acarbose treatment in patients with type 2 diabetes not controlled with diet alone. Feinbock C. et al.

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality Study event Relative Anticipated absolute effects (studies) bias bias of evidence rates (%) effect Follow up (95% CI) With With Risk with Risk difference with Acarbose Acarbose (95% CI)

Mean difference in HbA1c for Acarbose vs. Placebo (IMPORTANT OUTCOME; measured with: %; Better indicated by lower values)

2831 no no serious no serious no serious undetected ⊕⊕⊕⊕ 1442 1389 - The mean mean difference in The mean mean difference (28 studies1) serious inconsistency indirectness imprecision HIGH hba1c for acarbose vs. in hba1c for acarbose vs. 16 to 52 risk of placebo ranged across placebo in the intervention weeks bias control groups from groups was -1.61 to 1.6 % 0.77 lower (0.9 to 0.64 lower)

Mean difference in HbA1c for Acarbose vs SFU (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

596 no serious2 no serious no serious undetected ⊕⊕⊕⊝ 304 292 - The mean mean difference in The mean mean difference (8 studies) serious indirectness imprecision MODERATE2 hba1c for acarbose vs sfu in hba1c for acarbose vs sfu 16 to 30 risk of due to ranged across control groups in the intervention groups weeks bias inconsistency from was -2.16 to -0.2 % 0.38 higher (0.02 lower to 0.77 higher)

Mean weight difference for Acarbose vs SFU (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

397 no serious3 no serious no serious undetected ⊕⊕⊕⊝ 200 197 - The mean mean weight The mean mean weight (5 studies) serious indirectness imprecision MODERATE3 difference for acarbose vs sfu difference for acarbose vs 16 to 24 risk of due to ranged across control groups sfu in the intervention weeks bias inconsistency from groups was -0.59 to 1.84 kg 1.90 lower (4.01 lower to 0.21 higher)

Mean difference in LDL for Acarbose vs SFU (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

Page 31 of 37

312 no no serious no serious no serious undetected ⊕⊕⊕⊕ 162 150 - The mean mean difference in The mean mean difference (4 studies) serious inconsistency indirectness imprecision HIGH ldl for acarbose vs sfu ranged in ldl for acarbose vs sfu in 24 to 30 risk of across control groups from the intervention groups was weeks bias -0.42 to -0.07 mg/dL 0.10 higher (0.07 lower to 0.27 higher)

Mean difference in HDL for Acarbose vs SFU (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by higher values)

485 no no serious no serious no serious undetected ⊕⊕⊕⊕ 246 239 - The mean mean difference in The mean mean difference (7 studies) serious inconsistency indirectness imprecision HIGH hdl for acarbose vs sfu in hdl for acarbose vs sfu in 16 to 30 risk of ranged across control groups the intervention groups was weeks bias from 0.02 higher -0.07 to 0.1 mg/dL (0.02 lower to 0.06 higher)

Mean difference in triglycerides for Acarbose vs SFU (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

591 no no serious no serious no serious undetected ⊕⊕⊕⊕ 300 291 - The mean mean difference in The mean mean difference (8 studies) serious inconsistency indirectness imprecision HIGH triglycerides for acarbose vs in triglycerides for acarbose 16 to 30 risk of sfu ranged across control vs sfu in the intervention weeks bias groups from groups was -0.44 to 0.17 mg/dL 0.01 higher (0.18 lower to 0.2 higher)

Disease Related Deaths for Acarbose vs. SFU (CRITICAL OUTCOME; assessed with: Number of events)

133 no no serious no serious serious5 undetected ⊕⊕⊕⊝ 1/66 0/67 OR 0.32 2 per 100 1 fewer per 100 (1 study) serious inconsistency4 indirectness MODERATE4,5 (1.5%) (0%) (0.01 to (from 1 fewer to 10 more) 24 weeks risk of due to 8.08)6 bias imprecision

