Substance and claimed effects

Tempeh is a cake of cooked whole soybeans bound by the white mycelium of Rhizopus oligosporus (sometimes R. oryzae), a solid-state fermentation that originated in Java several centuries ago and remains a staple protein source across Indonesia Nout & Kiers 2005. The fermentation transforms the substrate: phytate is partially hydrolysed, oligosaccharides (raffinose, stachyose) drop, isoflavone glucosides are de-glycosylated into more bioavailable aglycones, and a fraction of soy storage proteins is pre-digested into smaller peptides — all while the beans remain visually whole and structurally intact, unlike tofu (curdled soy-milk gel with the okara discarded) or soy protein isolate (defatted, alkali-extracted) Astuti et al. 2000, Hutkins 2018.

Holistic claim set the entry must cover: (1) complete protein source — all nine essential amino acids in good ratios, with a PDCAAS only modestly below animal protein and improved over unfermented soy by Rhizopus-driven protein hydrolysis Wang et al. 1968; (2) intact dietary fibre from the whole-bean structure; (3) isoflavone content shifted toward the aglycone forms (genistein, daidzein, glycitein) Otieno & Shah 2008; (4) modest LDL-cholesterol reduction within the broader soy-protein evidence base Tokede et al. 2015; (5) lower postprandial glucose excursions than refined-carb meals of equivalent calories; (6) high satiety per calorie via the protein + fibre combination; (7) substantially improved mineral (Zn, Fe, Ca, Mg) bioavailability through phytate hydrolysis Egounlety & Aworh 2003; (8) gut-microbiome effects from intact fibre, oligosaccharides, isoflavones, and (when unpasteurised) live Rhizopus mycelium and incidental bacterial coloniser-derived metabolites Nout & Kiers 2005; (9) protein digestibility improvements over unfermented whole soy beans.

Evidence by addressing question

Mechanism

Rhizopus oligosporus is a heterotrophic mould that, during the 24–48 hour fermentation at ~30 °C, secretes a battery of hydrolytic enzymes: proteases that cleave 7S and 11S soy globulins into peptides and free amino acids; β-glucosidases that strip glucose from the major isoflavone glycosides (genistin → genistein, daidzin → daidzein), producing the lipophilic aglycones absorbed faster and at higher serum peaks than their glycosides Otieno & Shah 2008; phytases that hydrolyse myo-inositol hexakisphosphate (phytic acid, IP₆) progressively to lower-phosphate inositols, releasing the divalent cations it chelates Reddy & Sathe 2002; α-galactosidases that break down stachyose and raffinose, the indigestible oligosaccharides responsible for soy's flatulence reputation Egounlety & Aworh 2003. The mycelial mat itself binds the substrate into a sliceable cake and is consumed along with the beans, contributing fungal cell-wall β-glucans and chitin.

Cholesterol mechanism for the soy fraction: a mix of (a) direct displacement of saturated-fat-rich animal protein from the diet, (b) increased faecal bile-acid excretion via 7S globulin peptide binding, (c) upregulation of hepatic LDL receptors driven by specific soy peptides, and (d) modest contribution from isoflavones acting on lipid metabolism — the AHA 2006 advisory landed on the first as the dominant lever, the latter three as smaller but real contributors Sacks et al. 2006. Postprandial-glucose mechanism: low glycaemic load (intact protein matrix + ~7g fibre per 100g + ~9g fat per 100g slows gastric emptying), plus a documented incretin response to soy protein that flattens the glucose curve. Satiety mechanism: protein-induced GLP-1 and PYY release plus the mechanical bulk of intact bean fibre.

Evidence

LDL cholesterol. The Anderson 1995 meta-analysis of 38 controlled trials reported a mean LDL-C reduction of ~13% with an average ~47g/day soy-protein replacement of animal protein Anderson et al. 1995, the basis for the FDA's 1999 health claim allowing soy-protein products to carry coronary-heart-disease language at intakes ≥25g/day soy protein FDA 1999. The AHA's 2006 reassessment, on 22 newer trials, scaled the estimate sharply down — ~3% LDL reduction at ~50g/day soy protein, comparable to but smaller than the original — and concluded that soy's main cardiovascular benefit operates through replacement of animal protein rather than a unique soy effect Sacks et al. 2006. The Tokede 2015 meta-analysis of 35 RCTs, restricted to whole soy foods (tempeh, tofu, soy milk, soybeans) rather than isolated soy protein, found LDL reductions of ~5 mg/dL and total cholesterol ~6 mg/dL, with greater effect in hypercholesterolaemic baseline Tokede et al. 2015. The FDA in 2017 proposed revoking the unqualified health claim citing inconsistent newer evidence; the claim's downgrade does not contest the effect's existence, only its magnitude. The Jenkins 2002 crossover trial in hyperlipidaemic adults compared high- and low-isoflavone soy foods and found similar LDL reduction in both arms, evidence that the protein fraction carries most of the lipid effect with isoflavones contributing little incrementally Jenkins et al. 2002.

