Effect of Tight Blood Glucose Control Versus Conventional Control in Patients with Type 2 Diabetes Mellitus: A Systematic Review with Meta-Analysis of Randomized Controlled Trials

Authors


Anna Maria Buehler, R. Abílio Soares, 250 12° andar, Paraíso, São Paulo, CEP: 04005-000, Brazil. 
Tel.: 55-11-3053-66-11 Ext 8209; 
Fax: 55-11-3886-96-45; 
E-mail: abuehler@hcor.com.br

Summary 

Tight control of blood glucose reduces cardiovascular events and total mortality is conflicting. To summarize clinical effects of tight versus conventional glucose control in patients with type 2 diabetes. We systematically searched MEDLINE, EMBASE, Cochrane Library, and ISI Web of Knowledge with no limits of language and time. Further trials were searched from the reference lists of identified studies. We included randomized controlled comparing different levels of blood glucose control intensity in type 2 diabetic patients. Two independent reviewers extracted data of eligible studies using standard case report forms. We investigated total mortality, cardiovascular and microvascular events, and hypoglycemia in patients with type 2 diabetes. We used random-effects models to obtain relative risks (RR) with 95% confidence intervals (CI). We included 6 trials involving 27,654 patients. There was no significant effect of tight blood glucose control on all-cause mortality (RR 1.03; 95% CI 0.90–1.17) or cardiovascular mortality (RR 1.04; 95% CI 0.83–1.29). Tight glucose control reduced the risk for nonfatal MI (RR 0.85; 95% CI 0.76–0.95), although had no effect on the incidence of nonfatal stroke (RR 1.02; 95% CI 0.88–1.17). For microvascular events, tight glucose control reduced the risk progression of retinopathy (RR 0.80; 95% CI 0.71–0.91), incidence of peripheral neuropathy (RR 0.94; 95% CI 0.89–0.99), and progression of nephropathy (RR 0.55; 95% CI 0.37–0.80), but had not significant effect on the incidence of nephropathy (RR 0.69; 95% CI 0.42–1.14). The risk of severe hypoglycemia increased with tight glucose control (RR 2.39; 95% CI 1.79–3.18). Tight blood glucose control reduces the risk for some macrovascular and microvascular events, without effect on all-cause mortality and cardiovascular mortality. Tight glucose control increases the risk of severe hypoglycemia.

Background

Type 2 diabetes mellitus is a metabolic disorder characterized by high blood glucose levels as a consequence of insulin resistance and its relative deficiency. It is well known that this metabolic dysfunction contributes to atherosclerosis process and endothelium-dependent vasodilation impairment [1,2].

Both macro- and microvascular complications are important causes of morbidity and mortality worldwide [3–7]. Tight blood glucose control has been recommended as a therapeutic strategy to decrease the incidence of vascular complications [8]. Evidence from randomized controlled trials (RCTs) supports the adoption of tight glucose control for reducing microvascular complications [5,9]. However, evidences regarding cardiovascular events are conflicting [10,11]. Thus, a systematic review is required to summarize and critically appraise the available evidence.

This systematic review with meta-analysis of RCTs aimed to examine the effects of tight versus conventional glucose control in total mortality, cardiovascular and microvascular events, and hypoglycemia in patients with type 2 diabetes.

Methods

Criteria for Considering Studies for this Review

Types of Studies

This meta-analysis included all RCTs that studied the effects of tight versus conventional glucose control in total and cardiovascular mortality, macro- and microvascular events, and hypoglycemia in subjects with type 2 diabetes mellitus. Studies were considered for inclusion regardless of their publication status, language, or size.

Types of Participants

Trials enrolling patients ≥18 years old with type 2 diabetes were included in this meta-analysis.

To be eligible, the trial should have randomized patients to, at least, two arms with different blood glucoses targets (tight vs. conventional control). In addition, a minimal follow-up of 12 months was required.

The studies which defined a tight group as multiinterventional strategies for blood glucose, blood pressure, and blood cholesterol levels were excluded.

Types of Outcome Measures

The main outcomes of interest were: total mortality; macrovascular events (cardiovascular mortality, nonfatal myocardial infarction, nonfatal stroke, and limb amputation); microvascular events (retinopathies: incidence, progression, maculopathies, proliferative retinopathies, and retinal photocoagulation; vitreous hemorrhage; cataract extraction; visual deterioration; peripheral and autonomic neuropathies; and nephropathies: incidence, progression, microalbuminuria, and need for dialysis); and episodes of severe hypoglycemia.

