My Nature Cure
My Nature Cure
My Nature Cure
Background
Effective Health Care Program
Type 2 diabetes is characterized by insulin
The Effective Health Care Program
resistance accompanied by progressive
was initiated in 2005 to provide valid
deficiency in insulin secretion. Type 2
evidence about the comparative
diabetes is an increasingly common disease
effectiveness of different medical
that is closely associated with obesity. In
interventions. The object is to help
2005, the prevalence of Americans with
consumers, health care providers, and
diagnosed type 2 diabetes was 2.4 percent
others in making informed choices
for adults aged 20-39 years, 10 percent for
among treatment alternatives. Through
adults aged 40-59 years, and 21 percent for
its Comparative Effectiveness Reviews,
adults aged 60 years or over. From 1980
the program supports systematic
through 2004, the number of Americans
appraisals of existing scientific
diagnosed with diabetes more than
evidence regarding treatments for
doubled, from 5.8 million to 14.7 million.
high-priority health conditions. It also
Observational studies and clinical trials
promotes and generates new scientific
show that improved glycemic control
evidence by identifying gaps in
reduces microvascular complications (e.g.,
existing scientific evidence and
complications involving the eyes, kidneys,
supporting new research. The program
or nerves) and may reduce macrovascular
puts special emphasis on translating
complications (e.g., heart attack); however,
findings into a variety of useful
the effects of specific oral diabetes
formats for different stakeholders,
medications on these outcomes are less
including consumers.
certain.
The full report and this summary are
As new classes of medications have
available at www.effectivehealthcare.
become available for the treatment of
ahrq.gov/reports/final.cfm
diabetes, clinicians and patients have faced
a bewildering array of oral medications
with different mechanisms of action. The
first oral diabetes medications were
sulfonylureas, which were introduced into
the market in 1955. The second-generation
sulfonylureas, which are used today, were
Effective
Health Care
introduced in 1984. Metformin (a biguanide) was and pioglitazone had significant effects on the lipid
introduced in 1995, meglitinides in 1997, alpha- profile. Metformin at high doses and pioglitazone both
glucosidase inhibitors in 1998, and thiazolidinediones reduced triglycerides, while acarbose, rosiglitazone, and
in 1999. Although most experts consider the alpha- pioglitazone increased high-density lipoproteins. Lastly,
glucosidase inhibitors to be inferior to the other drug acarbose decreased low-density lipoproteins, while
classes in terms of efficacy, clinicians may find it rosiglitazone increased low-density lipoproteins.
difficult to choose between the other four drug classes Many outcomes besides HbA1c and lipid levels are
that are now in common use. Generally, clinicians must important when evaluating and comparing oral diabetes
choose between older, less expensive medications such medications, such as blood pressure control, weight
as a second-generation sulfonylurea or metformin and changes, microvascular and macrovascular disease,
the newer, more expensive medications such as a adverse events, and mortality. It is critical to evaluate
thiazolidinedione or meglitinide. In addition, clinicians adverse events, since these affect adherence as well as
must consider concerns about specific medications morbidity and mortality. Additionally, certain diabetes
conferring excess cardiovascular risks when compared medications may be less safe for patients with
with other oral diabetes medications or placebo. comorbid conditions. For instance, biguanides such as
The well-known United Kingdom Prospective Diabetes metformin are contraindicated in patients with renal or
Study (UKPDS) demonstrated that oral diabetes liver failure because of a potentially higher risk of lactic
medications may have similar effects on cardiovascular acidosis. To date, no study has evaluated proximal
morbidity and mortality when they have similar effects clinical measures, long-term effects, and adverse events
on glycemic control. However, the UKPDS was among oral diabetes medications used in the United
conducted prior to the emergence of thiazolidinediones States. If they could compare the short- and long-term
and statins. effects as well as the adverse effects of these
Several systematic reviews of oral diabetes medications medications, clinicians might have a better sense of
shed light on differences in short-term and long-term when to use which oral diabetes medication. This
outcomes. However, only two reviews have compared review will be helpful as new classes of oral diabetes
all of the oral diabetes medications used commonly in medications, such as the dipeptidyl peptidase IV (DPP-
the United States. IV) inhibitors, emerge on the market. Furthermore, it
may help policymakers and insurers to have better
In 2002, Inzucchi and colleagues from Yale University insight when deciding on policies relating to medication
found that: (1) most diabetes medications lower coverage.
