If I Had Type 2 Diabetes

Dr Edward Leatham
Jul 11
17 min read

What I would do if I had type 2 diabetes — and, just as much, what I would do to avoid it. NHS NICE guidance is engineered for the best value across a whole population. That is not always the same question as: what would I do for myself, if money were no object? Here is my honest answer — with the evidence set out in full.

For busy people, or to tune in when on the move, a Google NotebookLM audio podcast is available as a story beneath.

Patients ask me a version of this all the time: "If it were you, doctor — what would you actually do?" It is a fair question, and the real answer is often not the same as the printed guideline. So let me answer it plainly, in the order I would tackle it — and this time, with the references attached.

Fix the food first, then move, then build muscle — and only then reach for drugs, in a sequence chosen to shrink visceral fat and flatten the glucose curve rather than simply to satisfy a cost-effectiveness threshold.

Why "best value" is not the same as "best possible"

NICE does an extraordinary job of a genuinely difficult task: deciding how to spend a finite NHS budget so that the most people benefit for each pound. That is the right question for a health system. But it is a population question — one that must weigh cost-effectiveness alongside safety, access and the average case. Ask instead, "what is the best I could do for this one body, setting those constraints aside?" and the answer can look different — earlier, more layered, and more aggressive about the thing that, in most people, actually drives type 2 diabetes — especially in white Europeans and anyone whose waist exceeds half their height: visceral fat.

In fairness, that gap has narrowed. Since NICE NG28 was updated in February 2026 it recommends modified-release metformin together with an SGLT2 inhibitor as first-line therapy for most adults with type 2 diabetes, continues these agents for cardiovascular and renal protection independent of the glucose reading, and brings GLP-1 agonists or tirzepatide forward for selected higher-risk groups. The distinction I am drawing, then, is not that NICE is behind the science — it now explicitly balances individualised care, cardiorenal protection, obesity and patient preference, not merely cost — but that a private, individualised strategy can act on the same evidence with fewer cost and eligibility constraints.

So this is not a criticism of NICE. It is simply what I would choose if cost and eligibility thresholds were not part of the calculation.

One distinction decides whether any of this will work, because type 2 diabetes is not a single disease but a spectrum. At one end sits the metabolic-overload form — an insulin-resistant body, usually carrying excess visceral fat, whose pancreas is labouring against a load it was never built to carry (in the research literature, the severe insulin-resistant cluster, SIRD). That is the form this article is written for, and the one most likely to be pushed into remission once the load comes off.

At the other end is the lean, insulin-deficient form — the severe insulin-deficient cluster (SIDD) — diabetes that appears despite little visceral fat, driven by a pancreas genuinely failing to make enough insulin, as in type 1, LADA (latent autoimmune diabetes of adults), or the insulin-deficient end of type 2. Here the limit is the organ, not the load, and it would be wrong to imply diet and exercise alone can beat it; in that setting insulin is not optional.

Most people are a blend of the two, and the balance can move: years of overload can gradually exhaust the beta cells, shifting someone from the resistant end toward the deficient one — though there is an early window in which shedding ectopic fat can still restore function. Which end dominates, and which way you are travelling, is what should shape the plan.

A simple bedside check points the way: a waist more than half your height — a waist-to-height ratio above 0.5 — makes visceral fat the likely driver and places you toward the modifiable, metabolic-overload end where this plan does most. Ethnicity shifts the odds, too — as population tendencies, not individual verdicts. In a white European man carrying central fat, the visceral-fat story is the more probable one; South and East Asian backgrounds carry a greater innate limit on beta-cell reserve, so diabetes tends to arrive younger and leaner and to sit further toward the deficient pole — the same plan still helps, but full remission can be harder and insulin support may come sooner. Hispanic and Latino backgrounds, by contrast, tend toward strong insulin resistance and liver fat, so the visceral-fat strategy stays squarely on target. I set the ethnic and mechanistic detail out more fully in my own companion white paper.

And you do not need a diagnosis to use most of this. Nearly everything that follows matters as much for people trying to avoid diabetes — anyone with a widening waist, a family history, or a "pre-diabetic" HbA1c — as for those who already have it. The biology is a continuum, so the early levers are identical; what mainly changes is the pharmacology, which dials down as you move from treatment toward prevention. The table near the end sets out who does what.

