Weight Loss.

To lose stored fat you have to run a sustained energy deficit — fewer calories in than out, for long enough. This is not in scientific dispute: in metabolic-ward studies where every calorie is weighed, a deficit produces fat loss every time, and no macronutrient trick reverses it. Anything promising fat loss without a deficit is wrong on physics.

‘Eat less, move more’ is true but nearly useless as an instruction, because both sides of the equation push back. Cut intake and the body burns less (metabolic adaptation) and raises hunger (appetite hormones). Add exercise and you unconsciously eat more and move less the rest of the day. The accounting is correct; the inputs are not fixed levers you fully control.

Kevin Hall’s validated dynamic model (the engine behind the NIH Body Weight Planner) puts the long-run exchange rate at roughly 24 kcal/day of sustained change per ~1 kg of eventual body weight — about half the change reached in a year, ~95% within three. Loss is slow and self-limiting, not linear. The old ‘3,500 kcal = 1 lb, so a 500/day deficit loses 1 lb/week forever’ rule was formally dropped in 2012 because it ignores this and badly overpredicts loss. Evidence ●●●●●

A smaller body burns fewer calories, so a fixed calorie cut becomes a smaller and smaller deficit as weight falls — until intake and your new, lower expenditure re-balance and weight stops moving. The plateau is not a failure of effort; it is the arithmetic of energy balance reaching a new equilibrium. This is why every diet, whatever its name, tends to peak around six months.

At the plateau, the metabolic slowdown is modest (<200 kcal/day) while appetite has climbed 400–600 kcal/day since the start. Per kilogram lost, expenditure falls only ~20–30 kcal/day but hunger rises ~100 kcal/day — roughly a 3–5× stronger signal. Intake creeping back up, largely below conscious awareness, drives most plateaus. Evidence ●●●●○

Each gram of stored glycogen is held with ~3–4 g of water. A well-fed adult carries ~400–500 g of glycogen binding ~1.5–2 kg of water, so the fast ‘whoosh’ of early low-carb or fasting weight loss is largely water, not fat — and it returns when carbs do. A classic estimate: ~70% of the first few days’ loss is water and glycogen, ~25% fat, ~5% protein.

Daily body weight swings ~±1–2 kg from fluid, food in transit, sodium, carbohydrate, and hormonal water retention — none of it fat. Weigh at the same time under the same conditions and judge a weekly or multi-week average, not any single morning. A stall on the scale with a shrinking waist is fat loss hidden by water.

After weight loss, resting metabolism falls by more than the loss of body mass predicts — the extra, unexplained slowdown is called adaptive thermogenesis. It is real and detected in the large majority of studies. It is leptin-mediated: restoring leptin to pre-diet levels in the lab reverses much of the drop, evidence the body is actively defending its former weight.

For a typical moderate weight loss the extra slowdown is modest — on the order of 50–180 kcal/day (~6–7% below predicted), largest during active dieting and often shrinking once weight stabilises. The famous ~500 kcal/day figure comes from the Biggest Loser follow-up (n=14, extreme rapid loss) and is an outlier, not the norm. Better long-term studies find it does not reliably predict who regains. Evidence ●●●○○ on magnitude

In a landmark study (Sumithran 2011), people lost ~13.5 kg on a 10-week diet. Their satiety signals (leptin, peptide YY, CCK, amylin, insulin) fell and hunger signal ghrelin rose — a coordinated ‘eat more’ state. Crucially, at 62 weeks these changes had still not normalised, and subjective hunger remained significantly above baseline. A year on, people were hungrier than before they started, despite being lighter. Evidence ●●●●○

Leptin, made by fat cells, is the brain’s ‘fuel-gauge’. When you lose fat, leptin falls, and the hypothalamus reads falling leptin as a starvation threat — a system evolution built far more machinery to defend against than against overeating. It responds by ramping hunger, raising the reward value of food, and trimming energy spent. This is physiology, and it operates below conscious awareness.