hemoglobin A1c (HbA1c) by an absolute reduction of
1-2 percent,1 with equivalent efficacy across This report summarizes the available evidence
medications, except for alpha-glucosidase inhibitors, comparing the efficacy and safety of oral diabetes
which decrease HbA1c by 0.5-1 percent; (2) medications in the treatment of type 2 diabetes. The
medications in combination confer additional glycemic report addresses the following key questions:
benefits; (3) long-term micro- or macrovascular risk 1. Do oral diabetes medications for the treatment of
reduction was demonstrated only with sulfonylureas adults with type 2 diabetes differ in their ability to
and metformin. affect the following proximal clinical outcomes:
In 2004, Buse and colleagues from the University of glycated hemoglobin, weight, blood pressure,
North Carolina compared effects on serum lipid levels serum lipid levels, and 2-hour postprandial glucose
among all the oral diabetes medications. They found (PPG) levels?
that only metformin, acarbose, voglibose, rosiglitazone, 2. Do oral diabetes medications for the treatment of
adults with type 2 diabetes differ in their ability to
1
One characteristic of type 2 diabetes is an elevation of the affect distal diabetes-related complications
proportion of HbA1c in the blood from a normal level of 6.5 including mortality and the following
to 7 percent to an elevated level of >6.5 to 7 percent (e.g., 10 macrovascular and microvascular complications:
percent). In this report, an “absolute” reduction of 1 percent coronary artery disease, myocardial infarction,
means a reduction of one percentage point in that proportion stroke, transient ischemic attack, arrhythmia,
(e.g., from 10 percent to 9 percent).
coronary artery stenosis and in-stent restenosis,
2
retinopathy, nephropathy, neuropathy, and sufficient data to make comparisons. In each column of
peripheral arterial disease? the summary table, we indicate the medication that had
3. Do oral diabetes medications for the treatment of a better effect on the listed outcome or note when there
adults with type 2 diabetes differ in their ability to were at least a moderate number of studies in which no
influence other health outcomes, including quality apparent difference was detected.
of life and functional status? The text below summarizes the conclusions regarding
4. Do oral diabetes medications for the treatment of the main comparisons of interest by outcome, and
adults with type 2 diabetes differ in terms of risk of qualifies points noted in the summary tables.
the following life-threatening adverse events: life- Comparisons of effects of oral diabetes
threatening hypoglycemia leading to emergency medications
care or death; liver failure; congestive heart failure
(CHF); severe lactic acidosis; cancer; anemia, Glycemic control (hemoglobin A1c). Based on direct
thrombocytopenia, or leucopenia requiring data from randomized controlled trials, most oral
transfusion; and allergic reactions leading to diabetes medications (thiazolidinediones, second-
hospitalization or death? generation sulfonylureas, metformin, and repaglinide)
had similar reductions in hemoglobin A1c (~1-percent
5. Do oral diabetes medications for the treatment of absolute reduction) compared with one another as
adults with type 2 diabetes differ in their safety for monotherapy. Indirect data, in addition to a few head-to-
the following adverse events that are not life head trials, showed that nateglinide and alpha-
threatening: hypoglycemia requiring any assistance; glucosidase inhibitors were less efficacious in reducing
elevated aminotransferase levels; pedal edema; hemoglobin A1c as monotherapy (~0.5-percent absolute
hypervolemia; anemia, thrombocytopenia, and reduction). Combination therapies had an additive effect
leucopenia not requiring transfusion; mild lactic and were better at reducing HbA1c compared with
acidosis; and gastrointestinal (GI) problems? monotherapy regimens (~1-percent absolute reduction).
