Annals of Internal Medicine
Systematic Review: Comparative Effectiveness and Safety of Oral
Medications for Type 2 Diabetes Mellitus
Shari Bolen, MD, MPH; Leonard Feldman, MD; Jason Vassy, MD, MPH; Lisa Wilson, BS, ScM; Hsin-Chieh Yeh, PhD;
Spyridon Marinopoulos, MD, MBA; Crystal Wiley, MD, MPH; Elizabeth Selvin, PhD; Renee Wilson, MS; Eric B. Bass, MD, MPH;
and Frederick L. Brancati, MD, MHS

Background: As newer oral diabetes agents continue to emerge on
had a beneficial effect on high-density lipoprotein cholesterol levels the market, comparative evidence is urgently required to guide (mean relative increase, 0.08 to 0.13 mmol/L [3 to 5 mg/dL]) but a harmful effect on low-density lipoprotein (LDL) cholesterol levels(mean relative increase, 0.26 mmol/L [10 mg/dL]) compared with Purpose: To summarize the English-language literature on the ben-
other oral agents. Metformin decreased LDL cholesterol levels by efits and harms of oral agents (second-generation sulfonylureas, about 0.26 mmol/L (10 mg/dL), whereas other oral agents had no biguanides, thiazolidinediones, meglitinides, and ␣-glucosidase in- obvious effects on LDL cholesterol levels. Most agents other than hibitors) in the treatment of adults with type 2 diabetes mellitus.
metformin increased body weight by 1 to 5 kg. Sulfonylureas andrepaglinide were associated with greater risk for hypoglycemia, Data Sources: The MEDLINE, EMBASE, and Cochrane Central Reg-
thiazolidinediones with greater risk for heart failure, and metformin ister of Controlled Trials databases were searched from inception with greater risk for gastrointestinal problems compared with other through January 2006 for original articles and through November oral agents. Lactic acidosis was no more common in metformin 2005 for systematic reviews. Unpublished U.S. Food and Drug recipients without comorbid conditions than in recipients of other Administration and industry data were also searched.
oral diabetes agents.
Study Selection: 216 controlled trials and cohort studies and 2
Limitations: Data on major clinical end points were limited. Studies
systematic reviews that addressed benefits and harms of oral dia- inconsistently reported adverse events other than hypoglycemia, betes drug classes available in the United States.
and definitions of adverse events varied across studies. Some harmsnot assessed in trials or observational studies may have been over- Data Extraction: Using standardized protocols, 2 reviewers serially
abstracted data for each article.
Conclusions: Compared with newer, more expensive agents (thia-
Data Synthesis: Evidence from clinical trials was inconclusive on
zolidinediones, ␣-glucosidase inhibitors, and meglitinides), older major clinical end points, such as cardiovascular mortality. There- agents (second-generation sulfonylureas and metformin) have sim- fore, the review was limited mainly to studies of intermediate end ilar or superior effects on glycemic control, lipids, and other inter- points. Most oral agents (thiazolidinediones, metformin, and repa- mediate end points. Large, long-term comparative studies are glinide) improved glycemic control to the same degree as sulfonyl- needed to determine the comparative effects of oral diabetes ureas (absolute decrease in hemoglobin A1c level of about 1 per- agents on hard clinical end points.
centage point). Nateglinide and ␣-glucosidase inhibitors may haveslightly weaker effects, on the basis of indirect comparisons of Ann Intern Med. 2007;147:386-399.
placebo-controlled trials. Thiazolidinediones were the only class that For author affiliations, see end of text.
The prevalence and morbidity associated with type 2 on clinical outcomes are even less certain. As newer oral
diabetes mellitus continue to increase in the United agents, such as thiazolidinediones and meglitinides, are in- States and elsewhere (1, 2). Several studies of the treatment creasingly marketed, clinicians and patients must decide of type 2 diabetes suggest that improved glycemic control whether they prefer these generally more costly medica- reduces microvascular risks (3–7). In contrast, the effects of tions over older agents, such as sulfonylureas and met- treatment on macrovascular risk are more controversial (3, 4, 8, 9), and the comparative effects of oral diabetes agents Systematic reviews and meta-analyses of oral diabetes agents have attempted to fill this gap (10 –19), but fewhave compared all agents with one another (18, 19). The few investigations that have compared all oral agents fo-cused narrowly on individual outcomes, such as hemoglo- bin A1c level (18) or serum lipid levels (19). No systematic review has summarized all available head-to-head compar- isons with regard to the full range of intermediate end points (including hemoglobin A1c level, lipid levels, and body weight) and other clinically important outcomes, Conversion of graphics into slides such as adverse effects and macrovascular risks. Therefore, the Agency for Healthcare Research and Quality commis-sioned a systematic review to summarize the comparative 386 2007 American College of Physicians
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review benefits and harms of oral agents that are used to treat type technical report (available at http://effectivehealthcare.ahrq .gov/repFiles/OralFullReport.pdf) provides a more detaileddescription of the study methods (20).
Data Extraction and Quality Assessment
Data Sources and Selection
One investigator used standardized forms to abstract We searched MEDLINE, EMBASE, and the Coch- data about study samples, interventions, designs, and out- rane Central Register of Controlled Trials from inception comes, and a second investigator confirmed the abstracted to January 2006 for original articles. We also searched data. Two investigators independently applied the Jadad these databases until November 2005 for systematic re- scale to assess some aspects of the quality of randomized views. We reviewed reference lists of related reviews and trials (21). We considered observational studies and non- original data articles, hand-searched recent issues of 15 randomized trials to provide weaker evidence than ran- medical journals, invited experts to provide additional ci- domized trials, and we did not use a standardized scoring tations, reviewed selected medications from the U.S. Food system to assess their quality (22). We used the GRADE and Drug Administration (FDA) Web site, and reviewed (Grading of Recommendations Assessment, Development unpublished data from several pharmaceutical companies and Evaluation) working group definitions to grade the and public registries of clinical trials. Our search strategy overall strength of the evidence as high, moderate, low, for the bibliographic databases combined terms for type 2 very low, or insufficient (23).
diabetes and oral diabetes agents and was limited to En-glish-language articles and studies in adults. The search for Data Synthesis and Analysis
systematic reviews was similar but included terms for study We first performed a qualitative synthesis based on design as well.
scientific rigor and type of end point. In general, we de- We selected studies that included original data on scribed the UKPDS (United Kingdom Prospective Diabe- adults with type 2 diabetes and assessed benefits or harms tes Study) separately, because this large randomized, con- of FDA-approved oral diabetes agents that were available trolled trial differed from other trials in design, end points, in the United States as of January 2006. To facilitate head- and duration.
to-head comparisons of drug classes, we included drugs not When data were sufficient (that is, obtained from at on the U.S. market if members of their class were in use least 2 randomized, controlled trials) and studies were and had not been banned (voglibose, gliclazide, and glib- relatively homogeneous in sample characteristics, study enclamide). We also included studies of combinations of duration, and drug dose, we conducted meta-analyses for the therapies that are commonly used, such as combinations of following intermediate outcomes and adverse effects: hemo- metformin, second-generation sulfonylureas, and thiazo- globin A1c level, weight, systolic blood pressure, LDL cho- lidinediones. We excluded studies that evaluated combina- lesterol level, HDL cholesterol level, triglyceride level, and tions of 3 oral diabetes agents, and we also excluded first- hypoglycemia. For trials with more than 1 dosing group, we generation sulfonylureas, because few clinicians prescribe chose the dose that was most comparable with other trials these medications.
and most clinically relevant. We combined drugs into drug We sought studies that reported on major clinical out- classes only when similar results were found across individual comes (for example, all-cause mortality, cardiovascular drugs. We could not perform formal meta-analyses for mi- morbidity and mortality, and microvascular outcomes) or crovascular or macrovascular outcomes, mortality, and ad- any of the following intermediate end points or adverse verse events other than hypoglycemia because of method- events: hemoglobin A1c level, body weight, systolic and ological diversity among the trials or insufficient numbers of diastolic blood pressure, high-density lipoprotein (HDL) cholesterol level, low-density lipoprotein (LDL) cholesterol We used a random-effects model with the DerSimo- level, triglyceride level, hypoglycemia, gastrointestinal nian and Laird formula to derive pooled estimates (post- problems, congestive heart failure, edema or hypervolemia, treatment weighted mean differences for intermediate out- lactic acidosis, elevated aminotransferase levels, liver fail- comes and posttreatment absolute risk differences for ure, anemia, leukopenia, thrombocytopenia, allergic reac- adverse events) (24). We tested for heterogeneity among tions requiring hospitalization or causing death, and other the trials by using a chi-square test with ␣ set to 0.10 or serious adverse events. For intermediate end points, we less and an I2 statistic greater than 50% (25). If heteroge- included only randomized, controlled trials, which were neity was found, we conducted meta-regression analyses by abundant. For major clinical end points and adverse using study-level characteristics of double-blinding, study events, we considered observational studies as well as trials, duration, and dose ratio (calculated as the dose given in the because fewer randomized trials assessed these end points.
study divided by the maximum approved dose of drug).