Search Methods for Identification of Studies

Electronic Searches

We searched for references on MEDLINE, Cochrane library, EMBASE, and ISI Web of Knowledge from January 1966 to May 2011 without language restrictions. A standard protocol for this search was developed and whenever possible, controlled vocabulary (Mesh term for MEDLINE and Cochrane and EMTREE for EMBASE) were used. Keywords and their synonymous were used to sensitize the search (See Appendix A for search terms used).

For the identification of RCTs in PUBMED/MEDLINE, the optimally sensitive strategy developed for the Cochrane Collaboration was used [12]. To identify the RCTs in EMBASE, a search strategy using similar terms was adopted.

In ISI Web of knowledge database, a different search strategy was adopted. A key study which, probably, all similar studies would cite as a reference was chosen. We chose UKPDS 33 in this meta-analysis [9]. In the next step, all the references that cited UKPDS 33 were selected.

Other Information Sources

All eligible articles for this meta-analysis had their references lists analyzed to detect other potentially eligible studies. For ongoing studies or when the confirmation of any data or additional information was needed, the authors were contacted by e-mail.

Data Collection and Analysis

Assessment of Study Eligibility

The previously described search strategy was used to obtain titles and abstracts of studies that might be relevant for this review. Each abstract identified in the research was independently evaluated by two authors. If at least one of the authors considered one reference eligible, the full text was obtained for complete assessment.

In a similar fashion, two authors independently evaluated full text articles for eligibility and filled inclusion and exclusion criteria in a standard form. A standardized data extraction form was used to inclusion and exclusion criteria. In case of any disagreement, the authors discussed the reasons for their decisions and a final decision was made by consensus. Studies reported in non-English language were translated before assessment.

Data Extraction

Two authors independently extracted data from the published reports using standard data extraction forms. Disagreements were resolved by one of the authors (A.B.C. or A.M.B.). Duplicate publications or substudies of included studies were listed under the primary reference as they may have provided information on relevant outcomes not available in the original publication. Any further information required from the original author was requested by e-mail.

Risk of Bias of Included Studies

The risk of bias of included studies was assessed independently by two authors. The following criteria were assessed: allocation concealment, blinding (participants, investigators, and outcome assessors) intention-to-treat analysis, and completeness of follow-up (see Appendix B for definitions).

Quality of Meta-Analysis Evidence

The quality of evidence generated by this meta-analysis was classified using the GRADE System as high, moderate, low, or very low [13]. The level of evidence from randomized trials were initially set as high and were downgraded when methodological limitations, inconsistencies, indirectness, imprecision, or high risk of reporting bias was present.

Statistical Assessment

In each trial, the relation between tight and conventional glucose control and risk for all investigated outcomes was presented as relative risk (RR) and their respective 95% confidence interval (CI). Then, these estimates were pooled using Mantel–Haenszel random effects analysis. Statistical heterogeneity across trials was evaluated with χ2 and I2 statistics. The probability of publication bias, analyzed by funnel plot, was not performed because of few studies included. Analyses were performed with Review Manager Version 5.0 (Cochrane IMS, Oxford, UK).

Five subgroup analyses were performed, respectively:

  • 1Studies that included a small proportion (<20%) of patients with cardiovascular disease (UKPDS and Kumamoto) versus studies that included a greater proportion (>20%) of patients with cardiovascular disease (ACCORD, ADVANCE, VADT, and VACSDM).
  • 2Studies which included patients with newly diagnosed type 2 diabetes (UKPDS) versus studies which included patients with long-lasting type 2 diabetes (ACCORD, ADVANCE, VADT, VACSDM, and Kumamoto).
  • 3Studies those achieved a HbA1c < 7.0% in the tight glucose control group (ACCORD and ADVANCE) versus those studies which achieved a mean HbA1c ≥ 7.0% in the tight glucose control group (VADT, VACSDM, UKPDS, and Kumamoto).
  • 4Studies which included patients <60 years (UKPDS and Kumamoto) versus studies that included patients who were ≥60 years (ACCORD, ADVANCE, VADT, and VACSDM).
  • 5Studies that used rosiglitazone for most patients in the tight glucose control group (ACCORD and VADT) versus studies that did not use rosiglitazone for most patients in the tight glucose control group (ADVANCE, VACSDM, Kumamoto, and UKPDS).

