Current Diabetes Reports

, Volume 10, Issue 1, pp 24–31

Does Choice of Antidiabetes Therapy Influence Macrovascular Outcomes?


    • Division of Cardiovascular MedicineBridgeport Hospital

DOI: 10.1007/s11892-009-0083-9

Cite this article as:
Zarich, S.W. Curr Diab Rep (2010) 10: 24. doi:10.1007/s11892-009-0083-9


Despite a clear epidemiologic relationship between hemoglobin A1c levels and the risk of cardiovascular (CV) disease in patients with type 2 diabetes mellitus (T2DM), prospective studies examining the benefit of intensive glucose lowering in reducing CV events have yielded conflicting results. Controversy over the choice of antidiabetic therapy for lowering macrovascular events has existed for nearly four decades, beginning with the potential risk of increased CV mortality with sulfonylurea use. Although sulfonylureas were subsequently felt to be safe, a more recent controversy was raised as to whether rosiglitazone use was associated with an increased risk of CV events. Additionally, early positive results for metformin in reducing macrovascular events have not been clearly substantiated. Because a typical patient with T2DM may live 20 to 40 years with the disease, long-term prevention of CV events is very important. An evidenced-based review of choice of antidiabetic therapy to reduce CV events in T2DM is discussed below.


Cardiovascular diseaseAntidiabetic medicationsType 2 diabetes mellitus

Clinical Trial Acronyms


Action to Control Cardiovascular Risk in Diabetes


Bypass Angioplasty Revascularization Investigation 2 Diabetes


Hyperglycemia and Its Effect After Acute Myocardial Infarction on Cardiovascular Outcomes in Patients with Type 2 Diabetes Mellitus


Prospective Pioglitazone Clinical Trial in Macrovascular Events


Rosiglitazone Evaluation for Cardiac Outcomes and Regulation of Glycaemia in Diabetics

4 T

Treat To Target in Type 2 Diabetes


University Group Diabetes Program


United Kingdom Prospective Diabetes Study


Veteran Affairs Diabetes Trial


Type 2 diabetes mellitus (T2DM) remains a leading cause of morbidity and mortality in the United States. Individuals with diabetes mellitus have a two- to fourfold increased incidence of cardiovascular (CV) disease, as well as nearly twice the mortality from myocardial infarction, compared to individuals without diabetes mellitus [1, 2]. T2DM is now considered a “CV risk equivalent” [3] and its prevalence is expected to double worldwide by the year 2030 [4]. Because up to 80% of individuals with diabetes mellitus will die of CV causes [5], this rapidly increasing burden of T2DM underscores the need for CV disease reduction by continued adherence to evidence-based therapies.

Sulfonylureas and Metformin

Although trials of glucose lowering showed a definitive reduction in microvascular complications with strict glycemic control, the effect on macrovascular events remains controversial. Prospective epidemiologic studies showed a direct relationship exists between hemoglobin A1c (HbA1c) levels and CV events and death (an 18% increased risk of CV events and a 12% to 14% increase in mortality for each 1% increase in HbA1c levels) [6]. However, the sulfonylurea (SU) tolbutamide was associated with a higher CV mortality in the UGDP study compared with insulin or placebo use [7]. Subsequently, a boxed warning for potential increased CV mortality was included in the labeling of SUs by the US Food and Drug Administration (FDA), despite flaws in study design and patient selection in the UGDP. Concern over the CV safety of SUs has focused on a potential adverse effect on myocardial ischemic preconditioning, a process by which repetitive ischemic episodes result in protection of the myocardium from subsequent ischemia. Experimental and clinical data have documented inhibition of ischemic preconditioning with predominantly early-generation SUs (tolbutamide, glibenclamide, and glipizide) [8]. Additionally, glyburide has been shown to increase peripheral vascular tone and to block the vasodilating effect of diazoxide [9].