We excluded studies that followed patients for less than 3 The full report contains data on indirect comparisons, in months (the conventional threshold for determining effects which 2 interventions are compared through their relative on hemoglobin A1c) or had fewer than 40 patients. Figure
effect against a common comparator (20). We tested for 1 shows the search and selection process, and the full
publication bias by using the tests of Begg and Mazumdar 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 387
Review Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Figure 1. Study flow diagram.
*Numbers add up to more than the number of abstracts or articles excluded because there may have been more than 1 reason for exclusion. †More thantwo thirds of the articles that were excluded for having fewer than 40 participants would have been excluded for other reasons as well. ‡The numbers ofarticles for intermediate outcomes, adverse events, microvascular and macrovascular outcomes, and mortality are not mutually exclusive.
(26) and Egger and colleagues (27). All statistical analyses (Table 1). For each head-to-head comparison on specific
were done by using STATA Intercooled, version 8.0 (Stata, outcomes, the number of randomized trials (ⱕ3 trials) and College Station, Texas).
the absolute number of events were small (20). The few Role of the Funding Source
observational studies were limited in quantity, consistency, The Agency for Healthcare Research and Quality sug- and adjustment for key confounders.
gested the initial questions and provided copyright release Since our review, 2 high-profile comparative random- for this manuscript but did not participate in the literature ized trials with about 4 years of follow-up have been pub- search, data analysis, or interpretation of the results.
lished, providing data on cardiovascular outcomes (28, 29).
In ADOPT (A Diabetes Outcome Progression Trial) (28), the incidence of cardiovascular events was lower with gly- Comparative Effectiveness of Oral Diabetes Agents in
buride than with rosiglitazone or metformin (1.8%, 3.4%, Reducing the Risk for Microvascular and Macrovascular
and 3.2%, respectively; P ⬍ 0.05). This effect was mainly Outcomes and Death
driven by fewer congestive heart failure events and a lower We found no definitive evidence about the compara- rate of nonfatal myocardial infarction events in the gly- tive effectiveness of oral diabetes agents on all-cause mor- buride group. Loss to follow-up was high (40%) and was tality, cardiovascular mortality or morbidity, peripheral ar- disproportionate among the groups and therefore may ac- terial disease, neuropathy, retinopathy, or nephropathy count for some differences among groups.
388 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review Table 1. Evidence of the Comparative Effectiveness of Oral Diabetes Medications on Mortality, Microvascular and Macrovascular
Outcomes, and Intermediate End Points*

Level of Evidence†
All-cause mortality It was unclear whether mortality differed when metformin plus a sulfonylurea was compared with sulfonylurea or metforminmonotherapy or when metformin was compared with sulfonylureas.
Data were insufficient to compare other oral diabetes medications.
Cardiovascular disease mortality It was unclear whether cardiovascular mortality differed when metformin plus a sulfonylurea was compared with sulfonylurea or metforminmonotherapy.
It was unclear whether the effects on cardiovascular mortality differed between metformin and sulfonylureas.
Data were insufficient to compare other oral diabetes medications.
Cardiovascular morbidity (nonfatal There were too few studies to support conclusions about how myocardial infarction and stroke) cardiovascular morbidity differed between the medications, except thatthe risk for congestive heart failure is increased with thiazolidinedionescompared with other oral agents.
Peripheral vascular disease No evidence exists for a difference between oral diabetes medications in effects on peripheral vascular disease.
Microvascular outcomes Too few comparisons were made to draw firm comparative conclusions on microvascular outcomes.
Most oral diabetes medications (thiazolidinediones, second-generation sulfonylureas, and metformin) produced similar absolute reductions inHbA1c level (approximately 1%) compared with one another asmonotherapy.
Repaglinide produced similar reductions in HbA1c level when compared directly with sulfonylureas.
Combination therapies were better at reducing the HbA1c level than was monotherapy by about 1% (absolute difference).
Repaglinide produced similar reductions in HbA1c level when compared indirectly with thiazolidinediones and metformin.
Indirect data and data from a few head-to-head trials showed that nateglinide and ␣-glucosidase inhibitors were less efficacious inreducing HbA1c levels (approximately 0.5%–1% absolute difference).
Systolic and diastolic blood pressure Moderate to low for most Most oral diabetes medications (thiazolidinediones, metformin, and sulfonylureas) had similarly minimal effects on systolic and diastolicblood pressure (⬍5 mm Hg).
Too few studies compared meglitinides with oral diabetes medications other than sulfonylureas to permit firm conclusions.
LDL cholesterol level Moderate for most Thiazolidinedione monotherapy and rosiglitazone plus metformin or a sulfonylurea increased LDL cholesterol levels (approximately 0.26–0.31mmol/L [10–12 mg/dL]) compared with metformin orsecond-generation sulfonylurea monotherapy, which generallydecreased LDL cholesterol levels.
Rosiglitazone increased LDL cholesterol levels more than pioglitazone (approximately 0.26–0.39 mmol/L [10–15 mg/dL]), according toindirect comparisons and a few head-to-head comparisons.
Metformin decreased LDL cholesterol levels compared with second-generation sulfonylureas (approximately 10 mg/dL).
Metformin plus a sulfonylurea decreased LDL cholesterol levels compared with second-generation sulfonylurea monotherapy (approximately 0.21mmol/L [8 mg/dL]).
Metformin monotherapy compared with metformin plus a sulfonylurea had similar effects on LDL cholesterol levels.
Second-generation sulfonylureas had similar effects on LDL cholesterol levels compared with repaglinide.
␣-Glucosidase inhibitors had similar effects on LDL cholesterol levels compared with second-generation sulfonylureas.
Indirect comparisons of acarbose and metformin showed similar effects on LDL cholesterol levels. The one direct comparison favored acarboseat maximal doses over metformin at submaximal doses.
According to 1 head-to-head trial and mainly indirect comparisons, rosiglitazone increased LDL cholesterol levels more than acarbose(approximately 0.26–0.39 mmol/L [10–15 mg/dL]).
Too few studies compared meglitinides with other oral diabetes medications (other than sulfonylureas) to draw firm conclusions.
HDL cholesterol level Pioglitazone increased HDL cholesterol levels more than rosiglitazone, according to indirect and a few direct comparisons (approximately0.03–0.08 mmol/L [1–3 mg/dL]).
Continued on following page 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 389
Review Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Level of Evidence†
HDL cholesterol level (continued) Pioglitazone increased HDL cholesterol levels compared with metformin or second-generation sulfonylureas (approximately 0.08–0.13 mmol/L[3–5 mg/dL]).
The combination of rosiglitazone with metformin or a second-generation sulfonylurea increased HDL cholesterol levels slightly more thanmetformin or second-generation sulfonylureas alone (approximately0.08 mmol/L [3 mg/dL]).
Metformin, second-generation sulfonylureas, acarbose, and meglitinides had similarly minimal or no effect on HDL cholesterol levels.
Combination therapy with metformin plus a second-generation sulfonylurea did not differ in effect on HDL cholesterol levels frommonotherapy with either of the 2 classes.
Triglyceride level Indirect comparisons and a few head-to-head comparisons showed that pioglitazone decreased triglyceride levels (range, 0.17–0.59 mmol/L[15–52 mg/dL]) compared with rosiglitazone, which increasedtriglyceride levels (range, 0.07–0.15 mmol/L [6–13 mg/dL]).