Adverse effects for Acarbose vs SFU (Total) (IMPORTANT OUTCOME; assessed with: Number of events)

507 no no serious no serious no serious undetected ⊕⊕⊕⊕ 82/302 161/205 OR 3.95 272 per 1000 324 more per 1000 (7 studies) serious inconsistency7 indirectness imprecision8 HIGH7,8 (27.2%) (78.5%) (2 to 7.8) (from 156 more to 473 24 weeks risk of more) bias

Hypoglycemia Acarbose vs SFU (CRITICAL OUTCOME; assessed with: Number of events)

219 no no serious no serious no serious undetected ⊕⊕⊕⊕ 20/111 2/108 RR 0.1 180 per 1000 162 fewer per 1000 (1 study9) serious inconsistency indirectness imprecision HIGH (18%) (1.9%) (0 to 0) (from 180 fewer to 180 26 weeks risk of fewer) bias

Page 32 of 37

Adverse effects (Gastrointestinal) Acarbose vs SFU (IMPORTANT OUTCOME; assessed with: Number of events)

145 no no serious no serious no serious undetected ⊕⊕⊕⊕ 24/71 59/74 OR 7.70 338 per 1000 459 more per 1000 (1 study) serious inconsistency4 indirectness imprecision10 HIGH4,10 (33.8%) (79.7%) (3.64 to (from 312 more to 555 24 weeks risk of 16.31) more) bias

1 Bases on the review by Van De Laar, et al. Review by Bolen et al. found 4 additional trials comparing alpha-glucosidase inhibitors with placebo that showed similar results. 2 In majority of the trials in the analysis, the dosing for second-generation SFUs was submaximal. 3 I-squared value of 82% is reported with a mean difference ranging from -3.26 to -0.55 Kg between studies. 4 Given that there is only one study for this outcome, inconsistency in unknown. 5 Study has small N, with low number of outcomes and wide CI -- OR 0.32 [0.01, 8.08]. 6 Determined to be not statistically or clinically significant. 7 I-squared value of 63% is reported with an OR ranging from 1.20 up to 19.46. However, after excluding one study the OR ranges from 1.20 to 6.72. Majority of the studies favor SFUs, therefore, no points were deducted for inconsistency. 8 A mean effect of OR 3.95 is reported with a wide CI [2.00, 7.80]. Since most of the studies included in the analysis favor SFUs over acarbose and is consistent with the OR, no points were deducted. 9 Data from: Prospective multicentre trial comparing the efficacy of, and compliance with, glimepiride or acarbose treatment in patients with type 2 diabetes not controlled with diet alone. Feinbock C. et al. 10 The CI is wide [3.64, 16.31] with an OR of 7.70; however, this is supported by the overall adverse effect profile, therefore, no points were deducted.

Page 33 of 37

Table 12: Question: Should SFU vs meglitinides be used for Diabetes Mellitus, Type 2? Bibliography: Oral diabetes medications for adults with type 2 diabetes: An update. Bennett W. et al.

Quality assessment Summary of Findings

Participants Risk of Inconsistency Indirectness Imprecision Publication Overall quality Study event rates Relative Anticipated absolute effects (studies) bias bias of evidence (%) effect Follow up (95% CI) With With Risk with Meglitinides Risk difference with SFU (95% Meglitinides SFU CI)

Mean difference in HbA1c for SFU vs Repaglinide (CRITICAL OUTCOME; measured with: %; Better indicated by lower values)

1687 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 1058 629 - The mean mean difference The mean mean difference (7 studies) risk of bias inconsistency indirectness imprecision HIGH in hba1c for sfu vs in hba1c for sfu vs 12 to 52 repaglinide in the control repaglinide in the weeks groups was intervention groups was -0.2 % 0.07 higher (0.15 lower to 0.29 higher)

Mean difference in weight for SFU vs Repaglinide (NOT IMPORTANT OUTCOME; measured with: Kg; Better indicated by lower values)