Whether the body defends a fixed ‘set-point’ weight is genuinely contested. The set-point view (active feedback toward a target) explains why dieting triggers hunger and slows metabolism. The settling-point view says weight simply lands wherever intake and expenditure balance, with no target — which better explains decades of population-wide weight gain. A middle model (dual-intervention-point) proposes the body strongly defends a lower and an upper boundary but drifts freely between them.

You do not need to resolve the debate. Both camps agree on the actionable part: the body resists loss more than it resists gain, that resistance persists, and it is countered by ongoing habits (activity, protein, monitoring) rather than a one-off ‘diet’ with an end date. Evidence ●●●○○ (model is debated; the defended-loss phenomenon is not)

BMI cannot tell muscle from fat, and it is blind to where fat sits. It was devised as a population statistic, never for individual diagnosis, and its cut-offs came from older, largely white cohorts. One analysis estimated ~75 million US adults are mis-classified as healthy or unhealthy by BMI category alone. It is a rough screen, not a verdict.

Visceral (abdominal) fat is the metabolically dangerous kind, and simple waist measures capture it better than BMI. The most useful rule: keep your waist under half your height (waist-to-height ratio < 0.5). 0.5–0.59 signals increased central fat; ≥0.6 is high. In a 31-study meta-analysis, waist-to-height out-predicted both waist alone and BMI for cardiometabolic risk. Evidence ●●●●○

Absolute waist cut-offs for raised risk: roughly >94 cm (men) / >80 cm (women) for increased risk, and >102 / >88 cm for substantially increased risk — with lower thresholds (~90 / 80 cm) for South and East Asian backgrounds, who carry higher risk at lower BMI.

DEXA is the practical reference (error ~±1–3%). Home bio-impedance scales are the least reliable (~±4–10%, and thrown off by hydration, food, and exercise); skinfold calipers sit in between (~±3.5–5%, operator-dependent). Treat any single body-fat reading as a rough trend line, not a precise number.

Weight is one signal among many. Waist inches, blood pressure, fasting glucose / HbA1c, lipids, resting heart rate, strength and stamina, how clothes fit, and energy and sleep all track real progress — sometimes while the scale is flat. Fitness in particular predicts mortality strongly and partly independently of weight.

Guidelines converge on ~0.5–1 kg (1–2 lb) per week, from roughly a 500–600 kcal/day deficit — fastest at first, then slowing as the deficit shrinks. Aggressive crash loss buys a bigger early number but a larger share of muscle and a stronger metabolic and appetite rebound. Slower is not just safer; it holds better.

You do not need an ‘ideal weight’ to get most of the health benefit. A sustained 3–5% loss already improves triglycerides, blood glucose and diabetes risk; 5–10% meaningfully lowers blood pressure and improves lipids and insulin sensitivity. In the Diabetes Prevention Program, ~7% loss plus activity cut progression to type 2 diabetes by 58% — beating metformin. Evidence ●●●●●

DIETFITS (2018) randomised 609 adults to a healthy low-fat or healthy low-carb diet for a year, both coached toward whole foods. Result: −5.3 kg vs −6.0 kg — not a significant difference. It had pre-registered that a gene pattern and insulin level would predict who did best on which diet. Neither did. Individual results ranged from −30 kg to +10 kg within both arms, driven by adherence, not assignment.

Pooled across dozens of trials, low-carb vs low-fat differ by roughly ~1 kg at 12 months — clinically trivial next to the 5–10 kg swings adherence produces. The 2007 A-TO-Z trial gave Atkins a modest early edge over the weakest comparator, but that edge largely vanished once diet quality was equalised. The honest headline: pick a high-quality pattern you can sustain. Evidence ●●●●●

This is where low-carb is under-sold, not over-sold. In type 2 diabetes it produces the largest short-term drops in blood sugar of any dietary approach — one continuous-care program cut HbA1c from 7.6% to 6.3% with ~12% weight loss and got most patients off insulin. It also reliably lowers triglycerides and raises HDL. Caveat: in lean, metabolically healthy people, LDL cholesterol can rise sharply (the ‘lean-mass hyper-responder’ pattern) — reversible by adding carbs back, but worth monitoring. Evidence ●●●●○ for glycaemia

Both are excellent patterns whose headline benefits are not primarily about the scale. Mediterranean eating is strongest for cardiovascular risk and only modestly slimming (~1.8 kg vs control, and mainly with calorie restriction). DASH was built for blood pressure (it lowers it largely independently of weight). Choose them for those endpoints; expect weight loss only if you also hold a deficit.

