Reviews/Commentaries/ADA Statements PERSPECTIVES ON THE NEWS

The Avandia Debate

ZACHARY T. BLOOMGARDEN, MD the topic of whether TZDs should be used was addressed, Silvio Inzucchi (New Ha- ven, CT) agreeing with their use, while Jay Perspectives on the News commentaries are SmithKline (GSK), the manufacturer of Skyler (Miami, FL) spoke in opposition. now part of a new, free monthly CME activ- , and the Food and Drug Inzucchi began with a discussion of ity. The Mount Sinai School of Medicine, Administration (FDA) should have re- the importance of resistance and New York, New York, is designating this ac- leased similar data earlier, mentioned in- of the concept that TZDs treat the root tivity for 2.0 AMA PRA Category 1 credits. If vestigations being started in Congress, cause of type 2 , while sulfonyl- you wish to participate, review this article and quoted Nissen to have stated, “It’s a ureas and agents acting at the glucagon- and visit www.diabetes.procampus.net to huge risk,” with “tens of thousands of like peptide 1 receptor act on the ␤-cell, complete a posttest and receive a certificate. people” having had myocardial infarction the latter also with glucagon-suppressing The Mount Sinai School of Medicine is ac- caused by the treatment (3). Patients with effects, and acts principally in credited by the Accreditation Council for diabetes were understandably worried, reducing hepatic glucose production. The Continuing Medical Education (ACCME) to and their physicians were faced, as often only approach available to directly in- provide continuing medical education for is the case with such media events, with crease peripheral glucose uptake in skel- physicians. the dilemma of determining what should etal muscle and in adipose tissue is by be the correct advice at a time of limited increasing insulin action with TZDs. In- his is the first in a series of articles data availability. Hundreds of such news zucchi reviewed his study comparing based on presentations at the Amer- articles have appeared, the GSK share metformin, which decreased hepatic glu- T ican Diabetes Association’s (ADA’s) price dropped by nearly one-tenth, and cose production by 19%, with troglita- 67th Scientific Sessions in June 2007, in lawsuits are already being filed against zone, which increased peripheral glucose Chicago, reviewing aspects of thiazo- GSK on behalf of individuals with diabe- uptake, primarily in skeletal muscle, by lidinediones (TZDs) treatment, with fo- tes who have cardiovascular disease 54% (6). This, he suggested, reverses the cus on the controversy over possible (CVD) developing or worsening while re- primary defect in early diabetes. A further adverse effects of rosiglitazone. ceiving the drug. There is, however, un- argument for the TZDs comes from the On 21 May 2007, an article appeared certainty as to whether the analysis was important interplay between insulin re- on the New England Journal of Medicine correctly performed. Two large studies sistance and CVD, which may involve gly- Web site giving a meta-analysis of 42 tri- are being carried out with rosiglitazone cemia, hyperinsulinemia, dyslipidemia, als comparing rosiglitazone with placebo under the aegis of the National Institutes hypertension, hypercoagulability, and in- or other agents. Steven Nissen and Kathy of Health: the Bypass Angioplasty Revas- flammation. TZDs have been shown to Wolski reported a significant 43% in- cularization Investigation 2 Diabetes improve all of these abnormalities. TZDs crease in myocardial infarction and a (BARI 2D) and Action to Control Cardio- decrease vascular smooth muscle cell and trend almost achieving statistical signifi- vascular Risk in Diabetes (ACCORD) neointimal proliferation. A meta-analysis cance of a 64% increase in cardiovascular trials, including 2,368 and 10,251 indi- of seven clinical trials including 608 indi- mortality (1). In an accompanying edi- viduals with type 2 diabetes, respectively. viduals has shown a reduction in the rate torial, Bruce Psaty and Carl Furberg Both have Data Safety and Monitoring of restenosis with TZD treatment (7). strongly questioned the wisdom of choos- Boards, which have met and reported no Furthermore, recent evidence suggests ing treatments for diabetes based on “the increase in risk to patients in the trials and reduction in carotid intima-medial thick- single dimension of glycemic control,” recommended that the studies proceed ness (IMT) with TZDs (8–11). Although suggesting that although “high levels of unchanged (4). The Endocrine Society sim- Inzucchi acknowledged that IMT is a sur- glycated hemoglobin increase risk,” one ilarly expressed concern that the Nissen rogate measure, he suggested that there must “require proof of health benefits” be- study might be appropriately regarded as may then be a direct anti-atherosclerotic fore accepting that an agent lowering preliminary, given its “substantial limita- benefit with these agents. In his study of blood glucose benefits individuals with tions” (5). 24,953 diabetic individuals with acute diabetes (2). myocardial infarction and diabetes in a These articles were accompanied by Medicare hospitalization database, 2,231 tremendous publicity. On 22 May 2007, Debate: Should TZDs be primary were discharged receiving either met- an article in the New York Times entitled agents in the treatment of type 2 formin or a TZD and 6,641 received nei- “Heart Attack Risk Seen in Drug for Dia- diabetes? ther agent. There was a nonsignificant 8% betes” summarized the public concern. At a debate held on 10 April 2007 by the decrease in mortality and readmission in The article questioned whether Glaxo- New York Metropolitan Diabetes Society, the sensitizer group and a significant 48% ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● reduction in mortality among those re- Zachary T. Bloomgarden, MD, is a practicing endocrinologist in New York, New York, and is affiliated with ceiving both agents (12). the Division of Endocrinology, Mount Sinai School of Medicine, New York, New York. In the PROspective Abbreviations: ADA, American Diabetes Association; CVD, cardiovascular disease; DREAM, Diabetes Clinical Trial In macroVascular Events REduction Assessment with ramipril and rosiglitazone Medication; FDA, Food and Drug Administration; FFA, free fatty acid; GSK, GlaxoSmithKline; IMT, intima-medial thickness; TZD, . (PROactive), 5,238 diabetic individuals DOI: 10.2337/dc07-zb09 with CVD randomized to pioglitazone © 2007 by the American Diabetes Association. versus conventional hypoglycemic therapy

