The metabolic picture

Chapter 5 — How Fat Loss Actually Works

Energy balance is true, and on its own it is nearly useless. Both halves of that sentence need saying, because the first half is what every diet book asserts and the second half is what every honest reader discovers within a fortnight.

Fat is lost when energy taken in is less than energy spent, for long enough. Nobody has ever demonstrated otherwise, and the exceptions readers are told about — a slow thyroid, a broken metabolism, a body that "holds on" to fat — turn out on measurement to be small effects sitting inside a much larger accounting. So far so unhelpful. The problem is that the accounting is not fixed. Energy expenditure is not a constant that you subtract from; it is a variable that moves when you move it.

The most careful quantitative model of this comes from Kevin Hall and colleagues, who set out what actually happens when a population's intake changes (Hall et al., Lancet 2011). Three findings from that work should permanently alter how you read weight-loss arithmetic. First, adult body weight responds to a change in intake slowly, with a half-time of about one year — meaning that if you eat 200 kcal less per day starting today, you will be roughly halfway to your new steady weight in twelve months, not in six weeks. Second, the excess intake that produced the modern obesity epidemic corresponds to a persistent average imbalance of about 30 kJ per day, some 7 kcal — the energy in two cashew nuts. Third, and in the opposite direction, the maintenance energy gap — the extra food a heavier population must eat every day simply to stay heavy — is about 0.9 MJ, some 215 kcal per day.

Read together, those three numbers demolish the rule most readers were taught: that 3,500 kcal equals half a kilogram of fat, so a 500 kcal daily cut yields half a kilogram a week indefinitely. It does not, because as you get lighter you spend less — less to carry, less to heat, less to move. The deficit you designed on day one has shrunk by month six without you changing a thing. This is not sabotage; it is subtraction. The useful conclusion is not that energy balance is a myth, but that small persistent changes beat large temporary ones, and that the honest unit of a diet is not the week but the year.

Adherence beats the macronutrient argument, and this is settled

Three trials settle the question that occupies most of the diet aisle, and the book's entire licence to be a Kerala cookbook rests on them.

DIETFITS randomised 609 adults to a healthy low-fat or a healthy low-carbohydrate diet for twelve months, with 481 (79%) completing (Gardner et al., JAMA 2018). Weight change was −5.3 kg on low-fat and −6.0 kg on low-carbohydrate, a between-group difference of 0.7 kg with a confidence interval running from −0.2 to 1.6 kg — which is to say, a difference indistinguishable from none. The trial had been designed to find the people for whom one diet would suit better than the other, and it looked hard: there was no diet-by-genotype interaction (P=0.20) and no diet-by-insulin-secretion interaction (P=0.47). Neither the genetic pattern nor the insulin response predicted which arm a person would do well in. What the trial did find, within each arm, was a spread of individual outcomes running from about −25 kg to +5 kg. The variation between people inside one diet dwarfed the variation between the diets.

POUNDS LOST put 811 overweight adults through four different macronutrient combinations — fat at 20 or 40% of energy, protein at 15 or 25%, carbohydrate at 65 or 35% — and followed them for two years (Sacks et al., N Engl J Med 2009). At six months every group had lost about 6 kg. At two years every comparison was statistically indistinguishable, with P>0.20 throughout: 3.0 versus 3.6 kg for the two protein levels, 3.3 kg for both fat levels, 2.9 versus 3.4 kg for the two carbohydrate levels. Then comes the sentence that should be printed on the inside of every kitchen cupboard: attendance was strongly associated with weight loss, at 0.2 kg per counselling session attended. Turning up predicted the result. The macronutrients did not.

Johnston's network meta-analysis pooled 48 randomised trials and 7,286 participants (Johnston et al., JAMA 2014). Low-carbohydrate diets produced 8.73 kg at six months and 7.25 kg at twelve; low-fat diets 7.99 kg and 7.27 kg. The largest gap between any two named commercial diets was 1.71 kg, Atkins over Zone, at six months — a difference that had evaporated by a year. The authors' conclusion, verbatim: "This supports the practice of recommending any diet that a patient will adhere to in order to lose weight."