1431 no serious no serious no serious no serious undetected ⊕⊕⊕⊕ 883 548 - The mean mean difference The mean mean difference (6 studies) risk of bias inconsistency indirectness imprecision HIGH in weight for sfu vs in weight for sfu vs 12 to 52 repaglinide in the control repaglinide in the weeks groups was intervention groups was -0.1 Kg 0.01 higher (0.97 lower to 0.99 higher)

Mean difference in LDL for SFU vs Meglitinides1 (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

0 serious2 no serious no serious serious2 undetected ⊕⊕⊝⊝ - 0 -1 See comment See comment (2 studies) inconsistency indirectness LOW2 12 months due to risk of bias, imprecision

Mean difference in HDL for SFU vs Meglitinides (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by higher values)

1577 no serious serious3 no serious no serious undetected ⊕⊕⊕⊝ 995 582 - The mean mean difference The mean mean difference (6 studies) risk of bias indirectness imprecision MODERATE3 in hdl for sfu vs meglitinides in hdl for sfu vs meglitinides 12 to 52 due to in the control groups was in the intervention groups 1.1 mg/dL was

Page 34 of 37 weeks inconsistency 0.67 lower (2.07 lower to 0.74 higher)

Mean difference in triglycerides for SFU vs Meglitinides (NOT IMPORTANT OUTCOME; measured with: mg/dL; Better indicated by lower values)

958 serious4 no serious no serious no serious undetected ⊕⊕⊕⊝ 615 343 - The mean mean difference The mean mean difference (4 studies) inconsistency indirectness imprecision MODERATE4 in triglycerides for sfu vs in triglycerides for sfu vs 12 to 52 due to risk of meglitinides in the control meglitinides in the weeks bias groups was intervention groups was 1.0 mg/dL 0.20 higher (3.83 lower to 6.57 higher)

All-cause mortality for SFU vs. Meglitinides (CRITICAL OUTCOME; assessed with: Number of events)

544 serious5 no serious no serious serious2 undetected ⊕⊕⊝⊝ 3/362 1/182 RR 0.66 1 per 100 0 fewer per 100 (1 study) inconsistency6 indirectness LOW2,5,6 (0.83%) (0.55%) (0 to 0) (from 1 fewer to 1 fewer) 1 years due to risk of bias, imprecision

Hypoglycemia for SFU vs Meglitinides (CRITICAL OUTCOME; assessed with: Number of events)

1387 serious7 no serious no serious no serious undetected ⊕⊕⊕⊝ 89/866 61/521 OR 0.78 103 per 1000 21 fewer per 1000 (6 studies) inconsistency indirectness imprecision MODERATE7 (10.3%) (11.7%) (0.55 to (from 44 fewer to 11 more) 12 to 52 due to risk of 1.12) weeks bias

1 Per two RCTs comparing SFUs with repaglinide, the between-group differences were non-significant with a range from -1.5mg/dL to 1mg/dL. 2 Bennett et al reviewers rated 2 RCTs as Low due to medium risk of bias and imprecision. No source of bias or imprecision was identified. No meta-analysis or Forest plot is included for this outcome. Two points were deducted for bias and imprecision. 3 I-squared stastic is 95%. However, inconsistency is not identified by Bennett et al. A point was deducted for iconsistency. With GRADE, the strength of evidence is reduced to Moderate from High, as identified by Bennett et al. 4 Bennett et al reviewers rated 6 RCTs as Moderate due to medium risk of bias. No source of bias was identified.A point was deducted for bias. 5 Bennett et al reviewers rated 1 RCT as Low due to medium risk of bias and imprecision. No source of bias or imprecision was identified. The imprecision may stem from this outcome being based on 1 RCT with limited outcomes. The review did not provide a meta-analysis of these trial, therefore, the reveiwers determination of bias and imprecision is accepted, 6 Given only 1 RCT, consistency is unknown. No points deducted. 7 Bennett et al reviewers rated 8 RCTs as Low due to medium risk of bias. No source of bias was identified. A point was deducted for bias. Using GRADE, the strength of evidence was increased to Moderate from Low.