Pioglitazone decreased triglyceride levels more than metformin (approximately 0.29 mmol/L [26 mg/dL]), and decreases were similarcompared with sulfonylureas. However, the pooled estimate suggesteda potential statistically nonsignificant difference of approximately 0.33mmol/L (29 mg/dL) compared with sulfonylureas.
Metformin decreased triglyceride levels more than second-generation sulfonylureas and more than metformin plus rosiglitazone(approximately 0.11 mmol/L [10 mg/dL]).
Metformin plus a second-generation sulfonylurea decreased triglyceride levels more than sulfonylurea monotherapy (approximately 0.34mmol/L [30 mg/dL]) and produced a statistically nonsignificantdecrease in triglyceride levels compared with metformin monotherapy.
Second-generation sulfonylureas had similar effects on triglyceride levels compared with repaglinide and acarbose.
Indirect comparisons and 1 direct comparison showed that pioglitazone decreased triglyceride levels more than acarbose (approximately 0.34mmol/L [30 mg/dL]).
Rosiglitazone increased triglyceride levels when compared indirectly with metformin and acarbose, yet had similar effects on triglyceride levelswhen compared directly with metformin.
According to indirect and a few direct comparisons, metformin showed similar effects on triglyceride levels when compared with acarbose.
Too few comparisons were available for nateglinide to draw conclusions.
Thiazolidinediones, second-generation sulfonylureas, and combinations of metformin plus second-generation sulfonylureas consistently increasedbody weight by 1 to 5 kg when compared directly with metformin,which was weight-neutral in placebo-controlled trials.
Repaglinide had similar effects on body weight compared with second-generation sulfonylureas. There were too few comparisons ofrepaglinide with other oral diabetes medications to draw conclusions.
Thiazolidinediones and second-generation sulfonylureas caused similar weight gain (approximately 3 kg) when used as monotherapy or incombination therapy with other oral diabetes medications.
Thiazolidinediones caused weight gain (approximately 3 kg) compared with acarbose and repaglinide, according to indirect comparisons ofplacebo-controlled trials and a few direct comparisons.
Acarbose compared with sulfonylureas showed no statistically significant differences in weight, but there was a suggestion of differencesbetween groups in the direct comparisons. The indirect comparisonsshowed that sulfonylureas were associated with weight gain comparedwith acarbose, which was weight-neutral.
According to a few head-to-head comparisons and indirect comparisons, acarbose had similar weight effects compared with metformin.
There were too few comparisons of nateglinide with other oral diabetes medication to evaluate its effect on weight.
*HbA ⫽ hemoglobin A ; HDL ⫽ high-density lipoprotein; LDL ⫽ low-density lipoprotein; RCT ⫽ randomized, controlled trial.
† Evidence was rated as follows: high ⫽ further research is very unlikely to change our confidence in the estimates; moderate ⫽ further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate; low ⫽ further research is likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate; very low ⫽ any estimate of effect is very uncertain; insufficient ⫽ not graded if too few comparisons (⬍3 studies) and not a key comparison of interest.
‡ Evidence was graded as very low for the following comparisons related to blood pressure effects: metformin versus metformin plus sulfonylurea, sulfonylurea versussulfonylurea plus thiazolidinedione, meglitinides versus sulfonylureas, and ␣-glucosidase inhibitors versus all other oral diabetes medications.
§ Evidence was graded as moderate to low for rosiglitazone plus metformin and for second-generation sulfonylureas compared with monotherapy. The rest of the comparisonswere graded as moderate.
390 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review The interim analysis of the RECORD (Rosiglitazone mainly middle-aged, overweight or obese adults of Euro- Evaluated for Cardiac Outcomes and Regulation of Gly- pean ancestry who had had diabetes for more than 2 years caemia in Diabetes) study reported that rosiglitazone plus and no major comorbid conditions. Mean baseline hemo- metformin or a sulfonylurea compared with metformin globin A1c levels ranged from 6% to 12% but were typi- plus a sulfonylurea had a hazard ratio of 1.08 (95% CI, cally between 7% and 9%. About two thirds of studies 0.89 to 1.31) for the primary end point of hospitalization received pharmaceutical industry support. Only 22 (16%) or death from cardiovascular disease (29). The hazard ratio trials described their randomization techniques, and 83 was driven by more congestive heart failure in the rosigli- (61%) reported double-blinding. In 33 (24%) studies, tazone plus metformin or sulfonylurea group than in the losses to follow-up and reasons for withdrawals were not control group of metformin plus sulfonylurea (absolute risk, 1.7% vs. 0.8%, respectively). In Kaplan–Meier curves, Hemoglobin A1c Level. Figure 2 shows the comparative
the risk for hospitalization or death from myocardial in- effects of oral diabetes agents on hemoglobin A1c. Thiazol- farction was slightly lower in the control group than in the idinediones, second-generation sulfonylureas, and met- rosiglitazone group, but the difference was not statistically formin produced similar reductions in hemoglobin A1c lev- significant. A limitation of this interim analysis was the els when used as monotherapy (absolute reduction, about 1 lack of power to detect differences, owing to fewer cardio- percentage point). Repaglinide produced similar reductions vascular events than initially predicted.
in hemoglobin A1c levels compared with sulfonylureas.
Combination therapies had additive effects, producing an Comparative Effectiveness of Oral Diabetes Agents in
absolute reduction in hemoglobin A Improving Intermediate Outcomes
1c levels of about 1 percentage point more than monotherapy.
Summary of Evidence
The strength of evidence was moderate to high that The results of these meta-analyses were generally con- most oral agents (thiazolidinediones, metformin, and repa- sistent with results of the UKPDS, a multicenter random- glinide) improved glycemic control to the same degree as ized trial starting in 1977 that had minimal loss to follow-up sulfonylureas (decrease in hemoglobin A (3). After 3 months of dietary intervention, participants 1c level, about 1 absolute percentage point). Nateglinide and ␣-glucosidase were stratified by ideal body weight and randomly assigned inhibitors may have slightly weaker effects on hemoglobin to receive insulin, chlorpropamide, glibenclamide, or di- etary intervention alone. Overweight participants were also 1c levels on the basis of indirect comparisons of placebo- controlled trials (low strength of evidence). The strength of randomly allocated to metformin. All agents had similar evidence was moderate that, compared with most other effects on hemoglobin A1c levels. After 10 years, gliben- oral agents, thiazolidinediones had a beneficial effect on clamide and metformin had a statistically insignificant be- HDL cholesterol levels (relative mean increase, 0.08 to tween-group absolute difference of 0.3 percentage point (3, 0.13 mmol/L [3 to 5 mg/dL]) but a harmful effect on LDL cholesterol levels (relative mean increase, 0.26 mmol/L [10 Few head-to-head comparisons involved repaglinide, mg/dL]). Metformin decreased LDL cholesterol levels by nateglinide, or ␣-glucosidase inhibitors. To evaluate these about 0.26 mmol/L (10 mg/dL), whereas other oral agents agents, we therefore relied on indirect comparisons with had no obvious effect on LDL cholesterol levels. The placebo controls. Repaglinide produced similar reductions strength of evidence was moderate that thiazolidinediones, in hemoglobin A1c levels (about 1 absolute percentage second-generation sulfonylureas, and metformin had simi- point) when compared indirectly with thiazolidinediones larly minimal effects on systolic blood pressure. There was and metformin. In contrast, nateglinide and ␣-glucosidase moderate evidence that most agents other than metformin inhibitors produced weaker reductions in hemoglobin A1c increased body weight by about 1 to 5 kg. Metformin had levels (about 0.5 absolute percentage point). Appendix Ta-
no effect on body weight in placebo-controlled trials.
ble 1 (available at shows findings for pla-
Table 1 shows evidence grades and a summary of the
cebo-controlled trials and the full report on indirect com- comparative conclusions. These studies applied primarily parisons (20).
to patients with type 2 diabetes and no major comorbid Blood Pressure. Figure 2 shows the comparative effects
of oral diabetes agents on blood pressure. Thiazolidinedi-ones, second-generation sulfonylureas, and metformin hadsimilarly minimal effects on systolic blood pressure (mean Characteristics and Quality of Studies of Intermediate
decrease ⬍5 mm Hg). The greatest contrast was between thiazolidinediones and sulfonylureas—the former agent The full report (20) provides a list of references and produced a 3–mm Hg greater reduction— but this differ- detailed evidence tables. We found 136 randomized trials ence was not statistically significant. Too few comparisons that addressed intermediate outcomes and a systematic re- of meglitinides and acarbose with other oral diabetes agents view on acarbose versus other oral diabetes agents (20).
in terms of blood pressure were available to draw firm Study duration ranged from 12 weeks to 10 years, but conclusions. Results were similar for diastolic blood pres- most studies lasted 24 weeks or less. Participants were sure (data not shown) (20).
