Glucose Control and Cardiovascular Disease in Type-2 Diabetes: A Meta-Analysis

Abstract

Background

Results from clinical trials examining the effect of intensive glucose control on cardiovascular disease have been conflicting.

Purpose

To summarize clinical benefits and harms of intensive versus conventional glucose control for adults with type-2 diabetes.

Data Sources

Studies were retrieved by systematically searching the MEDLINE database (January 1950-April 2009) with no language restrictions.

Study Selection

Two independent reviewers screened abstracts or full text articles to identify randomized trials comparing clinical outcomes in type-2 diabetes patients treated with intensive compared to conventional glucose control.

Data Extraction

Two investigators independently abstracted data on study variables and outcomes including severe hypoglycemia, cardiovascular disease, and all-cause mortality.

Data Synthesis

Five trials involving 27,802 adults were included. Intensive glucose targets were lower in the three most recent trials. Summary analyses showed that, compared with conventional control, intensive glucose control reduced the risk of cardiovascular disease (relative risk (RR): 0.90, 95% confidence interval (CI): 0.83, 0.98; risk difference per 1,000 patients per 5 years (RD): -15, CI: -24, -5) but not cardiovascular death (RR: 0.97, CI: 0.76, 1.24; RD: -3, CI: -14, 7) or all-cause mortality (RR: 0.98, CI: 0.84, 1.15; RD: -4, CI: -17, 10) and increased the risk of severe hypoglycemia (RR: 2.03, CI: 1.46, 2.81; RD: 39, CI: 7, 71). Similar to overall analyses, intensive glucose control reduced risk of cardiovascular disease and increased risk of severe hypoglycemia in pooled findings from early and more recent trials.

Limitations

Summary rather than individual data were pooled across trials.

Conclusions

Intensive glucose control reduced risk for some cardiovascular disease (e.g., non-fatal myocardial infarction), but did not reduce risk for cardiovascular or all-cause mortality and increased risk of severe hypoglycemia.

Introduction

The prevalence of type-2 diabetes is increasing globally (1-3). Epidemiologic evidence indicates that diabetes is a major risk factor for cardiovascular disease (CVD), and recent data suggest that the CVD burden attributable to diabetes is on the rise (4-7). Clinical trials have demonstrated that intensive glucose control reduces the risk of microvascular complications among patients with type-2 diabetes, but its effect on CVD, including coronary heart disease (CHD), stroke, and peripheral arterial disease (PAD), is uncertain (8-10). Early data from the United Kingdom Prospective Diabetes Study (UKPDS) 34 suggested a protective effect of improved glucose control on CVD, CVD deaths and all-cause mortality (11). However, within the past year three large randomized controlled trials have reported conflicting results (12-14). While the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation (ADVANCE) trial and the Veterans Affairs Diabetes Trial (VADT) found no effect of intensive glucose control on major cardiovascular events (13, 14), the Action to Control Cardiovascular Disease in Diabetes (ACCORD) trial identified an increased risk of death from cardiovascular causes and total mortality associated with intensive glucose control (12). Based on these results, a recent perspective by Montori and colleagues suggested that additional research is needed to confirm or refute the importance of tight glucose control (15). Thus, recommendations for healthcare providers regarding optimal glycosylated hemoglobin (HbA1c) levels in patients with type-2 diabetes remain unclear.

Due to the early termination of the ACCORD trial and less events than anticipated in ADVANCE and VADT, there is real concern that these studies were under-powered to capture the true effects of intensive glucose control on CVD risk (12-14). Therefore, we conducted a meta-analysis of randomized controlled trials to examine the effects of intensive glucose control on CVD among patients with type-2 diabetes. Furthermore, we examined the separate effects of intensive glucose control on all-cause mortality, CVD mortality, CHD, congestive heart failure (CHF), stroke and PAD. In an effort to explain incongruities among trial results, we conducted subgroup analyses and examined the occurrence of severe hypoglycemia.

Methods

Data Sources and Searches

We developed and followed a standard protocol for all steps of the review. Investigators conducted a literature search of the MEDLINE database (from January 1950 through April 2009) using the Medical Subject Headings cardiovascular diseases, coronary disease, stroke, peripheral vascular diseases, hypoglycemic agents, and diabetes mellitus, type 2 as well as the keywords coronary heart disease, glucose control and glycemic control. We restricted the search to randomized controlled trials conducted among human adults (aged 19 years or older), with no language restrictions. We also performed a manual search of references cited by the published original studies and contacted experts in the area.