Low-carb diets move the scale fast in week one — but that is mostly glycogen and water (see section 1), not a metabolic head start. Motivating, but don’t mistake it for fat.

Across the whole literature, a durable fat-loss diet looks unremarkable: protein-forward, fiber-rich, built on mostly whole foods, in a modest deficit, following a pattern you can actually maintain. Then layer the specific-job logic: diabetes or high triglycerides → lean lower-carb; high blood pressure → DASH; cardiovascular risk → Mediterranean.

Protein helps fat loss three independent ways. Satiety: it raises the same fullness hormones the drugs target (GLP-1, PYY, CCK) and lowers ghrelin — in one trial, raising protein to 30% of calories cut spontaneous intake by ~440 kcal/day with no conscious restriction. Thermic effect: 20–30% of protein calories are burned just digesting it, vs ~5–10% for carbs and ~0–3% for fat. Muscle: it preserves lean mass in a deficit. Evidence ●●●●●

Roughly 1.6 g/kg body weight per day is the floor for holding muscle while training; pushing to 1.6–2.2 g/kg (or up to ~2.3–3.1 g/kg of lean mass) helps most when the deficit is steep, you’re lean, older, or on a GLP-1 drug. Spread across 3–4 meals of ~25–40 g. Powder is just convenient food to hit the target, not a fat-burner.

High protein does not harm healthy kidneys — a meta-analysis of 28 trials found no effect on filtration rate. The transient rise sometimes cited as ‘damage’ is normal, reversible adaptation. The genuine exception is pre-existing kidney disease, where protein should be managed with a clinician.

Viscous fibers form a gel that slows gastric emptying and blunts glucose spikes; fermentable fibers feed gut bacteria that produce short-chain fatty acids, which stimulate the satiety hormones GLP-1 and PYY. Net effect: fuller on fewer calories. Adding ~14 g/day of fiber is associated with ~10% lower energy intake and ~1.9 kg loss over four months.

Target is ~25 g/day (women) to ~38 g/day (men); typical intake is only ~15 g. It is one of the cheapest levers available. In one trial, the single instruction ‘eat ≥30 g fiber/day’ delivered ~78% of the weight loss of a far more complex prescription. Evidence ●●●●○

The strongest causal evidence here is a metabolic-ward study (Hall 2019): 20 adults ate freely from ultra-processed vs minimally-processed diets matched for calories, sugar, fat, salt and fiber. On the ultra-processed diet they ate ~500 kcal/day more and gained ~1 kg in two weeks; on whole foods they lost the same. The driver was eating rate and energy density, not the ‘processing’ label — a follow-up showed a diet that was >80% ultra-processed but low in energy density still caused weight loss.

Energy density (calories per gram) is the master dial for how much you eat by volume. People eat a fairly constant weight of food, so water- and fiber-rich foods let you eat more while taking in less. In a year-long trial, the group told to add water-rich vegetables ate ~25% more food by weight and lost more weight. Evidence ●●●●○

Fasting and time-restricted eating do cause weight loss — by cutting total intake (fewer hours means fewer snacks and less late-night grazing). But when calories are matched head-to-head against ordinary daily restriction, they are not superior (TREAT; a year-long alternate-day trial; a calorie-matched trial that added an eating window and saw no extra benefit). The value is behavioural simplicity, not metabolic magic. Evidence ●●●●○

Eating ‘6 small meals to stoke metabolism’ is false — the thermic effect scales with total calories, not meal count. Skipping breakfast does not wreck metabolism; in trials, people told to eat breakfast simply ate ~260 kcal/day more. One real caution: a fasting window with no protein target or resistance training can cost extra muscle — fixable by prioritising both.

