Deep dive #9 · 10 November 2026

Type 2 diabetes

For health professionals General practice World Diabetes Day, 14 November

Type 2 diabetes is the condition every GP manages and nobody finds glamorous. Yet the last decade rewrote it: drugs that cut deaths and kidney failure, not just glucose; trials showing early disease can go into remission; and a pipeline so crowded the 2015 algorithm looks like a museum piece. World Diabetes Day falls on 14 November, Banting's birthday. Here's the evidence, with the numbers.

At a glance

Burden: 589 million adults (20–79) have diabetes, over 90% type 2; 252 million (over 4 in 10) are undiagnosed (IDF Atlas 2025). Australia: 5.3%, 1.3 million (ABS 2022).
Mechanism: insulin resistance plus progressive beta-cell failure; ectopic liver and pancreatic fat is the reversible part (Counterpoint, DiRECT).
Diagnosis: HbA1c ≥48 mmol/mol (6.5%), fasting glucose ≥7.0 or 2-hour OGTT ≥11.1 mmol/L, confirmed if asymptomatic. Know when HbA1c lies.
Organ protection: SGLT2 inhibitors and GLP-1 RAs cut cardiovascular and kidney events (EMPA-REG CV death −38%; CREDENCE renal HR 0.70; FLOW HR 0.76).
Remission is real but fragile: 46% at 12 months and 36% at 24 months with intensive weight management; 13% at 5 years (DiRECT).
Age at diagnosis matters: diagnosed at 30, about 14 years of life lost; at 50, about 6 (US rates, ERFC 2023).

Biggest take-home: think organs, not just HbA1c. Pick drugs for the heart and kidneys, and treat weight as a disease-modifying target.

For patients: Type 2 diabetes, Prediabetes and the GLP-1 medicines guide.

Definition

What it is: hyperglycaemia from a progressive loss of adequate beta-cell insulin secretion, usually on a background of insulin resistance, without autoimmune beta-cell destruction.
Diagnostic thresholds (WHO/ADA): HbA1c ≥6.5% (48 mmol/mol), fasting plasma glucose ≥7.0 mmol/L, 2-hour OGTT glucose ≥11.1 mmol/L, or random ≥11.1 mmol/L with symptoms. Without symptoms, confirm with a second test.
Remission (ADA-led consensus 2021): HbA1c <6.5% (48 mmol/mol) at least 3 months after stopping glucose-lowering drugs. "Remission", not "cure", because it can return.

History

Ancient India, 1st millennium BCE, Sushruta: "madhumeha", honey-like urine that was sweet, sticky and drew ants.
2nd century CE, Aretaeus of Cappadocia: named it diabetes, from the Greek for "siphon" or "pass through".
1674, Thomas Willis: urine "wonderfully sweet like sugar or honey". In the 1770s, Matthew Dobson showed the sweetness was sugar, in urine and blood.
1870, Apollinaire Bouchardat: noticed glycosuria fell in his patients during the siege of Paris and prescribed eating less: an early dietary therapy.
1936, Harold Himsworth: split diabetes into "insulin-sensitive" and "insulin-insensitive" types in the Lancet: the seed of type 1 vs type 2.
1957, Jean Sterne: first reported metformin for diabetes (in a Moroccan journal), derived from goat's rue (Galega officinalis).
Late 1950s, Rosalyn Yalow and Solomon Berson: the insulin radioimmunoassay, showing type 2 is inefficient use of insulin, not lack of it. Yalow won the 1977 Nobel; Berson had died in 1972. In 1969 Dorothy Hodgkin solved insulin's 3D structure.
1986, Pan Xiao-Ren and Li Guangwei: the Da Qing study in China, among the first lifestyle-prevention trials. In 2024 Svetlana Mojsov, Joel Habener and Lotte Bjerre Knudsen won the Lasker Award for GLP-1 work.

Pathophysiology

Two defects, one clock: insulin resistance and a beta cell that can't keep up. The fat in the wrong places is the most reversible part.