Results

Description of Studies

We initially identified 9 221 citations, from which 7 potentially eligible articles were found by manual search (Figure 1). Of these, 2 659 were duplicates. Thus, the titles and abstracts of 6 562 citations were assessed. Full texts of 165 citations were selected for a thorough analysis. Two ongoing studies were also identified [14,15]. A study in Russian language was initially judged to be eligible, but was later excluded because the trial failed to achieve different levels of HbA1c [16]. Finally, 6 trials which included 27,654 participants were eligible for this review [9–11,17–20]. Results of these 6 trials are reported in 61 articles.

Figure 1.

Study selection process. RCT: indicates randomized controlled trials; DM: diabetes mellitus; FU: follow-up.

Baseline characteristics of the trials are listed in Table 1. There was substantial variation in characteristics across the studies. The follow-up periods ranged from 2.25 to 10 years. The UKPDS study [9,17] included patients with newly diagnosed diabetes mellitus, although in the other trials the average duration of diabetes ranged from 7.7 to 11.5 years. Most of the studies included overweight or obese patients, except the Kumamoto study [20] that included patient with a normal body mass index. For most of the studies, there was a significant proportion of the patients with previous cardiovascular diseases, except for the UKPDS [9,17] and Kumamoto [20] studies. Although all six trials aimed at similar levels of HbA1c, the median HbA1c achieved in the intervention group at the end of follow-up period was 6.4% in the ACCORD [10] and ADVANCE [11] studies, but closer to 7.0% in the other trials (Table 2).

Table 1. Baseline characteristics of patients in studies
 ACCORDKumamotoADVANCE
TightConventionalTightConventionalTightConventional
  1. aData presented in median. To change LDL colesterol from mg/dL to mmol/L, multiply by 0.02586. NR: not reported.

Randomized patients5 1285 12355555 5715 569
Age62.262.248516666
Men3 143 (61%)3 156 (62%)28 (51%)26 (47%)3 198 (57%)3 213 (58%)
Time of disease DMII (years)10a10a8.68.57.98.0
Previous cardiovascular disease1 826 (36%)1 783 (35%)001 794 (32%)1 796 (32%)
Smokers733 (14%)702 (14%)NRNR793 (14%)757 (123%)
Mean systolic blood pressure (mmHg)136.2136.5120.5122145145
LDL cholesterol (mg/dL)104.9104.9NRNR120.6120.3
IMC (Kg/m2)32.232.220.620.428.028.0
HbA1c8.38.39.48.97.517.52
Fasting glucose174.9 (56%)175.7 (56.5%)Not ReportedNR153.3 (50.1%)152.8 (49.7%)
 VADTVACSDMUKPDS
TightConventionalTightConventionalTightConventional
Randomized patients89289975783 0711 138
Age60.560.360.459.953.253.4
Men866 (97%)873 (97%)75 (100%)78 (100%)806 (26%)705 (62%)
Time of disease DMII (years)11.511.587.7NRNR
Previous cardiovascular disease355 (40%)368 (41%)31 (41%)27 (35%)00
Smokers154 (17%)145 (16%)17 (23%)10 (13%)844 (27%)353 (31%)
Mean systolic blood pressure (mmHg)131132136.1134.5135.6135.6
LDL cholesterol (mg/dL)107108136141136.1136.1
IMC (Kg/m2)31.331.230.731.32828.2
HbA1c9.49.49.39.57.097.05
Fasting glucoseNRNR206227145.9144.12
Table 2. HbA1c levels of glucose control after follow-up
StudiesFollow-up (months)HbA1c (%) of follow-upFasting glucose (mg/dL) of follow-up
TightConventionalTightConventional
  1. aMedian.

ACCORD426.4a7.5aNRNR
ADVANCE606.4a7.0a111.7a131.5a
VADT727.0a8.7aNRNR
VACSDM247.19.1125210
Kumamoto967.29.4122162
UKPDS1207.0a7.9aNRNR

Risk of Bias of Included Studies

Table 3 shows the risk of bias in individual studies. In general, there were not important limitations in methodological characteristics of the included RCTs, although information regarding randomization and allocation concealment was unclear in Kumamoto study [20]. The loss of follow-up was not considered as a significant variable in this meta-analysis (for reasons of loss of follow-up see Appendix C)

Table 3. Methodological characteristics of included studies
 ACCORDADVANCEVADTVACSDMUKPDSKumamoto
  1. aTight/conventional.