The defining problem is poor compensation. Given 450 kcal/day as jelly beans, people cut back elsewhere and held weight; given the same as soda, they did not, and gained. Liquid calories are added on top of what you’d eat anyway. In a masked children’s trial, swapping a daily sugary drink for a sugar-free one meant ~1 kg less weight over 18 months. Evidence ●●●●○

100% juice and smoothies are liquid sugar with better nutrition and the same low-satiety problem — eat the fruit instead. On diet drinks, the randomised trials favour swapping sugar for water or a diet drink for weight loss; the scary headlines come from observational studies distorted by reverse causation. Hierarchy: water > diet drinks > sugar. Diet drinks are a fine bridge off sugar, not a health food.

Alcohol is 7 kcal/g (nearly as dense as fat) and, like sugary drinks, poorly compensated. Worse, the body prioritises burning it and pauses fat oxidation ~36% while it does. And it disinhibits eating — a drinking occasion adds ~256 kcal of food-plus-drink on average, with no later cutback.

Population data on moderate drinking and body weight are genuinely mixed and confounded — so this is not fear-mongering. But for anyone actively in a deficit, the mechanisms are unambiguous: extra un-counted calories, suppressed fat-burning, and weakened restraint all pull against you. It makes the math harder, not impossible. Evidence ●●●○○ for population weight; strong for the mechanisms

Exercise alone typically produces only ~1.5–3 kg of loss over months — roughly 30–40% of what the calories-burned arithmetic predicts — because the body compensates by raising appetite and cutting spontaneous movement. Meaningful loss from exercise needs a lot of it: ACSM puts the threshold around 225–420 min/week. Diet drives the deficit far more reliably. Evidence ●●●●○

Doubly-labelled-water work (Pontzer) suggests the body partly ‘constrains’ total energy expenditure — you burn less extra than a treadmill readout claims (roughly 25–30% is offset on average). Whether expenditure is strictly capped is actively debated, with 2025 data on both sides. The safe statement: exercise burns real calories, just fewer net than advertised.

The volume of activity that barely moves the scale during weight loss is almost exactly what predicts maintaining it. People in the National Weight Control Registry — who’ve kept ~30 kg off for years — average about an hour a day of activity (~2,500–2,800 kcal/week), mostly walking, confirmed by accelerometer. Evidence ●●●●○ (partly self-selected; caveat below)

Most maintenance evidence is observational — people who choose to move more keep weight off better, which is not the same as proving the exercise caused it. Leaner maintainers also find activity easier. The association is robust and biologically plausible, but state the causal magnitude with humility.

Aerobic exercise preferentially strips visceral fat — the dangerous kind — often with little scale movement, and it improves blood pressure, lipids and insulin sensitivity largely independent of weight. Cardiorespiratory fitness is one of the strongest mortality predictors we have: each 1-MET higher is ~13% lower all-cause mortality, and being fit blunts much of the risk of a higher BMI. Evidence ●●●●●

Matched for time and effort, interval training and steady-state produce similar fat loss — HIIT’s real advantage is finishing in ~40% less time. Ignore two persistent myths: the ‘fat-burning zone’ (low intensity burns a higher percent of fat but fewer total calories) and the ‘afterburn’ (EPOC adds only ~50–200 kcal). Pick the cardio you’ll actually do.

Left unaddressed, roughly 20–30% of the weight lost in a calorie deficit is fat-free mass, not fat — higher with steeper deficits, rapid loss, and in leaner people. Losing some lean tissue is normal (a smaller body needs less), but the proportion is very much within your control. Evidence ●●●●●

Adding resistance training during weight loss preserves fat-free mass (pooled effect ~+0.4 SD) and increases fat loss, with essentially no change to total scale weight — it changes what you lose. In one hard-training, high-protein (2.4 g/kg) study, young men actually gained 1.2 kg of muscle while losing 4.8 kg of fat in a steep deficit. Keeping the deficit moderate (roughly ≤500 kcal/day) protects muscle further. Evidence ●●●●○

On semaglutide and tirzepatide, roughly 25–40% of weight lost is lean mass (the tirzepatide body-composition substudy came in near 25% — about the same fraction as any comparable diet). The drugs don’t preferentially destroy muscle; the risk is that blunted appetite makes it easy to under-eat protein exactly when you need it most, and many users are older and sedentary.

