Paper Cruxy
An ICU bedside at night, with the indicator lights of an infusion pump and a monitor glowing in the dim room
Column

The Rise and Fall of "Tight Glycaemic Control" — Reading 13 Trials Until the Question Itself Changed

Column

The Rise and Fall of "Tight Glycaemic Control" — Reading 13 Trials Until the Question Itself Changed

Glycaemic control Metabolism & Endocrinology ≈18 min read 14 trials

Bring blood glucose closer to normal, and patients should do better. This intuition was strongly supported once, both in outpatient care and in the ICU — and then refuted in both settings, in 2008–2009. From 2015 onward, the question itself was replaced: not "how far down" but "with what." Lining up 13 trials from DCCT through NICE-SUGAR to EMPA-REG and LEADER, across three acts, reveals why "making the glucose number look better" and "making the patient's outcome better" are not the same thing.

Guidelines this column draws on

The text has been checked against the statements below. Please read the originals, and the current editions, before acting on them.

The 14 trials behind it

Intensive insulin therapy dramatically reduces microvascular complications in type 1 diabetes N Engl J Med, 1993 Intensive glycaemic control in type 2 diabetes reduces microvascular complications Lancet, 1998 Tight glycaemic control (HbA1c<6%) actually increased mortality N Engl J Med, 2008 Slow, gradual intensive glycaemic control reduced nephropathy without increasing mortality N Engl J Med, 2008 Intensive glycaemic control did not reduce cardiovascular events in long-standing type 2 diabetes N Engl J Med, 2009 Intensive insulin therapy in critically ill patients (Leuven/Van den Berghe) N Engl J Med Tight glucose control in the ICU (81–108 mg/dL) increased mortality N Engl J Med, 2009 EMPA-REG OUTCOME trial: the first diabetes drug to reduce cardiovascular death N Engl J Med, 2015 Liraglutide reduces cardiovascular events and death in type 2 diabetes N Engl J Med, 2016 Semaglutide reduces cardiovascular events (but watch for retinopathy) N Engl J Med, 2016 Dulaglutide reduces cardiovascular events across a broad population of type 2 diabetes, including primary prevention Lancet, 2019 Canagliflozin reduced MACE but increased lower-limb amputation and fractures N Engl J Med, 2017 Dapagliflozin did not reduce MACE, but did reduce the composite of heart failure hospitalisation and CV death N Engl J Med, 2019 Insulin degludec is CV noninferior to glargine, with significantly less severe hypoglycaemia N Engl J Med, 2017
01The question

"Making the glucose number look better" and "making the patient better" are not the same thing

Hyperglycaemia is bad. This is hard to dispute. Infections increase, wound healing slows, and osmotic diuresis disturbs haemodynamics. So the reasoning goes: bring glucose closer to normal, and the patient should improve. This reasoning was, at one point, strongly supported — both in outpatient care and in the ICU.

But between 2008 and 2009, that reasoning was refuted in both settings. In each case, it was the largest trial to date that did the refuting.

And from 2015 onward, the question itself was replaced — not "how far down" but "with what." Tracing these three acts leaves a lesson that extends well beyond glycaemic control.

02Act I

Act I: Lower it, and you protect the "microvasculature"

DCCT (1993, type 1 diabetes) lowered HbA1c from approximately 9% to approximately 7.2% with intensive insulin therapy, and over a mean follow-up of 6.5 years reduced the onset of retinopathy by 76% and its progression by 54%. Nephropathy and neuropathy were also substantially reduced. However, severe hypoglycaemia increased roughly threefold. It is worth remembering that the benefit and the trade-off were presented together from this very first trial.

UKPDS 33 (1998, newly diagnosed type 2 diabetes) points in the same direction. At a median of 10 years, HbA1c of 7.0% vs 7.9% reduced microvascular complications by approximately 25%, driven mainly by retinopathy.

However, UKPDS 33 found no significant difference in macrovascular disease or mortality. In other words, what was established at this point was only that "lowering glucose protects the microvasculature." "It also protects the macrovasculature" was not shown.

This distinction became the starting point for the next decade of debate.

3 Stages: Glycemic ControlAct1 1993–1998↓Glucose→saves microvascDCCT1993 Retin. −76%UKPDS 331998 Microvasc −25%Act2 2001–2009Macrovasc benefit refuted, both settingsLeuven2001 ICU mort 8.0→4.6%ACCORD2008 death +22%ADVANCE2008 Renal −21%VADT2009 No diffNICE-SUGAR2009 90d death +2.6ptAct3 2015–From 'How low?' to 'With what?'EMPA-REG2015: CV & ACM ↓LEADER2016・MACE HR 0.87SUSTAIN-62016・MACE HR 0.74REWIND2019・MACE HR 0.88Values = primary result per trial; see papers for details/CIs
Figure 1: Overview of the three acts. Act One established benefit for microvascular disease; Act Two attempted to extrapolate this to macrovascular disease and failed. Act Three represents a shift in the very framing of the question.
03Act II

Act II: The failure of extrapolation — surely lower would be better still

The remaining question was a natural one. If UKPDS showed no difference in macrovascular outcomes, was it because the lowering hadn't gone far enough? So why not lower it further?

ACCORD (2008) tested this. In type 2 diabetes with high cardiovascular risk, the trial aimed for HbA1c below 6%. The result: HbA1c reached 6.4% vs 7.5%, closer to target, but total mortality increased by 22% (HR 1.22) and cardiovascular death rose by 35%. Major cardiovascular events were not reduced. The intensive arm was stopped early.

ADVANCE, published the same year, used a gentler intensification (HbA1c target 6.5%, built on modified-release gliclazide) and reduced the primary composite by 10%, driven mainly by a 21% reduction in nephropathy, with no difference in macrovascular disease or mortality. Critically, unlike ACCORD, mortality did not increase.

VADT (2009) enrolled veterans with long-standing, poorly controlled diabetes (HbA1c approximately 9.4%). Even with HbA1c of 6.9% vs 8.4%, major cardiovascular events were not reduced (HR 0.88, P=0.14), mortality was unchanged, and severe hypoglycaemia increased.

The conclusion across the three trials is consistent. In established type 2 diabetes, keeping glucose lower for its own sake yields little to no cardiovascular benefit.

Sign up free to read the rest

6 more sections, 3 figures, drawing on 14 trials, follow.
An e-mail address and a password. No card required.

Sign up free and read on Log in

This page is educational commentary on the medical literature and is not a substitute for clinical judgement in an individual patient. The figures given are those reported in the original papers. Decisions on whether to treat, on dose and on targets must rest on the original papers and on the guidelines that apply where you practise.