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2401 Perspectives on the News showed a 16% relative and 2% absolute re- TZDs, and given the 20–30% of diabetic over time between the control and inter- duction in the “principal secondary end individuals intolerant of metformin, he vention groups (22). point” of nonfatal myocardial infarction, suggested that TZDs and inhibitors of The concept of the DREAM Trial was nonfatal stroke, or CVD death (13), which dipeptidyl peptidase-4 should be favored based on the premise that rosiglitazone Inzucchi suggested would be the typical over and that, recognizing would offer benefit to individuals with end point chosen for a study of cardiovas- that all pharmaceutical agents have con- impaired glucose tolerance, but in fact, cular end points. Given this overall benefit, traindications and adverse effects, TZDs Skyler suggested that the sevenfold in- he suggested that subgroup analysis was ap- should be considered excellent agents for crease in heart failure and trend to overall propriate, with the striking findings of a the treatment of type 2 diabetes. increase in CVD was worrisome. He also 28% reduction in recurrent myocardial in- Jay Skyler took the opposing posi- commented unfavorably on the 8-lb farction among the 2,445 individuals who tion, reviewing many of the same studies weight gain over the study period for ros- had prior myocardial infarction (14) and a but suggesting they implied that TZDs iglitazone-treated patients. 48% reduction in recurrent stroke among should rarely be used. In analyzing the ADOPT Study, al- 984 with prior stroke (15). He pointed out He proposed that in the PROactive though he acknowledged that rosiglita- that the duration of the PROactive Study Study, the primary end point represents a zone was better than both glyburide and was only 3 years, while that of most other reasonable collection of macrovascular metformin in reducing fasting glucose, he important clinical trials of strategies for complications of diabetes. This end point pointed out that metformin maintained prevention of atherosclerotic events, such failed to show a statistically significant im- the glucose target for 45 months, not as the Heart Protection Study, CARE, and provement, leading Skyler to conclude much different from the 57 months with MICRO-HOPE studies, was 5–6 years. that “none of the secondary end points rosiglitazone, and with weight loss rather He compared Kaplan-Meier cumulative count.” He noted that the principal sec- than weight gain. He further suggested survival plots of these studies with PRO- ondary end point cited by Inzucchi was that “glyburide is the worst of the sulfo- active, suggesting that over a longer time not included in the original methods pa- nylureas” but that despite this, the mean even more favorable outcome would be per. Furthermore, he noted that subgroup A1C during the first 3 years of treatment seen. Despite the higher rate of hospital- analysis of this study showed benefit only was actually lower with this agent than ization for congestive heart failure of for individuals not treated with statins, with either metformin or TZDs, so, per- 5.7% with pioglitazone compared with which he considered might be an impor- haps, “starting with metformin and a sul- 4.1% in individuals allocated placebo, In- tant potential reason to consider pioglita- fonylurea is not necessarily a bad thing.” zucchi suggested that the prevention of Furthermore, he noted that adverse zone not actually to be effective in an events justifies the use of this treatment. events occurred more frequently with optimal treatment approach. Another Furthermore, he presented a popula- rosiglitazone, with evidence of CVD in weakness of the study was, he stated, the tion analysis showing that among 16,417 3.4% of those receiving this agent, in 3% improved level of glycemic control among Medicare beneficiaries with diabetes dis- of those receiving metformin, and in 1.8% individuals randomized to pioglitazone, charged after hospitalization with princi- of those treated with glyburide; edema in pal discharge diagnosis of heart failure, as well as the lower blood pressure and 14, 7, and 8%; and with doubling of frac- those receiving TZDs, like those receiving triglyceride and higher HDL cholesterol ture rates, confirmed by subsequent anal- metformin, had a 13% lower multivari- in this group, suggesting indirect mecha- yses of the GSK and Takeda databases, for ate-adjusted 1-year CVD mortality than nisms of benefit. Furthermore, he con- rosiglitazone and pioglitazone. those not receiving insulin sensitizer trasted the 58 fewer primary end points He commented that the improvement treatment (16). with the 115 more heart failure events, in carotid IMT with pioglitazone in the There is a ␤-cell protective effect of suggesting that the latter should be taken Chicago study appeared to be transient, TZD treatment, with evidence that these into account, and he commented unfa- with levels returning to baseline at 18 agents prevent diabetes, including the vorably about the association of pioglita- months, and further suggested that ros- 58% reduction in the in Pre- zone with weight gain. iglitazone appears to adversely affect LDL vention of Diabetes (TRIPOD) Study (17), Analyzing the DREAM Trial, he sug- cholesterol levels, leading him to con- the 75% 1-year reduction in the Diabetes gested that, at best, rosiglitazone led to a clude that the TZDs give no better glyce- Prevention Program (DPP) (18), and the 1-year delay in progression to diabetes, mic control and no better ␤-cell function 62% reduction in the Diabetes REduction predictable from the 10 and 20 mg/dl re- than other agents, while, he suggested, Assessment with ramipril and rosiglita- ductions in fasting and 2-h glucose in being associated with substantial weight zone Medication (DREAM) Trial (19). treated individuals, so that rather than gain, edema, and increase in LDL choles- The ADOPT Study of individuals with preventing diabetes, he suggested that terol, with increased risk of congestive type 2 diabetes receiving glyburide, met- there was merely masking of its develop- heart failure and of fractures, bolstering formin, and rosiglitazone showed reduc- ment. Furthermore, the washout study his contention that use of the agents tion in monotherapy failure at 5 years showed identical progression after rosigli- should be avoided. with the latter agent (20), further evi- tazone discontinuation, similar to what dence of benefit. was seen in the DPP after discontinuation Debate: Is rosiglitazone particularly Inzucchi concluded by citing the ex- of troglitazone, leading to his conclusion associated with adverse outcome? tensive evidence that individuals with di- that the effect was simply one of pharma- Steven Nissen (Cleveland, OH) debated abetes show progressively worsening cologic glucose lowering. He contrasted Philip Home (Newcastle, U.K.) on the glycemia over time, as particularly shown this with the effect of lifestyle, with fol- specific question of adverse cardiac effects in the UK Prospective Diabetes Study low-up of the Finnish Diabetes Preven- of rosiglitazone. Nissen began his presen- (21). Given the evidence of benefit of tion Study showing increasing separation tation 25 June 2007 at the ADA meeting