Same two levers: progressive resistance training plus deliberately protein-first meals (~1.2–1.6 g/kg and up). Dedicated trials in GLP-1 users are still reporting, so current advice extrapolates from the strong deficit literature — but the direction is not in doubt. Evidence ●●●○○ in GLP-1 users specifically (emerging)

The strongest evidence is randomised: giving chronically short sleepers about one extra hour a night cut their intake by ~270 kcal/day (Tasali 2022, measured by the gold-standard doubly-labelled-water method), and they lost weight over two weeks without dieting. Mechanistically, short sleep lowers the satiety hormone leptin, raises hunger hormone ghrelin, and makes junk food more rewarding. Evidence ●●●●○

Dieting on short sleep wrecks the quality of loss. In a controlled trial, the same calorie deficit under 5.5 vs 8.5 hours of sleep cut fat loss by 55% and increased muscle loss — the fat share of weight lost fell from 56% to 25%. Same diet, same scale drop, opposite body composition. Aim for ~7–9 hours.

Cortisol clearly drives central fat only when pathologically high (Cushing’s syndrome — rare, ~13 per million). For the everyday stress most people have, the cortisol–abdominal-fat correlation is small (r ≈ 0.10, ~1% of the variance) and partly runs backwards, since carrying more fat itself raises cortisol turnover. ‘Cortisol belly’ as a mass diagnosis overstates the evidence. Evidence ●●○○○ for everyday stress → fat

Where stress genuinely touches weight is behavioural: it shifts a susceptible subset (~35–40% of people) toward palatable high-fat, high-sugar food — while a similar fraction eat less, so ‘stress makes you fat’ is not universal. Cortisol-blocking weight-loss supplements are unsupported: the marketers of CortiSlim were sued by the US FTC and settled for $12 million.

The single strongest behavioural predictor of weight loss. In a landmark review, all 15 studies of food logging found it linked to more loss, in a clear dose-response — more consistent tracking, more loss. Frequent self-weighing tracks with success too (in a 10,000-user cohort, only daily weighing was associated with loss). The point is awareness and early course-correction, not surveillance. Evidence ●●●●●

New habits take a median ~66 days to feel automatic (range 18–254), and missing one day doesn’t reset the clock — so slow progress is normal, not failure. Structured programs reliably beat going it alone: intensive programs (≥12–16 contacts) roughly double the odds of a ≥5% loss, and longer programs beat shorter ones. For environment design, smaller bowls and serving spoons beat the over-hyped ‘small plate’.

They mimic gut incretin hormones. Semaglutide (Wegovy for weight, Ozempic for diabetes) is a GLP-1 receptor agonist; tirzepatide (Zepbound / Mounjaro) is a dual GIP + GLP-1 agonist. Both act on appetite centres in the hypothalamus and hindbrain to cut hunger and food intake, slow gastric emptying, and boost glucose-dependent insulin release. The appetite effect is the main driver of weight loss. Both are once-weekly injections; liraglutide (Saxenda) is an older, once-daily GLP-1.

Over ~68–72 weeks, mean weight loss runs about 15% for semaglutide 2.4 mg (STEP 1), up to ~21% for tirzepatide 15 mg (SURMOUNT-1), and ~8% for liraglutide. Head-to-head, tirzepatide beat semaglutide (−20.2% vs −13.7%, SURMOUNT-5). About half of people on the top tirzepatide dose lose ≥20% — results previously seen only with surgery. Evidence ●●●●●

Semaglutide cut major cardiovascular events by 20% in a 17,600-person trial of people with heart disease and obesity (SELECT), and tirzepatide is now approved for moderate-to-severe obstructive sleep apnoea. The benefits extend past the scale.