That is a licence, and it should be read as one. If adherence is the active ingredient, then building a diet out of food you already love, already know how to cook, and will still be cooking in five years is not a soft option or a cultural concession. It is the mechanism.

What "adherence" actually means here Not willpower. Not discipline. Adherence in these trials meant still doing it at twelve months — still shopping that way, still cooking that way, still turning up. The single strongest predictor in POUNDS LOST was attendance at a group session, a behaviour with no metabolic content whatsoever. Design for continuation, not for intensity.

Protein: the one structural edit that matters most in a rice cuisine

For the general reader in a deficit, the target is 1.2–1.6 g of protein per kilogram of body weight per day, distributed as roughly 25–30 g per meal (Leidy et al., Am J Clin Nutr 2015). Higher-protein energy restriction produces greater weight loss, greater fat-mass loss and better preservation of lean tissue than lower-protein energy restriction in controlled feeding, along with reductions in triglycerides, blood pressure and waist circumference.

Be honest about the satiety claim, though, because it is routinely oversold. Acute trials show a modest effect — greater reported fullness, higher satiety hormone concentrations — but they do not show reduced intake at the next eating occasion. Long-term findings are mixed, and the reviewers' own explanation for the inconsistency is compliance rather than physiology. Protein helps; it does not do the work by itself.

Two boundaries. A lean, resistance-trained reader in an aggressive deficit needs more — 2.3–3.1 g per kilogram of fat-free mass, scaled up with the severity of restriction and with leanness (Helms et al., Int J Sport Nutr Exerc Metab 2014) — which is an upper bound for a specific person, not a target for a general one. An older reader needs a floor rather than a ceiling: at least 1.0–1.2 g/kg/day, at least 1.2 g/kg/day if active, and 1.2–1.5 g/kg/day during acute or chronic illness (Bauer et al., PROT-AGE, J Am Med Dir Assoc 2013). The stated exception is severe kidney disease — an eGFR below 30 and not on dialysis — where protein may need restricting, and that is a decision for a nephrologist, not a cookbook.

Here is why this matters more in Kerala than almost anywhere. A rice-and-coconut-forward day is easy to eat at 0.6–0.8 g/kg of protein; two meals of rice with a vegetable thoran and a spoon of pickle will land you there without your noticing. Getting from 0.7 to 1.3 g/kg is the single most consequential structural change this book makes to the traditional plate, and it is entirely achievable inside the cuisine: fish, egg, chicken, mussel, prawn, and on the vegetarian side vanpayar, kadala, muthira, cherupayar, thuvara parippu, curd and buttermilk. The pulse rule in Chapter 11 exists partly for glycaemia and partly for this.

Energy density, and the most useful sentence in this book

Barbara Rolls's synthesis of two decades of feeding studies contains one finding that, if you take nothing else from Part One, will change your plate more than any other (Rolls, Physiol Behav 2009). It is this: people tend to eat a roughly consistent weight of food. So when the energy density of the available food falls, energy intake falls with it.

Not the calories. The weight. Grams on the plate, not kilojoules in the gram. The effect holds across weight status, across sex, across behavioural type, and in children as young as three. Both population studies and long-term clinical trials indicate it persists rather than fading.

The demonstrations are worth listing because their size is startling. A low-energy-dense soup eaten before lunch reduced energy intake at that lunch by about 26%; a whole apple before lunch, about 15%. Reducing the energy density of the available food by roughly 30%, while holding the weight of food constant, produced a roughly 30% cumulative reduction in energy intake over two days. In Ello-Martin's twelve-month trial of 97 women, the group told to add fruit and vegetables lost 7.9 kg against 6.4 kg in the group told to restrict portions, while eating about 25% more food and reporting less hunger. In PREMIER, the participants who reduced energy density most lost about 5.9 kg against 2.3 kg over six months. Run the same physics in reverse and over six years 186 young women eating a higher-energy-density diet gained about 2.5 times more weight — 6.4 kg against 2.5 kg (Savage et al., as reviewed in Rolls 2009).