18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 391
Review Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Figure 2. Weighted mean difference in blood pressure, laboratory values, and body weight with use of oral medications for type 2
diabetes mellitus.

Drug 1 More Beneficial
Drug 1 Less Beneficial
Drug 1 Less Beneficial
Drug 1 More Beneficial
Glyb vs. other SU
–0.03 (–0.13 to 0.07)
Met vs. SU
0.20 (–0.44 to 0.83)
12 (1749)
TZD vs. SU
–0.05 (–0.13 to 0.02)
11 (2828)
Met vs. Met + SU
0.38 (–1.35 to 2.1)
TZD vs. Met
–0.04 (–0.23 to 0.15)
Met + SU vs. SU
0.58 (–0.58 to 1.74)
Repag vs. SU
–0.06 (–0.30 to 0.18)
SU vs. Met
–0.09 (–0.30 to 0.10)
18 (2494)
Repag vs. SU
0.60 (–0.77 to 2.09)
SU vs. Acarbose
–0.38 (–0.77 to 0.02)
Acarbose vs. SU
0.78 (–0.78 to 2.34)
Met + TZD vs. Met
–0.62 (–1.0 to –0.23)
Pio vs. Met
3.06 (2.17 to 3.95)
SU + TZD vs. SU
–1.0 (–1.30 to –0.69)
Met + Rosi vs. Met
3.20 (2.14 to 4.26)
Met + SU vs. Met
–1.0 (–1.34 to –0.76)
11 (2139)
Met + SU vs. SU
–1.0 (–1.34 to –0.67)
11 (2335)
Pio vs. SU
4.95 (4.05 to 5.86)
–1.5 –1.0 –0.5 Weighted Mean Difference in
Weighted Mean Difference in HDL
Hemoglobin A Value, %
Cholesterol Level, mg/dL
Drug 1 More Beneficial
Drug 1 Less Beneficial
Drug 1 More Beneficial
Drug 1 Less Beneficial
TZD vs. Met
–0.1 (–2.7 to 2.5)
Met + Rosi vs. Met
11.3 (6.8 to 15.8)
Repag vs. SU
–0.2 (–4.3 to 3.9)
SU vs. Acarbose
–0.9 (–17.8 to 16.0)
Met + SU vs. SU
–1.1 (–3.5 to 1.4)
Met + SU vs. Met
–6.9 (–14.9 to 1.2)
Met vs. SU
–9.1 (–16.0 to –2.1)
13 (1806)
Met vs. SU
–1.7 (–5.0 to 1.5)
Pio vs. SU
–24.2 (–45.7 to –2.6)
Pio vs. Met
–25.5 (–29.8 to –29.1)
TZD vs. SU
–3.1 (–6.6 to 0.4)
Met + SU vs. SU
–30.1 (–44.9 to –15.3)
Weighted Mean Difference in
Weighted Mean Difference in
Systolic Blood Pressure, mm Hg
Triglyceride Level, mg/dL
Drug 1 More Beneficial
Drug 1 Less Beneficial
Drug 1 More Beneficial
Drug 1 Less Beneficial
SU vs. Met
Met + Rosi vs. Met
14.6 (13.3 to 15.8)
(RCTs 24 wk)
3.5 (3.0 to 4.0)
Pio vs. Met
12.5 (8.8 to 16.2)
Met + SU vs. Met
2.4 (1.1 to 3.6)
SU vs. Met
Pio vs. SU
10.4 (7.3 to 13.6)
(RCTs <24 wk )
1.9 (1.4 to 2.4)
Met vs. Met + SU
–1.6 (–6.6 to 3.3)
TZD vs. Met
1.9 (0.5 to 3.3)
SU vs. Acarbose
1.9 (0.2 to 4.0)
SU vs. Acarbose
–3.9 (–10.5 to 2.7)
TZD vs. SU
1.1 (–0.9 to 3.1)
Met + SU vs. SU
–8.1 (–13.1 to –3.1)
SU vs. Met + SU
0.05 (–0.5 to 0.6)
Met vs. SU
–10 (–13.1 to –6.9)
SU vs. Repag
0.03 (–1.0 to 1.0)
10 (2006)
Weighted Mean Difference in LDL
Weighted Mean Difference in
Cholesterol Level, mg/dL
Body Weight, kg
Error bars represent 95% CIs. To convert cholesterol and triglyceride values to mmol/L, multiply by 0.0259 and 0.0113, respectively. Glyb ⫽ glyburide; HDL ⫽ high-density lipoprotein; LDL ⫽ low-density lipoprotein; Met ⫽ metformin; Pio ⫽ pioglitazone; RCT ⫽ randomized, controlled trial; Repag ⫽ repaglinide; Rosi ⫽ rosiglitazone; SU ⫽ sulfonylurea; TZD ⫽ thiazolidinedione.
392 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review Plasma Lipid Levels. Figure 2 shows the comparative
tinal problems (flatus, nausea, vomiting, and abdominal effects of oral diabetes agents on plasma lipid levels. Met- pain) than were most other oral diabetes agents. However, formin decreased LDL cholesterol levels by about 0.26 rates of lactic acidosis were similar between metformin and mmol/L (10 mg/dL), whereas thiazolidinediones consis- other oral diabetes agents, according to a systematic review tently increased LDL cholesterol levels by a relative mean of 176 comparative trials (35).
of 0.26 mmol/L (10 mg/dL). Sulfonylureas had similar In many randomized trials, thiazolidinediones were as- minimal effects on LDL cholesterol compared with acar- sociated with higher risk for edema than were sulfonylureas bose or repaglinide (33, 34).
or metformin (absolute risk difference, 2% to 21%). Other Thiazolidinediones increased HDL cholesterol levels than edema and hypoglycemia, we had difficulty assessing by a mean of 0.08 to 0.13 mmol/L (3 to 5 mg/dL) com- harms associated with thiazolidinediones because there pared with metformin or second-generation sulfonylureas; were few trials and events. In addition, cohort studies often these latter agents had little effect on HDL cholesterol.
did not adjust for key confounders. Thiazolidinediones ap- Combination therapy with thiazolidinediones increased peared to confer a higher risk for congestive heart failure HDL cholesterol levels similarly to monotherapy with thia- (although absolute risks were small— generally 1% to 3%) zolidinediones. Repaglinide and acarbose had little effect and higher risk for mild anemia yet produced similarly low on HDL cholesterol compared with second-generation rates of elevated aminotransferase levels (⬍1%) compared with sulfonylureas and metformin.
Only rosiglitazone increased triglyceride levels, by a Few studies compared the effect of meglitinides with mean of 0.11 mmol/L (10 mg/dL) in placebo-controlled that of other oral diabetes agents for outcomes other than trials (data not shown). Pioglitazone decreased triglyceride hypoglycemia. Most studies on adverse effects were appli- levels more than metformin, by a mean of 0.29 mmol/L cable to persons without major cardiovascular, renal, or (26 mg/dL), and metformin decreased triglyceride levelsmore than second-generation sulfonylureas, by a mean of hepatic comorbid conditions.
0.11 mmol/L (10 mg/dL). Repaglinide and acarbose pro-duced similar reductions in triglyceride levels, by a mean of0.11 to 0.34 mmol/L (10 to 30 mg/dL) compared with Characteristics and Quality of Studies on Adverse Events
Overall, 167 original articles and 2 Cochrane system- Data on nateglinide were too sparse to draw conclu- atic reviews evaluated adverse events (the full report pro- sions about its comparative effects on lipid levels.
vides a list of references and detailed evidence tables) (20).
Body Weight. Compared with sulfonylureas, thiazol- About two thirds of the studies were randomized, con- idinediones and repaglinide produced similar gains in body trolled trials, and the rest were observational. Most were weight (1 to 5 kg). Metformin produced no weight gain based in the United States or Europe and had industry compared with most other oral agents or placebo (Figure 2
support. Study duration varied from 3 months to more and Appendix Table 2), and acarbose produced no weight
than 10 years. Most participants were middle-aged to older gain compared with placebo (Appendix Table 2).
adults of European ancestry who were overweight or obese.