Many organs: DeFronzo's "ominous octet" (2009): muscle, liver and beta cell, plus fat, gut incretins, alpha cell, kidney and brain.
Beta-cell decline: function estimated at about 50% of normal at diagnosis (HOMA model), falling progressively on all therapies (UKPDS 16).
A late, steep turn: in Whitehall II (6,538 people), fasting glucose rose steeply only in the 3 years before diagnosis, after insulin sensitivity had fallen for 5 (Tabák 2009).
Twin cycles of ectopic fat: on 600 kcal/day, liver fat fell 30% in 1 week and fasting glucose normalised (9.2 to 5.9 mmol/L); pancreatic fat fell and first-phase insulin response approached normal by 8 weeks (Counterpoint, n=11; Lim 2011).
Genetics: 1,289 independent signals at 611 loci from 2.5 million people (39.7% non-European), falling into 8 mechanistic clusters (Suzuki 2024).
Beta cells may go quiet, not die: in mice, beta-cell "loss" was dedifferentiation (Talchai 2012). Human relevance is unproven.

Type 2 diabetes from genes and ectopic fat to hyperglycaemia, and the reversible window (simplified; Suzuki 2024, Lim 2011, UKPDS 16, Lean 2018). Download this diagram

Epidemiology

Global

IDF Atlas 11th edition (2025): 589 million adults aged 20–79, rising to 853 million by 2050; over 3.4 million deaths a year; spending over USD 1 trillion.
NCD-RisC (2024): 828 million adults (18+) in 2022, up 630 million since 1990; 445 million (59%) of those aged 30+ untreated. WHO: adult prevalence doubled from 7% to 14% (1990–2022).
Youth-onset: in the TODAY cohort, 60.1% had at least one complication after 13 years, at a mean age of 26 (diabetic kidney disease 54.8%, retinal disease 51.0%).

Australia

Prevalence: 5.3%, 1.3 million people, in 2022; 87.6% type 2 (ABS National Health Survey).
First Nations adults: 2.9 times as likely to have diabetes as non-Indigenous adults, age-adjusted (AIHW, 2018–19 data).
Deaths and burden: underlying cause of about 6,000 deaths and associated with 11% of all deaths in 2022; type 2 was 2.2% of total disease burden in 2023 (AIHW).
Kidneys: diabetic kidney disease caused 36% of new kidney replacement therapy starts in 2023, the leading cause (ANZDATA).

Causes and risk factors

Adiposity and inactivity: the main modifiable drivers; weight loss reverses early disease (DiRECT).
Genes and ancestry: 611 loci; risk is higher in many non-European populations and in First Nations Australians.
Gestational diabetes: nearly 10-fold risk of later type 2 (RR 9.51, 95% CI 7.14–12.67; 1.3 million women; Vounzoulaki 2020).
Air pollution: HR 1.15 per 10 μg/m³ PM2.5; about 3.2 million incident cases in 2016 (Bowe 2018). GBD 2019: about one-fifth of type 2 burden.
Shift work: OR 1.09 (1.05–1.12), highest with rotating shifts (Gan 2015, 12 studies).
Prenatal famine: late-gestation exposure in the Dutch Hunger Winter: 2-hour glucose +0.5 mmol/L in adulthood (Ravelli 1998).

Preventable factors

Lifestyle in impaired glucose tolerance: −58% incidence in both the Finnish DPS (n=522) and US DPP (n=3,234; NNT 6.9 over 3 years). Metformin −31% in DPP.
Durable: DPPOS at 15 years, lifestyle −27%, metformin −18%.
Hard outcomes: Da Qing, 30 years: diabetes delayed a median 3.96 years, CVD events −26%, CV death −33%, all-cause death −26%, life expectancy +1.44 years.
Drugs (trial data): tirzepatide in obesity with prediabetes: type 2 in 1.3% vs 13.3% over 176 weeks (HR 0.07); HR 0.12 after 17 weeks off drug (SURMOUNT-1).
After gestational diabetes: the RR of 9.51 makes this one of the best-defined groups for recall and testing.