Study Selection

Two investigators independently reviewed the contents of 341 abstracts or full-text manuscripts identified through the literature search to determine whether they met the eligibility criteria. Studies were eligible for inclusion if: 1) the study design was a randomized controlled trial; 2) the study compared intensive glucose control to conventional treatment, with a priori specification of glycemic goals for the intensive and conventional glucose control groups; 3) clinical CVD was the primary endpoint; 4) the study sample size was ≥ 500; and 5) the study was conducted among participants with type-2 diabetes mellitus. Reviewers resolved disagreements regarding study inclusion or exclusion by consensus and reference of the original reports.

Data Extraction and Quality Assessment

Study investigators (TNK and LAB) independently abstracted data in duplicate using a standardized data-collection form. Reviewers did not contact authors to request additional information. Reviewers abstracted characteristics of each trial and its participants. Reviewers critically appraised methodologic characteristics of trials, such as randomization procedures, blinded assessment of outcomes, adjudication procedures for outcomes, and follow-up rates but did not use a scoring system to formally rate study quality of the individual trials (Appendix Table 1).

Reviewers recorded the number of clinical CVD, CHD, stroke, and CHF events, along with cardiovascular deaths and all-cause mortality for the intensive and conventional glucose control groups as the main outcomes of interest. Reviewers also recorded single endpoints including non-fatal MI, fatal MI, non-fatal stroke, fatal stroke, and PAD. In addition, reviewers recorded the number of severe hypoglycemic events for each trial arm. Because definitions of certain composite outcomes varied between trials, the definitions of each outcome are presented for each trial in Appendix Table 2.

Data Synthesis and Analysis

We examined the relationship between intensive glucose control and risk of all study outcomes using relative risk and risk difference measures. We calculated the relative risks for each trial based on the number of events in the intensive glucose control and conventional treatment groups and used these estimates for pooling analyses. In order to estimate the risk difference, we first calculated the annual absolute risk of event for participants in each trial arm by dividing the number of events in each trial arm by the corresponding number of person-years (estimated as median treatment time × number of participants in the trial arm). We then multiplied the annual absolute risk by 5 to estimate the 5 year risk for participants in each trial arm. We calculated the risk difference for each trial by subtracting the 5 year risk in the conventional glucose control group from the 5 year risk in the intensive glucose control group. We logarithmically transformed the relative risks and risk differences and their corresponding standard errors to stabilize the variance and normalize their distribution. We pooled relative risks and risk differences using both fixed-effects and DerSimonian and Laird random-effects models (16). We used inverse variance weighting to calculate fixed- and random-effects summary estimates. We assessed heterogeneity formally using Dersimonian and Laird's Q test, considering any p-value < 0.100 as evidence of heterogeneity, and by examination of the I² quantity. While fixed- and random-effects models yielded similar findings, we detected between study heterogeneity for several study outcomes (severe hypoglycemia, cardiovascular deaths, all-cause mortality, and fatal MI). Due to this heterogeneity and trial differences in median diabetes duration of participants, achieved HbA1c levels, and therapeutic regimens, we present results from the random-effects models.

We conducted a pre-stated subgroup analysis to examine the effects of intensive glucose control on all study outcomes. We then compared the relative risks of CVD, CHD, CHF, stroke, cardiovascular deaths, all-cause mortality and severe hypoglycemia, as well as fatal and non-fatal MI, fatal and non-fatal stroke, and PAD between the early UKPDS trials (8, 11) and the three more recent ACCORD, ADVANCE and VADT (12-14). We conducted all analyses in STATA version 9.2.

Role of the funding source

This study was funded in part by career development award 1K08HL091108 from the National Heart, Lung, and Blood Institute and by Award Number K12HD043451 from the Eunice Kennedy Shriver National Institute Of Child Health & Human Development. The funding sources played no role in the study design, collection, analysis and interpretation of the data, in the writing of the report, and in the decision to submit the paper for publication.