Three UKPDS articles reported weight changes that The duration of diabetes ranged from 1 year to 15 years, were consistent with these results favoring metformin over and mean baseline hemoglobin A1c levels were typically sulfonylurea (mean relative decrease, 2 kg at 10 years of between 7% and 9%. Most randomized, controlled trials follow-up) (3, 30, 32). Most of the weight gain in the excluded people with major cardiovascular, hepatic, or re- glibenclamide group occurred in the first 2 years, whereas nal disease.
the metformin group maintained body weight in the first 2 Eighty-five percent (105 of 123) of the randomized, years and then experienced weight gain (3).
controlled trials with data relevant to adverse events didnot describe the randomization technique in sufficient de- Comparative Risk for Adverse Events with Oral Diabetes
tail to determine whether the randomization was appropri- ate. About two thirds (66%) of these trials were reported as Summary of Evidence
double-blind. However, 90 (73%) of these trials did not Several randomized, controlled trials and some obser- describe the masking procedure. Twenty-two (18%) trials vational studies consistently demonstrate that minor and did not report on withdrawals or losses to follow-up.
major hypoglycemic episodes are more frequent in adults Hypoglycemia. Minor and major hypoglycemic epi- receiving second-generation sulfonylureas (especially gly- sodes were more frequent in patients receiving second- buride) than in those receiving metformin or thiazol- generation sulfonylureas (especially glyburide) than in idinediones. Repaglinide and second-generation sulfonyl- those receiving metformin or thiazolidinediones. Absolute ureas conferred similar risks for hypoglycemia.
risk differences between groups ranged from 4% to 9% In many trials and a few observational studies, met- when sulfonylureas were compared with metformin or formin was almost always associated with more gastrointes- thiazolidinediones in short-term randomized trials, al- 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 393
Review Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Figure 3. Pooled hypoglycemia results for randomized trials, by drug comparison.
Drug 1 Less Harmful
Drug 1 More Harmful
Met vs. Met + TZD
0.00 (–0.01 to 0.01)
SU vs. Repag
0.02 (–0.02 to 0.05)
Glyb vs. Other SU
0.03 (0.00 to 0.05)
SU vs. Met
0.04 (0.00 to 0.09)
SU + TZD vs. SU
0.08 (0.00 to 0.16)
SU vs. TZD
0.09 (0.03 to 0.15)
SU + Met vs. SU
0.11 (0.07 to 0.14)
SU + Met vs. Met
0.14 (0.07 to 0.21)
Weighted Absolute Risk Difference
Error bars represent 95% CIs. Glyb ⫽ glyburide; Met ⫽ metformin; Repag ⫽ repaglinide; SU ⫽ sulfonylurea; TZD ⫽ thiazolidinedione.
though reported levels of hypoglycemic risk ranged widely parative studies were available on nateglinide to draw firm across studies: 0% to 36% for second-generation sulfonyl- conclusions (Table 2).
ureas, 0% to 21% for metformin, and 0% to 24% for Elevated Aminotransferase Levels and Liver Failure. Currently marketed thiazolidinediones, second-generation The 10-year follow-up from UKPDS reported the an- sulfonylureas, and metformin had similarly low rates (gen- nual rates of minor and major hypoglycemia as 17.5% and erally ⬍1%) of clinically significant elevated aminotrans- 2.5%, respectively, in the glibenclamide group (obese and ferase levels (⬎1.5 to 2 times the upper limit of normal).
nonobese persons) and 4.2% and 0%, respectively, in the An insufficient number of studies evaluated or reported on metformin group (obese persons only). Results from obser- the effects of meglitinides on serum aminotransferase lev- vational studies were consistent with those of the UKPDS.
els, but they appeared to be similar to the effects of other Glyburide and glibenclamide conferred a slightly oral diabetes agents (Table 2). Liver failure was so rare that
higher risk for hypoglycemia compared with other second- agents could not be compared for this outcome by using generation sulfonylureas (absolute risk difference, about 2% in trials of short duration). Repaglinide and second- Congestive Heart Failure. Risk for congestive heart fail- generation sulfonylureas conferred similar risks for hypo- ure was greater with thiazolidinediones as monotherapy or glycemia. Comparative data on acarbose and nateglinide combination therapy than with metformin or sulfonylureas were sparse. The incidence of minor and major hypoglyce- (range of absolute risk differences, 0.7% to 2.2% in head- mia was higher with combinations that included sulfonyl- to-head, short-duration randomized trials). The absolute ureas compared with metformin or sulfonylurea mono- risk for congestive heart failure in the trials ranged from therapy (absolute risk differences of 8% to 14% for short- 0.8% to 3.6% for thiazolidinediones and 0% to 2.6% for duration trials) (Figure 3).
nonthiazolidinediones. In contrast, neither metformin nor Gastrointestinal Problems. Metformin produced more second-generation sulfonylureas were associated with con- gastrointestinal symptoms (range, 2% to 63%) than most gestive heart failure risk in 2 of 3 observational studies and other oral diabetes agents (range, 0% to 36% for thiazo- 2 of 2 placebo-controlled trials. Congestive heart failure lidinediones, 0% to 32% for second-generation sulfonyl- was reported mostly in cohort studies that did not adjust ureas, and 8% to 11% for repaglinide). The absolute risk for key confounders, such as duration of diabetes, hemo- differences among groups ranged from 0% to 31%, al- globin A1c level, blood pressure, and medication adher- though most were between 5% and 15%. Acarbose pro- ence. However, the cohort studies were consistent with one duced percentages of gastrointestinal symptoms (range, another and with limited data from randomized trials 15% to 30%) similar to those with metformin and higher (Table 2).
than those with thiazolidinediones and sulfonylureas in a Peripheral Edema. Edema was more frequent in pa- few trials (⬍3 trials for each comparison). Too few com- tients receiving thiazolidinediones as monotherapy or com- 394 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review bination therapy (range, 0% to 26%) than in patients re- fatal lactic acidosis in any medication group. The estimated ceiving second-generation sulfonylureas (range, 0% to 8%) hypothetical upper limit of the underlying incidence of or metformin (range, 0% to 4%). The absolute risk differ- lactic acidosis was 8.4 cases per 100 000 patient-years in ences ranged from 2% to 21% in head-to-head random- the metformin group and 9 cases per 100 000 patient-years ized trials (Table 2).
in the nonmetformin group (35). We found 8 additional Lactic Acidosis. We found a systematic review that re- studies with data on lactic acidosis (3 randomized trials and ported similar rates of lactic acidosis between metformin 5 cohort studies). All showed little or no elevated risk for and other oral diabetes agents (35). In this review, pooled lactic acidosis in metformin recipients (Table 2).
data from 176 comparative trials and cohort studies total- Anemia, Leukopenia, and Thrombocytopenia. Six head- ing 35 619 patient-years revealed no cases of fatal or non- to-head randomized trials, 7 placebo-controlled random- Table 2. Adverse Effects Related to Oral Diabetes Medications in Head-to-Head Comparisons*
Study Type
Range in Risk
Congestive heart failure
TZD vs. sulfonylurea Sulfonylurea ⫹ TZD vs. sulfonylurea TZD vs. metformin TZD vs. sulfonylurea TZD vs. meglitinides Metformin ⫹ TZD vs. metformin Sulfonylurea ⫹ TZD vs. sulfonylurea Metformin vs. TZD ⫺26% (metformin vs. pioglitazone) to 10.4% (metformin vs. rosiglitazone)¶ Metformin vs. sulfonylurea Sulfonylurea vs. TZD** Metformin vs. meglitinides Metformin vs. metformin ⫹ sulfonylurea Metformin ⫹ sulfonylurea vs. sulfonylurea Aminotransferase levels >1.5 times the
upper limit of normal
TZD vs. metformin Sulfonylurea vs. TZD Meglitinides vs. TZD Metformin vs. metformin ⫹ TZD Sulfonylurea vs. sulfonylurea ⫹ TZD Metformin vs. nonmetformin Systematic review * Head-to-head comparisons for which more than 1 study was available are included. RCT ⫽ randomized, controlled trial; TZD ⫽ thiazolidinedione.