It is worth being clear why I keep returning to visceral fat, because it is far more than a proxy for insulin resistance. The fat packed around the organs is metabolically active and, frankly, hostile. It pours out inflammatory cytokines — some of them increasingly linked to the long-term risk of dementia — and it drains a constant stream of free fatty acids straight into the portal vein feeding the liver. The liver packages that flood into VLDL, and the downstream consequence is a shift toward small, dense LDL: the particle most prone to oxidise and lodge in the artery wall, which is much of why visceral fat ties so tightly to coronary disease. And the compensatory high insulin of the resistant years is itself no innocent bystander — decades of hyperinsulinaemia do their own quiet damage.

This also explains why high visceral fat does not condemn everyone to diabetes. It is a principal driver of insulin resistance, but whether that ever tips into a raised HbA1c depends on the pancreas behind it. Someone with generous beta-cell reserve may carry a large visceral load for years with a normal glucose — paying the cardiovascular and inflammatory price all the while, but never earning the label. Someone nearer the insulin-deficient end — the pattern common in many East Asian populations — exhausts that reserve sooner, and years of visceral overload will march them toward pancreatic failure. Most people sit somewhere between the two.

The practical conclusion is the same whichever way you lean: high visceral fat challenges the whole system, so lowering it helps across the board — and if the aim is to head off pancreatic failure, it has to be tackled early, long before the glucose budges. That is why the humblest instrument in the kit, a tape measure, earns its place from the age of thirty. It is the cheapest diabetes-prevention tool we have.

First — before any drug — I would fix the food

If my HbA1c were creeping into pre-diabetic territory (42–47 mmol/mol, roughly 6.0–6.4%) or across the diabetic line (≥48 mmol/mol, 6.5%), the first lever is the plate, not the prescription.

I would cut the carbohydrate load, and in particular hunt down the ultra-processed foods where sugar hides in plain sight — sauces, breads, "healthy" cereals, low-fat products that replace fat with starch. Then I would measure rather than guess. I would wear a continuous glucose sensor and aim to keep my response to each meal as flat as possible.

The Lingo app logs your meals and then feeds back which foods and which timings drive your glucose excursions — turning a vague sense of "I should eat better" into a personal, meal-by-meal map of what your own body actually does with each food.

Armed with that map, I would do one of two things with each offending food: eliminate it, or defuse it — pairing it with fat, protein and fibre, and eating it last in the meal, so the same food produces a gentler rise. Here is the mechanistic reason it works: dietary fat provokes almost no insulin response, and protein only a modest one. It is carbohydrate — especially refined carbohydrate and hidden sugar — that drives the large glucose and insulin swings. Build the plate around protein, fat and fibre, keep the starch small and eat it last, and the curve stays flat.

To make this practical I would sit a food app on top of the sensor. SNAQ is the one I would reach for: you photograph a meal and it estimates the carbohydrate, protein and fat, then lays that meal directly onto your glucose curve so cause and effect become impossible to miss. Crucially for what comes later, it lets me set a protein target per meal and see whether I actually hit it — and if I switch on Apple Health, my steps and my resistance-training sessions flow onto the same dashboard, so food, glucose and exercise finally sit in one place.

It is worth pausing on why cutting carbohydrate works at all, because the idea is far older than any modern fad. Before insulin arrived in 1921, the only treatment that bought a person with diabetes any time was carbohydrate restriction — the so-called "animal diet" of fatty meat and green vegetables, with the starch quite literally boiled away. That lineage runs straight through the low-carbohydrate diet popularised by the cardiologist Robert Atkins to today's ketogenic approaches, and it is the backbone of Mark Hyman's Eat Fat, Get Thin.

There is a mechanistic reason the idea endures: for most of the body, glucose is not the default fuel. The heart in particular runs largely on free fatty acids and ketones. Indeed, one leading hypothesis for why SGLT2 inhibitors help failing hearts is that the modest rise in ketones they produce hands the myocardium a more oxygen-efficient fuel — β-hydroxybutyrate reaches the Krebs cycle in fewer steps and yields more energy per unit of oxygen burned. The heart-failure benefit of these drugs is itself well established, including in people without diabetes; the fuel-switch remains one hypothesis among several for how it comes about, but a compelling one.

Follow that logic and glucose starts to look less like a staple and more like a specialist fuel. It is genuinely obligatory for only a handful of tissues — red blood cells, parts of the kidney and eye — and it powers part of the brain, though even the brain runs substantially on ketones once they are available. The small amount the body truly needs, the liver can manufacture itself, which is why there is no dietary requirement for carbohydrate at all. The other moment glucose earns its keep is anaerobic, oxygen-starved effort — the sprint, the maximal lift — where fat cannot be burned fast enough to keep up.