Now look at what a Kerala kitchen already makes. Thoran, mezhukkupuratti, olan, avial, sambar, moru curry, kanji, a fish simmered in a thin sour gravy, an ishtu loaded with vegetables — these are naturally low-energy-density preparations. Grated coconut in quantity, coconut oil, parotta, banana chips, achappam, unniyappam and payasam are naturally high. The cuisine contains both poles, and always did. The job of this book is to shift the daily ratio between them without exiling the festival foods, and the reason that works is Rolls's finding: you will not need to eat less food. You will need to eat food that weighs the same and carries less energy.

That is why the plate in Chapter 11 is drawn in fractions of area rather than in calories. Half the plate as vegetables is not a moral position. It is an energy-density instrument.

Fibre, liquid energy, and meal timing

Fibre earns its place in the plate mostly through the two mechanisms already described — it comes attached to food that is bulky and low in energy density, and it comes attached to pulses, which do measurable things to a rice meal. Mohan's work in Chennai compared three five-day diets against each other: white rice, brown rice, and brown rice with 50 g of legumes a day. Against white rice, the glycaemic incremental area under the curve was 19.8% lower on brown rice and 22.9% lower on brown rice with the legumes (Mohan et al., Diabetes Technol Ther 2014). The 22.9% belongs to the grain and the pulse together, not to the pulse instead of the grain — but the pulse is what took the arm past the grain swap on its own. Pooled across 26 randomised trials, pulses at a median 130 g/day lowered LDL cholesterol by 0.17 mmol/L (Ha et al., CMAJ 2014). Those are real, replicated numbers, and they are the reason a pulse goes on every plate in this book. What I will not give you is a clean effect size for fibre and appetite specifically.

Liquid energy is the most invisible surplus in a Malayali day. Two to four occasions of sweetened tea or coffee, sometimes with a fried item alongside, carry energy that has no meal status in anyone's mental accounting. In the Nurses' Health Study II, women who increased sugar-sweetened soft drink intake from one a week or fewer to one a day or more gained 4.69 kg over four years, against 1.34 kg in those who decreased intake, and had a relative risk of 1.83 for developing type 2 diabetes compared with those drinking fewer than one a month (Schulze et al., JAMA 2004). The finding that matters most here is the one about fruit punch, at RR 2.00 — homemade sweetened fruit drinks are not exempt. Note that Rolls's soup preload shows liquid running the other way when it is dilute and savoury: sambharam is on the same side of that line as soup, and sharbat is not.

On meal frequency — three meals against five, grazing against not grazing — I have no trial evidence to offer you, and I would rather say so than fill the gap with plausible physiology. Eat on a schedule that lets you cook properly and stop deliberately. That is a craft recommendation, not a metabolic one.

Time-restricted eating, told straight

This is where a lot of readers have invested hope, so it deserves the actual numbers.

TREAT randomised 116 adults to 16:8 time-restricted eating or three structured meals a day for twelve weeks, with 105 completing (Lowe et al., JAMA Intern Med 2020). The time-restricted arm lost 0.94 kg; the three-meal arm lost 0.68 kg; the difference between them was −0.26 kg, with a confidence interval from −1.30 to +0.78 and P=0.63. There is no effect there. And there is a cost worth knowing about: appendicular lean mass index favoured the control arm by 0.16 kg/m² (P=0.005) — the fasting group lost slightly more muscle. The authors' own conclusion: time-restricted eating, in the absence of other interventions, is not more effective for weight loss than eating throughout the day.

Liu's trial ran the harder test. 139 adults with obesity, all on the same prescribed energy intake — 1,500–1,800 kcal/day for men, 1,200–1,500 for women — randomised to take that food inside an eight-hour window from 8 a.m. to 4 p.m., or across the day, for twelve months with 84.9% completion (Liu et al., N Engl J Med 2022). Result: −8.0 kg with the eating window against −6.3 kg without, a net difference of −1.8 kg with a confidence interval from −4.0 to +0.4 and P=0.11. Waist circumference, BMI, body fat, lean mass, blood pressure and the metabolic risk factors were all consistent with that null.