† Studies with available data on risk estimates (differences in percentage of adverse events).
‡ Reported as the number of study participants, unless otherwise indicated.
§ Reported as the percentage of risk difference, unless otherwise indicated.
㛳 Odds ratio.
¶ One study showed both pioglitazone versus metformin and rosiglitazone versus metformin, which had different results.
** In 2 of these trials, metformin was added to both the TZD and sulfonylurea groups.
†† Patient-years. There were 8.4 cases per 100 000 patient-years.
18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 395
Review Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus ized trials, and 1 cohort study evaluated anemia as an out- points. Thiazolidinediones, although they pose a lower risk come. Thiazolidinediones may be associated with an for hypoglycemia and a slight beneficial effect on HDL increased risk for anemia compared with other oral diabe- cholesterol level, showed no advantage in glucose-lowering tes agents (posttreatment absolute risk differences, 1% to effect and were associated with adverse effects on LDL 5%). The mean decrease in hemoglobin level was small cholesterol level, body weight, and risk for congestive heart (ⱕ1 g/dL). Only 1 study reported an adverse event of thrombocytopenia and leukopenia.
These findings support the current American Diabetes Serious Allergic Reactions. No study reported an allergic Association and International Diabetes Federation recom- reaction to oral diabetes medications that led to hospital- mendations that favor metformin as initial pharmacother- ization or death.
apy for type 2 diabetes (36, 37). They are also consistentwith the 2007 American College of Endocrinology guide- Unpublished Data on Harms
lines that suggest choosing an oral diabetes agent on the In addition to data published in peer-reviewed jour- basis of the individual patient's burden of comorbid con- nals, we reviewed data from the FDA, unpublished trials ditions (38). Of course, optimal glycemic control often conducted by industry, and clinical trial registries. The requires multidrug therapy. Our review confirms that a only new finding was that pioglitazone was associated with second agent is additive both in terms of improved glyce- an increased risk for hospitalization for acute cholecystitis mic control and increased risk for adverse events, unless (12 patients) compared with placebo (1 patient) in a both agents are used at lower doses. Although they are not pooled analysis of 1526 patients (20). Otherwise, unpub- clearly superior to newer agents, sulfonylureas remain a lished data were consistent with those from the published reasonable alternative as second-line therapy, especially if cost is an issue.
Our findings are generally consistent with those of previous reviews of the effects of oral diabetes agents on intermediate outcomes, such as hemoglobin A We did not find strong evidence of possible publica- levels, and body weight (10, 12, 14, 16, 18, 19, 39, 40).
tion bias. Only 2 drug comparisons, from studies of hypo- Inzucchi (18) conducted a systematic review of the effect of glycemia, had statistically significant results for publication oral diabetes agents and placebo on hemoglobin A bias (P ⬍ 0.05) according to the less conservative test of drew conclusions similar to ours. Our study adds to this Egger and colleagues (27): metformin versus second-gener- research by including more recent articles, comparisons in- ation sulfonylureas (8 studies; P ⫽ 0.04) and repaglinide volving meglitinides, and meta-analyses of head-to-head versus placebo (3 studies; P ⫽ 0.035). The 3 largest studies comparisons. In a 2002 systematic review (without quan- in the comparison of metformin with sulfonylureas had titative meta-analyses) on the lipid effects of oral diabetes smaller absolute risk differences than the smaller studies; medications, Buse and coworkers (19) reported findings however, all studies showed that metformin is associated similar to ours. Our investigation updates their review and with less hypoglycemia than sulfonylureas. There were too adds more detail on differences between drugs from formal few studies in the comparison of repaglinide versus placebo meta-analyses. The main contribution of our review is its to draw conclusions about publication bias. For all other comprehensiveness: We included a broad range of clini- comparisons, the funnel plots appeared to be roughly sym- cally relevant outcomes and adverse effects across all avail- metrical, and results of the tests of Begg and Mazumdar able drug classes.
(26) and Egger and colleagues (27) were not statistically Nissen and Wolski (11) recently reported results of a meta-analysis suggesting a relationship between use of ros-iglitazone and risk for myocardial infarction. When they analyzed specific drug– drug or drug–placebo comparisons, Ideally, oral diabetes agents should improve micro- however, their results were not statistically significant.
vascular and macrovascular outcomes and mortality. We Likewise, we found no statistically significant differences found no definitive comparative evidence on these out- between specific oral diabetes medications in terms of car- comes. Because of this uncertainty, we evaluated medica- diovascular outcomes other than congestive heart failure.
tion effects on intermediate outcomes and other adverse Limitations of Nissen and Wolski's study included the events. By these criteria, we found that metformin was small number of largely unadjudicated events and the fact similar to, or better than, other currently available oral that cardiovascular events were not the primary outcome.
agents. Second-generation sulfonylureas also fared well An additional limitation that influenced their conclusions against other agents, apart from the increased risk for hypo- was the decision to include studies with 2 diverse patient glycemia. Compared with newer agents, metformin and samples: nondiabetic persons, in whom the risk-to-benefit second-generation sulfonylureas share 3 additional advan- ratio of an oral diabetes agent may differ greatly from that tages: lower cost, longer use in practice, and more intensive in their diabetic counterparts, and diabetic persons with scrutiny in long-term trials with clinically relevant end congestive heart failure, for whom rosiglitazone is contra- 396 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review indicated. The decision to include these studies may have Our study has limitations. First, most of the trials, biased the meta-analysis toward showing harm. Finally, ex- especially those of newer agents, were short-term trials, clusion of studies with no cardiovascular events in either generally lasting less than 1 year. Ideally, therapeutic deci- group introduced a small bias against finding no difference sion making should be based on long-term effectiveness.
in cardiovascular risk. Given the limitations of Nissen and Second, head-to-head data were limited in many instances.
Wolski's analysis, the effects of rosiglitazone on cardiovas- This was especially true for multidrug regimens now in cular mortality and myocardial infarction are still uncer- common use and for some of the newer agents, such as tain. A recently published interim analysis from the rosiglitazone, nateglinide, and miglitol. Third, although al- RECORD study showed no statistically significant eleva- most all studies reported the incidence of hypoglycemia, tion in cardiovascular risk (besides congestive heart failure) reporting of other adverse events was inconsistent, and the related to rosiglitazone compared with metformin and sul- definitions of adverse events varied across studies. For in- fonylureas (20). Overall, these recent findings are consis- stance, gastrointestinal events could include nausea, vom- tent with ours: We found no conclusive evidence of worse iting, abdominal pain, flatulence, or a combination of these cardiovascular morbidity or mortality with oral diabetes events, making comparisons across studies difficult. Few agents, other than the higher risk for congestive heart fail- trials reported data on elevated liver aminotransferase lev- ure with thiazolidinediones than with other oral medications.
els, congestive heart failure, anemia, and allergic reactions; Several adverse events merit further discussion. First, therefore, we relied on cohort studies for many of these because of concerns about lactic acidosis, metformin is outcomes. The available cohort studies, however, were lim- contraindicated in patients with impaired renal function or ited by their ability to adjust for key confounding factors, congestive heart failure. However, neither our review nor such as hemoglobin A1c level, blood pressure, duration of that of Salpeter and colleagues (35) found evidence of an diabetes, adherence to medications, and medication dose.
elevated risk for lactic acidosis in patients taking met- Finally, we focused on safety issues by making an a priori formin compared with other oral diabetes agents. The ev- hypothesis of potential harm, and we may have missed idence for metformin-induced lactic acidosis stems mainly harms reported only in case reports or those that were not from about 300 case reports. We did not consider case assessed in trials or observational studies.
reports in our review because they pose problems in deter- Compared with newer, more expensive agents (thia- mining causality and provide no clear denominator for risk zolidinediones, ␣-glucosidase inhibitors, and meglitinides), estimation. Underlying comorbid conditions, such as older agents (second-generation sulfonylureas and met- chronic kidney disease or myocardial infarction, are well- formin) have similar or superior effects on glycemic control established risk factors for lactic acidosis; therefore, attrib- and other cardiovascular risk factors (blood pressure, lipid uting lactic acidosis to metformin use versus an underlying levels, and body weight). Each oral diabetes agent is asso- comorbid condition is often difficult. Most reported cases ciated with adverse events that counterbalance its benefits.