2402 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 Bloomgarden on the issue of possible adverse effects of gives relevant information, but Nissen ate randomized clinical trial and . . . the rosiglitazone by acknowledging, “I know disparaged this as an open-label study, al- data on the company’s Web site consti- I’ve made life more difficult for a lot of you though following 4,400 patients for 6 tuted the data we used . . . we avoided . . . and I’d like to tell you why.” In response years. Furthermore, he said that the study publication bias . . . because we had ev- to questions at the conclusion of the ses- was performed at the request of the Euro- erything.” He acknowledged a number of sion, he said further, “We didn’t call for pean regulatory agency to address the weaknesses, that there were “no patient- the withdrawal of the drug, and we didn’t CVD issue, implying that it should have level data; [that] time-to-event data call for regulatory action. We simply said been performed earlier. [were] not available, precluding Kaplan- that we want you and others who treat In the DREAM Trial, 2,635 individu- Meier curves; [that] cardiovascular end these patients to be aware of the findings. als treated with rosiglitazone had 15 myo- points were not the primary end point in There was a lot of criticism—should this cardial infarctions compared with 9 of any of the trials; . . . [that] events were not have been published or shouldn’t it—but 2,634 individuals receiving placebo. adjudicated; . . . [and that] small numbers let me say to you the alternative to us was Twelve vs. 10 cardiovascular deaths oc- of events were observed, resulting in wide unacceptable. The alternative would be to curred; there were 24 vs. 20 new cases of confidence intervals.” keep the scientific community in the angina, 35 vs. 27 revascularizations, and He concluded that rosiglitazone is as- dark, to not tell you that a pooled analysis 14 vs. 2 individuals developed congestive sociated with a significant increase in risk of all these data showed a pretty substan- heart failure—a composite 75 vs. 55 car- of myocardial infarction and of death, tial increase in the risk of the most serious diovascular events. Nissen recalled that commenting, “Patients and providers complication of diabetes.” he began to worry about the agent at that should consider the potential for serious In May 1999, he recalled that there time, writing a letter to Lancet suggesting adverse cardiovascular effects of treat- were major concerns about the hepatic that there might be harm (25). “In ment with rosiglitazone for type 2 diabe- toxicity of troglitazone. In that context, as ADOPT,” he said, “the same thing tes.” Furthermore, he stated that in 2005 both rosiglitazone and pioglitazone ap- happened.” There were 2 fatal and 25 and 2006, “we learned that GSK had per- peared free of this life-threatening side ef- nonfatal myocardial infarctions with ros- formed a similar study . . . which showed fect, he stated that the FDA agreed to put iglitazone but 2 and 21 with metformin statistically significant 31% greater rate of aside initial studies of rosiglitazone, and 3 and 15 with glyburide, respectively. myocardial ischemic events . . . FDA an- showing 1.2% ischemic heart disease Nissen termed this a “consistent pattern of nounced that they had conducted their rates in individuals treated with this agent excess myocardial infarctions,” which he own independent meta-analysis, which compared with rates of 0.5% with pla- found “very worrisome.” He stated that an showed a 40% increase in events.” He as- cebo, 0.6% with sulfonylureas, and 1.3% out of court settlement of a lawsuit by serted that GSK subsequently commis- with metformin. (Home subsequently re- New York Attorney General Elliot Spitzer sioned “a retrospective observational marked on this rate with the agent we cur- required GSK to publicly disclose all clin- study . . . using an insurance claims data- rently recommend as first choice for ical trial results. From this, Nissen found base of myocardial infarction and revas- treatment of diabetes.) Overall, Nissen 42 randomized studies of at least 24 cularization with incomplete assessment stated that there was a 1.8-fold increase in weeks’ duration, the majority of which of covariates such as smoking or aspirin, cardiovascular risk with rosiglitazone in were unpublished, comparing rosiglita- showing no increased CVD.” He criticized its registration trials. The approval letter zone with other agents or placebo, which this study, however, stating that few for this agent was issued on 25 May 1999, he used for the meta-analysis (1). In the events were observed, that there were Nissen stated, and requested that a long- small trials combined there was, he said, a wide CIs and short follow-up, and that no term study be performed, including as- 1.45-fold increase, for the DREAM Trial a comparison was made with pioglitazone. sessment of cardiovascular safety. The 1.65-fold increase, and for the ADOPT “What,” he asked, “about the initial studies showed that rosiglitazone Study a 1.33-fold increase in myocardial RECORD trial?” He maintained that “ma- increased LDL cholesterol by 18.6%, a infarction rate. Nissen performed a meta- jor errors in design have largely rendered further concern, with Nissen recognizing analysis to combine the different studies, this study futile.” He criticized the pri- John Buse, the ADA’s President-Elect, as finding overall a 1.43-fold increase in mary end point of death plus cardiovas- having been vocal in discussing this at sci- myocardial infarctions. He noted, “Much cular hospitalization. He observed that entific meetings and in writing to the has been said, pro and con, about meta- the study was designed based on a postu- company and to the FDA. Despite this, he analysis, and I think we did a good job of lated 11% annual event rate but has had a said that no major CVD outcome trials outlining strengths and weaknesses.” rate of just 2.5%, terming this “a major have yet been completed. Many small Myocardial infarctions occurred 14% miscalculation . . . [which] results in a short-term studies have been conducted. more often with rosiglitazone compared huge problem . . . only about 45% power The large DREAM (19) and ADOPT (20) with metformin, 24% more often than for the primary end point .... Bottom studies were conducted, but Nissen stated with sulfonylureas, 80% more often than line, RECORD will not give us an answer.” that they did not adjudicate cardiovascu- with placebo, and 178% more often than The primary end point was 11% more lar events. (This may be an erroneous with insulin, leading Nissen to remark, likely and myocardial infarction 23% statement; such adjudication was actually “When you see this kind of consistency in more likely, both with wide confidence performed in a blinded fashion, as dis- data, it makes you feel there’s something limits including unity, with only heart cussed in Appendix A of the Dream Trial going on.” He continued, “One strength is failure being significantly more likely online supplementary appendix [23].) that there was a large group of trials [and with rosiglitazone, occurring 115% more Only the Rosiglitazone Evaluated for Car- that] we used hard end points . . . [of] often than in the comparator group. diac Outcomes and Regulation of Glycae- myocardial infarction and cardiovascular Rosiglitazone “is not,” Nissen contin- mia in Diabetes (RECORD) Study (24) death ....Weincluded every appropri- ued, “the first agent to show a hazard in