The fair statement is this: an eating window is a legitimate tool for producing a deficit, and it is not a metabolic shortcut. If not eating before eleven or after seven makes your day simpler, use it. But the deficit is doing the work, the clock is not, and in the one trial that measured lean mass carefully the fasting arm came out slightly worse. For South Asian readers, who carry lower muscle reserves at any given weight, that last point is not a footnote. Treat time-restricted eating as optional scaffolding. Never present it to yourself as the mechanism.

Sleep, resistance training and the movement you don't notice

Three levers here are genuinely important, and I can give you the mechanism without giving you a number, because the research underpinning this book did not secure primary citations for their effect sizes.

Resistance training matters in a deficit for one reason: what you lose should be fat, and lean tissue is expensive to lose, particularly for a reader in their fifties whose next thirty years depend on being able to carry shopping up stairs. Two sessions a week of progressive loaded movement is the standard prescription.

Daily movement outside exercise — walking to the shop, standing while the sambar simmers, taking the stairs, cooking rather than ordering — accumulates far more energy over a week than a gym session does, and it is the component that quietly collapses when a person moves to a car-dependent suburb or a Gulf desk job.

Sleep is not a diet variable in the usual sense, but short sleep is consistently associated with higher body weight and appears to move appetite regulation in the unhelpful direction. What I can source is the behaviour of people who succeed: members of the National Weight Control Registry, who have kept off at least 13.6 kg for at least a year, average around 2,800 kcal a week of physical activity, and about 90% of them exercise daily.

Metabolic adaptation, and the thing almost everybody gets backwards

Lose weight and your resting metabolic rate falls further than your smaller body alone accounts for. This is real, it is measurable, and it persists. The clearest documentation comes from following 14 of the 16 original "Biggest Loser" competitors for six years (Fothergill et al., Obesity 2016). At the end of the 30-week competition they had lost 58.3 kg on average and resting metabolic rate had fallen 610 kcal/day. Six years later, 41.0 kg had come back, resting rate was 704 kcal/day below baseline, and the adaptation component — the shortfall beyond what their new body size predicted — was −499 kcal/day.

Now the part that gets misreported. Regain was not correlated with the degree of metabolic adaptation at the end of the competition (r = −0.1, P = 0.75). And the competitors who had kept the most weight off showed the greatest concurrent metabolic slowing (r = 0.59, P = 0.025). Adaptation is not a curse that drags you back up. It is a proportional response to being lighter, and the people carrying the most of it were the people who had succeeded most.

So the physiology of regain is not principally metabolic. It is environmental and behavioural: the deficit stops being maintained, the food environment reasserts itself, the measuring stops. That is a more hopeful diagnosis than the folk version, because it points at things you can change.

What a realistic rate actually looks like

Anchor to trial data, not to advertising. A full year of supported, well-designed conventional dieting produced 5.3–6.0 kg in DIETFITS. POUNDS LOST produced about 6 kg at six months and about 4 kg at two years. DiRECT — an 825–853 kcal/day total diet replacement with structured clinical support, which is about as aggressive as non-surgical intervention realistically gets — produced 10.0 kg at twelve months and 7.6 kg at twenty-four.

Set against that: 0.5–1.0% of body weight per week is the defensible range, which for a 75 kg reader is 0.4–0.75 kg a week, and the slower end of that is the end that survives. Any book promising more than about 10 kg in a year without formula replacement is promising something the literature does not deliver.

Waist behaves differently and better, for reasons Chapter 4 sets out — the visceral and hepatic depots empty early, so the tape measure moves before the mirror does and the blood tests move before the tape. For a floor rather than a ceiling: Kerala's own peer-led community programme moved mean waist circumference from 89.5 to 87.5 cm over twelve months with no formal diet at all (Ravindranath et al., Transl Behav Med 2020). A reader running an actual deficit should expect more than 2 cm a year.