of metformin-related lactic acidosis were associated with Overall, metformin seemed to have the best profile of ben- severe underlying illnesses (41, 42). Because of lingering efit to risk. Large, long-term comparative studies on major fears about biguanides (phenformin was unequivocally re- clinical end points, such as myocardial infarction, chronic lated to risk for lactic acidosis), we suspect that apparent kidney disease, and cardiovascular mortality, are needed to cases of "metformin-induced lactic acidosis" may have determine definitively the comparative effects of the oral been overreported. However, we could not rule out the diabetes agents, especially in light of recent controversy possibility that metformin conferred additional risk in the presence of severe underlying cardiac or renal disease, be-cause these conditions were excluded in most randomized From Johns Hopkins University School of Medicine, Johns Hopkins trials and were too uncommon in cohort studies to allow Bloomberg School of Public Health, Evidence-based Practice Center, and Welch Center for Prevention, Epidemiology, and Clinical Research,Johns Hopkins University, Baltimore, Maryland, and Washington Uni- Second, macular edema has been mentioned as an ad- versity School of Medicine, St. Louis, Missouri.
verse event related to use of rosiglitazone only in case re-ports (43), which we excluded from our review. Third, the Disclaimer: The authors of this article are responsible for its contents,
ADOPT study (published after our review was completed) including any clinical or treatment recommendations. No statement in reported an increase in fracture risk in women taking ros- this article should be construed as an official position of the Agency for iglitazone compared with metformin or sulfonylureas (28).
Healthcare Research and Quality or of the U.S. Department of Health No cases were reported in the studies from our review, but and Human Services.
this will need further investigation. Finally, repaglinide Acknowledgment: The authors thank Steven Fox for his help as the
may be associated with less serious hypoglycemia compared Task Order Officer.
with second-generation sulfonylureas, as was seen in 1study of elderly persons (44), and in patients who skip Grant Support: This article is based on research conducted by the Johns
meals, as was seen in 1 randomized trial not included in Hopkins Evidence-based Practice Center under contract number 290- our review (because it was ⬍3 months in duration) (45).
02-0018 with the Agency for Healthcare Research and Quality. Dr.
18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 397
Review Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Brancati was supported by a mid-career investigator award for patient- 17. van Wijk JP, de Koning EJ, Martens EP, Rabelink TJ. Thiazolidinediones
oriented research in diabetes from the National Institute of Diabetes and and blood lipids in type 2 diabetes. Arterioscler Thromb Vasc Biol. 2003;23: Digestive and Kidney Diseases (5 K24 DK062222-05).
1744-9. [PMID: 12907465]
18. Inzucchi SE. Oral antihyperglycemic therapy for type 2 diabetes: scientific
review. JAMA. 2002;287:360-72. [PMID: 11790216]
Potential Financial Conflicts of Interest: None disclosed.
19. Buse JB, Tan MH, Prince MJ, Erickson PP. The effects of oral anti-hyper-
glycaemic medications on serum lipid profiles in patients with type 2 diabetes.
Requests for Single Reprints: Shari Bolen, MD, MPH, Welch Center
Diabetes Obes Metab. 2004;6:133-56. [PMID: 14746579] for Prevention, Epidemiology, and Clinical Research, Johns Hopkins 20. Bolen S, Wilson L, Vassy J, Feldman L, Yeh J, Marinopoulos S, et al.
University, 2024 East Monument Street, Suite 2-600, Room 2-615, Comparative Effectiveness and Safety of Oral Diabetes Medications for Adults Baltimore, MD 21205; e-mail,
with Type 2 Diabetes. Rockville, MD: Agency for Healthcare Research andQuality; 2007. AHRQ Publication no. 07-EHC010-EF. Available at
Current author addresses are available at
21. Jadad AR, Moore RA, Carroll D, Jenkinson C, Reynolds DJ, Gavaghan DJ,
et al.
Assessing the quality of reports of randomized clinical trials: is blinding
necessary? Control Clin Trials. 1996;17:1-12. [PMID: 8721797]
22. Deeks JJ, Dinnes J, D'Amico R, Sowden AJ, Sakarovitch C, Song F, et al;
1. Engelgau MM, Geiss LS, Saaddine JB, Boyle JP, Benjamin SM, Gregg EW,
International Stroke Trial Collaborative Group. Evaluating non-randomised
et al. The evolving diabetes burden in the United States. Ann Intern Med.
intervention studies. Health Technol Assess. 2003;7:iii-x, 1-173. [PMID: 2004;140:945-50. [PMID: 15172919] 2. World Health Organization. Diabetes Programme. Facts and Figures. Ac-
23. Atkins D, Best D, Briss PA, Eccles M, Falck-Ytter Y, Flottorp S, et al;
cessed at on 11 June 2007.
GRADE Working Group. Grading quality of evidence and strength of recom-
3. Effect of intensive blood-glucose control with metformin on complications in mendations. BMJ. 2004;328:1490. [PMID: 15205295] overweight patients with type 2 diabetes (UKPDS 34). UK Prospective Diabetes 24. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials.
Study (UKPDS) Group. Lancet. 1998;352:854-65. [PMID: 9742977] 1986;7:177-88. [PMID: 3802833] 4. Intensive blood-glucose control with sulphonylureas or insulin compared with 25. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency
conventional treatment and risk of complications in patients with type 2 diabetes(UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet. 1998; in meta-analyses. BMJ. 2003;327:557-60. [PMID: 12958120] 352:837-53. [PMID: 9742976] 26. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test
5. Gaster B, Hirsch IB. The effects of improved glycemic control on complica-
for publication bias. Biometrics. 1994;50:1088-101. [PMID: 7786990] tions in type 2 diabetes. Arch Intern Med. 1998;158:134-40. [PMID: 9448551] 27. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis
6. Ohkubo Y, Kishikawa H, Araki E, Miyata T, Isami S, Motoyoshi S, et al.
detected by a simple, graphical test. BMJ. 1997;315:629-34. [PMID: 9310563] Intensive insulin therapy prevents the progression of diabetic microvascular com- 28. Kahn SE, Haffner SM, Heise MA, Herman WH, Holman RR, Jones NP
plications in Japanese patients with non-insulin-dependent diabetes mellitus: a et al; ADOPT Study Group. Glycemic durability of rosiglitazone, metformin, or
randomized prospective 6-year study. Diabetes Res Clin Pract. 1995;28:103-17.
glyburide monotherapy. N Engl J Med. 2006;355:2427-43. [PMID: 17145742] 29. Home PD, Pocock SJ, Beck-Nielsen H, Gomis R, Hanefeld M, Jones NP,
7. Vijan S, Hofer TP, Hayward RA. Estimated benefits of glycemic control in
et al. Rosiglitazone evaluated for cardiovascular outcomes—an interim analysis.
microvascular complications in type 2 diabetes. Ann Intern Med. 1997;127:788- N Engl J Med. 2007;357:28-38. [PMID: 17551159] 95. [PMID: 9382399] 30. United Kingdom Prospective Diabetes Study Group. United Kingdom Pro-
8. Pitale SU, Abraira C, Emanuele NV, McCarren M, Henderson WG, Pacold
spective Diabetes Study 24: a 6-year, randomized, controlled trial comparing I, et al. Two years of intensive glycemic control and left ventricular function in
sulfonylurea, insulin, and metformin therapy in patients with newly diagnosed the Veterans Affairs Cooperative Study in Type 2 Diabetes Mellitus (VA type 2 diabetes that could not be controlled with diet therapy. Ann Intern Med.
CSDM). Diabetes Care. 2000;23:1316-20. [PMID: 10977025] 1998;128:165-75. [PMID: 9454524] 9. Stettler C, Allemann S, Ju¨ni P, Cull CA, Holman RR, Egger M, et al.
31. U.K. Prospective Diabetes Study. II. Reduction in HbA1c with basal insulin Glycemic control and macrovascular disease in types 1 and 2 diabetes mellitus: supplement, sulfonylurea, or biguanide therapy in maturity-onset diabetes. A meta-analysis of randomized trials. Am Heart J. 2006;152:27-38. [PMID: multicenter study. Diabetes. 1985;34:793-8. [PMID: 2862087] 32. United Kingdom Prospective Diabetes Study (UKPDS). 13: Relative efficacy 10. Saenz A, Fernandez-Esteban I, Mataix A, Ausejo M, Roque M, Moher D.
of randomly allocated diet, sulphonylurea, insulin, or metformin in patients with Metformin monotherapy for type 2 diabetes mellitus. Cochrane Database Syst newly diagnosed non-insulin dependent diabetes followed for three years. BMJ.
