Identification of the treatment target is the most important prerequisite to effective metabolic management. In the context of blood glucose management, the first evidence for such a target was provided by the UKPDS trial. In the UKPDS trial, a reduction of HbA1c by 0.9% from baseline translated into impressive microvascular benefits [3]. However, the microvascular endpoints which led to such impressive findings were driven by a reduction in rates of cataract extraction and albuminuria. The recent outcome studies, as well as recommendation committees, do not consider cataract extraction and microalbuminuria as standard components of microvascular outcomes [22].
The next spate of trials to define intensive glycaemic control was conducted a decade later. Of these, the results of the STENO-2 and ADVANCE trials attempted lowering the HbA1c levels to < 6.5%, whereas ACCORD and VADT suggested a therapeutic target of < 6.0% [6,7,8,9]. Although, standardised outcome measures were analysed in these trials, there was considerable heterogeneity with respect to patient population, the duration of diabetes as well as the associated co-morbidities. This phenomenon likely explains the encouraging results obtained from the STENO-2 study, disappointing results from the ACCORD trial and neutral results from ADVANCE and VADT.
Despite differing baseline characteristics and inconsistent outcomes, different meta-analyses grouped all these studies together to determine the potential benefits of intensive glycaemic control over less stringent control.
However, the most important objection to grouping all these trials together is the divergent EOS HbA1c achieved in these trials. Most of the meta-analyses adopted the target HbA1c set at trial initiation as the definition of intensive control [23,24,25]. However, the EOS HbA1C values in all these studies were very different from the target HbA1C. For example, STENO-2 achieved a mean EOS HbA1c of 7.7% (target was ≤ 6.5%), and VADT achieved a mean HbA1c of 7.0% at EOS (target was ≤ 6.0%) Intuitively, this analytical strategy would seem inaccurate.
This problem is further compounded by the different HbA1c cut-off values proposed by the different diabetes guidelines. The 2018 ACP diabetes guidelines, which recommended a HbA1c range of 7.0–8.0% for most non-pregnant adults with T2DM, caused huge confusion amongst the physicians [4] because, in contrast, the 2021 ADA guidelines recommend a target HbA1c level of < 7.0%, and the 2020 AACE guideline recommends < 6.5% for similar types of patients with diabetes [26, 27]. The problem seems to stem from the fact that most of these analyses combined the results of only UKPDS, ACCORD, ADVANCE and VADT to arrive at a conclusion. However, there were many other trials during the same period and later which were not included in these analyses.
Identification of the appropriate and accurate HbA1c level, a target which has a positive impact on micro- and macrovascular complications, necessitates assessment of different HbA1c ranges.
We conducted this meta-analysis in an attempt to overcome some of these shortcomings. As an initial strategy, we mimicked the pattern followed by previous meta-analyses and guidelines. The only difference was that we expanded the search to include all prospective trials to date and did not restrict our search up to 2008. Several smaller trials such as Kumamoto, Veterans Affairs and HOME were also included to increase the evidence base and to arrive at a more robust conclusion [12, 14, 17]. Even trials which primarily studied anti-diabetic agents, but divided the two comparative groups into intensive and conventional arms (PROactive and RECORD), were included to improve the yield [18, 19]. Thus, a total of 15 studies were included in the meta-analysis in contrast to the much smaller numbers in previous meta-analyses. The result suggested significant improvement in new-onset or progressive retinopathy, new-onset or persistent macroalbuminuria, ESRD and NFMI. This outcome with intensive glucose-lowering was encountered in a HbA1c target range of < 6.0–7.6%, differing from an EOS HbA1C of 6.3–7.7%. The question then is: how are the glycaemic targets set at < 6.5% or even 7.0% in T2D?
To answer this question, we divided the HbA1c achieved at EOS into different ranges (≤ 6.5%, 6.6–7.0% and 7.1–7.7%). Is the HbA1c < 6.5% justified as suggested by some diabetes guidelines? The only endpoint significantly impacted by targeting a HbA1c < 6.5% was new-onset or persistent macroalbuminuria (24% reduction). However, we need to appreciate the fact that only 2 out of the 15 studies achieved the targeted HbA1c ≤ 6.5% (ADVANCE and ACCORD). Targeting HbA1c between 6.6 and 7.0% resulted in a significant 23% reduction in new-onset or progressive retinopathy. However, the maximum benefit seems to be derived from a target HbA1c range of 7.1–7.7% (46% reduction in new-onset or progressive retinopathy, 52% reduction in new-onset or persistent macroalbuminuria, 36% reduction in NFS and 22% reduction in ACM).