What predicts maintenance

Self-monitoring is the best-evidenced single behaviour in this entire literature, and it is free. Across 449 adults in a sixteen-week programme, participants weighed themselves on 80.9% of days, and the proportion of days weighed correlated with weight lost at r = −0.56 (Shetty and Ross, Obes Sci Pract 2025). During maintenance the pattern shifts in an instructive way: consistency mattered more than frequency. The number of weeks in which a person weighed on six or seven days predicted less regain, while raw overall frequency did not (P=0.141) (Brockmann et al., Obesity 2020). And this works outside clinics and outside the West: a rural Thai community trial using twice-daily self-weighing plus weekly volunteer counselling produced −1.2 kg against +0.3 kg over twenty weeks, a between-group difference of 1.0 kg (P=0.015) (Liampeng et al., BMC Prim Care 2023).

Alongside that, the National Weight Control Registry's members — averaging about 30 kg lost and held for 5.5 years — report a consistent cluster: most eat breakfast daily, about three-quarters weigh themselves at least weekly, most are highly active, and crucially they keep the same eating pattern from day to day and from weekday to weekend. Consistency again, in a different guise.

Why this book gives you no calorie target

You will not find a daily calorie number anywhere in these pages, and the omission is deliberate.

Counting works, for the people it works for. But the evidence above says the variable that decides outcomes at twelve months is not the precision of the arithmetic; it is whether you are still doing it. And counting has a specific failure mode in a cuisine like this one: a coconut-based curry, a hand-ground masala and a leaf of rice served by somebody else's ladle are all extremely hard to count and extremely easy to abandon counting over.

So this book substitutes structure for arithmetic. The plate does the portion work — half vegetables by area is an energy-density instrument, and Rolls tells you why it functions without conscious restriction. The pulse does the glycaemic and protein work — 50 g of cooked pulse on every plate, every day, non-negotiable. The measured oil does the fat work — a teaspoon levelled rather than poured, which is the difference between 45 and 180 kcal in a single pan. The weekly weigh-in and the monthly tape do the monitoring work, at the frequency and consistency the maintenance data support.

Those four things are countable, checkable and repeatable in a Malayali kitchen. A calorie ceiling is none of the three. And the whole force of DIETFITS, POUNDS LOST and Johnston's 48 trials is that a method you will still be running next Onam beats a better method you abandoned in March.

Chapter 6 — What Can Actually Be Reversed

Some of what has gone wrong in your blood results is reversible. Some of it is preventable. Some of it is neither, and this chapter will tell you which is which, because a hope chapter that overpromises does more damage than a cautious one.

Prevention: three trials, and one uncomfortable comparison

The Diabetes Prevention Program randomised 3,234 American adults with elevated fasting and post-load glucose to placebo, metformin at 850 mg twice daily, or an intensive lifestyle programme, and followed them for a mean of 2.8 years (Knowler et al., N Engl J Med 2002). Diabetes incidence came out at 11.0, 7.8 and 4.8 cases per 100 person-years respectively — a 58% risk reduction with lifestyle (95% CI 48–66) and 31% with metformin (17–43). Mean weight loss was 0.1, 2.1 and 5.6 kg. The number needed to treat over three years was 6.9 for lifestyle and 13.9 for metformin: treat seven people with food and movement for three years, prevent one case.

The Finnish Diabetes Prevention Study ran the same logic in 522 middle-aged adults with impaired glucose tolerance, mean BMI 31, mean age 55 (Tuomilehto et al., N Engl J Med 2001). Individualised counselling on weight, total fat, saturated fat, fibre and activity produced 4.2 kg of weight loss against 0.8 kg in the first year, and a four-year cumulative diabetes incidence of 11% against 23% — again a 58% reduction.

Then IDPP-1, in 531 native Asian Indians with impaired glucose tolerance, and the numbers that should make every Malayali reader sit up (Ramachandran et al., Diabetologia 2006). Three-year cumulative incidence was 55.0% in the control arm, 39.3% with lifestyle modification, 40.5% with metformin and 39.5% with both — a relative risk reduction of 28.5% for lifestyle (95% CI 20.5–37.3), with a number needed to treat of 6.4 and, notably, no additional benefit from combining lifestyle change with metformin.