6. Do safety and effectiveness of oral diabetes Weight. Weight increased by 1-5 kg with most of the
medications for the treatment of adults with type 2 oral diabetes medications (thiazolidinediones, second-
diabetes differ across particular adult populations, generation sulfonylureas, and repaglinide), but not for
such as those based on demographic factors (e.g., metformin and acarbose, which had no effect on weight
race/ethnicity, age greater than 65 years, or gender) in placebo-controlled trials. In direct comparisons with
or comorbid conditions (e.g., renal insufficiency, thiazolidinediones and second-generation sulfonylureas,
CHF, liver disease, obesity, depression, or metformin caused relative weight loss. However, this
schizophrenia)? might be an artifact reflecting the withdrawal from a
prior sulfonylurea (and withdrawal of its weight-
Conclusions increasing effect) that often occurred in head-to-head
trials. There were too few comparative studies of
Summary Table A presents the main conclusions from
nateglinide to draw conclusions.
published evidence regarding the comparative
effectiveness of oral diabetes medications, organized by Different types of weight gain (central vs. peripheral)
key question and type of outcome. The summary table may have different effects on morbidity, with central
also includes our rating of the level of evidence that adiposity considered to have greater prediction of
supports each conclusion. Meta-regression was cardiovascular outcomes. Only a few studies evaluated
conducted using study-level characteristics such as dose whether weight gain was related to increases in visceral
of medication, study duration, and study quality. When adipose tissue, subcutaneous fat, or plasma volume.
important differences arose based on these Therefore, it is unclear whether the weight gains caused
characteristics, we reported them in the table. by the different medications are physiologically
equivalent.
In Summary Table B we present a short synopsis of the
comparative effectiveness of the oral diabetes
medications used most often and for which there were
3
Systolic and diastolic blood pressure. Most oral present despite randomization. Repaglinide and
diabetes medications (thiazolidinediones, second- acarbose had similar reductions in triglycerides when
generation sulfonylureas, and metformin) had similarly compared with second-generation sulfonylureas. There
minimal effects on systolic and diastolic blood pressure were too few comparisons for nateglinide to draw
(<5 mm Hg). Too few studies compared meglitinides conclusions.
and acarbose with other oral diabetes medications to All-cause mortality. There were too few studies to
draw firm conclusions. There was a suggestion of support any conclusions about how mortality differed
decreased blood pressure in the thiazolidinedione group between the medications. It was unclear whether effects
when compared with second-generation sulfonylureas on mortality differed between the combination of
and acarbose. However, the clinical relevance of these metformin with a sulfonylurea and monotherapy with a
small nonsignificant between-group differences of 3-5 sulfonylurea or metformin, due to lack of adjustment
mmHg is questionable. for key confounders in cohort studies and lack of
Low-density lipoprotein. Only thiazolidinediones studies evaluating this combination. Other comparisons
consistently increased low-density lipoprotein (by between drugs had too few studies to draw conclusions.
about 10 mg/dL), while only metformin consistently Cardiovascular mortality and morbidity. There were
decreased low-density lipoprotein (by about 10 too few studies to support any conclusions about how
mg/dL). Of the two thiazolidinediones, rosiglitazone cardiovascular morbidity or mortality differed between
increased low-density lipoprotein cholesterol more the medications. It was unclear whether cardiovascular
than pioglitazone (difference of about 10-15 mg/dL). In mortality differed between the combination of
addition, second-generation sulfonylureas showed metformin with a sulfonylurea and monotherapy with a
similar minimal effects on low-density lipoprotein sulfonylurea or metformin, due to lack of adjustment
cholesterol when compared with repaglinide and alpha- for key confounders in cohort studies and lack of
glucosidase inhibitors. Too few studies on nateglinide studies evaluating this combination. Only pioglitazone
were available to draw conclusions. and metformin improved cardiovascular morbidity
High-density lipoprotein. Only thiazolidinediones when compared with placebo or diet (one study each,
increased high-density lipoprotein. Thiazolidinediones PROactive and UKPDS).
increased high-density lipoprotein by about 3-5 mg/dL, Peripheral vascular disease. Only two randomized
compared with metformin or second-generation controlled trials reported information on peripheral
sulfonylureas, which had little effect on high-density vascular disease, making it difficult to draw
lipoprotein. Meglitinides had little effect on high- conclusions. In the largest of the two trials (PROactive),
density lipoprotein, but there were too few trials to pioglitazone had no effect on peripheral vascular
draw comparative conclusions. Combination therapy disease when compared with placebo in subjects with a
with thiazolidinediones increased high-density history of cardiovascular disease.