Approved for adults with BMI ≥30, or ≥27 with a weight-related condition (e.g. hypertension, type 2 diabetes, dyslipidaemia), plus diet and activity — and for adolescents 12+ with obesity. Semaglutide also carries a cardiovascular-risk-reduction indication, tirzepatide a sleep-apnoea one. These are the same BMI thresholds as older anti-obesity drugs.

Wegovy escalates over 16+ weeks, four weeks per step: 0.25 → 0.5 → 1.0 → 1.7 → 2.4 mg weekly, then holds at 2.4 mg (a higher 7.2 mg option now exists for tolerant patients). Ozempic is the same molecule for diabetes and tops out at 2 mg, so off-label weight use underdoses relative to Wegovy. An oral semaglutide for weight has also been approved.

Steps up in 2.5 mg increments no sooner than every four weeks: 2.5 (a starter dose, not therapeutic) → 5 → 7.5 → 10 → 12.5 → 15 mg weekly. Approved maintenance doses are 5, 10 or 15 mg. The slow climb exists to limit gastrointestinal side effects during escalation.

The next wave, none yet approved for obesity: retatrutide, a triple GLP-1/GIP/glucagon agonist, hit ~24% in phase 2 and ~28% in early phase 3; orforglipron, an oral once-daily pill, ~12%; CagriSema (cagrilintide + semaglutide) ~22.7%; plus survodutide and once-monthly MariTide. Cross-trial comparisons are indirect — different populations and endpoints — and press-release figures precede full peer review. Evidence ●●●○○ (early / not yet approved)

Nausea (reported in ~14–58% across trials, commonly ~44% early), plus diarrhoea, constipation and vomiting. They are dose-dependent, peak during titration, and usually fade with time. They are the main reason people stop — but only ~1–4% actually discontinue for GI reasons. Slow titration, smaller meals and hydration help. Evidence ●●●●●

Gallstones and gallbladder disease are genuinely raised (~1.4×, higher with weight-loss doses and rapid loss — partly a consequence of fast weight loss itself). Pancreatitis is on the label, but pooled randomised data have not confirmed a real increase; the warning is precautionary. Severe, persistent abdominal pain warrants stopping and prompt review.

The black-box warning comes entirely from rodent studies, where these drugs caused thyroid C-cell tumours; human C-cells carry far fewer receptors and human causality is unproven, with large studies not showing raised thyroid cancer. Still, they are firmly contraindicated with a personal or family history of medullary thyroid carcinoma or MEN 2. Evidence ●○○○○ for human risk

Because they slow the stomach, there are rare reports of ileus (bowel stall) and occasional persistent gastroparesis, and retained stomach contents can raise aspiration risk under sedation. Tell any surgeon or anaesthetist you are on a GLP-1; current guidance individualises whether to pause it rather than always stopping.

Cardiovascular safety is favourable — semaglutide cut major cardiac events 20% (SELECT); tirzepatide is CV-neutral to positive. A suicidality signal was investigated and not confirmed, and regulators asked for the warning to be removed. ‘Ozempic face’ (facial hollowing) is just the normal fat loss of any rapid weight loss, not a specific drug effect. The real substantive issue is muscle loss — covered in the muscle section. Evidence ●●●●○

When treatment stops, the appetite biology reasserts itself. A year after stopping semaglutide, people regained about two-thirds of the lost weight (STEP 1 extension). In a tirzepatide withdrawal trial, those switched to placebo regained ~14% while those continuing kept losing; ~90% who continued held onto their loss vs ~17% who stopped. Blood pressure, glucose and lipid gains reverse in step. Evidence ●●●●●

This behaves like blood-pressure or cholesterol medication: it works while taken, and the numbers drift back when it stops — not because the patient failed, but because the underlying drive was being suppressed, not removed. Current practice treats anti-obesity medication as indefinite therapy paired with the lifestyle foundation, not a time-limited course.

US list prices run ~$1,000–1,350/month, but real cash prices have fallen sharply via manufacturer direct-pay programs — roughly $350–500/month for self-pay in late 2025. These figures move fast and vary by dose and country; treat any single number as a snapshot, not a fixed price.

