DiRECT (Lancet 2018) is the first registered trial to set remission of type 2 diabetes as a primary outcome and to deliver the intervention in routine primary care rather than a research centre. It supplies two things the fabric did not hold: a quantified total-diet-replacement (TDR) protocol with a structured maintenance phase, and a remission-by-weight-loss dose-response — the %-body-weight -> outcome gradient that grounds the weight-loss lever with a magnitude.
Its pathophysiological premise is Taylor’s twin cycle hypothesis — that T2D of limited duration is driven by excess fat in liver and pancreas and is reversible by negative energy balance. Earlier mechanistic work had normalised liver insulin resistance within 7 days and pancreatic function over 8 weeks on a 600-700 kcal/day diet; DiRECT tests whether this is practicable at scale. (Lean et al., 2018) The premise is now held directly: «people with recent onset T2DM could regain normal glucose control and normal β-cell function when the fat content of the liver and the pancreas was decreased by a weight loss diet- ary regimen» — the depot-drawdown mechanism the remission gradient rides on -> Ectopic Fat and Depot-Specific Risk. (Taylor & Holman, 2014)
The stratum DiRECT’s BMI cut-off excludes — normal-weight T2D (a decision-change). DiRECT required BMI 27-45, so its evidence is silent on the ~third of T2D patients who are normal-weight, and such patients are routinely not offered weight loss. Taylor’s personal-fat-threshold frame extends the lever there: T2D is one condition across BMI (36% of newly-diagnosed UKPDS patients had BMI <25), and reversal is «achievable equally readily by people with lower initial BMI» — indeed at a smaller required loss (UKPDS: ~13% at normal BMI vs ~21% for the whole cohort to reach normoglycaemia). So a normal-weight person with short-duration T2D is a candidate for the same energy-deficit lever, sub-threshold BMI notwithstanding — a recommendation DiRECT’s own population cannot license but the mechanism does. (Taylor & Holman, 2014) (inferred from Taylor & Holman, 2014)
The intervention — an energy target, NOT a macronutrient one
The Counterweight-Plus programme, in three phases:
- Total diet replacement: a low-energy formula diet, “825-853 kcal/day; 59% carbohydrate, 13% fat, 26% protein, 2% fibre”, for 3 months (extendable to 5). Note the 59% carbohydrate — this is a high-carbohydrate, low-energy diet, not a low-carbohydrate one.
- Stepped food reintroduction: 2-8 weeks, “about 50% carbohydrate, 35% total fat, and 15% protein”.
- Structured weight-loss maintenance: ongoing monthly visits, behaviour-change methods (including CBT elements), a step-count target.
«All oral antidiabetic and antihypertensive drugs were discontinued on day 1», reintroduced only per protocol if glucose or blood pressure required it. (Lean et al., 2018)
Population and design: aged 20-65, T2D diagnosed within 6 years, BMI 27-45, not on insulin. Open-label, cluster-randomised at 49 primary-care practices (the practice, not the patient, was the randomisation unit); 149 per group in the ITT population. Co-primary outcomes, analysed hierarchically: “weight loss of 15 kg or more, and remission of diabetes, defined as glycated haemoglobin (HbA1c) of less than 6.5% (<48 mmol/mol) after at least 2 months off all antidiabetic medications, from baseline to 12 months”. (Lean et al., 2018)
The headline result
| Outcome (12 months, ITT) | Intervention | Control | Effect |
|---|---|---|---|
| Weight loss >=15 kg | 36 (24%) | 0 | p<0.0001 |
| Diabetes remission | 68 (46%) | 6 (4%) | OR 19.7 (95% CI 7.8-49.8) |
| Mean weight change | -10.0 kg (SD 8.0) | -1.0 kg (3.7) | adj diff -8.8 kg (-10.3 to -7.3) |
| HbA1c change | -0.9% | +0.1% | adj diff -0.85% (-1.10 to -0.59) |
| QoL (EQ-5D VAS) | +7.2 | -2.9 | adj diff +6.4 (2.5 to 10.3) |