RandomizationAppropriateAppropriateAppropriateNot clearAppropriateNot clear
Concealment allocationAppropriateAppropriateAppropriateAppropriateAppropriateNot clear
Blinding of investigatorsNoNoNoNoNoNo
Blinding of participantsNoNoNoNoNoNo
Outcome assessesYesYesYesYesYesYes
Intention to treatYesYesYesYesYesYes
Loss of follow-upa110 (2%)/102 (2%)245 (4%)/295 (5%)120 (13%)/139 (15%)2 (2%)/0 (0%)122 (4%)/45 (4%)2 (4%)/2 (4%)

Effects of Interventions

Figure 2 presents the individual and pooled relative risks of main outcomes. This meta-analysis demonstrates no effect of tight glucose control on all-cause mortality (RR 1.03; 95% CI [0.90–1.17]; I2= 50%) and cardiovascular mortality (1.04; 95% CI [0.83–1.29]; I2= 60%). Tight glucose control was able to decrease the relative risk of nonfatal myocardial infarction by 15% (RR 0.85; 95% CI [0.76–0.95]; I2= 0%) and had no effect on nonfatal stroke (RR 1.02; 95% CI [0.88–1.17]; I2= 0%) and limb amputation (RR 0.69; 95% CI [0.44–1.08]; I2= 0%). Considering microvascular events, tight glucose control had no effect on incidence of retinopathy (RR 0.75; 95% CI [0.37–1.55]; I2= 65%) but decreased the relative risk of progression of retinopathy (RR 0.80; 95% CI [0.71–0.91]; I2= 0%). Regarding neuropathies, the tight glucose control had no effect on autonomic neuropathy (RR 0.55; 95% CI [0.37–0.80]; I2= 0%) but reduced relative risk of peripheral neuropathy by 6% (RR 0.94; 95% CI [0.89–0.99]; I2= 2%). Moreover, tight glucose control did not reduce incidence of nephropathy (RR 0.69; 95% CI [0.42–1.14]; I2= 73%) but reduced relative risk of progression of nephropathy by 45% (RR 0.55; 95% CI [0.37–0.80]; I2= 0%) and microalbuminuria by 22% (RR 0.78; 95% CI [0.65–0.92]; I2= 79%). Tight glucose control had no effect on other described outcomes (See Appendix D for results). Finally, tight glucose control increases the relative risk of severe episode of hypoglycemia by 2.4 times compared to conventional glucose control (RR 2.39; 95% CI [1.79–3.18]; I2= 62%).

Figure 2.

Meta-analysis of main outcomes.

Sensitivity Analysis

There was high heterogeneity between estimates of effect from the included studies for some outcomes. We performed some subgroup analyses aiming to explore sources of heterogeneity of effect across studies on cardiovascular mortality. For cardiovascular mortality, none of the subgroups analyses were able to explain heterogeneity. Taking into account studies whose patients showed previous cardiovascular diseases (more than 20%), the subgroup analysis found a high heterogeneity (I2= 72%, P= 0.01). The heterogeneity was also high for studies whose patients achieved HbA1c less than 7.0% (I2= 90%, P= 0.002) and for studies whose patients’ mean age was more or equal to 60 years (I2= 72%, P= 0.01). The exception was the subgroup analysis of studies that mainly used rosiglitazone in the tight glucose control group versus studies that did not. In those studies using rosiglitazone in the tight control group there was an increase in mortality (RR 1.38; 95% CI [1.10–1.73]; I2= 0%, P= 0.55), whereas in studies not using this drug there was no effect on this outcome (RR 0.88; 95% CI [0.78–1.00]; I2= 0%, P= 0.84; Figure 3).

Figure 3.

Sub-group analysis for cardiovascular mortality.

Discussion

Tight glucose control did not change the risk of all-cause or cardiovascular death in type 2 diabetes patients. On the other hand, it reduced the risk of nonfatal myocardial infarction by 15%, but had no statistically significant effect on the risk of nonfatal stroke and limb amputation. Effects of tight glucose control on microvascular events were also mixed: there was reduction in the risk of retinopathy progression, incidence of peripheral neuropathy, and progression of nephropathy, without a significant effect on the incidence of nephropathy and other microvascular outcomes. The risk of hypoglycemia increased 2.4 times compared to conventional glucose control.