Results

A flow chart depicts the study selection process (Figure 1). We excluded two trials, the Veterans Affairs (VA) Feasibility trial (N=153) and the Kumamoto study (N=110), due to small sample sizes (9, 17). The VA Feasibility trial was a pilot study that examined whether intensive glucose control could be effectively sustained in patients with type-2 diabetes and was a precursor to the subsequent VADT. The Kumamoto study examined the effects of intensive glucose control on microvascular complications of diabetes. The current meta-analysis included a total of five trials conducted among 27,802 participants (8, 11-14). Table 1 presents the characteristics of the five randomized controlled trials and trial participants. The number of trial participants ranged from 753 to 11,140, while intervention duration ranged from 3.4 to 10.7 years. The UKPDS 33 and 34 recruited participants with newly diagnosed diabetes, which differed from the ADVANCE, ACCORD and VADT, whose participants had an average duration of diabetes ranging from 7.9 to 11.5 years at the time of trial enrollment. Although the VADT did not provide data on aspirin use, it appeared to be more common in recent trials compared to the earlier UKPDS 33 and 34.

Figure 1.

Flow of study selection process.

Table 1.

Characteristics of 5 randomized controlled trials of intensive glucose control.

	UKPDS 33 (1998)	UKPDS 34 (1998)	ACCORD (2008)	ADVANCE (2008)	VADT (2009)
Participants, number	3,867	753	10,251	11,140	1,791
Intervention duration, median, y	10.0	10.7	3.4	5.0	5.6
Treatment
Intensive glucose control	Sulfonylurea or insulin	Metformin	≥ 2 classes of hypoglycemic agents plus other drugs	Gliclazide plus other drugs	Glimperide or metformin, plus rosiglitazone, or insulin
Conventional glucose control	Diet	Diet	Diet and/or pharmacological treatment	Continue current therapy, if necessary. Those taking gliclazide substituted the drug with another sulphonylurea.	Glimperide or metformin, plus rosiglitazone, or insulin
Treatment goal
Intensive glucose control	FPG<6.0 mmol/L	FPG<6.0 mmol/L	HbA1c<6.0%	HbA1c≤6.5%	HbA1 < 6% and 1.5 percentage points less than conventional
Conventional glucose control	FPG: 6.1-15.0 mmol/L	FPG: 6.1-15.0 mmol/L	HbA1c: 7.0-7.9%	Local Standards	HbA1c <9% and HbA1c 1.5 percentage points higher than intensive
Age, mean, y	53.3	53.0	62.2	66.0	60.4
Men, %	61%	47%	61%	58%	97%
Race/Ethnicity, %
White	81%	86%	64%	…	62%
Asian	10%	5%	…	…	…
Black	8%	8%	19%	…	17%
Hispanic	…	…	7%	…	16%
Other	1%	1%	…	…	5%
Diabetes duration, mean, y	0.0*	0.0*	10.0†	7.9	11.5
Aspirin use, %	2%	2%	55%	44%	…
History of cardiovascular disease,%	…	…	35%	32%	40%

UKPDS = United Kingdom Prospective Diabetes Study

ACCORD = Action to Control Cardiovascular Risk in Diabetes

ADVANCE = Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation

VADT = Veterans Affairs Diabetes Trial

y = years

FPG = fasting plasma glucose

HbA1c = glycosylated hemoglobin

^*The UKPDS 33 and 34 trials recruited participants with newly diagnosed diabetes.

^†Median

Table 2 shows the average pre- and post-intervention values of key CVD risk factors in trial participants. On average, trial participants were overweight, with mean baseline BMI values ranging from 28 to 32 kg/m². Post-intervention weight in the ACCORD, ADVANCE and VADT was higher among those in the intensive groups compared to those in the conventional groups. Systolic blood pressure seemed to decrease between the pre-intervention and post-trial period in ACCORD, ADVANCE and VADT, while average diastolic blood pressure decreased in all studies. In general, average high-density lipoprotein levels did not change from baseline to the end of the study, while both low-density lipoproteins and triglycerides decreased in participants of all trials. HbA1c values decreased from pre- to post-intervention in the ACCORD, ADVANCE and VADT trials, while they increased over the trial periods of the UKPDS 33 and 34. Post-intervention HbA1c levels in the intensive groups of the UKPDS 33 and 34 were higher than those in the conventional groups of the ACCORD, ADVANCE and VADT. All trials showed lower post-intervention HbA1c levels in the intensive compared to the conventional glucose control group, with median differences ranging from -0.5 to -1.4 percentage points. The sample-size weighted overall difference in median Hba1c levels was -0.8 percentage points.

Table 2.