Two things stand out. The first is that the Indian effect size was half the American and Finnish one — 28.5% against 58% — against a control-arm incidence more than twice as high. The second is the comparison that matters most for this book. IDPP-1's participants entered the trial at a mean BMI of 25.8 and a mean age of 45.9. The Finnish participants entered at a mean BMI of 31 and a mean age of 55. Indian adults were crossing into diabetes nine years younger and five BMI units lighter than Finnish adults were. If you have been reassured by a normal BMI, that comparison is the reassurance being withdrawn.

DiRECT, and the gradient that answers "how much do I have to lose?"

DiRECT recruited 306 adults from 49 UK primary-care practices — aged 20–65, type 2 diabetes of under six years' duration, BMI 27–45, not on insulin (Lean et al., Lancet 2018; PMID 29221645). The intervention withdrew antidiabetic and antihypertensive drugs under supervision, replaced all food with a 825–853 kcal/day formula for three to five months, reintroduced food over two to eight weeks, then provided structured maintenance support. Remission was defined as HbA1c below 6.5% after at least two months off all antidiabetic medication.

At twelve months, 68 of 149 in the intervention group (46%) were in remission against 6 of 149 controls (4%), an odds ratio of 19.7 (95% CI 7.8–49.8). Mean weight change was −10.0 kg against −1.0 kg.

But the finding to write on the wall is not the headline. It is what happened when the investigators sorted every participant in the trial, from both arms, by how much weight they had actually lost:

Read down that column slowly. This is the most useful table in the diabetes-diet literature, and it is useful precisely because it is a dose-response, not a verdict. Nobody who gained weight went into remission. Losing under 5 kg bought a 7% chance. Losing 5–10 kg — which is what a reasonable year of the plate in this book might produce — bought better than one in three. Losing 15 kg bought better than five in six.

Durability was tested twice more. At twenty-four months, 53 of 149 (36%) were still in remission against 5 controls (3%), and of the 36 intervention participants who had maintained more than 10 kg of loss, 29 (81%) were in remission (Lean et al., Lancet Diabetes Endocrinol 2019). At five years, participants who took up three further years of low-intensity support had held 6.1 kg off with 13% in remission, and against those who did not continue they had far more visits off glucose-lowering medication (62% against 30%) and in remission (34% against 12%), with serious adverse events roughly halved (Lean et al., Lancet Diabetes Endocrinol 2024).

ReTUNE: the trial for the reader whose BMI is "normal"

Every number above comes from people with a BMI of 27 or more. If you are a Malayali reader at BMI 24 with a 92 cm waist and an HbA1c of 6.8% — which is an extremely common shape for this to take — none of those trials was about you. One was.

ReTUNE studied people with type 2 diabetes and a BMI of 21–27 kg/m², taking them through stepwise weight loss in three 5% decrements (Taylor et al., Clin Sci 2023, as reviewed in Taylor, Diabetologia 2025). Seventy per cent achieved remission at one year. The weight loss required spanned 5.5% to 10.2% of body weight, with a median of 6.5% — for many readers of this book, four to seven kilograms. Liver fat, which had been elevated roughly threefold at baseline, normalised as the weight came off. And maintenance was better than in heavier participants: there was no mean weight regain between the end of the weight-loss phase and the twelve-month assessment.

That is the paragraph to come back to. The message for a reader who is not visibly heavy is not "you need to lose 15 kg." It is: you may need to lose 6 to 10% of your body weight, and there is roughly a 70% chance that is enough. Nothing else in this literature is simultaneously this motivating and this defensible.

The personal fat threshold, in plain language

Roy Taylor's explanation of why the same weight loss does such different things to different people has two moving parts (Taylor, Diabetologia 2025).

The first is the twin cycle. Sustained positive energy balance puts fat into the liver. A fatty liver responds poorly to insulin and starts manufacturing more fat from carbohydrate on its own account, which puts more fat into the liver — the first, self-reinforcing cycle. That liver exports the excess as triglyceride, some of which lands in the pancreas. Fat in the pancreas suppresses the beta cells' ability to release insulin in a first, fast burst — the second cycle. The two together produce the rising glucose that eventually gets a name.