lipoprotein similarly to monotherapy with
thiazolidinediones, while combination therapies without Microvascular outcomes. Few studies examined how
thiazolidinediones had little effect on high-density microvascular outcomes differed between the
lipoprotein levels. medications, but some differences were reported. In the
UKPDS, glibenclamide decreased the need for
Triglyceride levels. Most oral diabetes medications photocoagulation and had a protective effect on
(pioglitazone, metformin, second-generation combined microvascular outcomes (retinopathy plus
sulfonylureas, acarbose, and repaglinide) decreased nephropathy) compared with the conventional arm
triglycerides, except for rosiglitazone, which generally (diet), while metformin showed no effect on retinopathy
increased triglycerides. Pioglitazone decreased compared with the conventional arm. Pioglitazone may
triglycerides more than metformin (difference of about be better at reducing short-term nephropathy compared
26 mg/dL), and metformin decreased triglycerides to a with metformin, based on two short-duration
greater degree than second-generation sulfonylureas randomized controlled trials.
(difference of about 10 mg/dL). These small differences
in triglyceride reduction may reflect differences Quality of life and functional status. No conclusions
between groups in initial triglyceride levels that were could be drawn regarding the comparative effects of the
medications on quality of life and functional status
4
because of a limited number of studies and differences Elevated aminotransferase levels/liver failure.
in the questionnaires used to assess quality of life. Several oral diabetes medications (thiazolidinediones,
Hypoglycemia. Minor and major hypoglycemic second-generation sulfonylureas, and metformin) had
episodes were more frequent in subjects taking second- similarly low rates (less than 1 percent) of clinically
generation sulfonylureas (especially glyburide) than in relevant elevated aminotransferase levels (greater than
subjects taking other oral diabetes medications except 1.5 to 2 times the upper limit of normal). Insufficient
repaglinide. Reported percentages of subjects studies evaluated or reported on the effects of
experiencing minor or major hypoglycemic episodes meglitinides on serum aminotransferase levels, but their
ranged from 0 to 58 percent for second-generation effects appeared to be similar to the effects of other oral
sulfonylureas vs. 0 to 21 percent for metformin and 0 to diabetes medications. The evidence was insufficient to
24 percent for thiazolidinediones. The absolute risk compare oral diabetes medications on the outcome of
difference was 5-10 percent when comparing second- liver failure since there were too few events.
generation sulfonylureas with metformin or Congestive heart failure. Thiazolidinediones were
thiazolidinediones. Glyburide/glibenclamide had a associated with greater risk of CHF compared with
higher risk of hypoglycemia compared with other metformin or sulfonylureas (two head-to-head
second-generation sulfonylureas (absolute risk randomized controlled trials with absolute risk
difference of ~2 percent). Repaglinide and second- differences of 1-2 percent; cohort studies had a range in
generation sulfonylureas had a similar incidence of odds ratios of 1.06-2.27, which was significant in four
subjects with hypoglycemia. However, repaglinide may of five head-to-head studies). Metformin and second-
be associated with less serious hypoglycemia in the generation sulfonylureas had similarly little impact on
elderly and in people who skip meals. Data were sparse the incidence of CHF. CHF was reported mostly in
on the comparisons between acarbose or nateglinide cohort studies that did not adjust for key confounders,
and other oral diabetes medications. The incidence of such as duration of diabetes, HbA1c level, blood
minor and major hypoglycemia was higher with pressure level, and medication adherence. However, the
combinations that included sulfonylureas, compared cohort studies were consistent with one another and
with metformin or sulfonylurea monotherapy (absolute were consistent with the data found in the randomized
risk differences of 8-14 percent). The combination of controlled trials, making these conclusions likely to be
metformin plus rosiglitazone had a similar risk of accurate.
minor hypoglycemia compared with metformin Edema. Edema was more frequent in subjects taking
monotherapy, and no serious events occurred in either thiazolidinediones (range 0-26 percent) than in subjects
of these treatment groups. taking second-generation sulfonylureas (range 0-8
Gastrointestinal adverse events/problems. Metformin percent) or metformin (range 0-4 percent). The
and acarbose were generally associated with a higher absolute risk differences ranged from 2 to 21 percent
percent of subjects with GI adverse events (range 2-63 when comparing thiazolidinediones with second-
percent and 15-30 percent, respectively) compared with generation sulfonylureas or metformin. No cases of
other oral diabetes medications (thiazolidinediones: macular edema were identified in the studies reviewed;
range 0-36 percent, second-generation sulfonylureas: however, case reports were excluded from the review,
range 0-32 percent, and meglitinides: range 8-11 and this is where most macular edema cases have been
percent). The absolute risk differences ranged from 5 identified. Cohort studies are needed to explore the
to 15 percent when comparing metformin or acarbose issue of macular edema further. Data were too sparse to
with these other oral diabetes medications draw conclusions about how the incidence of edema
(thiazolidinediones, second-generation sulfonylureas, or may differ between other oral diabetes medications.