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2403 Perspectives on the News the PPAR [peroxisome proliferator– of Nissen, pointing out that by 2000, medical and prescription benefit coverage activated receptor] class,” recalling his re- these agents were shown effective in glu- in a large U.S. managed care organization. port that muriglitazar showed mortality cose lowering, complementing existing Home also cited a study of 200 individu- plus nonfatal myocardial infarction plus treatment approaches, and improving in- als with metabolic syndrome randomized stroke rates significantly increased by sulin sensitivity—although fluid reten- to rosiglitazone or placebo after undergo- 123% (26), leading him to suspect a class tion and occasional cardiac failure were ing revascularization, extending the effect. He stated that more than 50 inves- known, as well as the mixed effects on meta-analysis cited by Inzucchi (7) with a tigational new drug applications for per- lipid profile. European regulators re- postintervention myocardial infarction oxisome proliferator–activated receptor quired cardiovascular outcomes studies, rate trending to favor rosiglitazone (30). ligands have been filed over the past 7 leading to the performance of the PROac- Home remarked that the Nissen years and that most of these agents were tive and RECORD studies. Concerns were meta-analysis failed to give a study hy- rejected because of toxicity. He did state, related to heart failure rather than to clas- pothesis, leading to what he considered however, that the PROactive Study with sic cardiovascular outcomes. Subsequent “data snooping on quite a large scale.” If, pioglitazone showed that the likelihood evidence accumulated on improvement in fact, the study hypothesis was that of the primary end point was reduced in cardiovascular risk markers, in blood myocardial infarction was increased by nonsignificantly by 10% and cited favor- pressure, and in surrogate cardiovascular rosiglitazone and if this hypothesis was ably that the combined principal second- outcomes, such as carotid IMT. The ADA based on Nissen’s observations in his let- ary end point of death, myocardial and European Association for the Study of ter to Lancet about the DREAM Trial, then infarction, and stroke decreased signifi- Diabetes published a consensus algo- the inclusion of the DREAM Trial data in cantly by 16% (13), leading Nissen to rithm for initiation and adjustment of the meta-analysis actually would repre- conclude, “we must interpret this trial as treatment, including TZDs alongside in- sent a flawed approach, in essence using neutral rather than favorable, but it was sulin and sulfonylureas largely because of the same information twice. Home trending . . . [to suggest that] pioglitazone the safety basis of the agents not causing pointed out further that many of the stud- is a very different drug . . . [with] more hypoglycemia (28). ies in Nissen’s meta-analysis included favorable effect on lipids . . . the jury is Home pointed out that rather than zero or only one report and that events still out but so far the trends are going in a there being an effort to obscure the 31% were not end points but were rather based favorable direction.” It may be relevant to increase in CVD seen in the retrospective on nonadjudicated investigator reports, Nissen’s enthusiasm for this agent, also analysis, this was included in the revision that there was no time-to-event analysis, expressed in the meta-analysis article (1), to the European rosiglitazone labeling in and that in the fashion reported “cardio- that Nissen is the principal investigator of September 2006, along with results of a vascular death is not a hard end point ei- the Pioglitazone Effect on Regression of large observational study to which Nissen ther,” as deaths may be said to be of Intravascular Sonographic Coronary Ob- referred unfavorably. Home reviewed this cardiac origin without actual documenta- struction Prospective Evaluation Study study in more detail. It gave 39,132 per- tion. (27). son-years of follow-up of 33,363 type 2 Given Nissen’s reservations about Nissen concluded that the FDA diabetic patients receiving rosiglitazone RECORD, Home went into some detail in “rushed” to approve new TZDs in 1999 or comparator agents (29). The study describing the study. It was primarily de- because of concern about troglitazone showed a trend to lower risk of hospital- signed as a cardiovascular safety study. toxicity, stating, “In my view, that was a ization for myocardial infarction and cor- Expectations in 1999–2000 were of a regulatory mistake ....Youdon’t know onary revascularization in individuals higher event rate than actually seen, per- from a limited exposure what’s going to receiving rosiglitazone than in those haps reflecting improvements in outcome happen later on . . . . A strong safety sig- treated with sulfonylureas, with still due to the widespread current use of st- nal of excess ischemic cardiovascular lower risk among individuals receiving atins and other risk-modifying agents. events was ignored.” Furthermore, he metformin. Contradicting Nissen’s state- The primary end point is a composite of termed blood glucose a “surrogate end ment that few events were observed, cardiac and vascular events, and the study point . . . [which is] not acceptable given Home stated that annual event rates were was designed with an active comparator the adverse cardiovascular signal.” He 1.6%, 1.9%, and 2.4% among individuals group, with rosiglitazone used in combi- characterized the postmarketing clinical receiving monotherapy, oral combination nation therapy and assessed for noninfe- trial program as including few studies that therapy, and combination therapy with riority, “mak[ing] it more relevant to your examined important health outcomes. insulin, respectively. One-hundred twenty practice.” He explained that the interim Finally, he reiterated his thought that the myocardial infarctions occurred in the analysis was necessitated by the immense limitations of RECORD are such that 14,975 patient-years of follow-up of indi- publicity generated by the Nissen/Wolski