The second is the personal fat threshold. Everyone has a private tolerance for fat stored inside organs rather than under the skin, and it varies enormously between individuals. Two people at the same weight, the same height and the same age can sit on opposite sides of their own thresholds. This is why the question "am I heavy enough for this to be my problem?" has no population answer — and it is exactly why Chapter 4's finding matters, that South Asian men carry +0.56 standardised units more liver fat than white European men at a lower BMI (Iliodromiti et al., Diabetologia 2023). A lower storage capacity under the skin means the overflow arrives in the liver sooner.

The mechanism is reversible, and fast. In Taylor's Counterpoint study, twelve people with type 2 diabetes on about 800 kcal/day saw liver fat fall by roughly a third within seven days, with fasting plasma glucose normalising at day 7 — before most of the weight had gone. Over eight weeks they lost 15.3 kg and HbA1c fell from 7.4% to 6.0%. Counterbalance then showed what limits it: responders had a mean diabetes duration of 3.8 years against 9.8 years in non-responders. Liver fat normalised in both groups. It was beta-cell recovery that depended on duration. Which is the least comfortable and most important sentence in this chapter: the sooner after diagnosis you act, the more of your pancreas is still available to recover.

"Remission" is not "cure", and the distinction is not pedantic

An international expert group convened by the American Diabetes Association settled the wording, and this book uses theirs exactly. The agreed term is remission — not cure, not reversal — and the usual diagnostic criterion is HbA1c below 6.5% (48 mmol/mol) measured at least three months after stopping all glucose-lowering medication (Riddle et al., Diabetes Care / Diabetologia 2021). DiRECT itself used two months; the consensus later standardised on three.

Say the rest bluntly. Remission means the numbers have come back to a non-diabetic range and stayed there without drugs. It does not mean the susceptibility has gone. It does not mean the threshold has moved. It means you are currently below your threshold and you will remain below it only for as long as the weight stays off — which is why the ADA statement pairs remission with continued active observation, and why DiRECT's five-year data show remission slipping as weight returns. Regain the weight and the glucose follows it back up. That is not failure of the concept; it is the concept working in the other direction.

One more thing, and it is not negotiable. In DiRECT, medication was withdrawn by the trial's clinicians, under monitoring, as part of the protocol. Nothing in this book is an instruction to reduce or stop any medicine. If your numbers improve — and they may improve quickly, as Counterpoint's day-7 finding shows — the person who changes your prescription is your doctor, and the change should be made with your glucose being watched.

Blood pressure: roughly a millimetre per kilogram

The arithmetic here is unusually clean. Pooling 25 randomised trials and 4,874 participants, a net weight reduction of 5.1 kg lowered systolic pressure by 4.44 mmHg and diastolic by 3.57 mmHg — which works out to −1.05 mmHg systolic and −0.92 mmHg diastolic per kilogram lost (Neter et al., Hypertension 2003).

Two refinements are worth having. The benefit is not linear: populations losing more than 5 kg saw −6.63 mmHg systolic against −2.70 mmHg for those losing less. And the diastolic effect was larger in people already taking antihypertensive drugs. So a reader who loses 8 kg is not buying 8 mmHg by rote — but 6 to 9 mmHg is a reasonable expectation, and it arrives on top of whatever the dietary pattern itself delivers, which Chapter 10 shows can be another 11 mmHg in someone with hypertension.

Cholesterol: the arithmetic of a swap

The Mensink equations, from a meta-analysis of 60 controlled feeding trials, let you calculate a fat swap before you make it (Mensink et al., Am J Clin Nutr 2003). Replacing 1% of daily energy from saturated fat lowers LDL cholesterol by:

• 1.8 mg/dL if you replace it with polyunsaturated fat
• 1.3 mg/dL if you replace it with monounsaturated fat
• 1.2 mg/dL if you replace it with carbohydrate

And the qualifier that decides the Kerala argument: the total-to-HDL cholesterol ratio did not improve when saturated fat was replaced by carbohydrate, but did improve significantly when it was replaced by unsaturated fat, especially polyunsaturated. Swapping ghee for rice is not the same trade as swapping ghee for sesame oil.