Newer SUs, such as glimepiride, have lesser effects on extrapancreatic ATP-dependent potassium channels than conventional agents and do not appear to inhibit ischemic preconditioning significantly [10]. Subsequently, the UKPDS 33 (a 10-year trial that compared an intensive glycemic control approach with a conventional approach using SUs and insulin) showed no adverse effect on CV disease outcomes with use of first- or second-generation SU agents [11]. When the intensive group was further subgrouped into SU versus conventional therapy, there was a borderline 22% reduction in myocardial infarction with SU use (P = 0.06). SUs have since remained tier 1 agents as first-line therapy in T2DM, as recently reiterated in the European Association for the Study of Diabetes (EASD)/American Diabetes Association (ADA) consensus paper [12]. Unfortunately, the UKPDS trial found only a borderline association between intensive glycemic control and decreased risk of myocardial infarction (16% risk reduction; P = 0.052). Each 1% reduction in HbA1c levels resulted in a 14% decreased risk of myocardial infarction compared with a 37% reduction in the risk of microvascular events [13]. However, the study was performed in patients with newly diagnosed T2DM and excluded those with CV disease, was underpowered (4.1% incidence of myocardial infarction over 6 years in 5,102 participants), and the level of glycemic control was modest and lacked durability.

In contrast to the modest effect of intensive glycemic control on macrovascular events with SU and insulin in the UKPDS 33, myocardial infarction risk was reduced 39% and death was reduced by 36% in the UKPDS 34 in 753 obese patients randomized to intensive treatment with metformin for 10.7 years (P = 0.01), despite only a modest hypoglycemic effect (HbA1c reduced by 0.6%) [14]. The results in this small group of patients (342 allocated to metformin) have never been replicated and are in contrast to the negative results of the UGDP study with phenformin (although this biguanide has several fundamental pharmacologic differences from metformin and has been withdrawn from the market) [7]. A recent meta-analysis of metformin use in T2DM was unable to confirm any benefit of metformin on CV death or all-cause mortality compared with placebo or comparator therapy, despite a 26% reduction in myocardial infarction [15]. Similarly, metformin was not associated with superior CV outcomes compared with SU or rosiglitazone in a 5-year study of drug-naïve subjects with T2DM [16].

The early addition of metformin to SUs in the UKPDS also resulted in an unexpected 60% increased risk of death, compared with the addition of metformin or insulin to SU only in the setting of persistent profound hyperglycemia [14]. A subsequent meta-analysis of combination therapy with metformin and SU suggested a 43% relative increase in a composite end point of CV hospitalization and death with combination therapy, irrespective of comparator group (diet, metformin monotherapy, or SU monotherapy) [17]. There were also nonsignificant trends for excess CV and total mortality with combination therapy with metformin and SU. Finally, with the addition of SU to metformin monotherapy failures, although glycemic control is initially improved, deterioration in glycemic control may resume as early as 6 months [18]. In 1 year, nearly half of all patients on metformin/SU combination therapy had an HbA1c greater than 8.0%.

Despite the above concerns, longer-term follow-up of the UKPDS trial has confirmed the macrovascular benefit of glycemic control with SU, insulin, or metformin therapy [19••]. After active treatment had ended, 95% of the UKPDS participants were followed up for an additional 8.5 years. The results were re-analyzed, although group differences in glycated hemoglobin levels were lost within 1 year. Patients in the intensive SU-insulin group experienced a significant 15% decrease in the risk of myocardial infarction at 10 years (P = 0.01) and death from any cause of 13% (P = 0.007) compared with the conventional arm. Sustained benefits were also seen in the obese arm with intensive treatment with metformin (33% reduction in myocardial infarction and 27% reduction in all-cause mortality; P = 0.005 and 0.002, respectively). This recent 10-year follow-up of the UKPDS survivors suggests that long-term therapy with SUs, insulin, and metformin reduces macrovascular events. No separate between-group comparisons were performed to suggest which antidiabetes therapy is superior. Additionally, it appears that many years may need to elapse before any salutatory effects of antidiabetic agents are seen in macrovascular events.

Thus, metformin remains the cornerstone of first-line therapy in T2DM. Hypoglycemia is rare with metformin and modest weight loss usually ensues. Although renal dysfunction is considered a contraindication because of the risk of lactic acidosis, recent studies suggest that metformin use is safe with estimated glomerular filtration rates greater than 30 mL/min [20]. Additionally, congestive heart failure (CHF) is no longer considered a contraindication to its use. Two large retrospective analyses of Canadian and US databases showed reduced morbidity and mortality with metformin compared with SUs in patients with T2DM with incident CHF [21, 22]. Interestingly, combination metformin and thiazolidinedione (TZD) use was associated with even greater risk reduction.