Diagnosis essentials

Australian screening (RACGP Red Book): AUSDRISK every 3 years from 40; score ≥12 means high risk: test with fasting glucose or HbA1c, then every 3 years (yearly with IFG or IGT). Aboriginal and Torres Strait Islander people: yearly blood test from 18.
Medicare: HbA1c for diagnosis in high-risk asymptomatic people (item 66841), once per 12 months.
When HbA1c misleads (ADS 2015): haemoglobinopathies, raised HbF, haemolysis, splenomegaly, increased red cell turnover (iron, B12 or folate therapy, EPO, chronic liver disease), dapsone, ribavirin, antiretrovirals, high-dose vitamins C and E. Use glucose-based tests instead.
Is it really type 2? LADA is 2–12% of adult-onset diabetes. In UKPDS 25, GAD antibodies were found in 34% aged 25–34 vs 7% aged 55–65, and 84% of antibody-positive under-35s needed insulin within 6 years. Test antibodies if young, lean, or failing fast.
At diagnosis: baseline kidney (eGFR, urine ACR), eyes, feet, lipids, BP and cardiovascular risk.

Current treatment

Targets

RACGP: HbA1c generally ≤7% (53 mmol/mol), individualised.
Too low can harm: targeting <6.0% raised mortality (HR 1.22, 1.01–1.46; ACCORD, n=10,251).
Treat everything: intensive multifactorial care in Steno-2 cut deaths from 50% to 30% at 13 years (HR 0.54) and added a median 7.9 years of life at 21 years.

Medicines: effect sizes

Metformin: all-cause mortality −36% in overweight patients (UKPDS 34, n=753); MI −33% and death −27% persisted 10 years after the trial (UKPDS 80). Caveat: early metformin added to a sulfonylurea raised diabetes-related death in UKPDS 34.
SGLT2 inhibitors: empagliflozin MACE HR 0.86, CV death −38%, HF hospitalisation −35%, all-cause death −32% (EMPA-REG, n=7,020). Canagliflozin renal composite HR 0.70 (CREDENCE). KDIGO 2022: start if eGFR ≥20.
GLP-1 RAs: liraglutide MACE HR 0.87 (LEADER); semaglutide HR 0.74 (SUSTAIN-6; retinopathy complications HR 1.76); semaglutide kidney events HR 0.76, CV death HR 0.71 (FLOW).
Tirzepatide: HbA1c −2.30 vs −1.86 points and weight −11.2 vs −5.7 kg vs semaglutide 1 mg (SURPASS-2). Noninferior, not superior, to dulaglutide for MACE (HR 0.92, 0.83–1.01; SURPASS-CVOT).
Durability: over 5 years, glargine and liraglutide held targets best, sitagliptin worst (HbA1c ≥7% at 26.5 and 26.1 vs 38.1 per 100 person-years; GRADE).
Early combination: metformin plus vildagliptin halved early treatment failure vs stepwise metformin (HR 0.51; VERIFY).

Weight, remission and surgery

DiRECT (n=298, primary care, under 6 years since diagnosis): remission 46% vs 4% at 12 months; 86% in those losing ≥15 kg; 36% at 24 months; 13% at 5 years.
Metabolic surgery: HbA1c ≤6.0% at 5 years in 29% after gastric bypass and 23% after sleeve vs 5% with medical therapy (STAMPEDE, n=150).