Page 35 of 37

References:

1. ADA, Standards of medical care in diabetes--2012, American Diabetes Association. Diabetes Care, 2012. 35 Suppl 1: p. S11-63. 2. Qaseem, A., et al., Oral pharmacologic treatment of type 2 diabetes mellitus: a clinical practice guideline from the American College of Physicians. Ann Intern Med, 2012. 156(3): p. 218-31. 3. WHO. WHO Diabetes Fact Sheet, August 2011. 2011 July 21, 2012]; Available from: http://www.who.int/mediacentre/factsheets/fs312/en/index.html. 4. WHO, Global status report on noncommunicable disease 2010. 2010. 5. Alexander, G.C., et al., National trends in treatment of type 2 diabetes mellitus, 1994-2007. Arch Intern Med, 2008. 168(19): p. 2088-94. 6. WHO, The Selection and Use of Essential Medicines - Report of the WHO Expert Committee, 2011. 2011. 7. Bennett, W.L., et al., Oral Diabetes Medications for Adults With Type 2 Diabetes: An Update, 2011: Rockville (MD). 8. Black, C., et al., Meglitinide analogues for type 2 diabetes mellitus. Cochrane Database Syst Rev, 2007(2): p. CD004654. 9. Bolen, S., et al., Comparative Effectiveness and Safety of Oral Diabetes Medications for Adults With Type 2 Diabetes, 2007: Rockville (MD). 10. Richter, B., et al., Pioglitazone for type 2 diabetes mellitus. Cochrane Database Syst Rev, 2006(4): p. CD006060. 11. Richter, B., et al., Rosiglitazone for type 2 diabetes mellitus. Cochrane Database Syst Rev, 2007(3): p. CD006063. 12. Richter, B., et al., Dipeptidyl peptidase-4 (DPP-4) inhibitors for type 2 diabetes mellitus. Cochrane Database Syst Rev, 2008(2): p. CD006739. 13. Saenz, A., et al., Metformin monotherapy for type 2 diabetes mellitus. Cochrane Database Syst Rev, 2005(3): p. CD002966. 14. Van de Laar, F.A., et al., Alpha-glucosidase inhibitors for type 2 diabetes mellitus. Cochrane Database Syst Rev, 2005(2): p. CD003639. 15. WHO. National Medicines List/Formulary/Standard Treatment Guidelines. 2012 July 7th, 2012]; Available from: http://www.who.int/selection_medicines/country_lists/en/index.html. 16. (MSH), M.S.f.H., International Drug Price Indicator Guide. 2010. 17. Lexicomp, Lexicomp online database, 2012. 18. Al Sifri, S., et al., The incidence of hypoglycaemia in Muslim patients with type 2 diabetes treated with sitagliptin or a sulphonylurea during Ramadan: a randomised trial. Int J Clin Pract, 2011. 65(11): p. 1132-40. 19. Arechavaleta, R., et al., Efficacy and safety of treatment with sitagliptin or glimepiride in patients with type 2 diabetes inadequately controlled on metformin monotherapy: a randomized, double-blind, non-inferiority trial. Diabetes Obes Metab, 2011. 13(2): p. 160-8. 20. Goke, B., et al., Saxagliptin is non-inferior to glipizide in patients with type 2 diabetes mellitus inadequately controlled on metformin alone: a 52-week randomised controlled trial. Int J Clin Pract, 2010. 64(12): p. 1619-31. 21. Nowicki, M., et al., Long-term treatment with the dipeptidyl peptidase-4 inhibitor saxagliptin in patients with type 2 diabetes mellitus and renal impairment: a randomised controlled 52-week efficacy and safety study. Int J Clin Pract, 2011. 65(12): p. 1230-9. 22. Bellomo Damato, A., et al., Nateglinide provides tighter glycaemic control than glyburide in patients with Type 2 diabetes with prevalent postprandial hyperglycaemia. Diabet Med, 2011. 28(5): p. 560-6. 23. Zhang, H., et al., Effect of repaglinide and gliclazide on glycaemic control, early-phase insulin secretion and lipid profiles in newly diagnosed type 2 diabetics. Chin Med J (Engl), 2011. 124(2): p. 172-6. 24. Aschner, P., et al., Efficacy and safety of monotherapy of sitagliptin compared with metformin in patients with type 2 diabetes. Diabetes Obes Metab, 2010. 12(3): p. 252-61. 25. Okayasu, S., et al., The evaluation of risk factors associated with adverse drug reactions by metformin in type 2 diabetes mellitus. Biol Pharm Bull, 2012. 35(6): p. 933-7. 26. Kahn, S.E., et al., Glycemic durability of rosiglitazone, metformin, or glyburide monotherapy. N Engl J Med, 2006. 355(23): p. 2427-43. Page 36 of 37