TZDs seemed to hold particular promise in reducing CV events owing to their salutary nonhypoglycemic effects on endothelial function, adipocytokine production, and proinflammatory and prothrombotic CV risk factors [23]. Experimentally, rosiglitazone therapy reduces infarct size and improves left ventricular remodeling [24, 25], whereas clinically TZD use was associated with an odds ratio for myocardial infarction of 0.40 (95% CI, 0.2–0.8) [26]. Randomized trials of the effect of TZDs on restenosis after coronary stenting also suggested significant angiographic and clinical benefit [27, 28], as did trials of TZDs that examined atherosclerosis progression by serial imaging to assess carotid intima-medial thickness [29, 30]. Encouraging observations for atherosclerosis stabilization with TZDs were noted in a trial using serial intravascular ultrasound imaging that compared use of pioglitazone or glimepiride in T2DM subjects with nonobstructive coronary artery disease [31].

The PROactive trial studied the effect of pioglitazone on macrovascular event reduction in patients with T2DM with established CV disease compared with placebo [32, 33]. Although the primary composite macrovascular end point in PROactive was not reduced significantly with pioglitazone—10% relative reduction in events (P = 0.095)—the prespecified secondary composite end point of risk of time to death, myocardial infarction, or stroke was significantly reduced in the pioglitazone arm by 16% (P = 0.027). The risk of recurrent fatal or nonfatal myocardial infarction among individuals with a history of prior myocardial infarction was reduced by 28% with pioglitazone (P = 0.045), with a similar 47% decline in recurrent cerebrovascular accident (P = 0.009).

Conversely, a meta-analysis of 42 short-term (mean duration, 6 months) studies suggested an increased risk of myocardial ischemic events with rosiglitazone (odds ratio, 1.43; P = 0.03), creating a controversy over the safety of rosiglitazone and potentially the safety of TZDs as a class [34]. However, the meta-analysis did not meet many of the requisites of scientific rigor (adequate statistical power, adequate study duration to assess CV events, appropriate adjudication of events or inclusion of similar patient populations). The CV event rate (~ 0.60% myocardial infarction rate in both groups) was extremely low, with many of the trials having only 0 to 1 events, making statistical analysis difficult. Although a significant effect of rosiglitazone on myocardial infarction risk was shown by the Peto fixed-effects statistical method, no significant effect on myocardial infarction or CV death was seen when seven other statistical methods were used [35].

Nevertheless, meta-analyses by the manufacturer, GlaxoSmithKline (Clifton, NJ), and the FDA, as well as a subsequent meta-analysis of randomized trials [36], also suggested an increased risk of myocardial infarction with rosiglitazone. The meta-analysis of randomized trials included patients with mild CHF (a known relative contraindication to TZDs), which may have biased the results. A similar meta-analysis of randomized trials comparing pioglitazone with control subjects showed a reduced incidence of myocardial infarction driven by the PROactive trial, which contributed nearly 80% of all events (although a similar trend was seen in the remaining trials) [37].

An interim analysis of the RECORD trial was subsequently performed at 3.75 years into a planned 6-year study [38]. The RECORD trial was specifically designed to measure CV outcomes in 4,447 patients with T2DM treated with rosiglitazone combined with SU or metformin compared with SU and metformin. No evidence for increased myocardial infarction, CV death, or total mortality with rosiglitazone was found. However, the results were felt to be inconclusive owing to the low CV event rate (composite CV end point of 2.5% per year compared with an expected rate of 11%).

The final results of the RECORD trial were similarly deemed inconclusive regarding rosiglitazone and myocardial infarction risk owing to the occurrence of fewer than expected CV events during the 5.5 years of mean follow-up [39••]. Rosiglitazone again was not associated with an increased risk of CV morbidity or mortality compared with conventional therapy with SU and metformin (although the efficacy of this combination has been questioned as mentioned above, it remains a well-validated tier 1 core therapy by recent consensus). However, the confidence limits for the primary end point’s hazard ratio excluded the predefined 20% excess risk, thus satisfying the criteria of noninferiority. In total, an excess of eight cases of myocardial infarction with rosiglitazone use were more than offset by the occurrence of 17 fewer stokes and 21 fewer deaths with rosiglitazone compared with SU and metformin. Blood glucose control was also superior at 5 years with rosiglitazone compared with both metformin and SU. However, a nearly twofold increased risk of CHF with rosiglitazone was confirmed in this study, which is a recognized class effect of TZDs.