Taking into account the quality of evidence using GRADE system, we considered the evidence from all-cause mortality as “moderate” because of inconsistency across the studies and the evidence from cardiovascular mortality as “low” because of inconsistency and imprecision of results across the studies. For nonfatal myocardial infarction and nonfatal stroke the quality of evidence was classified as “high.” For limb amputation, we classified the quality of evidence as “moderate” because of imprecision of result, because there were few events and confidence interval was wide. For microvascular events, the quality of evidence was considered as “low” for incidence of retinopathy because of inconsistency and imprecision, and as “high” for progression of retinopathy. We considered the quality of evidence as “low” for autonomic neuropathy because only two studies reported data with a serious imprecision and inconsistency results across the studies. However, for peripheral neuropathy, we classified the quality of evidence as “high,” despite a modest relative risk reduction effect. For incidence of nephropathy, we consider the quality of evidence as “low” because of imprecision and inconsistency across the studies and as “moderate” for progression of nephropathy because of imprecision results across the studies (See Appendix D for the quality of evidence for the other outcomes).

Our results are consistent with previous meta-analyses [21–27], despite some differences have to be highlighted. First, most of them [23,24,26,27] included the PROactive study [28] (pioglitazone with placebo). The PROactive study was excluded from this meta-analysis because it did not evaluate the effects of two blood glucose control intensity levels. Also, one previous meta-analysis [23] included the Steno-2 study [29] that aimed to compare a strategy of more intensive versus less intensive control of several targets (cholesterol, blood pressure, and blood glucose).

Other additional point has to be clarified. The control group of the UKPDS 34 study is, in fact, part of the control group of the previously reported UKPDS 33. This information was not accounted for in some previous reviews [21,22,26], which counted patients twice in their analyses.

This systematic review has some other highlights. We followed the mainly methodological characteristics of systematic reviews and meta-analyses. In this regard, both the eligibility and data extraction were assessed with standard forms and conducted independently. We also assessed risk of bias in included studies and graduated the quality of evidence from meta-analyses using the GRADE system [13].

We did not find important methodological limitations in individual studies, but they were very heterogeneous considering baseline characteristics. These different patients’ profiles might have contributed to moderate-to-high inconsistencies of some results across the studies. The included studies also differ in drugs and strategies used to achieve target blood glucose levels. This meta-analysis did not examine differences in glucose-lowering drug, class, or combination. We performed some subgroup analyses aiming to explore sources of heterogeneity of effect across studies on cardiovascular mortality. For cardiovascular mortality, none of the subgroups analyses were able to explain heterogeneity, except for the subgroup analysis of studies that used rosiglitazone for most patients in the tight glucose control group versus studies that did not. In those studies using rosiglitazone in the tight control group there was an increase in mortality, whereas in studies not using this drug there was no effect on this outcome. This finding has to be interpreted carefully, because it is a subgroup analysis, with no power and number of events enough to address this question. In this regarding, we cannot exclude a potential favorable effect of tight glucose control in patients whose principal treatment scheme does not use rosiglitazone.