27. (FDA), F.a.D.A. Food and Drug Administration (FDA). 2012 July 18th, 2012]; Available from: www.fda.gov. 28. GlaxoSmithKline, Prescribing Information and Medication Guide, 2011. 29. Takeda Pharmaceuticals America, I., Pioglitazone prescribing information and medication guide, 2012. 30. Willms, B. and D. Ruge, Comparison of acarbose and metformin in patients with Type 2 diabetes mellitus insufficiently controlled with diet and sulphonylureas: a randomized, placebo-controlled study. Diabet Med, 1999. 16(9): p. 755-61. 31. Horton, E.S., et al., Nateglinide alone and in combination with metformin improves glycemic control by reducing mealtime glucose levels in type 2 diabetes. Diabetes Care, 2000. 23(11): p. 1660-5. 32. Schramm, T.K., et al., Mortality and cardiovascular risk associated with different insulin secretagogues compared with metformin in type 2 diabetes, with or without a previous myocardial infarction: a nationwide study. Eur Heart J, 2011. 32(15): p. 1900-8. 33. Lund, S.S., et al., Targeting hyperglycaemia with either metformin or repaglinide in non-obese patients with type 2 diabetes: results from a randomized crossover trial. Diabetes Obes Metab, 2007. 9(3): p. 394-407. 34. Seck, T., et al., Safety and efficacy of treatment with sitagliptin or glipizide in patients with type 2 diabetes inadequately controlled on metformin: a 2-year study. Int J Clin Pract, 2010. 64(5): p. 562-76. 35. Scott, R., et al., Efficacy and tolerability of the dipeptidyl peptidase-4 inhibitor sitagliptin as monotherapy over 12 weeks in patients with type 2 diabetes. Int J Clin Pract, 2007. 61(1): p. 171-80. 36. Feinbock, C., et al., Prospective multicentre trial comparing the efficacy of, and compliance with, glimepiride or acarbose treatment in patients with type 2 diabetes not controlled with diet alone. Diabetes Nutr Metab, 2003. 16(4): p. 214-21. 37. Mafauzy, M., Repaglinide versus glibenclamide treatment of Type 2 diabetes during Ramadan fasting. Diabetes Res Clin Pract, 2002. 58(1): p. 45-53. 38. Madsbad, S., et al., Comparison between repaglinide and glipizide in Type 2 diabetes mellitus: a 1-year multicentre study. Diabet Med, 2001. 18(5): p. 395-401. 39. Marbury, T., et al., Repaglinide versus glyburide: a one-year comparison trial. Diabetes Res Clin Pract, 1999. 43(3): p. 155-66. 40. Amylin Pharmaceuticals, I., Symlin Prescribing Information, 2008. 41. (TGA), T.G.A. Therapeutic Goods Administration (TGA). 2012 July 18th, 2012]; Available from: www.tga.gov.au.

Page 37 of 37