The reviewer-anchor study to know cold: Anderson 1995 NEJM — 38 RCTs, ~13% LDL drop on ~47g/day soy protein, the basis for the FDA health claim. AHA 2006 — 22 newer RCTs, revised to ~3% LDL at ~50g/day. Tokede 2015 (whole soy foods) sits between at ~5 mg/dL LDL. The effect is real; the magnitude was inflated by older trials and the substitution component dominates the unique soy component.

Postprandial glucose and incident T2D. Mechanistic crossover trials show that substituting soy protein for white-rice carbohydrate at a meal flattens the glucose excursion and improves insulin sensitivity over the following hours. Prospective cohort evidence is strongest in Asian populations with high background intake: the Shanghai Women's Health Study followed 64,227 women without diabetes for 4.6 years and reported a relative risk of incident T2D of 0.62 in the highest quintile of legume/soy intake versus the lowest Villegas et al. 2008. Western cohorts with lower exposure ranges have produced weaker or null associations — likely a power/range issue rather than absence of effect Messina 2016.

Mineral bioavailability via phytate reduction. Egounlety & Aworh 2003 measured phytate before and after 36 hours of R. oligosporus fermentation of soaked and cooked soybeans: phytic acid dropped from ~14 mg/g dry matter to ~3 mg/g, a roughly 80% reduction; trypsin inhibitor was inactivated by the cook + fermentation combination; raffinose-family oligosaccharides fell by ~90% Egounlety & Aworh 2003. The bioavailability implication is concrete: at a typical animal-protein-replacement intake, the iron and zinc you'd predict from food composition tables (which assume unfermented soy phytate levels) underestimates actual absorption by a meaningful factor. The fermentation is the difference: soaked-and-cooked soybeans alone reduce phytate by only ~30%.

Protein quality. The Wang 1968 rat-feeding study established that tempeh's protein efficiency ratio modestly exceeds raw or unfermented cooked soybean, attributable to partial proteolysis (free-amino-acid content rises 50-fold over the substrate) and inactivation of trypsin inhibitors and lectins Wang et al. 1968. Modern PDCAAS analyses place tempeh in the 0.78–0.92 range depending on processing — below whey or egg (1.0) but above most other plant proteins (lentils 0.52, peanuts 0.52).

Bioactive peptides. Rhizopus proteolysis releases ACE-inhibitor, antioxidant, and immunomodulatory peptide sequences not present in unfermented soy; mechanistic and small-trial data exist, large clinical-endpoint trials do not Mani & Ming 2014.

Mortality (plant-protein substitution). NIH-AARP cohort, 416,104 adults followed 16 years: each 3% increase in caloric intake from plant protein in place of animal protein associated with 10% lower all-cause mortality and 11–12% lower cardiovascular mortality Huang et al. 2020. Tempeh sits in that substitution space; the cohort doesn't isolate tempeh specifically.

Protocol

A practical "regular protein source" intake is one or two servings most days of the week, where a serving is ~85–100g of tempeh (one quarter of a typical 400g block, ~16–20g protein, ~6–7g fibre, ~190 kcal). This places weekly soy-protein intake roughly in the 50–100g range — overlapping the 25g/day floor of the FDA health-claim regime and comfortably under the ~80g/day where any disputed concerns surface. Preparation needs no soaking; the cake slices like halloumi and takes any browning method: pan-sear, oven-bake, steam-then-glaze, crumble for "ground" applications. A 10-minute simmer in salted water before searing improves texture and reduces the residual bitterness some palates notice. Cooking does not destroy the protein, fibre, or mineral-bioavailability benefits; aglycone isoflavones are heat-stable to ordinary cooking temperatures. Deep-frying defeats the calorie density and saturated-fat profile that make tempeh worth eating.

Contraindications

True contraindications are narrow:

• Soy allergy. Absolute. Tempeh is whole soybean — protein content is the food.
• Levothyroxine timing. Soy reduces oral levothyroxine absorption; a documented case-series and small interaction trials show that co-ingestion can necessitate a dose increase to maintain TSH targets Bell & Ovalle 2000. Standard endocrinology advice: separate levothyroxine and a soy-containing meal by at least 4 hours, take the tablet on an empty stomach. The interaction is timing, not categorical avoidance.
• Hypothyroidism without iodine sufficiency. Soy isoflavones inhibit thyroid peroxidase in vitro; in iodine-sufficient euthyroid adults the meta-analysis of 18 trials found no clinically significant TSH or T₄ change Otun et al. 2019. The signal is real only at the intersection of subclinical hypothyroidism and frank iodine deficiency.
• MAOI users (rare). Tempeh, like other fermented foods, contains some tyramine; reported levels are modest compared to aged cheeses but high tempeh intake on classical MAOIs is reason for caution.

Concerns that do not rise to contraindication: men's reproductive hormones (Hamilton-Reeves 2010 meta-analysis of 51 studies, no effect on testosterone, free T, SHBG, or estradiol at intakes up to ~70g/day soy protein or ~240 mg/day isoflavones, in clinical trials and cohorts) Hamilton-Reeves et al. 2010; breast cancer (the JPHC cohort and follow-on meta-analyses find an inverse association in Asian populations, neutral in Western; soy is neither a risk nor a protective factor strong enough to drive a decision in either direction in low-baseline-intake Westerners) Yamamoto et al. 2003, Messina 2016; female reproductive hormones in healthy women — small and not clinically meaningful in pooled trials Hooper et al. 2009.