US Medicare is barred by statute from covering drugs for weight loss (it can cover them for a separate approved indication such as cardiovascular risk or sleep apnoea); commercial coverage is patchy, with prior authorisation common. The mass market in cheaper compounded semaglutide/tirzepatide lost its legal footing when the FDA declared the shortages over in 2025 — and the agency has flagged wrong salt forms and dosing errors (some causing hospitalisation) in compounded products.

Placebo-subtracted weight loss, roughly: phentermine ~3–5% (a stimulant, labelled short-term, raises heart rate/BP); phentermine/topiramate (Qsymia) ~9–10% (best of the older orals, but carries an oral-cleft pregnancy risk and a restricted program); naltrexone/bupropion (Contrave) ~5% (boxed suicidality warning from the bupropion); orlistat ~3% (blocks fat absorption — hence the GI effects — OTC, and cut diabetes progression 37% over four years). Evidence ●●●●○

Metformin gives a modest but durable ~2–2.5%, best in insulin resistance, PCOS or antipsychotic-related gain. SGLT2 inhibitors lose ~2–3% (a urinary-calorie leak that plateaus as appetite compensates). Setmelanotide is dramatic but only for confirmed rare genetic obesity, at ~$400,000/year. None approaches the incretins for magnitude.

Weight-loss drugs have a long graveyard: fen-phen (heart-valve damage, pulled 1997), sibutramine (raised heart attack and stroke, 2010), lorcaserin (a cancer signal, 2020), and rimonabant (depression and suicidality, never approved in the US). The pattern isn’t that these drugs are inherently dangerous — it’s that appetite is wired into receptors that also govern heart valves, blood vessels and mood. Each failure is why modern drugs now face cardiovascular, cancer and psychiatric surveillance as routine.

Sleeve gastrectomy (now the most common operation) removes ~75–80% of the stomach; Roux-en-Y gastric bypass reroutes the gut. Neither is merely ‘a smaller stomach’ — both drive large, favourable gut-hormone changes (including the body’s own GLP-1). Total weight loss is roughly 25–30% (sleeve) and 30–35% (bypass), durable to 12 years, with strong diabetes remission and a long-term reduction in mortality, heart attacks and strokes. Evidence ●●●●●

The adjustable band is largely abandoned (weak, high re-operation). Duodenal-switch variants give the most weight loss (~35–45%) and the best diabetes remission, but the greatest nutritional risk. Endoscopic, incision-free options — endoscopic sleeve gastroplasty (~15–20%) and the temporary intragastric balloon (~10–12%, usually regained after removal) — are less invasive and lower-magnitude.

The 2022 professional-society update lowered thresholds to BMI ≥35 regardless of other conditions, or ≥30 with metabolic disease (and lower still for Asian populations) — the biggest expansion in the field’s history, though insurance often lags behind it.

Modern bariatric surgery is remarkably safe (30-day mortality ~0.1–0.3%), but it demands lifelong vitamin and mineral supplementation and monitoring (B12, iron, calcium, vitamin D, protein), and carries risks of dumping syndrome, ulcers, gallstones, bone loss, and a raised rate of alcohol-use disorder after bypass. The tradeoff is durability: no drug yet matches surgery’s decade-plus staying power without continued treatment.

Psyllium has the best case — roughly ~2 kg over months at ~10 g before meals, plus real satiety and glucose benefits, and a good safety profile (take with plenty of water). Glucomannan is weaker (the best meta-analysis was null) and carries a genuine choking/obstruction hazard — never dry, and tablets are best avoided. Evidence ●●●○○ psyllium; ●●○○○ glucomannan

Not a fat-burner — a convenient way to hit a protein target that aids satiety and protects muscle in a deficit (see the protein section). If you already get enough protein from food, powder adds only convenience. Graded strong, but as a food tool, not a supplement with independent magic. Evidence ●●●●○

The nickname is marketing, not mechanism. Berberine works mainly through AMPK, not the GLP-1 receptor, and its effect on body weight in trials is not statistically significant (its real signal is a modest glucose and lipid improvement, from low-quality studies). It is a mild metabolic optimiser, not a weight-loss drug — and shares none of semaglutide’s ~15% effect. Evidence ●●○○○ for weight