meglitinides). Metformin monotherapy was associated Lactic acidosis. Despite traditional concerns, the rate
with more frequent adverse events compared with the of lactic acidosis was similar between metformin and
combination of metformin plus a second-generation other oral diabetes medications or placebo (8.4 vs. 9
sulfonylurea or metformin plus a thiazolidinedione if cases per 100,000 patient-years). We did not have
the metformin component was at a lower dose than the enough information on subjects taking metformin with
metformin monotherapy arm. chronic conditions such as chronic renal insufficiency,
5
chronic liver disease, congestive heart failure, or severe More head-to-head monotherapy trials of
pulmonary disease; therefore, we were unable to rosiglitazone with metformin and sulfonylurea
determine the safety of taking metformin in the monotherapy are needed to better assess potential
presence of comorbid conditions that predispose differences in lipid effects.
subjects to lactic acidosis. Future studies on weight should attend to effects
Anemia, thrombocytopenia, and leucopenia. on body composition and partition effects on
Thiazolidinediones may be associated with an increased weight or body mass index as an increase in fluid,
risk of anemia (range 0-7 percent) compared with other subcutaneous tissue, or visceral adipose tissue, as
oral diabetes medications (range 0-3 percent). The these may have different effects on health. If
absolute risk differences ranged from ~1-5 percent possible, investigators should then link these with
when comparing thiazolidinediones with other oral hard outcomes, such as morbidity and mortality.
diabetes medications. The decrease in hematocrit was Furthermore, since sulfonylureas and
small (1 g/dL) and would not be clinically relevant thiazolidinediones increase weight as
except for subjects with severe or borderline severe monotherapy, future studies need to identify
anemia. Only one study reported an adverse event of whether there would be an additive or synergistic
thrombocytopenia and leucopenia, making comparisons effect on weight for combinations of sulfonylureas
between medications impossible. with thiazolidinediones.
6
To improve understanding of the effects of oral medications in the elderly and in subjects with and
diabetes medications on nephropathy, studies without renal disease, congestive heart failure,
should evaluate long-term clinically relevant liver disease, or psychiatric disease.
nephropathy outcomes (such as time to dialysis) as
well as short-term proteinuria outcomes. Other general issues
Future observational studies could improve
For Key Question 3 (quality of life) understanding of the effects of oral diabetes
More studies should examine the effects of oral medications on adverse events and distal outcomes
diabetes medications on health-related quality of if they carefully assess key confounders, such as
life using standardized, validated questionnaires, duration of diabetes, adherence to medications,
especially since quality of life may affect whether dosing of medications, hemoglobin A1c levels, and
patients adhere to medications. blood pressure levels.
For Key Questions 4 and 5 (adverse Studies need to report consistently between-group
effects) changes from baseline, as well as measures of
dispersion such as standard errors.
Studies on oral diabetes medications need to report
consistently withdrawals and reasons for Further head-to-head trials are needed to compare
withdrawals to improve understanding of potential (1) nateglinide with all other oral diabetes
differences in adverse effects. medications and (2) repaglinide with other oral
diabetes medications besides second-generation
Studies on oral diabetes medications need to report sulfonylureas.
their definitions of adverse events more
thoroughly, and consistently report all adverse More studies should compare one combination of
events (not using aggregated events). oral diabetes medications directly with another
combination (specifically metformin,
Additional observational studies of metformin sulfonylureas, and thiazolidinediones in dual
compared with other oral diabetes medications in combinations as starting therapy) for all outcomes,
subjects prone to lactic acidosis would help as many clinicians have started using combinations
determine the safety of using this medication in as initial treatment in persons with diabetes.
populations with comorbid diseases.