Rev. 2005:CD002966. [PMID: 16034881] 1995;310:83-8. [PMID: 7833731] 11. Nissen SE, Wolski K. Effect of rosiglitazone on the risk of myocardial in-
33. Derosa G, Mugellini A, Ciccarelli L, Crescenzi G, Fogari R. Comparison
farction and death from cardiovascular causes. N Engl J Med. 2007;356:2457-71.
between repaglinide and glimepiride in patients with type 2 diabetes mellitus: a one-year, randomized, double-blind assessment of metabolic parameters and car- 12. Chilcott J, Tappenden P, Jones ML, Wight JP. A systematic review of the
diovascular risk factors. Clin Ther. 2003;25:472-84. [PMID: 12749508] clinical effectiveness of pioglitazone in the treatment of type 2 diabetes mellitus.
34. Marbury T, Huang WC, Strange P, Lebovitz H. Repaglinide versus gly-
Clin Ther. 2001;23:1792-823; discussion 1791. [PMID: 11768834] buride: a one-year comparison trial. Diabetes Res Clin Pract. 1999;43:155-66.
13. Chandler C, Chou R, Helfand M. Drug Class Review on Oral Hypogly-
cemics. May 2005. Accessed at 35. Salpeter S, Greyber E, Pasternak G, Salpeter E. Risk of fatal and nonfatal
/documents/OH%20Final%20Report%20Update%2021.pdf on 22 Septem- lactic acidosis with metformin use in type 2 diabetes mellitus. Cochrane Database Syst Rev. 2003:CD002967. [PMID: 12804446] 14. Chiquette E, Ramirez G, Defronzo R. A meta-analysis comparing the effect
36. American Diabetes Association. Standards of medical care in diabetes—
of thiazolidinediones on cardiovascular risk factors. Arch Intern Med. 2004;164: 2007. Diabetes Care. 2007;30 Suppl 1:S4-S41. [PMID: 17192377] 2097-104. [PMID: 15505122] 37. Clinical Guidelines Task Force, International Diabetes Federation. Glucose
15. van de Laar FA, Lucassen PL, Akkermans RP, van de Lisdonk EH, Rutten
control: oral therapy. In: Global Guideline for Type 2 Diabetes. Brussels, Bel- GE, van Weel C. Alpha-glucosidase inhibitors for patients with type 2 diabetes:
gium: International Diabetes Federation; 2005. Accessed at results from a Cochrane systematic review and meta-analysis. Diabetes Care.
/docs/GGT2D%2009%20Oral%20therapy.pdf on 19 April 2007.
2005;28:154-63. [PMID: 15616251] 38. American Association of Clinical Endocrinologists, American College of
16. Wulffele´ MG, Kooy A, de Zeeuw D, Stehouwer CD, Gansevoort RT. The
Endocrinology. The American Association of Clinical Endocrinologists Medical
effect of metformin on blood pressure, plasma cholesterol and triglycerides in type Guidelines for Clinical Practice for the Management of Diabetes Mellitus. May/ 2 diabetes mellitus: a systematic review. J Intern Med. 2004;256:1-14. [PMID: June 2007. Accessed at DMGuidelines2007 .pdf on 2 July 2007.
398 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6
Effectiveness and Safety of Oral Medications for Type 2 Diabetes Mellitus Review 39. Qayyum R, Adomaityte J. A meta-analysis of the effect of thiazolidinediones
43. GlaxoSmithKline. [Letter to health care providers]. 2005. Accessed at www
on blood pressure. J Clin Hypertens (Greenwich). 2006;8:19-28. [PMID: on 18 October 40. Norris S, Carson S, Roberts C. Drug Class Review on Thiazolidinediones.
44. Papa G, Fedele V, Rizzo MR, Fioravanti M, Leotta C, Solerte SB, et al.
Final Report. May 2006. Accessed at Safety of type 2 diabetes treatment with repaglinide compared with glibenclamide /documents/_TZDs%20Final%20Report.pdf on 22 September 2006.
in elderly people: A randomized, open-label, two-period, cross-over trial. Diabetes 41. Brown JB, Pedula K, Barzilay J, Herson MK, Latare P. Lactic acidosis rates
Care. 2006;29:1918-20. [PMID: 16873803] in type 2 diabetes. Diabetes Care. 1998;21:1659-63. [PMID: 9773726] 45. Damsbo P, Clauson P, Marbury TC, Windfeld K. A double-blind random-
42. Misbin RI, Green L, Stadel BV, Gueriguian JL, Gubbi A, Fleming GA.
ized comparison of meal-related glycemic control by repaglinide and glyburide in Lactic acidosis in patients with diabetes treated with metformin [Letter]. N Engl well-controlled type 2 diabetic patients. Diabetes Care. 1999;22:789-94. [PMID: J Med. 1998;338:265-6. [PMID: 9441244] 18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 399
Annals of Internal Medicine
Current Author Addresses: Drs. Bolen, Yeh, Selvin, and Brancati:
Welch Center for Prevention, Epidemiology, and Clinical Research,
Johns Hopkins University, 2024 East Monument Street, Suite 2-600,
Baltimore, MD 21205.
Dr. Feldman: Johns Hopkins University, Jefferson Building, 600 North
Wolfe Street, Room 242, Baltimore, MD 21287.
Dr. Vassy: University of Pennsylvania Health System, 3400 Spruce
Street, Philadelphia, PA 19104.
Ms. L. Wilson, Ms. R. Wilson, and Drs. Wiley and Bass: Johns Hopkins
University, 1830 East Monument Street, Eighth Floor, Baltimore, MD
Dr. Marinopoulos: University Health Services, Johns Hopkins Univer-
sity, 401 North Caroline Street, Baltimore, MD 21231.
Appendix Table 1. Summary Measures: Weighted Mean Absolute Difference in Hemoglobin A
Level between Groups for
Randomized, Controlled Trials Comparing Oral Diabetes Medications with Placebo or Diet
Studies with Data on
Weighted Mean Absolute Difference
Mean Differences, n
in Hemoglobin A1c Level between
Groups (95% CI), %

Pioglitazone vs. control ⫺0.97 (⫺1.18 to ⫺0.75) Rosiglitazone vs. control ⫺1.16 (⫺1.39 to ⫺0.92) Metformin vs. control ⫺1.14 (⫺1.4 to ⫺0.87) Sulfonylureas vs. control ⫺1.52 (⫺1.75 to ⫺1.28) Repaglinide vs. control ⫺1.32 (⫺1.9 to ⫺0.8) Nateglinide vs. control ⫺0.54 (⫺0.8 to ⫺0.27) Acarbose vs. control ⫺0.77 (⫺0.9 to ⫺0.64) * The control group consisted of placebo or diet.
Appendix Table 2. Summary Measures: Weighted Mean Absolute Difference in Body Weight between Groups for Randomized,
Controlled Trials Comparing Oral Diabetes Medications with Placebo or Diet

Studies with Data on
Weighted Mean Absolute Difference
Mean Differences, n
in Body Weight between Groups
(95% CI), kg

Pioglitazone vs. control Rosiglitazone vs. control Metformin vs. control 0.3 (⫺0.3 to 0.9) Sulfonylureas vs. control Meglitinides vs. control Acarbose vs. control ⫺0.1 (⫺0.5 to 0.2) * The control group consisted of placebo or diet.
18 September 2007 Annals of Internal Medicine Volume 147 • Number 6 W-77


ON SALE 10TH - 23RD MARCH 2016 Target Zero Hear urn with Nexium® 24HR 1 T elling Fr HOUR rn Brand Worldwide* ithout a prescrPiRp PROTECTION PROTECTION Your pharmacist's advice is required. S-26 Gold« *IMS Global Data MAT July 2015, OTC PPIs. Always read the label. Use only

Management Report In the terms of legal and Bylaws dispositions, the Management of Hypermarcas S.A. (‘Company' or ‘Hypermarcas') submits to its shareholders the Management Report and the individual and consolidated Standard Financial Statements of the Company, as well as the independent auditors report, regarding the fiscal year ended December 31, 2014.