Cardiovascular disease risk factors of trial participants at pre- and post-intervention.

	UKPDS 33		UKPDS 34		ACCORD		ADVANCE		VADT
	Conventional	Intensive	Conventional	Intensive	Conventional	Intensive	Conventional	Intensive	Conventional	Intensive
Weight, mean, kg
Pre-intervention	78.1	77.3	87.0	87.0	93.6	93.5	78.0	78.2	96.3‖	96.3‖
Post-intervention	79.0*	80.0*	87.0*	86.0*	94.0§	97.0§	77.0	78.1	100.4‖	104.4‖
Body mass index, mean, kg/m2
Pre-intervention	28	28	32	32	32	32	28	28	31	31
Post-intervention	29	29	32	32	…	…	28	28	32	34
Blood pressure
Systolic, mean, mm Hg
Pre-intervention	135	135	140	140	137	136	145	145	132	131
Post-intervention	138	139	139	141	127	126	138	136	125	127
Diastolic, mean, mm Hg
Pre-intervention	82	83	86	85	75	75	81	81	76	76
Post-intervention	77	77	77	78	68	67	74	74	69	68
Lipids, mmol/L (mg/dL)
Triglyceride, median
Pre-intervention	2.31 (204¶)‡	2.37 (210¶)‡	2.96 (262¶)‡	2.79 (247¶)‡	1.74¶ (154)	1.76¶ (156)	1.64 (145¶)	1.60 (142¶)	2.52¶ (223)	2.27¶ (201)
Post-intervention	1.45¶ (128)	1.45¶ (127)	1.62¶ (143)	1.77¶ (157)	…	…	1.59 (141¶)	1.45 (128¶)	1.80¶ (159)	1.71¶ (151)
HDL, mean
Pre-intervention	1.08 (42**)	1.07 (41**)	1.04 (40**)	1.06 (41**)	1.09** (42)	1.09** (42)	1.25 (48**)	1.26 (49**)	0.93** (36)	0.93** (36)
Post-intervention	1.11** (43)	1.09** (42)	1.04** (40)	1.11** (42)	…	…	1.25 (48**)	1.24 (48**)	1.06** (41)	1.04** (40)
LDL, mean
Pre-intervention	3.5 (135**)	3.5 (135**)	3.66 (141**)	3.67 (142**)	2.72** (105)	2.72** (105)	3.11 (120**)	3.12 (121**)	2.80** (108)	2.77** (107)
Post-intervention	3.26** (126)	3.26** (126)	3.34** (129)	3.37** (130)	2.36** (91)	2.36** (91)	2.65 (102**)	2.64 (102**)	2.07** (80)	2.07** (80)
Glycosylated Hemoglobin, median, %
Pre-intervention	6.9††	7.0††	7.0††	7.0††	8.1	8.1	7.2	7.2	9.4†	9.4†
Post-intervention	8.5	7.9	8.9	8.4	7.2	6.2	7.0	6.3	8.5	7.1

UKPDS = United Kingdom Prospective Diabetes Study

ACCORD = Action to Control Cardiovascular Risk in Diabetes

ADVANCE = Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation

VADT = Veterans Affairs Diabetes Trial

kg = kilograms

kg/m² = kilograms per meters squared

mm Hg = millimeters mercury

mmol/L = millimoles per liter

mg/dL = milligrams per decileter

HDL = high-density lipoprotein

LDL = low-density lipoprotein

^*Median

^†Mean

^‡Geometric Mean

^§Calculated based on net change in weight over the study period.

^‖Estimated based on reported weight in pounds using a conversion factor of 0.45.

^¶Estimated by multiplying (dividing) by a conversion factor of 0.0113.

^**Estimated by multiplying (dividing) by a conversion factor of 0.0259.