when completed, in 2009, 10 years after viduals receiving rosiglitazone, while 203 meta-analysis, leading many study partic- rosiglitazone was launched, we may still myocardial infarctions occurred in the ipants to withdraw after “members of not know whether this agent benefits or 24,522 patient-years of follow-up of indi- Congress leaked data to the New York harms individuals with type 2 diabetes, viduals receiving metformin and/or sulfo- Times.” Home characterized the interim so that “each physician must decide for nylureas, for 0.8% annual rates in each analysis as the “lesser of the evils.” A “fire- himself.” group. Observation times were up to 36 walled” approach was used, however, to Phillip Home (Newcastle, U.K.) pre- months, with the lack of comparison with maintain data integrity. sented what he termed “a balanced per- pioglitazone due to low prescription rates The RECORD Study enrolled individ- spective” on the relationship between for this agent in the Ingenix Research Da- uals receiving either a or rosiglitazone and CVD, giving a rather tabase, a proprietary research database of metformin, randomized to open-label different historical perspective from that commercial enrollees who have both addition of either rosiglitazone or the

2404 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 Bloomgarden other agent, leading to comparison of ered the statistical power of the study, as The use of a fixed effects rather than a 2,220 type 2 diabetic individuals receiv- did the 10% loss to follow-up. The com- random effects method in the meta- ing rosiglitazone in combination with ei- parator agents may have had differing ef- analysis, although described by Nissen in ther metformin or a sulfonylurea, with ficacy, but no confounding was found the article as being appropriate, is contro- 2,227 receiving metformin and a sulfo- differing effects of sulfonylureas from versial. The use of fixed effects calcula- nylurea in combination. The approach those of metformin. tions does not consider the variability used and the level of glycemic control Home concluded that the Nissen among studies. A classic methodological achieved through 18 months in the first study is “a poor basis for making deci- article with authors including the late 1,122 participants has been reported sions, . . . [that] TZDs have a continuing Thomas Chalmers, a founder of the use of (31). One weakness is loss to follow-up of role to play in glucose-lowering therapy, meta-analysis for quantitative compari- 218 and 223 of the participants with and [and] that their role needs to be redefined son of multiple randomized controlled without rosiglitazone, respectively. In ad- in the guidelines [as] individual physi- trials, showed the fixed effect model to dition, 140 and 224 transferred to insulin cians should not be relied upon to make lead to more errors in analyses including (rescue with insulin was required by these assessments.” He ended by noting both larger and smaller studies (34). The study design for A1C Ͼ8.5%), and 74 and an observation that “worried me and still Cochrane Collaboration “open learning 80 died, respectively. The interim com- worries me.” FDA licensing files show car- material” states that “fixed effect meta- parison was of all individuals receiving diovascular events with metformin vs. analysis are based on the mathematical rosiglitazone compared with all in the comparators occurring at rates of 1.3 vs. assumption that a single common (or comparator group. Their baseline age was 0.8%, with rosiglitazone at rates of 2.0 vs. ‘fixed’) effect underlies every study in the 59 years, 51% were male, diabetes dura- 1.5%, with at rates of 0.8 vs. meta-analysis . . . . A random effects anal- tion was 7 years, and A1C averaged 7.9%. 0.3%, and in registration ysis makes the assumption that individual Home presented the results of the in- studies there were five vs. no deaths; “so studies are estimating different treatment terim analysis of adjudicated events at there does appear to be something we effects .... Thedebateisnotabout 3.75 years (24). The primary end point of don’t understand from the early short- whether the underlying assumption of a CVD hospitalization (for acute myocar- term studies.” fixed effect is likely (clearly it isn’t)” (35). dial infarction, congestive heart failure, At the conclusion of the debate, a The random effect approach, then, is con- stroke, unstable angina pectoris, transient panel comprised of John Buse (Chapel sidered the more conservative approach. ischemic attack, unplanned cardiovas- Hill, NC), Barry Goldstein (Philadelphia, If as suggested in the Cochrane docu- cular revascularization, amputation of PA), and David Nathan (Boston, MA) ment, one analyzes the data in Table 3 of extremities, or any other definite cardio- agreed that for individuals treated with Nissen’s study with both fixed and ran- vascular reason) or death from cardiovas- rosiglitazone in good control of diabetes dom effect approaches, the fixed effect cular causes (including heart failure, and lipids that “there would be more risk analysis relative risk is 1.42 with 95% CI acute myocardial infarction, sudden in switching.” Although Buse suggested 1.03–1.96 (P ϭ 0.033), agreeing with the death, and death caused by acute vascular discontinuing for individuals not in good article, while the random effect relative events including stroke) occurred in 217 control, Goldstein expressed a “reluc- risk is 1.30 with 95% CI 0.94–1.79 (P ϭ individuals receiving rosiglitazone and tance to abandon these agents . . . [which 0.110) (36). 202 control subjects—not an unexpected are] related to the pathogenesis of diabe- Nissen’s analysis used the number of finding given the inclusion of a number of tes with inflammation and cytokines and events rather than time to first event. He heart failure end points. Cardiovascular increased FFA [free fatty acid] levels,” justified this approach based on the lack death occurred in 29 vs. 35 individuals, commenting that “a signal of cardiovascu- of what he acknowledged would be the and total mortality rates were 74 vs. 80. lar damage . . . is surprising.” Nathan sug- more accurate time-in-study data. This The combined end point of cardiovascu- gested that “there are insufficient data to approach, however, biases against longer lar death, myocardial infarction, and come to any conclusion that would con- trials and weights toward the short-term stroke, a standard approach used in many vince us all.” studies. Furthermore, it appears that trials, occurred in 93 vs. 96 individuals; rosiglitazone, as a particularly effective there were 43 vs. 37 myocardial infarc- Further considerations glycemic treatment, is associated with tions; and heart failure occurred signifi- A number of additional considerations greater adherence. This would lead to cantly more frequently in 38 vs. 17 appear relevant