Emerging treatments and trials

Orforglipron (oral non-peptide GLP-1 RA): HbA1c −1.24 to −1.48 points at 40 weeks; weight −4.5% to −7.6% (ACHIEVE-1). FDA-approved for obesity in April 2026.
Retatrutide (GIP/GLP-1/glucagon): HbA1c −2.02 at 24 weeks in phase 2. Phase 3 topline (not yet peer-reviewed): TRANSCEND-T2D-2, HbA1c −2.2 vs −1.8 and weight −20.9% vs −9.4% vs semaglutide 2 mg at 80 weeks.
CagriSema: HbA1c −1.91 vs −1.76 and weight −14.2% vs −10.2% vs semaglutide 2.4 mg at 68 weeks (REIMAGINE 2).
Weekly basal insulin: icodec FDA-approved for type 2 in March 2026 and on the ARTG; efsitora FDA-approved in September 2026 (QWINT-2: HbA1c −1.26 vs −1.17 with degludec).
Stem-cell islets: one 59-year-old man received autologous stem-cell-derived islet tissue in Shanghai in 2021 and was insulin-free for 33 months (Wu 2024). One case.

Theoretical frontiers

Hypotheses, not recommendations.

Causes: theories

Hypothesis: ectopic fat causes type 2 (twin cycle). For: liver fat −30% in 1 week with normal fasting glucose (Lim 2011); remission tracks weight loss (DiRECT). Against: Counterpoint had 11 people; not everyone with type 2 has a high BMI. Test: weight-loss trials in normal-BMI type 2 with MR liver and pancreatic fat.
Hypothesis: hyperinsulinaemia comes first. For: insulin excess can drive insulin resistance (Shanik 2008; Corkey 2012). Against: Whitehall II shows a compensatory rise, then fall, in beta-cell function. Test: trials that lower insulin secretion in early dysglycaemia.
Hypothesis: beta cells dedifferentiate rather than die. For: FoxO1 mouse model (Talchai 2012); early function returns with weight loss (Lim 2011). Against: mostly mouse data. Test: human pancreas and biomarker studies before and after remission.
Hypothesis: the gut microbiome is causal. For: metformin-altered microbiota improved glucose tolerance in germ-free mice (Wu 2017); lean-donor FMT improved insulin sensitivity at 6 weeks (Vrieze 2012). Against: no effect by 18 weeks (Kootte 2017). Test: larger, longer FMT or defined-consortium RCTs.
Hypothesis: air pollution is a major cause. For: HR 1.15 per 10 μg/m³ (Bowe 2018); one-fifth of burden in GBD 2019. Against: observational, residual confounding. Test: natural experiments around clean-air policies.
Hypothesis: diabetes is programmed in utero. For: thrifty phenotype (Hales and Barker 1992); Dutch famine (Ravelli 1998). Against: small effect (+0.5 mmol/L). Test: maternal nutrition RCTs with offspring follow-up.
Hypothesis: endocrine-disrupting chemicals contribute. For: dioxins RR 1.91, PCBs 2.39, BPA 1.45 (Song 2015). Against: mostly cross-sectional (41 of 49 studies). Test: prospective biomonitoring cohorts.
Hypothesis: circadian disruption contributes. For: shift work OR 1.09, rotating shifts worst (Gan 2015). Against: small effect, confounding. Test: RCTs of shift-schedule redesign with glycaemic outcomes.

Prevention: theoretical options

Hypothesis: incretin drugs can prevent type 2 at scale. For: HR 0.07 on tirzepatide (SURMOUNT-1). Against: cost, and higher risk after stopping (HR 0.12 vs 0.07). Test: long pragmatic trials with hard outcomes and cost-effectiveness.
Hypothesis: lifestyle prevention extends life. For: Da Qing all-cause death −26%, +1.44 years. Against: one cluster-randomised trial (n=577). Test: long-term outcome follow-up of DPP and DPS.
Hypothesis: post-GDM programs prevent much type 2. For: RR 9.51 (Vounzoulaki 2020). Test: postpartum prevention RCTs with registry linkage.
Hypothesis: cleaner air prevents diabetes. For: 3.2 million incident cases attributed to PM2.5 (Bowe 2018). Test: stepped policy evaluations.
Hypothesis: time-restricted eating is as good as calorie counting. For: HbA1c −0.91 vs control, no different from calorie restriction (Pavlou 2023, n=75). Test: larger prevention trials in prediabetes.