Glucose is rocket fuel: only needed when I want to launch the rocket.

Then — move, in the ordinary cracks of the day

Next comes aerobic activity, and I mean the unglamorous kind that survives a busy week. Park further from work. Take the stairs instead of the lift or escalator. A short walk after a meal blunts its glucose peak. The target I would set myself is an average of 7,000–10,000 steps a day — not every single day, but on average across the week.

If progress stalls — or if you simply like the idea of a home gym — this is where I would spend on a machine you will actually use. A low-impact incline elliptical (the kind with an adjustable ramp that mimics walking uphill) is gentle on the joints, and in my experience the interactive, coached versions earn their keep through sheer motivation: the device that nudges you onto it beats the "better" device gathering dust. NordicTrack's incline ellipticals are one well-known example.

Then — build muscle, because muscle is a glucose sink

Skeletal muscle is where a large share of glucose is disposed of, so growing and keeping it is metabolic gold. Two levers here. First, protein: enough to trigger muscle protein synthesis, which in practice often means aiming for around 25–30 g of good-quality protein per meal (the per-meal target I would set and track in SNAQ) so that the leucine threshold is comfortably cleared. Second, progressive resistance training: as little as fifteen minutes, five days out of seven, with weights or resistance bands.

If it were me and it were affordable, I would front-load this with a personal trainer at a gym for as long as that stayed sustainable — to learn form and build the habit — then step down to a simple home routine I could maintain for the rest of my life. The maintenance phase is the point; the initial push is just how you get there.

Then — protect your sleep, because cortisol is a metabolic saboteur

Sleep is the pillar people skip, and it quietly undoes the rest. Short or broken sleep pushes cortisol up, and sustained cortisol worsens insulin resistance, nudges glucose higher and stokes appetite the following day. So I would treat sleep as a measurable target rather than a luxury. A sleep-tracking ring makes it visible: the Oura ring is the best-known, though there are now capable, subscription-free alternatives such as the Ultrahuman Ring Air, RingConn and the Samsung Galaxy Ring. The point is not the gadget — it is that what gets measured tends to get managed.

Only then — the pharmacology

Now the drugs — and notice how far down the page they sit. My trigger to escalate would be visceral fat, not the scales: a raised VAT reading on CT or DEXA, or the simplest proxy of all — a waist more than half my height. At that point I would put a monitoring system in place first, so I could actually see whether the food-and-exercise measures were working. If the numbers moved, I would stay the course. If they did not, I would layer in pharmacology deliberately:

Metformin — I would start at 500 mg twice daily and titrate up towards 2 g/day. It reliably turns down the liver's own glucose production (hepatic gluconeogenesis), and it is well-understood, cheap and time-tested. Here I am close to current guidance, which since 2026 typically starts modified-release metformin and an SGLT2 inhibitor together from diagnosis.

A GLP-1 receptor agonist, gently dosed — if visceral fat would not come down to target on lifestyle alone, this is where a private plan can move sooner than NHS eligibility allows. A low, carefully titrated dose (what people loosely call "microdosing") to bring the waist to target. Worth being clear: microdosing for this purpose is largely a private, off-licence approach. NICE has, since 2026, brought incretin therapy forward for selected higher-risk groups, but the NHS still applies specific eligibility thresholds — so this is less a disagreement with NICE than the same evidence acted on with fewer constraints.

Beyond that private use, GLP-1 agonists are now recommended right across type 2 diabetes — whether or not the person is already on insulin. Where insulin is being used, adding one usually lets the insulin dose come down substantially, which is welcome: less injected insulin means less of the weight gain it drives. Part of that is a direct insulin-sensitising effect, but I suspect the larger part is the drug's shrinking of visceral and ectopic fat — less fat around the liver means fewer free fatty acids flooding it, and so less insulin resistance in the periphery. In type 1 diabetes the drugs are not licensed, though I understand they are widely used off-label.

An SGLT2 inhibitor — one of the genuinely clever drugs, and now part of NICE's first-line pairing. It blocks the sodium-glucose co-transporter in the kidney's proximal tubule, lowering the threshold at which the kidney hangs on to glucose. Above that lowered threshold, glucose simply spills into the urine, pulling circulating glucose down and shaving the peaks — with worthwhile heart and kidney protection alongside. (It carries its own cautions, including a rare risk of ketoacidosis — a particular reason for care if you are eating low-carb — and the need for "sick-day rules," both conversations to have with your doctor.)