There are some evidences pointing for a specific drug effect of thiazolidinediones (TZD) instead of rosiglitazone. From a physiological point of view, both TZDs pioglitazone and rosiglitazone seem to be associated with oedema and an increase in signs of heart failure [30]. However, rosiglitazone increases LDL-C concentration and the number of atherogenic apo B100-containing particles, probably raising triglycerides [31,32]. On the other hand, pioglitazone, has little effect on LDL-C, tends to lower apo B100, decreases plasma triglyceride levels [31,32], and raises HDL-C significantly [33]. From trials evidences, the cardiovascular safety of pioglitazone has been established by the PROactive study, a double-blind, placebo-controlled, 3-year study designed to investigate the effect of pioglitazone on macrovascular events in 5 238 patients with type 2 diabetes and established cardiovascular disease [28]. There were no significant differences on primary composite endpoints (all-cause mortality, nonfatal myocardial infarction, stroke, acute coronary syndrome, intervention in the coronary or leg arteries, and amputation above the ankle), but there were significant reductions in the secondary composite endpoints of all-cause mortality, myocardium infarction, and stroke. Indeed, five meta-analyses [34–38] analyzed the effects of pioglitazone on cardiovascular events and all of them also performed subanalyses without PROactive data with results consistent with the overall analysis and with the results of PROactive itself. Regarding rosiglitazone, the RECORD study [39] assessed the noninferiority of rosiglitazone (plus metformin or sulfonylurea) compared with metformin plus sulfonylurea alone in terms of overall cardiovascular risk for a follow-up of 5.5 years. This study did not show differences between the treatments, despite more patients in rosiglitazone group experienced myocardium infarction. However, these results have to be interpreted with caution because there were significant lost of follow-up, the study did not reach statistical power to address cardiovascular event rate and the greater use of statin/diuretics in the rosiglitazone group [39]. From observational studies, an US cohort study with elderly patients treated with TZD, revealed a 15% greater mortality and 13% greater risk of heart failure for rosiglitazone compared to pioglitazone [40]. Meta-analytical evidences [37,41–44] for rosiglitazone suggest increased myocardial ischemia risk but the results are not conclusive because there was an important clinical, methodological, and statistical heterogeneity between included studies in all these meta-analyses. Ongoing studies like the Thiazolidinedione Intervention with vitamin D Evaluation (TIDE) study [45] that directly compares the effects of rosiglitazone and pioglitazone on clinical cardiovascular outcomes can contribute to define the safety profile of these drugs.

Other possible explanation for the increased mortality in ACCORD and VADT studies exists. In the ACCORD trial, patients gained more weight in tight group and decreased HbA1c to 6.4% within 12 months, although the ADVANCE trial took several years to decrease HbA1c to 6.4%. Also, the HbA1c target in the ACCORD trial was very low (<6.0%) and the treatment was really aggressive, with most of patients using 4–5 glucose-lowering drugs and a high percentage of patients using insulin, sulfonylurea, rosiglitazone, or even the combination of these drugs. The potential for the antidiabetic drugs to cause harm or benefit through individual pharmacologic effects, independent of their effects on glycemia, must be considered in this context. Considering VADT trial, although the differences between the treatments were not statistically significant, there were more cardiovascular deaths in the tight control arm. The included patients in VADT trial showed similar baseline characteristics and followed the same outcome trends of the ACCORD study.

Current guidelines for management of type 2 diabetes emphasize a rapid and aggressive goal of achieving an HbA1c level of less than 7.0%. However, taking into account this meta-analysis, tight glucose control (HbA1c < 6.5%) may not be appropriated for the subset of diabetes patients with a limited life expectancy or with severe comorbidities. The risks of tight therapy may outweigh any benefits on long-term diabetic complications, especially considering the way of achieving the target of HbA1c, weight gain, and severe hypoglycemia events in patients at high risk of cardiovascular disease.

It is important to note that hypoglycemia is a common adverse drug-related event that results in poor diabetes outcomes; reduced medication adherence for fear of hypoglycemia; increased risk for dementia, hospitalizations, and emergency department visits; reduced quality-of-life; and increased likelihood of death [10]. Also, Zoungas et al. [46] showed that in the population of the ADVANCE study, severe hypoglycemia was clearly associated with increased risks of macrovascular events, microvascular events, and death from both cardiovascular and noncardiovascular causes. These findings were supported by other studies, including the post hoc analyses of ACCORD study [47]. Then, the authors suggest that hypoglycemia is as likely to be a marker of vulnerability to a wide range of adverse clinical outcomes and prompt action to address this possibility.

On the other hand, it is not possible to exclude some addition benefits in more strict glucose control targeting HbA1c < 6.5% in patient who have the profile of the UKPDS study, especially regarding duration of disease. Moreover, the most significant benefits in microvascular events were found in the UKPDS study.

This way, the ORIGIN [14] ongoing study will be crucial to a deep understanding of the most tight glucose targets in newly diagnosed diabetes type 2 population.

Implications and Conclusion

Tight glucose control did not reduce all-cause and cardiovascular mortality but reduces relative risk of nonfatal myocardial infarction and some microvascular events (progression of retinopathy, peripheral neuropathy, nephropathy progression, and microalbuminuria). Also, tight glucose control increases relative risk of severe hypoglycemia.

Our findings suggest that outlining the strategy to control glycemia have to consider the way of lowering HbA1c levels according to different diabetes mellitus population, rather than the intensity of the HbA1c target itself.