The 46% remission rate “greatly exceeded” the 22% deemed a-priori clinically important. On medication: at 12 months 73.6% of the intervention group were on zero oral antidiabetic drugs (vs 18.2% of control), and 68% remained off antihypertensives with no rise in mean blood pressure. (Lean et al., 2018)
The jewel — remission scales with weight loss (dose-response)
Pooling both randomised groups, remission rises monotonically with the weight loss maintained at 12 months. “Remission varied with weight loss in the whole study population”:
| Weight change at 12 months | Remission achieved | Rate |
|---|---|---|
| Weight gained | 0 of 76 | 0% |
| 0-5 kg loss | 6 of 89 | 7% |
| 5-10 kg loss | 19 of 56 | 34% |
| 10-15 kg loss | 16 of 28 | 57% |
| >=15 kg loss | 31 of 36 | 86% |
The discussion states the composite the table implies: “86% of participants with at least 15 kg weight loss, and 73% of those with weight loss of 10 kg or more” achieved remission. (Lean et al., 2018)
A different design reaches the same per-kilogram gradient (type-F corroboration, independence CAPPED)
DiRECT’s dose-response is a within-trial association (responder confounding is live — see Limits). A
separate observational design corroborates the same monotone, diet-independent gradient — which
strengthens the direction without lifting the finding to [E-independent] (the independence is capped;
see below). Reported in the Churuangsuk umbrella review: ADDITION-Cambridge, a population cohort of 867 newly-diagnosed T2D patients
across 49 UK primary-care practices, followed 5 years — overall 30% remission — found «Every 1 kg of
weight loss was associated with 7% higher chance of remission at 5 years, regardless of specific diet
regimens or lifestyle interventions», and «Loss of >10% of baseline body weight in the first year after
diagnosis was associated with 70% higher chance of remission at 5 years».
(Churuangsuk et al., 2021)
Parameter table (op-weave 2a — the two are matched on the claim they jointly back, NOT on magnitude):
| Parameter | DiRECT (Lean 2018) | ADDITION-Cambridge (via Churuangsuk 2022) | Same quantity? |
|---|---|---|---|
| Design | within-RCT dose-response, both arms pooled | population cohort, 49 practices | NO — different design |
| Population | T2D <6 yr, not on insulin, ~all white | newly-diagnosed T2D, UK primary care | overlapping, both early-disease |
| Exposure | weight-loss band at 12 mo | per 1 kg / >10% loss in yr 1 | different metric |
| Outcome / horizon | remission (HbA1c<6.5% off meds), 12 mo | remission, 5 yr | same outcome, different horizon |
| The claim backed | remission rises monotonically with kg lost, irrespective of diet | same | YES — the directional claim |
The magnitudes are not pooled (12-mo band-rate vs 5-yr per-kg association); what is shared is the
direction and diet-independence of the gradient. There IS a real independence gain on two axes — a
randomised within-population contrast vs an observational cohort, and different datasets/institutions
(Newcastle/Glasgow DiRECT vs Cambridge ADDITION). But the [E-independent] token is withdrawn
(2026-08-08), for two reasons the earlier version missed:
- Shared confounding, not “different confounding structures.” Both gradients are driven by the SAME dominant threat — responder confounding (people who lose more weight differ biologically, so the weight->remission slope is partly reverse-causal in the RCT and the cohort). A second design that shares the incumbent’s main confounder does not neutralise it, so it cannot lift the finding the way a genuinely orthogonal design would.
- The relay runs through DiRECT’s own PI. ADDITION-Cambridge is held ONLY via the Churuangsuk umbrella,
whose byline is «…Simon J. Griffin … & Michael E. J. Lean» — Lean is DiRECT’s joint PI/first
author and Griffin an ADDITION investigator. So the “independent corroboration” is read through a
review co-written by the incumbent trial’s own lead; that is a lineage tell E forbids, disclosed nowhere
before. Same shared-authorship class as the Fibre SACN⟂Reynolds demotion.