Further research is needed on the effects of oral
Further observational studies should evaluate the diabetes medications on beta cell function over a
incidence of (1) macular edema with 3-5 year period or longer, using standardized
thiazolidinediones, (2) anemia requiring outcomes, such as c-peptide and insulin levels, and
transfusion or hospital admission for time to requiring insulin.
thiazolidinediones compared with other oral
agents, and (3) allergic reactions in all oral A systematic review of drug-drug interactions in
diabetes medications. subjects with diabetes would help clinicians with
treatment decisions.
Further observational studies should evaluate
cancer and allergic reactions for all oral diabetes Future studies comparing oral diabetes
medications. medications must consider any new oral diabetes
medications that may be placed on the market,
For Key Question 6 (differences across such as the dipeptidyl peptidase IV (DPP-IV)
specific populations) inhibitor sitagliptin, which has just been approved
To determine differences in medication by the Food and Drug Administration.
effectiveness based on comorbidity or Lastly, studies comparing combinations of older
demographics, analyses should be stratified or diabetes medications, such as sulfonylureas and
adjusted based on comorbidity or demographics. metformin, with combinations of newer oral
Specific areas to focus on would be effects of diabetes medications, such as thiazolidinediones in
7
combination with DPP-IV inhibitors or outcomes and safety to guide treatment decisions for
meglitinides, would be interesting, especially given oral diabetes medications. Physicians and patients can
the cost associated with newer oral diabetes feel comfortable using older medications such as
medications. metformin and second-generation sulfonylureas, as
monotherapy or in combination, before newer diabetes
Synopsis medications such as thiazolidinediones or meglitinides,
especially when cost is a factor. Future research should
Several clinical trials have investigated short-term focus on comparing combinations of newer medications
outcomes of various preparations of oral medications (DPP-IV inhibitors, meglitinides, and
for type 2 diabetes. Compared to newer medications, thiazolidinediones) with combinations of older
such as thiazolidinediones and meglitinides, metformin medications (metformin and second-generation
had similar or superior effects on a range of clinically sulfonylureas) with respect to long-term effectiveness
relevant short-term outcomes. For these same and safety.
outcomes, second-generation sulfonylureas generally
were comparable to thiazolidinediones and
meglitinides. In terms of safety, each medication was Addendum
associated with specific adverse events, although Two high-profile original studies and one meta-analysis
thiazolidinediones and second-generation sulfonylureas on this topic have been published since this review was
were associated with more serious adverse events, such completed.2,3,4 One 4-year double-blind randomized
as congestive heart failure and serious hypoglycemia, trial2 compared rosiglitazone monotherapy with
respectively. Repaglinide may be associated with less metformin or glyburide monotherapy and showed a
serious hypoglycemia in the elderly and in people who significant difference in HbA1c favoring rosiglitazone
skip meals. Lactic acidosis rates were similar for (between-group absolute difference of -0.42 percent for
metformin in comparison with other oral diabetes rosiglitazone vs. glyburide and -0.13 percent for
medications. Thus, metformin may be associated with rosiglitazone vs. metformin). However, the incidence of
less risk of serious adverse events than second- cardiovascular events was lower with glyburide than
generation sulfonylureas or thiazolidinediones. When with rosiglitazone or metformin (1.8 percent, 3.4
oral diabetes medications were combined, the effects percent, and 3.2 percent, respectively; p < 0.05). This
with respect to HbA1c levels and adverse events were effect was mainly driven by significantly fewer
generally additive. If each individual drug was used at a congestive heart failure events and a nonsignificantly
lower dose in the combination, fewer adverse events lower rate of nonfatal myocardial infarction events in
were seen. the glyburide group. The high loss to followup (40
Not much evidence exists that might enable one to percent) may account for some differences between
know a priori which medications are most likely to be groups, since the loss to followup was disproportionate
effective in identifiable subgroups of patients with between the groups. This study illustrates the
diabetes, nor does much evidence exist to predict which importance of having more long-term followup data on
particular patients may be most susceptible to the cardiovascular outcomes. At a minimum, clinicians
adverse events associated with particular drugs. should not assume that a small benefit measured in
8
terms of HbA1c reduction will be associated with an rosiglitazone would have fallen short of statistical
improvement in cardiovascular outcomes. Indeed, this significance. They also included a study in which the
study suggests that cardiovascular outcomes could be patients had a history of congestive heart failure, even
worse with rosiglitazone despite its having a more though rosiglitazone is currently contraindicated in
beneficial effect on HbA1c. these subjects. Inclusion of these data may have
Of note, the fracture rate among women was higher in produced a higher apparent risk than would be expected
the rosiglitazone group than in the metformin and in practice today. They also excluded six studies that
sulfonylurea groups (9.3 percent, 5.1 percent, and 3.5 reported no cardiovascular events in either group,
percent, respectively; p < 0.01).2 We did not find any thereby biasing their results against finding no
reported fractures in shorter duration trials, and this will difference. Given the limitations of the analysis, the
be an important area for future research. For other effects of rosiglitazone on cardiovascular mortality and
outcomes reported in this article, the results were myocardial infarction are still uncertain.