We have also included an EOS HbA1c of 7.1–7.7%. Strange as it may sound, this is not without reason. This is because the HbA1C levels achieved at EOS in all the studies were consistently ≤ 7.7%; hence, this was the basis of the upper limit of this range. A range of 7.1–7.5% and another range of > 7.5% could also have been chosen; however, in the absence of an adequate number of studies in these ranges, we decided to combine all the outcomes ≥ 7.1%.
However, diabetes duration is another key determinant of the metabolic target. Long diabetes duration is arbitrarily considered to be > 10 years owing to the significantly higher prevalence or risk of complications [15]. We included an additional strategy in our meta-analysis, i.e. a combination of HbA1c range and the diabetes duration. Diabetes duration < 10 years and achievement of HbA1c ≤ 7.0% were associated with a 24% lower risk of new-onset or persistent macroalbuminuria. Diabetes duration of < 10 years and HbA1c level in the range 7.1–7.7% were associated with a more impressive 46% reduction in new-onset or progressive retinopathy, 42% reduction in new-onset or persistent macroalbuminuria, 49% reduction in doubling of serum creatinine and 36% reduction in NFS. In contrast, those with aT2DM duration of ≥ 10 years and who had achieved HbA1c ≤ 7.0% represented a mixed bag. There was a significant 24% reduction in new-onset or progressive retinopathy, 30% reduction in new-onset or persistent macroalbuminuria and 17% reduction in NFMI, but at the cost of a significant 40% increase in CV death and a 20% increase in ACM. Since only one study provided data pertaining to diabetes duration ≥ 10 years and achievement of HbA1c ≥ 7.1–7.7%, no definitive conclusions could be drawn in this respect.
An observational study by Currie et al. aiming at treatment intensification with insulin documented a U-shaped curve for mortality [26]. In view of their findings the authors suggested a revision of guidelines to include a minimum HbA1c value. Our result and that of the observational study by Currie et al. seem to support the 2018 AFP guidelines as far as the desired HbA1c range for most non-pregnant adults with type 2 diabetes is concerned, in contrast to the ADA, AACE, IDF or NICE guidelines [4, 27,28,29,30].
Study Limitations
The different HbA1c ranges were not equally represented in the studies included in the analysis. However, this limitation is unavoidable since we had no other option but to work with the available and eligible trials. Hence, any sub-group analysis would be biased toward such an imbalance. Data pertaining to a few endpoints were reported by a single study; hence, it was difficult to arrive at a definitive conclusion. We would like to include those areas as part of our research recommendations. Inclusion of other risk factors apart from target HbA1c and duration of diabetes may have provided different results. This was not done in our meta-analysis. The main roadblock to such an approach was the limited number of prospective studies available for analysis. Any additional subgroups would have resulted in gross under-evaluation of the endpoints. The increase in CV death and ACM in the subgroup represented by DM duration ≥ 10 years and HbA1c ≤ 7.0% could be due to the increased number of hypoglycaemic episodes in the intensive arm documented in the ACCORD trial. This could have skewed the data in the wrong direction. However, since the aim of this meta-analysis was to analyse the impact of an intensive versus convention therapeutic approach on outcomes, the mechanistic basis was not explored. Finally, the method of assessing HbA1c (JDS) was not the same as in the other included citations (NGSP method), which could have confounded some of the outcomes. However, after correcting for the difference as per the standardised equation (NGSP (%) = 1.02 × JDS (%) + 0.250), the KUMOTO study remained in the same target category (HbA1c range of 7.1% to 7.7%). As a result, this did not alter the effect size analysis of the outcomes.
Strengths of the Meta-Analyses
One of the most prominent strengths of this meta-analysis was the inclusion of a large number of prospective studies. This allowed for analysis of one of the largest pools of data compared to most meta-analyses available to date. Another advantage was the inclusion of the HbA1c value achieved at EOS for analysis in contrast to the target HbA1c value used in some meta-analyses. We predominantly used the random effect model, which is one of the most conservative modes of analysis, for estimation of the effect size. This helped minimise the risk of over-estimation of the effect. Last but not the least, macrovascular endpoints were better represented in this meta-analysis, which is not surprising given the large number of trials that have focussed on CV safety and/or superiority.