A gliptin (DPP-4 inhibitor) — the gentler incretin option. Rather than flooding the system with a potent analogue, it simply slows the breakdown of your own GLP-1. Because it works on the same axis as a GLP-1 receptor agonist, I would treat it as an alternative to one — not something to stack on top (both NICE and the ADA now explicitly advise against pairing the two) — reserved for when a full GLP-1 drug is not wanted or tolerated.

An SGLT2 inhibitor and a GLP-1 agonist together, by contrast, are increasingly recommended as a pair for higher-risk patients — NICE now places semaglutide alongside metformin and an SGLT2 inhibitor as first-line triple therapy after established cardiovascular disease, and the ADA endorses using both. The reason is that they work on different organs and different endpoints — the SGLT2 inhibitor strongest for heart failure and the kidney, the GLP-1 agonist for atherosclerotic events and weight — so their benefits add up rather than overlap. It is the mirror image of the gliptin problem: two drugs that complement each other rather than duplicate.

And the two I would reach for last — sulfonylureas and insulin. Here my bias is openly mechanistic. Almost everything above is designed to lower the demand for insulin — to improve sensitivity so the pancreas has less to do. A sulfonylurea such as gliclazide does the opposite: it whips the beta cells into secreting more insulin regardless of the underlying resistance, at the price of hypoglycaemia and weight gain — and, over time, of the fastest loss of glycaemic durability of the common oral agents. To my mind that is flogging a tiring horse rather than lightening its load. Reassuringly, this is now the direction the guidelines themselves are travelling: agents with heart, kidney and weight benefits are preferred, and sulfonylureas are increasingly something to limit or de-prescribe. Indeed, in intensive dietary programmes they are typically the first drug withdrawn. Where they survive, it is often on grounds of cost — which returns us to the theme of this whole piece.

Insulin I regard as the option of genuine last resort in type 2 diabetes — not a tool to reach for early simply to make the numbers look tidier; current guidance itself prefers a GLP-1 receptor agonist to insulin where there is no true insulin deficiency. Adding insulin to an already insulin-resistant, hyperinsulinaemic body tends to drive fat storage, working against the goal. But this is where I hold the line carefully: insulin is essential and must never be withheld when it is truly needed — in genuine insulin deficiency, in severe or symptomatic hyperglycaemia, and to protect against ketoacidosis and end-organ damage. Used briefly and deliberately it can even rest an exhausted pancreas. The error I want to avoid is reaching for it first, before the workload has been addressed.

And even where insulin is required — including type 1 — the levers in this article are not wasted. Building muscle, shedding any visceral fat and lowering the carbohydrate load all make the body more insulin-sensitive, which is why many insulin-dependent patients (and the diabetes nurses who look after them) report that both the dose and the weight come down when they train and eat this way. Because injected insulin carries its own costs — weight gain, and the unphysiological way subcutaneous dosing floods the periphery — I would want the lowest effective dose whatever form of diabetes I had. Even if I ended up insulin-dependent myself, that is the approach I would pursue: the smallest dose that keeps me safe, reached carefully with my team, and never by withholding insulin my body genuinely needs.

Remember what carbohydrate is doing that fat and protein largely are not: it is carbohydrate that forces the big insulin surges. Build the plate accordingly, and much of the pharmacology becomes a smaller job.

None of this is fringe. Contemporary programmes such as Virta Health, and Gary Taubes's history Rethinking Diabetes, make a similar case: fix the diet, take the load off, and much of the medication — sulfonylureas and insulin first — can often be peeled back under supervision. My own emphasis simply sits one layer deeper: on visceral fat, muscle and long-term healthspan rather than the glucose number alone.

How I would investigate — and keep score

Cardiovascular disease is the leading cause of death in type 2 diabetes, so in a cardiologist's hands the work-up leans there — on defining the arterial risk, not merely tracking glucose. NICE largely defines the problem by the glucose number; I would want to see the disease itself. For cardiovascular risk I would image the coronary arteries directly with CT coronary angiography (CTCA), an approach shown in trials to cut heart attacks by changing the treatment people go on to receive. To that I would add the fat attenuation index (FAI), a CTCA-derived read-out of inflammation in the fat hugging the coronary arteries, which carries prognostic information well beyond the plaque you can already see. For metabolic risk I would quantify the visceral fat on a low-dose CT and read the glucose pattern with a CGM rather than a once-a-year HbA1c.