Also, it is important to consider that patient with shorter duration of diabetes mellitus may be more likely to benefit from this strategy for both cardiovascular and microvascular events.

Acknowledgment

This report was supported by the Brazilian Ministry of Health (Ministério da Saúde do Brasil).

Conflict of Interest

The authors declare no conflict of interest.

Appendices


Appendix A: Search Strategies

Database Block description Search strategy Number of paper retrieved
PUBMED/MEDLINE Type of patients #1 “Diabetes Mellitus”[Mesh]235,188
  #2 “Diabetes Mellitus, Type 2”[Mesh] or (Diabetes Mellitus, Ketosis-Resistant) or (Diabetes Mellitus, Ketosis Resistant) or (Ketosis-Resistant Diabetes Mellitus) or (Diabetes Mellitus, Maturity-Onset) or (Diabetes Mellitus, Maturity Onset) or (Diabetes Mellitus, Non Insulin Dependent) or (Diabetes Mellitus, Non-Insulin-Dependent) or (Non-Insulin-Dependent Diabetes Mellitus) or (Type 2 Diabetes Mellitus) or (Diabetes Mellitus, Slow-Onset) or (Diabetes Mellitus, Slow Onset) or (Slow-Onset Diabetes Mellitus) or (Diabetes Mellitus, Stable) or (Stable Diabetes Mellitus) or (Diabetes Mellitus, Type II) or (Maturity-Onset Diabetes Mellitus) or (Maturity Onset Diabetes Mellitus) or (MODY) or (NIDDM) or (Diabetes Mellitus, Adult-Onset) or (Adult-Onset Diabetes Mellitus) or (Diabetes Mellitus, Adult Onset) or (Diabetes Mellitus, 
  Noninsulin Dependent)75,669
   #3: #1 or #2 241203
  Intervention #4 “Blood Glucose”[Mesh] or (Blood Sugar) or (Sugar, 
  Blood) or (Glucose, Blood)167,800
  #4 “Hypoglycemic Agents”[Mesh] or (Agents, Hypoglycemic) or (Hypoglycemics) or (Hypoglycemic Drugs) or (Drugs, Hypoglycemic) or (Antidiabetics) or (Antidiabetic Drugs) or 
  (Drugs, Antidiabetic) or (Antidiabetic Agents) or (Agents, 
  Antidiabetic)161,152
   #6: #4 or #5 268,382
  Adjectives to characterize tight glucose control #7 Tight or intensive or Strict196,290
  Randomized Controlled trials #8 (randomized controlled trial [pt] or controlled clinical trial [pt] or randomized controlled trials [mh] or random allocation [mh] or double-blind method [mh] or single-blind method [mh] or clinical trial [pt] or clinical trials [mh] or (“clinical trial”[tw]) or ((singl*[tw] or doubl*[tw] or trebl*[tw] or tripl*[tw]) and (mask*[tw] or blind*[tw])) or (placebos [mh] or placebo*[tw] or random [tw] or research design [mh:noexp] or comparative study [pt] or “Evaluation Studies as Topic”[Mesh] or follow-up studies [mh] or prospective studies [mh] or control*[tw] or prospective*[tw] or 
  volunteer*[tw]) not (animals [mh] not humans [mh])3,919,835
  Final result #3 and #6 and #7 and #8 3,098
COCHRANE LIBRARY (central for randomized controlled trials) Type of patients #1 MeSH descriptor Diabetes Mellitus explode all trees10,762
  #2 MeSH descriptor Diabetes Mellitus, Type 2 explode all 
  trees5,249
   #3: #1 or #2 10,762
  Intervention #4 MeSH descriptor Blood Glucose explode all trees8,031
  #5 MeSH descriptor Hypoglycemic Agents explode all trees8,068
   #6: #4 or #5 11,189
  Adjectives to characterize tight glucose control #7 (intensive) or (strict) or (tight)17,806
  Final result #3 and #6 and #7 423
EMBASE Type of patients #1-‘diabetes mellitus’/exp and 
  [embase]/lim and [1974–2009]/py256,040
   #2-’non insulin dependent diabetes mellitus’/exp and [embase]/lim and [1974–2009]/py58,517
   #3: #1 or #2 256,040
  Intervention #4-’antidiabetic agent’/exp and [embase]/lim and [1974–2009]/py178,380
   #5-’glucose blood level’/exp and [embase]/lim and [1974–2009]/py65,495
   #6: #4 or #5 211,974
  Adjectives to characterize tight glucose control #7-intensive or tight or strict and [embase]/lim and [1974–2009]/py185,413
  Randomized Controlled trials #8 ‘randomized controlled trial’/exp and [embase]/lim and [1974–2009]/py145,127
  #9 ‘controlled study’/exp and [embase]/lim and [1974–2009]/py2,875,706
  #10 ‘randomization’/exp and [embase]/lim and [1974–2009]/py26,716
  #11 ‘double blind procedure’/exp and [embase]/lim and [1974–2009]/py74,372
  #12 single blind procedure’/exp and [embase]/lim and [1974–2009]/py8,108
  #13 ‘clinical trial’/exp and [embase]/lim and [19742009]/py520,287
  #14 ‘crossover procedure’/exp and [embase]/lim and [1974–2009]/py21,207
  #15 ‘evaluation and follow up’/exp and [embase]/lim and [1974–2009]/py535,126
  #16 ‘prospective study’/exp and [embase]/lim and [1974–2009]/py81,429
  #17 ‘placebo’/exp and [embase]/lim and [1974–2009]/py137,411
  #18 ‘comparative study’/exp and [embase]/lim and [1974–2009]/py374,088
  #19: #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 OR #183,698,052
  #20 ‘animal’/exp and [embase]/lim and [1974–2009]/py118,358
  #21 ‘nonhuman’/exp and [embase]/lim and [1974–2009]/py3,209,574
  #22 ‘human’/exp and [embase]/lim and [1974–2009]/py6,610,686
  #23:#20 or #213,229,681
  #24: 23 # NOT #22 
   # 25: #19 NOT #24 2,525,627
  Final result #3 and # 6 and # and #7 and #25 3,412