^††Estimated from figure

Figure 2 presents the individual and pooled relative risks and risk differences (per 1,000 patients over 5 years treatment) of CVD, CHD, stroke, CHF, cardiovascular deaths, and all-cause mortality for the 5 trials. Overall analyses indicated that patients randomized to intensive glucose control had reduced risks of CVD (relative risk: 0.90; 95% CI: 0.83, 0.98 and risk difference: -15; 95% CI: -24, -5) and CHD (relative risk: 0.89; 95% CI: 0.81, 0.96 and risk difference: -11; 95% CI: -17, -5) compared to participants in the conventional treatment groups, with similar findings from subgroup analyses of the early UKPDS and more recent ACCORD, ADVANCE and VADT. We observed no overall effect of intensive glucose control on cardiovascular mortality (relative risk: 0.97; 95% CI: 0.76, 1.24 and risk difference: -3; 95% CI: -14, 7) or all-cause mortality (relative risk: 0.98; 95% CI: 0.84, 1.15 and risk difference: -4; 95% CI: -17, 10), but identified possible heterogeneity between the results of subgroup analyses (p-values for heterogeneity between subgroups=0.095 and 0.105, respectively). Pooled findings from the early UKPDS trials showed non-statistically significant protective effects of intensive glucose control on cardiovascular and all-cause mortality. In contrast, summary data from ACCORD, ADVANCE and VADT indicated non-statistically significant increased risks of these outcomes in the intensive glucose control group. There were no reductions in the overall risk of stroke or CHF associated with intensive glucose control.

Figure 2.

Pooled relative risk and risk difference (per 1,000 patients over 5 years treatment) estimates (95% confidence interval) for main study outcomes by trial, early and more recent trial subgroups, and overall. A) Cardiovascular disease; B) Coronary heart disease; C) Stroke; D) Congestive heart failure; E) Cardiovascular deaths; F) All-cause mortality.

Figure 3 shows the pooled relative risks and risk differences of non-fatal MI, fatal MI, non-fatal stroke, fatal stroke, and PAD in the early and more recent trial subgroups and overall. The ACCORD study did not present results on PAD and pooled findings for this outcome represent the combined results of the 4 other trials. After pooling the relative risks across all 5 trials, we observed a 20% reduced risk of non-fatal MI associated with intensive glucose control in the UKPDS trials, 15% in ACCORD, ADVANCE and VADT, and 16% overall. We observed absolute risk reductions of 9 events per 1,000 patients over 5 years of treatment in the overall and subgroup analyses. In contrast, we observed no associations between intensive glucose control and fatal MI, non-fatal stroke, fatal stroke, or PAD in subgroup or overall analyses.

Figure 3.

Pooled relative risk and risk difference (per 1,000 patients over 5 years treatment) estimates of non-fatal myocardial infarction, fatal myocardial infarction, non-fatal stroke, fatal stroke, and peripheral arterial disease by early and more recent trial subgroups and overall.

Figure 4 shows the occurrence of severe hypoglycemia. Intensive glucose control was associated with a 2-fold increase, or an absolute increase of 39 events per 1,000 patients over 5 years, in severe hypoglycemia in the overall analysis, with no association in the early UKPDS studies and a 2.5-fold increase, or an absolute increase of 54 events per 1,000 patients over 5 years, in the more recent trials.

Figure 4.

Pooled relative risk and risk difference (per 1,000 patients over 5 years treatment) estimates of severe hypoglycemia by trial, early and more recent trial subgroups, and overall.

We conducted a sensitivity analysis to determine whether the two studies excluded due to small sample size would have changed the results of the current analysis (9, 17). Inclusion of these studies did not alter any of the main findings, with nearly identical relative risks of 0.91 (0.82, 1.00) for CVD, 0.89 (0.82, 0.96) for CHD, 0.98 (0.85, 1.13) for stroke, 1.01 (0.88, 1.16) for CHF, 0.96 (0.76, 1.21) for cardiovascular deaths, and 0.98 (0.85, 1.14) for total deaths.

Discussion

The current study, combining data from nearly 28,000 participants of 5 large, randomized controlled trials, documented a 10% reduction in the risk of CVD and 11% reduction in the risk of CHD associated with intensive glucose control, with corresponding absolute risk reductions of 15 and 11 events per 1,000 patients over 5 years of treatment. Subgroup analyses of the early UKPDS trials and the more recent ACCORD, ADVANCE, and VADT had similar findings. In addition, intensive glucose control decreased the risk of non-fatal MI by 16%, or an absolute reduction of 9 events per 1,000 patients over 5 years of treatment. This association persisted in subgroup analyses, with risk reductions of 20% (absolute reduction of 9 events per 1,000 patients over 5 years of treatment) in the UKPDS trials and 15% (absolute reduction of 9 events per 1,000 patients over 5 years of treatment) in ACCORD, ADVANCE and VADT. The protective effect of intensive glucose control on non-fatal MI is likely the driving force behind the observed decreases in overall CVD and CHD risk. We observed no overall effect of intensive glucose control on cardiovascular or all-cause mortality. However, the early UKPDS trials suggested that intensive glucose control might reduce mortality from CVD and all-causes. In contrast, the more recent ACCORD, ADVANCE, and VADT suggested that more stringent glucose control might increase mortality from CVD and all-causes. In addition, we observed a 2-fold increased risk of severe hypoglycemia (39 excess events per 1,000 patients over 5 years of treatment) associated with intensive glucose control. Our study does not support associations between intensive glucose control and reduced risks of CHF, fatal MI, fatal and non-fatal stroke, and PAD.