This is not abstract. If a scan showed soft plaque in my left main stem or proximal LAD, or an FAI implying something like a twenty per cent eight-year mortality, a reassuring glucose would be cold comfort — I would want the disease itself measured, and then driven down.

For the day-to-day, I would hand much of the tracking to the patient — but not as a flat dashboard of equal numbers, which mostly breeds anxiety and tempts people to tinker with their own treatment. I would rank the measurements by how much they actually mean. At the top sits the waist (a tape measure; keep the waist-to-height ratio under 0.5), the closest home proxy for visceral fat and the marker that defines success. Next comes functional strength — a simple grip dynamometer will do — because metabolic health and longevity track muscle strength more tightly than muscle mass, and strength is what protects you as the weight comes off. Then the CGM, with SNAQ logging meals against the curve, superb for changing behaviour week to week. Bioimpedance scales come last: useful for engagement, but a flattering correlation with a DEXA scan is not the same as agreement, so I would not let them drive decisions.

One measure I would not treat as diabetes-specific at all is blood pressure — it is the first of the four pillars of cardiometabolic health, and it matters for everyone. A week of averaged home readings beats any single clinic measurement; in diabetes, where so much cardiovascular and kidney risk rides on pressure, I would simply track it especially closely. Every one of these is a sub-£100, no-prescription home device; I keep the current list on the VAT Trap toolkit page.

The foundation for treating plaque — in diabetics and non-diabetics alike — is lowering LDL cholesterol and ApoB, so for anyone with established disease that comes first. I would not settle for a normal-looking LDL; I would drive ApoB — the truest count of atherogenic particles — to a stringent target of below 0.65 g/L (65 mg/dL), reaching first for a statin, then adding ezetimibe, and moving to a PCSK9-inhibitor injection if the target were still missed. That last step is a big escalation, rightly reserved for high risk — but on the NHS it is, for now, largely confined to people who have already had a first or second cardiovascular event. The whole point of defining my risk by imaging is to earn the right to use these bigger weapons where they belong: before a first event, not only after it.

With the lipids driven down, I would turn to what else the imaging reveals. Because the FAI is a measure of coronary inflammation, it hands me a target I could not otherwise see: I would want to confirm that the inflammation driving much of the residual, modifiable risk is genuinely suppressed. If it were not, I would add an anti-inflammatory — low-dose colchicine has the strongest evidence in stable coronary disease.

For anyone whose CTCA and FAI flag high risk, I would then track whether the disease is regressing — a low-dose visceral-fat CT to follow the fat, and a repeat CTCA every two to three years to confirm the plaque is stabilising rather than advancing. Almost none of this sits within current NHS or NICE provision; it is, unapologetically, what I would choose for myself.

One clarification the table cannot make on its own, because it matters: the medicines are not interchangeable across the two columns. SGLT2 inhibitors are heart-failure, kidney and diabetes drugs — their proven job is cardiorenal protection and glucose control, not the prevention of diabetes. I would not reach for one in simple prediabetes unless heart failure or kidney disease were already in play, or the glucose were climbing fast. A GLP-1 agonist is the more logical tool when visceral fat is the problem, and its cardiovascular benefit now reaches people with obesity and heart disease even before diabetes appears — used with diet and strength training, not instead of them. Metformin, likewise, has a genuine evidence base for delaying progression in higher-risk prediabetes, though lifestyle beat it head to head.

Key Takeaways

1. NICE optimises value across a population; a "money-no-object" plan for one person can reasonably be earlier and more layered.

2. Food first: cut carbs and hidden sugars, use a CGM to flatten your meal responses, and defuse spikes with fat, protein, fibre and food order.

3. Move in the cracks of the day (7,000–10,000 steps average) and build muscle (~25–30 g protein/meal + resistance training) — muscle is a glucose sink.

4. Protect sleep — poor sleep raises cortisol and worsens insulin resistance. A sleep-tracking ring (Oura, or subscription-free options like Ultrahuman, RingConn, Samsung Galaxy Ring) makes it a target you can manage.

5. Escalate on visceral fat, not the scales. Metformin first; then, if needed, a GLP-1 agonist and/or an SGLT2 inhibitor — with a gliptin as an alternative to (not an addition to) a GLP-1 drug.

6. Know which diabetes you have. This plan targets the visceral-fat-driven form; the lean, insulin-deficient form (type 1, LADA) still benefits from these levers but cannot be beaten without insulin.

VAT Trap