Treatment: theoretical options

Hypothesis: remission should be the default goal early. For: DiRECT 46% at 1 year. Against: 13% at 5 years. Test: trials of maintenance strategies, including incretins, after remission.
Hypothesis: early combination preserves beta cells. For: VERIFY HR 0.51; GRADE durability with liraglutide. Against: VERIFY measured treatment failure, not beta-cell mass. Test: trials with clamp-measured beta-cell function.
Hypothesis: anti-inflammatory therapy treats type 2. For: salsalate HbA1c −0.37% (TINSAL-T2D, n=286). Against: weight, LDL and urine albumin rose. Test: targeted anti-inflammatory trials with outcomes.
Hypothesis: microbiome modulation can treat it. For: Vrieze 2012. Against: effect gone by 18 weeks (Kootte 2017). Test: repeated-dose RCTs.
Hypothesis: cell replacement can cure insulin-requiring type 2. For: 33 insulin-free months in one man (Wu 2024). Against: n=1. Test: phase 1/2 trials.
Hypothesis: triple agonists make weight-based remission routine. For: weight −20.9% (TRANSCEND-T2D-2 topline). Against: unpublished, no outcome data. Test: remission-designed trials.

Gaps and open questions

The undiagnosed 252 million: over 4 in 10 adults with diabetes don't know they have it (IDF).
Untreated majority: 59% of people aged 30+ with diabetes get no treatment (NCD-RisC).
Remission maintenance: most remissions lapse by 5 years.
Hard outcomes for new agents: tirzepatide wasn't superior to dulaglutide; retatrutide and CagriSema have no CV outcome results yet.
Youth-onset disease: 60% have a complication by their mid-20s (TODAY). How best to treat them is unclear.

Prognosis and quality of life

Life expectancy: diagnosis at 30, 14 years shorter; at 40, 10; at 50, 6 (US rates; 13, 9, 5 with EU rates). About 3–4 years lost per decade of earlier diagnosis (ERFC 2023, 1.5 million people).
Multifactorial care pays: +7.9 years median life (Steno-2).
Kidneys: the leading cause of new kidney replacement therapy in Australia (36%).
Legacy effect: early good control still showed MI and mortality benefit 10 years after trial differences vanished (UKPDS 80).

Weird trivia

World Diabetes Day is Banting's birthday, 14 November; IDF and WHO created it in 1991 and the UN made it official in 2006.
Banting, Best and Collip sold their insulin patents to the University of Toronto for $1 each.
Leonard Thompson, 14, got the first insulin injection on 11 January 1922.
Glucophage means "glucose eater". Its plant ancestor, goat's rue, is now a noxious weed in many US states.
Exenatide comes from Gila monster venom: John Eng found exendin-4 in 1992 at the Bronx VA, where Rosalyn Yalow worked.
Bouchardat's prescription was "mangez le moins possible": eat as little as possible.
Hodgkin took 34 years from her first insulin X-ray photo (1935) to the structure (1969).

Guidelines

Key trials

UKPDS 34 (1998) and UKPDS 80 (2008): metformin and the legacy effect.
Finnish DPS (2001) and DPP (2002): lifestyle −58%.
ACCORD (2008): too-tight control raised mortality.
Steno-2 (2008, 2016): multifactorial care, +7.9 years.
EMPA-REG OUTCOME (2015), CREDENCE (2019): SGLT2 inhibitors.
LEADER, SUSTAIN-6 (2016), FLOW (2024): GLP-1 RAs.
DiRECT (2018–2024): remission in primary care.
VERIFY (2019), GRADE (2022): early combination and durability.
SURPASS-2 (2021), SURPASS-CVOT (2025): tirzepatide.

Sources

Information only. Check product information, PBS criteria and current Australian guidelines before prescribing. Some agents above are not approved in Australia.

Updated 10 November 2026. General education for health professionals; not a substitute for local protocols.

maladywise.com/deep-dives/type-2-diabetes/