to the question of whether shorter length of follow-up among indi- individuals. “Those are the findings,” the Nissen meta-analysis was properly viduals taking comparators, who would Home said, pointing out that “rosiglita- performed. It appears from the presenta- nevertheless be included in study results zone appears to behave similarly for car- tion and from review of the article that using the “last observation carried for- diovascular death, all-cause death, and there was no literature search for all rele- ward” approach, allowing clinical trials to cardiovascular composite . . . [This] sug- vant studies. The Cochrane meta-analysis use all enrolled individuals in analyzing gests that rosiglitazone should still have a protocols require such a literature search efficacy of various treatments. Such an ap- role . . . [although] increase in myocardial to establish the base for an analysis, and a proach, while being conservative in only infarction not causing death cannot be literature search readily shows random- comparing individuals responding to ruled out.” The strengths of RECORD are ized controlled trials of rosiglitazone treatment, leads to a shorter period of ob- its specific design for ascertainment of treatment in which myocardial infarc- servation for individuals taking the less cardiovascular outcomes, with a long- tions were ascertained in the treatment effective treatment. term trial in a large cohort, using active and control groups (32,33), including As noted by Nissen, it is not possible comparators. Home acknowledged that one such trial carried out in Nissen’s in- to ascertain the duration of follow-up for the lower than predicted event rate low- stitution (30). all the studies. The largest study, ADOPT,

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2405 Perspectives on the News does give interesting information along that a warning label will be required, and suggesting that pioglitazone might im- these lines. Figure 2 of this study shows it is uncertain whether this will apply only prove ␤-cell function either directly or by the number of individuals still receiving to a putative myocardial infarction risk of reducing lipo- or glucotoxicity, poten- rosiglitazone, metformin, and glyburide rosiglitazone or whether the well- tially changing the natural history of the at 1–5 years. Summing the numbers of established heart failure risks for both ros- disease. In 40 women with polycystic individuals enrolled at these time points iglitazone and pioglitazone will lead to ovary syndrome, Jude et al. (abstract 584) suggests that there were 4,410 person- such a label for both of these agents. compared the effects of 500 mg met- years of observation with rosiglitazone formin twice daily with those receiving 4 but 3,688 person-years of observation Additional TZD presentations mg rosiglitazone daily, finding a 2-kg with glyburide—a 16% lower adherence. As always at the ADA meetings with every weight loss vs. 1-kg weight gain after 3 Even only using the 1-year ADOPT fol- topic, there were numerous fascinating months but with greater falls in gonado- low-up data, there were 1,207 individuals studies on aspects of TZD treatment. Sev- tropins, fasting glucose, and insulin levels receiving rosiglitazone but 1,114 receiv- eral studies addressed mechanism of ac- with the latter agent. Fetuin-A is an in- ing glyburide—an 8% lower rate. From tion. Basu et al. (abstract 250) studied the flammation-related glycoprotein secreted Nissen’s Table 1, it appears that 10 of the insulin resistance produced in 31 type 2 by the liver, acting to inhibit soft-tissue smaller studies (excluding ADOPT and diabetic individuals after infusion of a fat calcification, appearing to reduce insulin DREAM) were placebo controlled and emulsion with heparin to increase FFA receptor tyrosine kinase activity and, that in 16 the control group received a levels during a hyperinsulinemic- hence, associated with insulin resistance. sulfonylurea. This observation alone euglycemic clamp, showing a reduction Mori et al. (abstract 611), however, re- might lead one to question the assump- in insulin-induced stimulation of glucose ported increased serum fetuin-A levels in tion of the fixed-effect analysis. Further- utilization (abstract numbers refer to the 10 type 2 diabetic patients receiving pio- more, if study adherence is lower by an American Diabetes Association Scientific glitazone for 6 months, while neither met- appreciable amount in the rosiglitazone Sessions, Diabetes 56 [Suppl. 1], 2007). formin (given to nine individuals) nor an comparator groups and if the time for After 4 months of treatment, the effect of exercise program (eight participants) observation of serious adverse events is FFA on peripheral glucose utilization ap- changed fetuin-A levels, despite similar reduced, the increase in observed myo- peared particularly lessened by met- falls in A1C. Fetuin-A may be mechanis- cardial infarctions with rosiglitazone formin, while pioglitazone both increased tically relevant to TZD action, both on might simply reflect longer follow-up glucose utilization and decreased hepatic glycemia and, as recently reported, on with this agent. glucose production. Busui et al. (abstract bone loss with these agents. Apparently, Nissen excluded either 600) administered 8 mg rosiglitazone Cardiovascular effects of TZDs were four or six studies not reporting events; it daily or 10 mg glyburide daily for 6 addressed in a number of studies. Dietlein is difficult to ascertain this from the arti- months to 27 type 2 diabetic individuals, et al. (abstract 2,125) and Bierwirth et al. cle. There are standard ways of dealing finding that only the former agent re- (abstract 494) reported findings in a 26- with a zero event rate, and inclusion of duced plasma nitrotyrosine (by 82%), as week, open, multicenter observational these studies would, of course, have low- well as decreased C-reactive protein and trial of 1,426 type 2 diabetic individuals ered the overall event rate. Finally, Nis- von Willenbrad antigen; glyburide was treated with 30 or 45 mg pioglitazone sen’s failure to define the hypothesis to be associated with a reduction in myocardial daily, finding that cartotid IMT decreased tested, as discussed by Home, leads to an blood flow both at rest and following 57% more with the higher dose. There analysis that is not prespecified and vio- stress testing. Aso et al. (abstract 333) was a 1% reduction in A1C both in 627 lates Cochrane protocol. One can imagine compared effects of pioglitazone with individuals receiving statins and in 752 that multiple end points could have been those of voglibose administered for 12 not on such treatment; the former had considered, such as myocardial infarc- weeks to 34 patients with type 2 diabetes, twice as great a reduction in mean carotid tion, stroke, cardiovascular death, pro- with comparable