With the accumulation of over 25,000 person-years of follow-up (a nearly fivefold greater exposure to rosiglitazone compared with the Nissen and Wolski meta-analysis [34]) alone in the RECORD trial, there was no increase in CV morbidity or mortality with rosiglitazone. In a subsequent meta-analysis of randomized long-term prospective trials of TZDs, no association was seen with TZDs and increased CV death; also, no in-class difference in CV mortality was found between the use of rosiglitazone or pioglitazone [40].

Long-term follow-up of patients receiving rosiglitazone was subsequently reported in recent trials investigating the role of intensive glycemic control compared with standard antidiabetic therapy [41, 42]. In the VADT and the ACCORD study group, rosiglitazone use was not associated with an increased risk of myocardial infarction or death compared with traditional oral agents or insulin. Both studies contained a disproportionately higher number of patients on rosiglitazone in the intensively treated than the conventional treatment arm. Combining these data with the findings of prior studies yields over 40,000 patient-years of exposure to rosiglitazone in long-term prospective, randomized controlled trials in which no excess of CV events is seen.

Most recently, the BARI 2D trial randomized T2DM patients with established stable coronary artery disease in a 2 × 2 factorial design to early revascularization with intensive medical therapy versus intensive medical therapy alone and to a strategy of insulin-sensitizing (metformin and TZD) versus insulin-providing (SU and insulin) therapy [43••]. Rosiglitazone was the TZD used in nearly 90% of patients. Overall, mortality and CV event rates were similar for the revascularization/medical therapy and insulin sensitizing/providing arms. In participants assigned to coronary artery bypass grafting, those allocated to the insulin-sensitization arm had fewer major adverse CV events than those in the insulin-providing group. Because the insulin-sensitizing arm was associated with better glycemic control with fewer episodes of hypoglycemia, higher high-density lipoprotein levels, and less weight gain than the insulin-providing arm, insulin sensitization was felt to be potentially preferable in patients with T2DM and stable coronary artery disease. Interestingly, no increase in CHF was seen in the insulin-sensitizing group using predominantly metformin and TZDs.

Unfortunately, no head-to-head trials assessing CV risk between the two TZDs exist presently, but a future trial is planned ( NCT008879970). The two specifically designed TZD CV outcome trials (RECORD and PROactive) differed in study design; RECORD was an active comparator trial, whereas PROactive was placebo-controlled. Using the UKPDS outcomes model, the macrovascular benefit of pioglitazone in PROactive was solely due to reduction in HbA1c and CV risk factors and was not due to any effect on novel coronary artery risk factors [44].

Because TZD use is complicated by an increased risk of CHF, the net CV benefit has been questioned. However, despite an increased risk of serious CHF with pioglitazone in the PROactive trial (5.6% vs 4.1% with placebo), pioglitazone use was associated with fewer primary and secondary events and a trend toward reduced mortality compared with placebo in patients with serious CHF [33]. In patients without pre-existing CHF, although TZD use increased the risk of subsequent CHF (risk ratio, 1.72), CV mortality was not affected compared with placebo or comparative therapy [40]. Similarly, in a meta-analysis of individuals with T2DM with pre-existing CHF [45], all-cause mortality was actually reduced by 17% (P = 0.02) with TZDs, despite an increased risk of CHF. As noted above, patients with T2DM who had coronary artery disease followed closely in the BARI 2D trial did not show an excess risk of CHF with use of metformin and TZD [43••]. Because rosiglitazone does not adversely affect left ventricular function or remodeling when patients with T2DM and class I or II CHF are followed by serial echocardiography, the predominant cause of CHF with TZDs appears related to fluid retention [46]. TZDs are currently contraindicated in patients with class III or IV CHF and are not recommended in less severe cases of CHF according to current FDA guidelines.