A real but tiny thermogenic effect — on the order of ~1 kg at best, mostly caffeine-driven, and tolerance erodes it. The more important note is safety: brewed green tea is fine, but concentrated green-tea extract can cause liver injury. Net: not a meaningful weight-loss tool. Evidence ●●○○○

Effectively null or trivial, and some unsafe: apple cider vinegar (hype; enamel/throat harm), chromium (safe but ~doesn’t work), Garcinia cambogia (null — and a cause of acute liver failure), CLA (~1 kg but worsens HDL and insulin), raspberry ketones (no human evidence, pure mouse-study extrapolation), bitter orange/synephrine and yohimbine (cardiovascular risk), capsaicin and L-carnitine (trivial). Evidence ●○○○○

The real hazard isn’t weak efficacy — it’s that supplements aren’t pre-approved for safety, and regulators have found hundreds of weight-loss products spiked with hidden pharmaceuticals (most often sibutramine, a drug pulled for causing heart attacks and strokes, plus a former laxative suspected of being carcinogenic and even prescription antidepressants). With a ‘proprietary blend’ you cannot know the dose, the true contents, or whether it contains a withdrawn drug.

The idea (Simpson & Raubenheimer) that we regulate protein tightly and keep eating protein-poor food until we hit our target — over-consuming calories on the way — has real experimental support (in a controlled trial, diluting protein raised total intake ~12%). It neatly explains why protein-poor ultra-processed diets drive overeating. A well-supported contributing mechanism, appropriately hedged by its own authors. Evidence ●●●●○

The core (associated with Dr Gabrielle Lyon) is legitimate and under-appreciated: muscle is the main site of glucose disposal (~70–80%), resistance training improves insulin sensitivity, and strength and fitness are strong mortality predictors. The caveats: raw muscle mass is a much weaker, non-causal predictor than strength, and some influencer protein targets (~1 g per pound) exceed what the evidence requires. Solid substance, mild over-packaging.

The genuine finding is that glucose responses to the same meal are highly individual — which really does undercut one-size-fits-all glycaemic advice. But the leap to ‘flatten your spikes and you’ll lose weight’ is not supported by any randomised trial in healthy people; spikes in metabolically healthy people are normal, device error (~10–13%) can be larger than the ‘spike’ you’re chasing, and clinicians warn about fostering food anxiety. Interesting self-experiment; not a proven weight tool. Evidence ●●○○○ for weight

Fiber, protein and fermented foods genuinely do raise your own GLP-1 — but the rise is small and brief, on the order of ~1,000× weaker than a GLP-1 drug, which is engineered for week-long, high-level receptor activation. Eating more fiber and protein is good advice; it is not a stand-in for semaglutide. Supplements branded ‘nature’s Ozempic’ or ‘GLP-1 boosters’ are the clearest hype in this list, now drawing consumer-fraud lawsuits. Evidence ●○○○○ as a drug substitute

Every load-bearing number comes from a named trial, meta-analysis, guideline or regulatory document, with the evidence dots reflecting study design, size, replication and whether findings survived scrutiny. Figures are group averages; individual results vary widely. Doses and thresholds are typical adult figures from the literature, not instructions.

Drugs: the STEP program and SELECT (semaglutide); SURMOUNT 1–5 and SURMOUNT-OSA (tirzepatide); the SCALE trials (liraglutide); the pivotal trials for phentermine/topiramate, naltrexone/bupropion, orlistat (XENDOS) and metformin (DPP). Surgery: STAMPEDE, the Swedish Obese Subjects cohort, and 12-year gastric-bypass outcomes. Biology & behaviour: Sumithran 2011 (appetite hormones), Fothergill 2016 (metabolic adaptation), Kevin Hall’s dynamic model and 2019 ultra-processed-food trial, DIETFITS 2018, Tasali 2022 (sleep), the National Weight Control Registry, Burke 2011 (self-monitoring), and the Ashwell waist-to-height meta-analysis.