improvements in glyce- IMT, certainly suggesting that there may cedures, hospitalizations, etc., so that mia. The high molecular weight–to–total be benefits to the combined treatment ap- without an a priori hypothesis the statis- adiponectin ratio increased only with pio- proach, contradicting Skyler’s assertion tical significance of an association be- glitazone, and in the pioglitazone-treated from the PROactive subset analysis. In a comes unknown. patients, adiponectin showed a negative study of the effects of pioglitazone, sim- correlation with the change in A1C. vastatin, and the combination in 125 non- FDA advisory panel Kutoh and Wajs (abstract 609) com- diabetic individuals (109 with insulin recommendations pared effects of 3 months of treatment resistance by the homeostasis model), A 23-member advisory panel met 30 July with pioglitazone in 26 individuals with Scho¨ndorf et al. (abstract 613) reported 2007 to review this information (37). Al- baseline fasting insulin Ͼ15 vs. 27 with that insulin sensitivity improved 43% though at the time of publication no for- levels Ͻ7 ␮U/ml. The two groups differed with pioglitazone, while not changing mal statement had been issued, their in baseline A1C (9.1 vs. 10.0%), triglyc- with simvastatin, with parallel changes in recommendations were widely reported. eride (224 vs. 166 mg/dl), and HDL cho- adiponectin, while C-reactive protein de- The panel voted 20 to 3 that there was lesterol (50 vs. 61 mg/dl). In the high creased with both agents, and retinol sufficient evidence of increased myocar- insulin group, A1C decreased to 7.5%, binding protein 4 failed to change with dial infarction risk for concern but at the with improvement in triglyceride and either. same time did not consider the evidence HDL cholesterol and with a 41% reduc- Spanheimer et al. (abstract 553) re- sufficient to withdraw the agent from clin- tion in fasting insulin, while the low insu- ported a significantly greater rate of hos- ical use, voting 22 to 1 that the drug con- lin group showed a fall in A1C to 8.5% pitalizations for cardiac ischemia in tinue to be marketed (38). It is anticipated and a 126% increase in fasting insulin, placebo- than in pioglitazone-treated par-

2406 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 Bloomgarden ticipants in the PROactive Study of 5,238 trol Cardiovascular Risk in Diabetes 14. Erdmann E, Dormandy JA, Charbonnel B, type 2 diabetic individuals with macro- (ACCORD) Trials [article online], 2007. Massi-Benedetti M, Moules IK, Skene AM; vascular disease, with 14.2 vs. 11.9% re- Available from http://www.nhlbi.nih.gov/ PROactive Investigators: The effect of pio- quiring such hospitalization. Borchert et new/press/07-rosi-qa.htm. Accessed 11 glitazone on recurrent myocardial infarc- al. (abstract 251) reported observations July 2007 tion in 2,445 patients with type 2 diabetes 5. The Endocrine Society: The Endocrine and previous myocardial infarction: re- from a randomized controlled trial of 45 Society statement to providers on the re- sults from the PROactive (PROactive 05) mg pioglitazone vs. glimepiride (mean port published in the New England Jour- Study. J Am Coll Cardiol 49:1772–1780, dose 2.7 mg) daily in 180 type 2 diabetic nal of Medicine on Avandia [article online], 2007 individuals, finding that 7 of the pioglita- 2007. Available from http://www.endo- 15. Wilcox R, Bousser MG, Betteridge DJ, zone-treated individuals developed evi- society.org/publicpolicy/policy/avandia. Schernthaner G, Pirags V, Kupfer S, Dor- dence of heart failure, all with elevated cfm. Accessed 11 July 2007 mandy J; PROactive Investigators: Effects baseline brain natriuric peptide and 5 6. Inzucchi SE, Maggs DG, Spollett GR, Page of pioglitazone in patients with type 2 di- with baseline levels Ͼ100 pg/ml. Edema, SL, Rife FS, Walton V, Shulman GI: Effi- abetes with or without previous stroke: in contrast, although occurring threefold cacy and metabolic effects of metformin results from PROactive (PROspective pio- more often in individuals treated with and troglitazone in type II diabetes melli- glitAzone Clinical Trial In macroVascular tus. N Engl J Med 338:867–872, 1998 Events 04). Stroke 38:865–873, 2007 pioglitazone, was not associated with the 7. Riche DM, Valderrama R, Henyan NN: 16. Masoudi FA, Inzucchi SE, Wang Y, baseline brain natriuric peptide level. and risk of repeat tar- Havranek EP, Foody JM, Krumholz HM: This may offer a useful approach to the get vessel revascularization following per- Thiazolidinediones, metformin, and out- determination of individuals for whom cutaneous coronary intervention: a meta- comes in older patients with diabetes and TZD treatment is not appropriate. Balas et analysis. Diabetes Care 30:384–388, 2007 heart failure: an observational study. Cir- al. (abstract 617) measured whole-body 8. Minamikawa J, Tanaka S, Yamauchi M, culation 111:583–590, 2005 fat and total body water with dual-energy Inoue D, Koshiyama H: Potent inhibitory 17. Buchanan TA, Xiang AH, Peters RK, Kjos X-ray absorptiometry, a tritiated water di- effect of troglitazone on carotid arterial SL, Marroquin A, Goico J, Ochoa C, Tan wall thickness in type 2 diabetes. J Clin S, Berkowitz K, Hodis HN, Azen SP: Pres- lution technique, and bioimpedance in ␤ 35 individuals with nonalcoholic steato- Endocrinol Metab 83:1818–1820, 1998 ervation of pancreatic -cell function and hepatitis and type 2 diabetes or impaired 9. Koshiyama H, Shimono D, Kuwamura N, prevention of type 2 diabetes by pharma- Minamikawa J, Nakamura Y: Inhibitory cological treatment of insulin resistance in glucose tolerance receiving pioglitazone effect of pioglitazone on carotid arterial high-risk Hispanic women. Diabetes 51: versus placebo, showing a 2.5-kg weight wall thickness in type 2 diabetes. J Clin 2796–2803, 2002 gain over 6 months with pioglitazone, Endocrinol Metab 86:3452–3456, 2001 18. Knowler WC, Hamman RF, Edelstein SL, without change in patients receiving pla- 10. Sidhu JS, Kaposzta Z, Markus HS, Kaski Barrett-Connor E, Ehrmann DA, Walker cebo. Neither total body nor extracellular JC: Effect of rosiglitazone on common ca- EA, Fowler SE, Nathan DM, Kahn SE; water changed, suggesting an increase in rotid intima-media thickness progression Diabetes Prevention Program Research fat mass, with this population not show- in coronary artery disease patients with- Group: Prevention of type 2 diabetes with ing evidence of fluid retention. Evange- out diabetes mellitus. Arterioscler Thromb troglitazone in the Diabetes Prevention lista et al. (abstract 2,286) randomized Vasc Biol 24:930–934, 2004 Program. Diabetes 54:1150–1156, 2005 50 type 2 diabetic individuals to a 16- 11. Mazzone T, Meyer PM, Feinstein SB, Da- 19. DREAM (Diabetes REduction Assessment vidson MH, Kondos GT, D’Agostino RB with ramipril and rosiglitazone Medica- week period of treatment with pioglita- Sr, Perez