similar to those included in our report. After the release of the meta-analysis on rosiglitazone
In the meta-analysis, the authors reported that, in
3 and cardiovascular risk,3 an interim analysis of the
comparison with other oral diabetes medications or RECORD study4 was published. This randomized trial
placebo, rosiglitazone was associated with a borderline- of subjects with uncontrolled type 2 diabetes compared
significant increased risk of myocardial infarction (odds addition of rosiglitazone to existing metformin or
ratio, 1.43; 95-percent confidence interval (CI), 1.03 to sulfonylurea monotherapy vs. the combination of
1.98) and a nonsignificant association with metformin plus sulfonylurea (control group). This
cardiovascular death (odds ratio, 1.64; 95-percent CI, analysis yielded a hazard ratio of 1.08 (95-percent CI,
0.98 to 2.74). When they analyzed specific drug-drug 0.89 to 1.31) for the primary end point of
or drug-placebo comparisons, their results were not hospitalization or death from cardiovascular disease
statistically significant. Similarly, our report did not after a mean followup of 3.7 years. The hazard ratio
find any statistically significant differences between was driven by more congestive heart failure in the
specific oral diabetes medications in cardiovascular rosiglitazone group than in the control group (absolute
outcomes other than congestive heart failure. risk, 1.7 percent vs. 0.8 percent). In Kaplan-Meier
curves, the risk of hospitalization or death from
The authors acknowledged several limitations of their myocardial infarction was slightly lower in the control
study: (1) there were small numbers of absolute events; group than in the rosiglitazone group, but the difference
(2) the primary outcomes of the short-term trials were was not statistically significant. One limitation of this
not cardiovascular events; and (3) the authors had no interim analysis was the lack of power to detect
access to original source data. Among additional differences because of lower numbers of cardiovascular
limitations that influenced their conclusions was their events than initially predicted. The RECORD study
decision to include studies with diverse patient may now have trouble reaching the desired power for
populations. They pooled studies that examined use of detecting a difference in cardiovascular risk if patients
rosiglitazone for conditions other than type 2 diabetes, withdraw from the rosiglitazone arm of the study.
including studies of patients with chronic psoriasis,
Alzheimer’s disease, type 2 diabetes, and impaired Overall, these recent reports are consistent with our
glucose tolerance. Had the authors excluded data from review, which found no conclusive evidence of worse
the DREAM trial,5 which was conducted in adults with cardiovascular morbidity or mortality, outside of the
prediabetes, the pooled estimate of risk associated with higher risk of congestive heart failure with
thiazolidinediones than with other oral medications.
5
These new studies substantiate our call for more
DREAM (Diabetes REduction Assessment with ramipril vigorous post-marketing surveillance and long-term
and rosiglitazone Medication) Trial Investigators. Effect of
comparative assessments of major clinical outcomes.
rosiglitazone on the frequency of diabetes in patients with
impaired glucose tolerance or impaired fasting glucose: a
For example, such studies should pay attention to the
randomised controlled trial. Lancet 2006 Sep risk of myocardial infarction with rosiglitazone
23;368(9541):1096-105. compared with other oral diabetes medications.
9
Full Report Healthcare Research and Quality. May 2007. Available
at: www.effectivehealthcare.ahrq.gov/reports/final.cfm.