Appendix B: Quality Assessment

a. Allocation concealment:
( ) Adequate: Randomization method described that would not allow investigator/participant to know or influence intervention group before eligible participant entered in the study. Examples are: allocation involving a central independent unit, on-site locked computer, identically appearing numbered drug bottles or containers prepared by an independent pharmacist or investigator, or sealed envelopes.
( ) Unclear: Randomization stated but no information on method used is available.
( ) Inadequate: Method of randomization used such as alternate medical record numbers or unsealed envelopes; any information in the study that indicated that investigators or participants could influence intervention group.
b. Blinding of investigators:
( ) Yes( ) No( ) Not stated
( ) Unclear or inadequate - if the trial was described as double blind, but the method of blinding was not described or is not compatible with blinding
c. Blinding of participants:
( ) Yes( ) No( ) Not stated
( ) Unclear or inadequate - if the trial was described as double blind, but the method of blinding was not described or is not compatible with blinding
d. Blinding of outcome assessors (adjudicators):
( ) Yes( ) No( ) Not stated
( ) Unclear or inadequate-if the trial was described as double blind, but the method of blinding was not described or is not compatible with blinding
f. Intention-to-treat (ITT) analysis*:
( ) Yes–Confirmation on study assessment that the number of participants randomized and the number analyzed are identical; except for patients lost to follow-up or who withdrew consent for study participation
( ) No–Confirmation on study assessment that patients who were randomized were not included in the analysis, because they did not receive the study intervention, were not included because of protocol violation or any other reason.
( ) Not stated
*Obs.: Our evaluation should be independent of authors’ claim of ITT analysis, i.e., a study may be considered by us as analyzed according to ITT principle even if there is no such statement as long as we confirm that on study assessment. The opposite is also true, a study is reported as being ITT but we may consider it not to be ITT depending on our evaluation.


Appendix C

StudiesReason for loss of follow-up
TightConventional
ACCORD26 were lost to follow-up24 were lost to follow-up
 84 withdrew consent78 withdrew consent
ADVANCE7 had unknown vital status10 had unknown vital status
 238 were not assessed at final visit285 were not assessed at final visit
VADT58 were lost to follow-up57 were lost to follow-up
 12 had other reason12 had other reason
 7 had adverse event3 had adverse event
 43 withdrew consent67 withdrew consent
VACSDMNot reportedNot applicable ( 0%)
Kumamoto2 patient moved to other city2 patient moved to other city
UKPDS57 had unknown vital status19 had unknown vital status
 65 were not assessed at final visit26 were not assessed at final visit

Appendix D

inline image

Ancillary