A, Provost JC, Haffner SM: Effect tion) Trial Investigators, Gerstein HC, zone versus metformin—the former of pioglitazone compared with glimepiride Yusuf S, Bosch J, Pogue J, Sheridan P, associated with greater falls in C-reactive on carotid intima-media thickness in type Dinccag N, Hanefeld M, Hoogwerf B, protein, hepatic transaminase levels, fast- 2 diabetes: a randomized trial. JAMA 296: Laakso M, Mohan V, Shaw J, Zinman B, ing plasma glucose, A1C, serum insulin, 2572–2581, 2006 Holman RR: Effect of rosiglitazone on the and E-selectin but with a significant in- 12. Inzucchi SE, Masoudi FA, Wang Y, Kosi- frequency of diabetes in patients with im- crease in total cholesterol levels. borod M, Foody JM, Setaro JF, Havranek paired glucose tolerance or impaired fast- EP, Krumholz HM: Insulin-sensitizing ing glucose: a randomised controlled trial. antihyperglycemic drugs and mortality Lancet 368:1096–1105, 2006 References after acute myocardial infarction: insights 20. Kahn SE, Haffner SM, Heise MA, Herman 1. Nissen SE, Wolski K: Effect of rosiglita- from the National Heart Care Project. Di- WH, Holman RR, Jones NP, Kravitz BG, zone on the risk of myocardial infarction abetes Care 28:1680–1689, 2005 Lachin JM, O’Neill MC, Zinman B, Viberti and death from cardiovascular causes. 13. Dormandy JA, Charbonnel B, Eckland DJ, G; ADOPT Study Group: Glycemic dura- N Engl J Med 356:2457–2471, 2007 Erdmann E, Massi-Benedetti M, Moules bility of rosiglitazone, metformin, or gly- 2. Psaty BM, Furberg CD: Rosiglitazone and IK, Skene AM, Tan MH, Lefebvre PJ, Mur- buride monotherapy. N Engl J Med 355: cardiovascular risk. N Engl J Med 356: ray GD, Standl E, Wilcox RG, Wilhelmsen 2427–2443, 2006 2522–2524, 2007 L, Betteridge J, Birkeland K, Golay A, 21. UK Prospective Diabetes Study (UKPDS) 3. Saul S: Heart attack risk seen in drug for Heine RJ, Koranyi L, Laakso M, Mokan M, Group: Intensive blood-glucose control diabetes [article online], 22 May 2007. Norkus A, Pirags V, Podar T, Scheen A, with sulphonylureas or insulin compared New York Times. Available from http:// Scherbaum W, Schernthaner G, Schmitz with conventional treatment and risk of www.nytimes.com/2007/05/22/business/ O, Skrha J, Smith U, Taton J; PROactive complications in patients with type 2 di- 22drug.html. Accessed 22 June 2007 Investigators: Secondary prevention of abetes (UKPDS 33). Lancet 352:837–853, 4. National Heart, Lung, and Blood Insti- macrovascular events in patients with 1998 tute: Questions and answers: use of ros- type 2 diabetes in the PROactive Study 22. Lindstrom J, Ilanne-Parikka P, Peltonen iglitazone (Avandia) in the National Heart, (PROspective pioglitAzone Clinical Trial M, Aunola S, Eriksson JG, Hemio K, Ha- Lung, and Blood Institute’s Bypass Angio- In macroVascular Events): a randomised malainen H, Harkonen P, Keinanen-Kiu- plasty Revascularization Investigation 2 controlled trial. Lancet 366:1279–1289, kaanniemi S, Laakso M, Louheranta A, Diabetes (BARI 2D) and Action to Con- 2005 Mannelin M, Paturi M, Sundvall J, Valle

DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007 2407 Perspectives on the News

TT, Uusitupa M, Tuomilehto J; Finnish consensus statement from the American plantation in nondiabetic patients with Diabetes Prevention Study Group: Sus- Diabetes Association and the European metabolic syndrome. Chin Med J (Engl) tained reduction in the incidence of type 2 Association for the Study of Diabetes. Di- 119:1171–1175, 2006 diabetes by lifestyle intervention: fol- abetes Care 29:1963–1972, 2006 34. Cappelleri JC, Ioannidis JP, Schmid CH, low-up of the Finnish Diabetes Preven- 29. McAfee AT, Koro C, Landon J, Ziyadeh N, de Ferranti SD, Aubert M, Chalmers TC, tion Study. Lancet 368:1673–1679, 2006 Walker AM: Coronary heart disease out- Lau J: Large trials vs meta-analysis of 23. DREAM Trial Investigators: Effect of comes in patients receiving antidiabetic smaller trials: how do their results com- ramipril on the incidence of diabetes. agents. Pharmacoepidemiol Drug Saf 16: pare? JAMA 276:1332–1338, 1996 N Engl J Med 355, 2006 DOI: 10.1056/ 711–725, 2007 35. The Cochrane Collaboration: Things you NEJMoa065061/DC/1/1. 30. Bhatt D, Chew D, Grines C, Mukherjee D, can do with diversity and heterogeneity 24. Home PD, Pocock SJ, Beck-Nielsen H, Leesar M, Gilchrist I, Corbelli J, Blanken- [article online], 2002. Available from http:// Gomis R, Hanefeld M, Jones NP, Komajda ship J, Eres A, Steinhubl S: Peroxisome www.cochrane-net.org/openlearning/ M, McMurray JJ; RECORD Study Group: proliferator–activated receptor gamma HTML/mod13-4.htm. Accessed 8 July Rosiglitazone evaluated for cardiovascu- agonists for the Prevention of Adverse 2007 lar outcomes–an interim analysis. N Engl events following percutaneous coronary 36. Cucherat M, Boissel JP, Leizorovicz A: J Med 357:28–38, 2007 Revascularization: results of the PPAR EasyMA: a program for the meta-analysis 25. Nissen SE: The DREAM trial. Lancet 368: Study. Am Heart J 154:137–143, 2007 of clinical trials: computer methods and 2049, 2006 31. Home PD, Jones NP, Pocock SJ, Beck- programs. Biomedicine 53:187–190, 1997 26. Nissen SE, Wolski K, Topol EJ: Effect of Nielsen H, Gomis R, Hanefeld M, Koma- 37. FDA briefing document: Division of Metabo- on death and major adverse jda M, Curtis P; RECORD Study Group: lism and Endocrine Products and Office of cardiovascular events in patients with Rosiglitazone RECORD study: glucose Surveillance and Epidemiology Joint meeting type 2 diabetes mellitus. JAMA 294:2581– control outcomes at 18 months. Diabet of the Endocrinologic and Metabolic Drugs 2586, 2005 Med 24:626–634, 2007 Advisory Committee and the Drug Safety 27. Jeffrey S: Chicago: pioglitazone slows pro- 32. Hallsten K, Virtanen KA, Lonnqvist F, and Risk Management Advisory Committee, gression of carotid atherosclerosis vs Janatuinen T, Turiceanu M, Ronnemaa T, July 30, 2007 [article online]. Available glimepiride [article online], 2007. Avail- Viikari J, Lehtimaki T, Knuuti J, Nuutila from http://www.fda.gov/ohrms/dockets/ able from http://www.medscape.com/ P: Enhancement of insulin-stimulated ac/07/briefing/2007-4308b1-02-fda-back viewarticle/547748. Accessed 8 July 2007 myocardial glucose uptake in patients grounder.pdf. Accessed 3 August 2007 28. Nathan DM, Buse JB, Davidson MB, Heine with type 2 diabetes treated with rosigli- 38. Harris G: Diabetes drug backed, but with RJ, Holman RR, Sherwin R, Zinman B: tazone. Diabet Med 21:1280–1287, 2004 warnings. New York Times, 31 July 2007. Management of hyperglycemia in type 2 33. Cao Z, Zhou YJ, Zhao YX, Liu YY, Guo YH, Available from http://www.nytimes.com/ diabetes: a consensus algorithm for the Cheng WJ: Rosiglitazone could improve 2007/07/31/health/31drug.html. Accessed initiation and adjustment of therapy: a clinical outcomes after coronary stent im- 2 August 2007

2408 DIABETES CARE, VOLUME 30, NUMBER 9, SEPTEMBER 2007