Therefore, the promise of TZDs to alter CV event rates above and beyond glycemic control has not been realized in large outcome trials. However, the preponderance of data has confirmed the safety of TZDs in patients with T2DM at high risk for CV disease. Additionally, rosiglitazone use is associated with more durable glycemic control. In 4,360 drug-naïve patients with T2DM, rosiglitazone was 32% more effective than metformin and 63% more effective than SUs in maintaining durable glycemic control over a 5-year period [16]. The superiority of rosiglitazone over conventional oral antidiabetic agents in sustaining glycemic control was confirmed in 5.5 years of follow-up in the RECORD trial. Thus, TZDs when used with appropriate caution remain a viable option, especially in combination with metformin, to achieve durable glycemic goals. However, TZDs in general are associated with an increased risk of long bone fractures in postmenopausal women, and rosiglitazone is relatively contraindicated for use with nitrates or insulin. The recent caution against the use of rosiglitazone in the EASD/ADA consensus paper appears premature, although the use of pioglitazone may be associated with a more beneficial effect on lipid parameters compared with rosiglitazone [12].

Other Agents

α-Glucosidase inhibitors are often overlooked because of a high incidence of gastrointestinal adverse effects (with an associated 25% to 45% discontinuation rate) and lesser effects on glucose lowering compared with metformin and SUs (HbA1c reductions of 0.5% to 0.8%). However, these agents reduce postprandial glucose levels effectively, which might provide unique CV benefits. Acarbose was found to reduce CV event rates in a diabetes prevention trial, although the trial was underpowered for CV events [47]. Confirmation of a potential beneficial effect on CV events with acarbose is awaiting confirmation in a larger trial. Until then, α-glucosidase inhibitors seem to be appropriate agents to treat patients with mild hyperglycemia, as add-on therapy, or for use when contraindications to other agents exist.

Glinides, similar to SUs, stimulate insulin secretion, but bind to a different site within the SU receptor and have shorter half-lives, necessitating more frequent administration targeted primarily for postprandial hyperglycemia. Hypoglycemia may be less frequent with glinides compared with SUs, but weight gain is similar. Repaglinide appears to be more potent in lowering HbA1c compared with nateglinide [48], but CV outcomes data are lacking for both. Nateglinide is currently being examined in a large outcomes trial.

CV outcomes data are similarly lacking for incretin mimetics (the glucagon-like peptide-1 [GLP-1] agonist exenatide) and dipeptidyl peptidase-4 inhibitors (DPP-4; sitagliptin and saxagliptin), which prevent the degradation of GLP-1 and glucose-dependent insulinotropic peptides of intestinal origin. GLP-1 receptors are present on cardiomyocytes and are reputed to have cardioprotective and positive inotropic effects [49]. Exenatide is associated with weight loss typically in the 2- to 3-kg range but is limited by up to a 45% incidence of nausea, vomiting, and diarrhea, as well as the need for twice-daily injections. Conversely, DPP-4 inhibitors are typically well tolerated and are weight neutral. Exenatide and DPP-4 inhibitors are not associated with edema or an appreciable incidence of hypoglycemia. To date, no prospective outcome trials of these agents have been reported, but a large outcomes trial with sitagliptin is planned. Postmarketing reports of pancreatitis, angioedema, and Stevens-Johnson syndrome have been reported. These agents are both expensive and only modestly effective in controlling glycemia (HbA1c reductions of 0.5% to 1.0%) and should be used cautiously until the CV safety profile of these drugs is established.


The role of insulin therapy in reducing macrovascular events in patients with T2DM was perhaps even more controversial than the use of oral antidiabetic medication in the past. Nevertheless, after 10 to 20 years of diabetes, most T2DM patients require insulin to maintain adequate glycemic control. Initial concerns centered on the potentiation of atherogenesis with exogenous insulin given the observations that endogenous hyperinsulinemia was associated with increased CV events. However, concerns over weight gain, hypoglycemia, and mitogenesis with insulin use have been tempered by evidence of salutatory nonmetabolic effects of insulin on CV risk factors as outlined by Dandona et al. [50]. Although insulin was associated with adverse CV outcomes in two observational studies published in 2008 [51, 52], concerns over adverse CV effects with insulin therapy were not borne out in the UKPDS trial or multiple trials of intensive glycemic control [41, 42]. Similarly, multiple injections of insulin reduced mortality in patients with T2DM suffering an acute myocardial infarction [53], and insulin use dramatically reduced CV events in long-term follow-up of patients with type 1 diabetes mellitus [54]. Because most patients with suboptimal glycemic control on oral agents will require insulin (> 50% in the UKPDS study), the CV safety of insulin is reassuring. An ongoing study is testing a strategy of insulin as first-line therapy, compared with a traditional strategy of insulin only after failure of response to oral medications [55].