Work an actual Kerala example. Say a household member eats 2,000 kcal a day and takes 20 g of coconut oil in it — a modest allowance by current practice, about four teaspoons across the day's tempering and frying. That is 180 kcal, 9% of energy, and coconut oil is 90.9% saturated (ICMR-NIN IFCT 2017), so it delivers roughly 8.2% of energy as saturated fat. Move that same 20 g to groundnut, sesame or rice bran oil at around 20% saturated, and the saturated contribution falls to about 1.8% of energy — a reduction of some 6.4 percentage points, replaced by mono- and polyunsaturated fat. Run it through Mensink at 1.3–1.8 mg/dL per point and the estimated LDL reduction is roughly 8 to 12 mg/dL, from one decision about which bottle sits by the stove. That is an estimate built from the equations, not a measured trial result, and it should be read as one — but the direction and the order of magnitude are solid, and they are consistent with the pooled coconut-oil trials, where coconut oil raised LDL by 10.47 mg/dL against non-tropical oils (Neelakantan et al., Circulation 2020).

Soluble fibre adds a second, independent lever. Across 28 trials at a median dose of about 10.2 g/day, psyllium lowered LDL cholesterol by 0.33 mmol/L, non-HDL cholesterol by 0.39 mmol/L and apolipoprotein B by 0.05 g/L (Jovanovski et al., Am J Clin Nutr 2018), with the apoB evidence graded high. An earlier trial in 404 adults already on a low-fat diet found LDL down 9% against a control cereal (Olson et al., J Nutr 1997).

Stack the levers deliberately and you arrive at the Portfolio pattern, which is the strongest food-only LDL intervention in existence: across seven trial comparisons in 439 people with hyperlipidaemia, it lowered LDL cholesterol by about 17% — a mean difference of 0.73 mmol/L — with significant falls also in non-HDL cholesterol, apoB, total cholesterol, triglycerides, blood pressure and CRP, graded high for LDL (Chiavaroli et al., Prog Cardiovasc Dis 2018; Jenkins's protocols). Its four components, per 2,000 kcal, are 42 g nuts, 50 g plant protein, 20 g viscous fibre and 2 g phytosterols, each contributing about 5–12% on its own. Under fully controlled feeding, Jenkins's team reached 28–35%; free-living, the pooled figure is 17%; over a year it settled at 12.8%, with over 20% in the people who adhered best.

The Kerala translation is almost embarrassingly direct. Viscous fibre: oats, barley, vendakka, aubergine, guava, orange, psyllium. Plant protein: cherupayar, vanpayar, kadala, muthira, thuvara parippu. Nuts: cashew in strict moderation, since it is energy-dense, plus peanut. Only the phytosterols need a fortified product. Everything else is a Tuesday.

What this book cannot do

Four honest limits.

It cannot change your lipoprotein(a). Elevated Lp(a) is disproportionately common in South Asians, largely genetically determined, and essentially unresponsive to diet. Its role in this book is to justify one instruction: get it measured once, and if it is high, lower everything that is modifiable harder.

It cannot move your beta-cell reserve back to where it was ten years ago. Counterbalance is explicit that liver fat normalises regardless of how long you have had diabetes, and beta-cell recovery does not. Duration matters, and no plate undoes it.

It cannot make you an exception to energy balance, and it will not pretend that a cuisine, however good, is protective on its own. Kerala's own numbers in Chapter 1 are the disproof of that idea.

And it cannot substitute for medicine. Food and medication are not competitors, and framing them that way harms people. Metformin cut diabetes incidence by 31% in DPP and by 26.4% in IDPP-1 — a real effect, achieved with a cheap and well-understood drug, and in IDPP-1 adding it to lifestyle change bought nothing extra, which is a fact that cuts in both directions. A statin lowers LDL further and faster than any plate. An antihypertensive works on the day you take it. What food does, that drugs do not, is act on the upstream cause: it empties the liver, it lowers the threshold burden, and in DiRECT and ReTUNE it took people off their medication entirely rather than adding another. Both tools, working together, supervised by someone who has seen your actual results. That is the whole position, and it is the last thing this chapter has to say.