So this is type-F corroboration by a different design, with independence capped — real but not
[E-independent]; and transport beyond UK early-disease weight-centric-remission populations is corroborated by neither. (Churuangsuk et al., 2021; inferred from Lean et al., 2018)
Where DiRECT sits in the umbrella’s certainty-graded remission map
Churuangsuk’s GRADE table places total diet replacement (the two low-RoB RCTs, DiRECT + DIADEM-I) at the single GRADE-HIGH cell of the whole remission map — median 54% remission (range 46-61) vs 4-12% standard care, N=445 — above formula meal replacement (11%, moderate), Mediterranean (15%, low), and ketogenic (20%, very low). (Churuangsuk et al., 2021) So DiRECT is not an isolated strong trial: it anchors the one high-certainty conclusion in the graded T2D remission literature, and that conclusion is a format (energy-controlling TDR), not a macronutrient class -> Diets for Weight Management in Type 2 Diabetes.
Why this is the high-value finding. A monotone gradient from 0% (weight gain) to 86% (>=15 kg), in a within-population comparison, is direct evidence that the amount of weight lost — not the diet’s composition — is the operative variable. It is a cross-group association (not the randomised contrast), so it carries confounding risk (a responder who loses more may differ), but the biological gradient is steep and coherent with the mechanistic twin-cycle work. Read it as name-the-curve: the curve is monotone over the studied range with no visible plateau below 15 kg — every increment of sustained loss buys more remission. Whether it plateaus above 15 kg is unresolved (the top category is open-ended and the mechanism has a floor — remission is bounded at 100%).
The maintenance phase — the structure the fabric lacked
DiRECT is unusual in designing for maintenance “from the outset”. Among engagers, weight fell sharply in TDR then partly regained: -14.5 kg (95% CI 13.4-15.5) during TDR, then +1.0 kg during food reintroduction and +1.9 kg during maintenance — a net ~-11.6 kg trajectory for completers, against the -10.0 kg ITT mean. (Lean et al., 2018)
Adherence: “79% completed the intensive total diet replacement phase”; the ~25% “dropout rate” (6 never engaged, 26 withdrew — 15 during TDR, 6 during reintroduction, 5 during maintenance) marks the programme’s non-acceptability ceiling. This is the missing-piece for the NICE TDR recommendation, which names a low-energy TDR but not the reintroduction+maintenance scaffold -> Diets for Weight Loss - What NICE Recommends. (DiRECT’s 825-853 kcal formula sits at the bottom of NICE’s stated 800-1200 kcal band.)
Synthesis — is remission driven by WEIGHT LOSS or by CARBOHYDRATE RESTRICTION?
The fabric holds a low-carbohydrate remission result (Carbohydrate Restriction and Type 2 Diabetes Remission, Goldenberg 2021). DiRECT achieves remission on a 59%-carbohydrate formula diet — so the two routes must be reconciled. Parameter table (op-weave 2a):
| Parameter | DiRECT (Lean 2018) | Goldenberg 2021 (BMJ MA) | Same quantity? |
|---|---|---|---|
| Outcome | remission = HbA1c <6.5%, off antidiabetic meds >=2 months, 12 mo | remission = HbA1c <6.5% (+/- medication, both defs reported), 6/12 mo | ~YES on the med-free definition; Goldenberg’s headline permits meds |
| Intervention | 825-853 kcal TDR formula, 59% carbohydrate | ”<130 g/day or <26%” carbohydrate | NO — opposite on carbohydrate |
| Mechanism claim | weight loss / negative energy balance (twin cycle) | carbohydrate restriction | NO — the contested variable |
| Weight loss | -10.0 kg (ITT), dose-response to remission | MD -3.46 kg at 6 mo, +0.29 kg by 12 mo (regained) | NO — much larger, sustained |
The fourth column is NO on intervention and mechanism — so this is NOT a like-for-like tension to file. It is an F-refinement, and it runs in DiRECT’s favour on the causal question. Both routes produce weight loss; both produce remission. But:
- DiRECT’s within-population dose-response ties remission directly to kilograms lost, irrespective of macronutrient composition — and it does so at 59% carbohydrate, the opposite of a low-carb diet. So carbohydrate restriction is not necessary for remission: it is achieved here without it. (This does not exclude an added glycaemic contribution from carb restriction beyond the weight it produces — that is untested here, insufficient-evidence, not shown absent.)
