The Portfolio dietary pattern bundles four cholesterol-lowering plant foods — nuts, plant protein (soy, pulses), viscous (soluble) fibre, and plant sterols — each carrying an FDA / Health Canada / EFSA LDL health claim. The question this page answers: how large is the pooled LDL/apoB effect, and how far does it travel toward patient-important outcomes? (Chiavaroli et al., 2018)
Effect estimate — on surrogates only
Pooled SR+MA of 7 controlled-trial comparisons, 439 hyperlipidemic participants, Portfolio pattern on an NCEP Step II background vs NCEP Step II alone, median ~4 weeks. (Chiavaroli et al., 2018)
- LDL-C: −0.73 mmol/L (~−28 mg/dL), −17%, 95% CI −0.89 to −0.56, GRADE HIGH. Studied range is short-term (>=3 wk), hyperlipidemic adults on a Step II background — do not extrapolate the magnitude to normolipidemic people or free-living self-selection without the adherence discount below. (Chiavaroli et al., 2018)
- apoB −15% (−0.19 g/L, CI −0.23 to −0.15) and non-HDL-C −14% (−0.83 mmol/L), both GRADE HIGH — the causal lipid targets, moving in concert with LDL-C. Total cholesterol −12%, triglycerides −16% (all HIGH). (Chiavaroli et al., 2018)
- No effect on HDL-C (+0.01 mmol/L, p=0.56) or body weight (−0.10 kg, p=0.59) — true nulls, GRADE moderate/high, not imprecision. (Chiavaroli et al., 2018)
- Secondary, GRADE moderate (downgraded for imprecision — CIs contained the pre-specified MID): SBP −1.75 mm Hg, DBP −1.36 mm Hg, CRP −0.53 mg/L (−32%), and a model-estimated 10-yr CHD risk −1.34% absolute (−13% relative). The CHD figure is computed from the surrogates, not observed events. (Chiavaroli et al., 2018)
Full table with I2/p-values on the source page.
The surrogate ceiling — the transmission is borrowed, not shown
Every Portfolio endpoint is a surrogate (blood lipids, BP, CRP) or a model-derived risk score; no trial has measured CVD events or atherosclerosis. The authors say so and name the awaited plaque-imaging RCT (PortfolioEx, NCT02481466 — enhanced Portfolio + exercise, carotid MRI plaque over 3 yr). (Chiavaroli et al., 2018)
«In this regard, there remains a need for large randomized trials of the effect of the portfolio dietary pattern on hard CV outcomes or surrogate end- points of atherosclerosis.» (Chiavaroli et al., 2018)
So the step from −0.73 mmol/L LDL to fewer heart attacks is not made by this trial base. It is borrowed from the LDL-causality fabric — the CTT/Mendelian evidence that lowering LDL/apoB by a given amount reduces events roughly in proportion to absolute LDL reduction and cumulative exposure -> LDL ApoB and Cumulative Exposure, Surrogate Outcomes. Under that borrowing, a sustained 0.73 mmol/L LDL reduction is a plausible ~15% relative MACE reduction over a decade of adherence, but that number is a transported inference, not a Portfolio finding, and it inherits the assumption that diet-lowered LDL transmits like drug-lowered or genetically-lowered LDL.
Adherence is inside the effect
Restricting to the efficacy trials (metabolically-controlled feeding) lifts LDL-C to −21% (−0.87 mmol/L) and zeroes the heterogeneity; the free-living advice-only trials pull the pooled estimate down to −17%. The gap is the effectiveness discount — the whole-diet magnitude a person actually gets depends on how completely they hold all four components. (Chiavaroli et al., 2018)
Layer-1 sizing — against the drug alternative
The authors put the whole-pattern LDL effect in the range of a starting low-intensity statin or ezetimibe:
«The expected reductions of the intended combination of a Portfolio dietary pattern and NCEP Step II diet of ~27% (32% in efficacy and 15% in effectiveness trials) are similar to those seen with the starting doses of low intensity statins (18 to 28%) or 10 mg of ezetimibe (15–20%). These reductions would be expected to translate into meaningful reductions in CVD risk.» (Chiavaroli et al., 2018)
(The ~27% here is the authors’ additive-framing figure; the meta-analytic pool vs NCEP Step II alone is the −17% / −21%-efficacy above — read the two consistently: full adherence approaches a low-intensity statin, free-living adherence falls to ~10-15%.) An early metabolic-ward head-to-head had the diet match 20 mg lovastatin (−28.6% vs −30.9%), dropping to 10-15% free-living. (Chiavaroli et al., 2018)
For the LDL-lowering channel specifically, a statin is a mature, low-cost, low-harm substitute that captures at least as much of the effect with far lower behavioural cost, so the Portfolio pattern’s marginal LDL rock shrinks against an available statin -> Statins for Primary Prevention and the Power of Zero CAC. Two guards keep the lever from being written off: (i) substitution is outcome-specific — the pattern also moves BP, TG, CRP and is a whole-food substitution with pleiotropic and structural value a single-channel LDL drug does not replicate; (ii) it removes rather than masks the driver and carries no drug side-effect/dependency profile. The choice between diet and drug is the person’s (Layer 3); Layer 1 only notes the LDL rock is partly pre-captured by an available statin.
Trust caveats
All 7 comparisons come from one investigator group — the diet’s inventors — an unblinded dietary intervention appraised by its originators, with component-food industry funding across the author list. The authors flag the single-group indirectness but did not downgrade for it (citing one n=351 multi-centre trial). Read the pooled magnitude as the ceiling an efficacy-optimised, non-independent base produces. (Chiavaroli et al., 2018)
Independence note: this source shares authors (Jenkins, Salas-Salvado, Sievenpiper) with held soy/pulse/Mediterranean sources, so it is not independent type-E corroboration for component or plant-pattern claims — it is the same school (see Soy Products and Health).
Open questions
— the wiki’s own gap inventory:
- Does the LDL/apoB reduction transmit to events or plaque? (PortfolioEx pending; the only Portfolio trial aimed at a near-patient-important outcome.)
- Does a Portfolio effect hold when trials are run by independent groups, in normolipidemic people, and under real free-living adherence (the effectiveness-trial magnitude)?
- G (needs aggregation): the pooled Portfolio-vs-statin head-to-head effect on events — a magnitude the wiki cannot compute from held sources.