Recently, two large prospective trials studied the role of different strategies of insulin utilization and their effects on glycemic control and CV outcomes [56, 57]. In the 4 T trial [56], 708 T2DM patients with suboptimal glycemic control on maximally tolerated doses of metformin and SUs were randomized to basal insulin detemir once daily, biphasic insulin aspart twice daily, or prandial insulin aspart three times daily for 3 years. Patients allocated to basal or prandial strategies had better HbA1c control than patients allocated to a biphasic strategy. Mean weight gain and the rate of hypoglycemia were highest in the prandial group, whereas weight gain and hypoglycemia were seen less often in the basal group. Although the study was not designed to assess CV outcomes, fewer subjects died from any cause or from a CV cause with a basal approach. Thus, the 4 T trial supports the concept of initiation of basal insulin to control fasting hyperglycemia. Fasting hyperglycemia appears to contribute more to poor glycemic control than postprandial hyperglycemia in longstanding T2DM (mean duration, 9 years). Unfortunately, however, less than 45% of patients in all three groups reached the HbA1c target of 6.5% once insulin was initiated.

The HEART2D trial [57] randomized 1,115 T2DM subjects to a strategy of prandial versus basal insulin after acute myocardial infarction. The trial was halted for statistical futility when interim analysis showed no difference in CV event rates between the prandial and basal arms. On-treatment HbA1c levels were similar in both arms, and less than expected differences in postprandial blood glucose levels were observed between groups. Unfortunately, the study was underpowered for CV events, and less than one third of both arms achieved an HbA1c level of less than 7%. Thus, the question of whether postprandial hyperglycemia is an independent risk factor for CV events remains unanswered. As glycemic control was suboptimal in both studies despite frequent need for combined basal and prandial insulin, in many cases a strategy aimed at reducing basal and postprandial hyperglycemia appears warranted.

In summary, insulin therapy has had favorable effects on CV outcomes in large, prospective randomized trials. Therefore, clinicians should feel comfortable with the addition of insulin therapy to maximize glycemic control rather than delaying the initiation of insulin for long periods of time in patients failing to maintain adequate glycemic control in response to first- and second-line pharmacotherapies. Frequent monitoring and thoughtful dosing of insulin are important to avoid hypoglycemia and its attendant adverse CV effects.

A Final Note

In two large prospective, randomized controlled trials examining the role of intensive glycemic control, no single drug or drug combination, including rosiglitazone, which was used in up to 80% of patients in two studies, was associated with excess CV morbidity or mortality [41, 42]. In prespecified subgroup analyses, patients without pre-existing CV disease and those with a baseline HbA1c less than 8% benefited from strict glycemic control, stressing the CV benefit of early, sustained glycemic control, in addition to aggressive treatment of hypertension and dyslipidemia in T2DM.


There is no convincing data that any one class of hypoglycemic agents or particular strategy (insulin sensitizing/providing) confers a significant advantage in the prevention of macrovascular events in T2DM. In recent large prospective trials examining intensive glycemic control, no single hypoglycemic agent or combination of agents was associated with increased CV events or mortality. Although caution was advised regarding the use of TZDs (in particular rosiglitazone) and the risk of myocardial ischemia in the EASD/ADA consensus paper, the results of the RECORD and BARI 2D trials suggest that TZD use (in particular rosiglitazone), as well as an insulin-sensitizing strategy in general, is safe in high-risk T2DM subjects. The ADA in conjunction with the EASD, the American College of Cardiology, and the American Heart Association continue to recommend achievement and maintenance of near normoglycemia (HbA1c < 7.0%) with lower HbA1c goals potentially considered for those with a short duration of disease, long life expectancy, and no significant CV disease. Durability of glycemic control is very important and currently appears best achieved with TZDs. Avoidance of hypoglycemia through prudent glucose monitoring and avoidance of overzealous glucose-lowering protocols seems to be justified. Monotherapy failures are common and insulin will be required in most patients with suboptimal control on oral therapy necessitating individually tailored therapy. CV outcomes data are currently lacking for many of the newer antidiabetic agents.


Dr. Stuart Zarich has received research grants, consulting, and honorarium from GlaxoSmithKline, Pfizer, Sanofi Aventis, and Novartis.

Copyright information

© Springer Science+Business Media, LLC 2010