- Goldenberg’s own data are consistent with weight-mediation, not competing with it: its low-carb remission and weight advantage both attenuate to null by 12 months as weight is regained (weight MD +0.29 kg at 12 mo; 12-mo medication-free remission RD -0.04, NS). A remission effect that decays in step with the weight advantage is consistent with a weight-loss effect delivered via a low-carb route, rather than a carbohydrate effect independent of weight.
The defensible composite: T2D remission (of short-duration disease) is driven by the magnitude of sustained weight loss; carbohydrate restriction is one lever for achieving that weight loss, with no evidence it adds a remission effect beyond the weight loss it produces. This refines — does not contradict — Goldenberg: the low-carb result is real, but its framing (“low-carb for remission”) is better read as “weight loss for remission, low-carb being one delivery route”. A person choosing a diet for remission should optimize for the weight loss they can sustain, by whichever route they will adhere to (Low-Carbohydrate vs Balanced-Carbohydrate Diets). (Goldenberg et al., 2021; inferred from Lean et al., 2018)
DiRECT states the same logic for surgery: “The essential mechanisms behind bariatric surgery are weight loss and decrease in body fat content, rather than any direct surgical effect” — and “The very large weight losses targeted by bariatric surgery are not essential for achievement of remission”. So across three routes (formula diet, low-carb, surgery) the common cause is weight/fat loss, not the route. (Lean et al., 2018)
Remission is a real benefit weight loss moves — but it is intermediate
DiRECT is a counterweight to the cardiovascular-event null (Does Weight Loss Reduce Cardiovascular Events): where Look AHEAD showed lifestyle weight loss did not cut CV events, DiRECT shows the same lever delivers a patient-important benefit it demonstrably does move — freedom from diabetes and its drugs, plus a measured QoL gain. Weight loss is indicated for the outcomes it moves; remission is squarely one of them.
But remission is an intermediate outcome, not a hard one. HbA1c normalisation is a strong surrogate for microvascular risk, but DiRECT was “not designed or powered to evaluate effects on complications” and remission can relapse. DiRECT itself reaches for the hard-outcome bridge only by citing a Look AHEAD post-hoc: “a 10% weight loss in the first year… was associated with a 21% decrease in occurrence of cardiovascular outcomes” — the same >=10%-responder signal that page holds as reported, not proven -> Does Weight Loss Reduce Cardiovascular Events. (Lean et al., 2018)
Limits and bias — read honestly
- Open-label, and allocation was visible to participants (cluster design). Baseline balance suggests limited bias, but blinding of a lifestyle programme is impossible.
- Asymmetric medication handling: antidiabetic drugs were stopped in the intervention arm but not control. The trial compares the whole programme vs standard care, so the remission definition (off-meds HbA1c) is favourable to the arm whose meds were deliberately withdrawn — a design choice, disclosed, that inflates the between-arm remission gap relative to a med-matched comparison.
- 12-month horizon; durability is unaddressed here (follow-up planned to >=4 years). Remission of short-duration T2D only — “remission is less likely with longer durations of disease”.
- Near-uniformly white Scottish/Tyneside population; no unqualified transport to e.g. South Asian groups who develop T2D at lower weight.
- The dose-response is a within-population association, not the randomised contrast — responder confounding is possible.
- The twin-cycle / personal-fat-threshold mechanism is a HYPOTHESIS, not a settled cause. Taylor’s depot-drawdown account is mechanistically corroborated (liver/pancreas fat falls with weight loss and tracks glucose recovery) but the personal fat threshold — that each person has an individual fat level above which T2D triggers — remains an inference under test, not an established quantity; the remission gradient is consistent with it but does not prove it.
- Sponsor / vendor COI: DiRECT’s formula diet (Counterweight-Plus) is a commercial product; the Counterweight programme and formula-diet vendors have a stake in the result. Disclosed here for symmetric standards — it does not overturn the RCT, but the honest-bias inventory must name it.
- Adverse events during TDR were common but mild (constipation, cold sensitivity, headache, dizziness); 9 serious AEs in 7 (4.5%) intervention participants, 2 possibly related (biliary colic + abdominal pain, same participant), none causing withdrawal.
- Coherence, not validity (R1): a strong single trial, not proof the effect transports or persists.
(inferred from Lean et al., 2018)