Important differences in therapeutic regimens and achieved HbA1c levels existed among the 5 trials included in our meta-analysis. Each trial used different combinations of diet, sulfonylureas, thiazolidinediones, metformin and/or insulin therapies to achieve target levels of glucose control. The UKPDS 33 and 34 limited participant recruitment to newly diagnosed diabetes patients and used diet as their primary method of treatment in the conventional glucose control group. In contrast, the more recent ACCORD, ADVANCE, and VADT studies, which recruited participants with much longer diabetes duration, relied primarily on pharmacological therapy for treatment in the conventional control group. In addition, differences in achieved HbA1c levels between the studies were substantial. We observed smaller differences in median HbA1c levels between the intensive and conventional glucose control groups in the UKPDS 33 and UKPDS 34 compared to the more recent trials. Furthermore, the UKPDS 33 and UKPDS 34 trials attained post-intervention median HbA1c levels in the intensive treatment group that were either similar to or higher than those achieved in the conventional treatment groups of ACCORD, ADVANCE, and VADT. By today's standards, the UKPDS 33 and 34 examined the benefits of conventional pharmacological treatment, initially and predominantly as monotherapy, while the later three trials investigated what is generally accepted as intensive glucose control. Due to these substantial differences, we examined the UKPDS trials separately from the ACCORD, ADVANCE, and VADT in subgroup analyses. Importantly, we consider these results in the interpretation of the data.

We observed protective effects of intensive glucose control on the risk of CVD, CHD, and non-fatal MI in the overall analysis, with similar trends supported in our subgroup examinations. Similar to our findings, a 2006 meta-analysis of randomized controlled trials by Stettler and colleagues identified an association between intensive glucose control and both cardiac events and any macrovascular event among patients with type-1 or type-2 diabetes (18). While we did not identify effects of intensive glucose control on other CVD endpoints, Stettler and colleagues found associations between intensive glucose control and PAD and cerebrovascular disease (18). Several differences between the two meta-analyses could explain the conflicting findings. The 2006 meta-analysis was conducted before the release of the ACCORD, ADVANCE and VADT studies and represent results from the UKPDS studies as well as the VA feasibility and Kumamoto trials, which were not powered to examine CVD endpoints (9, 17, 18). Inclusion of these two trials in a sensitivity analysis did not change our results. Moreover, methodological weaknesses, including the use of fixed-effects models to pool potentially heterogeneous studies, were evident. Our findings also contrast with observational studies, which have identified consistent, positive associations between HbA1c and PAD, CHF, fatal CHD, and stroke among patients with type-2 diabetes (19-21). Several explanations for these discrepancies exist. Importantly, results from observational studies are subject to confounding effects of unknown or poorly measured risk factors. It is possible that the observational designs did not adequately control for variables such as healthy lifestyle and access to healthcare, which are associated with glucose control. Furthermore, clinical trials typically have shorter duration than prospective observational studies which could contribute to discrepancies in their results.

The premature termination of ACCORD due to excess mortality in the trial's intensive treatment arm alarmed both clinicians and investigators alike (12, 22). Although summary findings of the current meta-analysis do not support these results, subgroup analyses of the more recent trials suggested that intensive glucose control might increase risks of cardiovascular and all-cause mortality, which is in part due to the contribution of findings from the ACCORD trial. In the ACCORD trial, much of the excess mortality in the intensive glucose control arm was due to cardiovascular causes, particularly fatal MI, CHF and ‘unexpected or presumed CVD’. The use of the thiazolidinedione rosiglitazone has been linked to an increased risk of MI and is known to precipitate CHF in susceptible patients (23, 24). This antihyperglycemic agent was more commonly used in the intensive compared to conventional treatment group (91.2% versus 57.5%) of the ACCORD trial and may explain some of the observed increases in MI and CHF deaths (12). In contrast, thiazolidinediones were not used in the UKPDS trials, and were used similarly in the intensive and conventional arms of ADVANCE and VADT (although higher maximal doses were used in the intensive treatment arm of VADT). In addition, it has been suggested that excess mortality in the ACCORD study resulted from deaths due to severe hypoglycemia (22). It may be important to explore whether deaths from severe hypoglycemia could have been incorrectly ascertained in this trial as ‘unexpected or presumed CVD’ deaths.