This executive summary is part of the following
document: Bolen S, Wilson L, Vassy J, Feldman L, Yeh
J, Marinopoulos S, Wilson R, Cheng D, Wiley C, Selvin For More Copies
E, Malaka D, Akpala C, Brancati F, Bass E. For more copies of Comparative Effectiveness and
Comparative Effectiveness and Safety of Oral Diabetes Safety of Oral Diabetes Medications for Adults With
Medications for Adults With Type 2 Diabetes. Type 2 Diabetes: Executive Summary. No. 8 (AHRQ
Comparative Effectiveness Review No. 8. (Prepared by Pub. No. 07-EHC010-1), please call the AHRQ
Johns Hopkins Evidence-based Practice Center under Clearinghouse at 1-800-358-9295 or e-mail
Contract No. 290-02-0018.) Rockville, MD: Agency for ahrqpubs@ahrq.gov.
10
Summary Table A. Evidence of comparative effectiveness of oral diabetes
medications (continued)
11
Summary Table A. Evidence of comparative effectiveness of oral diabetes
medications (continued)
12
Summary Table A. Evidence of comparative effectiveness of oral diabetes
medications (continued)
13
Summary Table A. Evidence of comparative effectiveness of oral diabetes
medications (continued)
14
Summary Table A. Evidence of comparative effectiveness of oral diabetes
medications (continued)
1
Definitions of levels of evidence: 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; NA = not applicable since there was no validated grading
system to determine level of evidence for adverse events.
2
The evidence was graded very low for the following comparisons related to blood pressure effects: metformin vs. metformin
plus sulfonylurea, sulfonylurea vs. sulfonylurea plus thiazolidinedione, meglitinides vs. sulfonylureas, and alpha-glucosidase
inhibitors vs. all other oral diabetes medications.
Abbreviations: CHF = congestive heart failure; FDA = Food and Drug Administration; GI = gastrointestinal;
HbA1c = hemoglobin A1c; HDL = high-density lipoprotein; LDL = low-density lipoprotein; RCT = randomized controlled
trial; TG = triglycerides.
15
Summary Table B. Synopsis of comparative effectiveness of oral diabetes medications
Outcomes SU vs. Met SU vs. TZD SU vs. Meg Met vs. TZD
HbA1c ND ND ND1 ND
Weight Met ND ND1 Met
SBP/DBP ND ND Insufficient ND
LDL Met SU ND1 Met
HDL ND TZD ND1 TZD
TG Met ND2 ND1 Pio3
All-cause mortality Insufficient Insufficient Insufficient Insufficient
CVD mortality/ morbidity Insufficient Insufficient Insufficient Insufficient
Peripheral vascular disease Insufficient Insufficient Insufficient Insufficient
Microvascular outcomes Insufficient Insufficient Insufficient Insufficient
Quality of life Insufficient Insufficient Insufficient Insufficient
Hypoglycemia Met TZD ND1 ND
GI SU Insufficient Insufficient TZD
Elevated aminotransferase levels/liver failure ND ND Insufficient ND
CHF ND SU Insufficient Met
Edema Insufficient SU Insufficient Met
Lactic acidosis ND Insufficient Insufficient ND
Anemia Insufficient SU Insufficient Met
1
These conclusions refer to sulfonylurea vs. repaglinide only. See the text for more information about the comparison of
sulfonylureas with nateglinide.
2
Pioglitazone decreased triglycerides, while rosiglitazone increased triglycerides; therefore, pioglitazone showed similar
effects on TG when compared with sulfonylurea, while rosiglitazone likely was worse than sulfonylureas but no direct
comparisons were available to draw firm conclusions.
3
Pioglitazone decreased triglycerides, while rosiglitazone increased triglycerides; therefore, pioglitazone was better than
metformin, while rosiglitazone was worse than metformin.
Abbreviations: CHF = congestive heart failure; CVD = cardiovascular disease; DBP = diastolic blood pressure;
GI = gastrointestinal; HbA1c = hemoglobin A1c; HDL = high-density lipoprotein; LDL= low-density lipoprotein;
Meg = meglitinides; Met = metformin; ND = no apparent difference; Pio = pioglitazone; SBP = systolic blood pressure;
SU = second-generation sulfonylurea; TG = triglyceride; TZD = thiazolidinediones.
16