We identified severe hypoglycemia as an adverse effect strongly associated with intensive glucose control in the present study. Subgroup results from ACCORD, ADVANCE, and VADT found a particularly pronounced treatment effect, with a 2.5-fold increased risk of hypoglycemia, or an absolute increase of 54 events per 1,000 patients over 5 years treatment, associated with intensive glucose control. ACCORD showed the largest relative risk of hypoglycemia, followed closely by VADT. Similar to ACCORD, VADT had an increased number of sudden deaths in the intensive compared to conventional glucose control groups, again calling attention to the possibility of incorrect ascertainment of hypoglycemia-related deaths. Secondary analyses examining the impact of lower HBA1c thresholds on mortality could provide important information on this topic.

With over 27,000 participants among the 5 trials, we had excellent power to detect small but clinically important effects of intensive glucose control on major cardiovascular endpoints and all-cause mortality. In contrast, the power of subgroup analyses to detect small effects of intensive glucose control was limited. A further limitation of the current study includes the use of summary data rather than individual patient data from the 5 included trials. In addition, the recent clinical trials of intensive therapy were of relatively shorter duration than the UKPDS and raise the issue of inadequate time for demonstration of some cardiovascular and total mortality benefits. The ACCORD study stopped intensive treatment after 3.5 years rather than the planned 5 years, and it may be unrealistic to expect a significant reduction in events over this relatively short time frame. This issue is relevant in light of the finding that MI and mortality were reduced on long term follow up of the UKPDS intensive therapy cohort (10, 11).

The results of this meta-analysis provide some evidence for a beneficial effect of intensive glucose control on CVD, particularly on non-fatal MI, but not on cardiovascular deaths and all-cause mortality in patients with type-2 diabetes. Similar to the current study, a recent meta-analysis by Ray and colleagues identified a protective effect of intensive glucose control on CHD and non-fatal MI, with no overall effect of intensive glucose control on stroke or all-cause mortality (25). Moreover, they identified important trial heterogeneity in all-cause mortality findings. We explored this inconsistency with subgroup analyses and add findings that suggest decreased risks of both cardiovascular and all-cause mortality in early trials compared to increased risks in the more recent trials with more stringent intensive glucose control. Furthermore, our results emphasize severe hypoglycemia as an important adverse effect of intensive glucose control. In light of these findings, it is important to consider how best to approach the prevention of CVD and death in this high risk population. Randomized trials have consistently shown that lipid-lowering and blood pressure reduction interventions are extremely effective in decreasing CVD and all-cause mortality among type-2 diabetes patients (26-29). Multifactorial interventions combining glucose regulation, blood pressure control, aspirin use, and lipid-lowering agents have been shown to decrease cardiovascular events by 59%, cardiovascular deaths by 57% and total deaths by 46% in a type-2 diabetes population (30, 31). Nevertheless, there remains a residual excess risk among diabetes patients after controlling blood pressure and lipids (6, 32, 33). Additional approaches are needed to reduce this risk, ones that do not increase risks of severe hypoglycemia and weight gain, as observed in some of the trials examined here. The recent Bypass Angioplasty Revascularization Investigation 2 Diabetes (BARI 2D) trial conducted among patients with type 2 diabetes and CHD indicated that treatment with insulin sensitization compared to insulin provision drugs resulted in fewer severe hypoglycemic episodes, less weight gain, higher HDL levels, and better glucose control among these patients (34). Since BARI 2D was not designed to distinguish between the effects of insulin sensitization agents, like thiazolidinediones and metformin, more research in this area will be needed. Until then, healthcare providers should focus their efforts on combining elements of lifestyle modification, glucose control that minimizes hypoglcyemia, blood pressure reduction, and lipid lowering to optimally curtail the risk of CVD in patients with type-2 diabetes.