Misconceptions

The persistent misreads, ordered by how often they are repeated:

• Soy feminises men. The phytoestrogen → estrogen conflation. Isoflavones are selective oestrogen-receptor modulators (SERMs), not estrogens; binding affinity is hundreds-fold lower than estradiol, and the receptor-subtype preference is β over α, the opposite of the receptor that drives feminisation. Hamilton-Reeves 2010 is the clean answer: no hormone shifts at high intakes in men Hamilton-Reeves et al. 2010. The case-report folklore (one heavy-soy-milk drinker with gynaecomastia) is at the ~360 mg/day isoflavone level — roughly 12 servings of tempeh per day.
• Soy wrecks the thyroid. Iodine-sufficient adults: no Otun et al. 2019. The "wrecks the thyroid" claim conflates the rat literature on soy-formula-fed infants in iodine-deficient regions with adult human food consumption.
• Tempeh is a reliable B₁₂ source for vegans. Wishful. Some tempeh contains detectable cobalamin from incidental bacterial contaminants (notably Klebsiella pneumoniae, Citrobacter freundii, Propionibacterium), but the molecule present is often pseudo-vitamin B₁₂, a cobamide humans cannot use; reported levels are inconsistent batch to batch Watanabe 2007. Vegans need a supplemented source.
• Fermented soy is "estrogenic enough to matter" while tofu isn't. Total isoflavone content per gram is similar across whole-soy foods; the fermented form has more aglycones (bioavailable faster) but the same daidzein + genistein + glycitein totals on a dry basis Rizzo & Baroni 2018.
• "Unfermented soy is toxic; only fermented forms are safe." The traditional-foods folklore. Phytate, oligosaccharides, and trypsin inhibitors are real, but soaking + cooking handles most of them; fermentation handles more. Tofu, soy milk, and edamame are not toxic — they are simply different preparations with different anti-nutrient residuals.

Alternatives

Within the plant-protein category: tofu shares the soy fraction's lipid and isoflavone effects but loses much of the fibre (okara discarded) and retains higher phytate (unfermented); edamame retains fibre but has lower protein density and a less interesting cooking profile; natto uses Bacillus subtilis rather than Rhizopus, has a divisive flavour, and is the dominant dietary source of vitamin K₂ as MK-7 (tempeh has much less); lentils and other legumes match on fibre and price but are not complete proteins and provide no isoflavones. Within animal protein: chicken breast beats on protein density and PDCAAS, loses on fibre, isoflavones, and the CVD-substitution math. The case for tempeh is the rare combination — complete protein, fibre, fermented-mineral-bioavailability, low SFA, plant-substitution mortality story — in one cheap shelf-stable block.

Failure modes

The common ways the tempeh attempt does not deliver:

• Deep-fried tempeh on a takeaway plate. Doubles the calorie density, replaces the soybean oil with whatever the fryer was last used for, and saturates the calorie envelope the substitution was supposed to free up. The Indonesian street form is fried; the health-substitution form is not.
• Tempeh "bacon" / "nuggets" with added sugar, salt, and stabilisers. Branded products often add enough sodium and processed ingredients to negate the food-quality story. Whole tempeh blocks are the cheap, simple form.
• Eating it once a month and expecting a measurable effect. The LDL meta-analyses are at intakes of 25–50g soy protein per day. One stir-fry per fortnight does nothing on lipid panels; it is a flavour novelty, not an intervention.
• Sourcing only pasteurised, vacuum-sealed product and assuming live microbiology. Most commercial tempeh in Western markets is pasteurised at packaging — the mould is dead. The mineral-bioavailability gains (banked at fermentation) and the protein/fibre/isoflavone profile remain; the live-microbe contribution to gut transit is gone.
• Levothyroxine taken with the same meal. See contraindications. Quietly produces creeping TSH rises and a "soy made me feel sluggish" attribution that is actually a drug-absorption miss.

Practicalities

Cost in major Western grocers as of 2025: ~$3–5 per ~225g block, working out to roughly $0.80–1.30 per ~85g serving — comparable to lentils, cheaper than chicken breast, far cheaper than salmon. Shelf life: 3–4 weeks refrigerated unopened; 5–7 days once cut; freezes well for months. Available in mainstream supermarkets in the US, UK, EU, Australia; specialty Asian grocers stock fresher product. Home fermentation is straightforward (hulled soybeans + starter culture + incubator at ~30 °C for 36–48 h) and brings the per-serving cost under $0.30 but requires the temperature setup. No prescription, no clinician involvement, no biomarker testing required to start.

History

Tempeh is documented in Javanese culinary records from at least the 19th century, with linguistic and ingredient evidence suggesting pre-colonial origin probably in central or eastern Java. The substrate is whole soybean (introduced from China centuries earlier), the inoculant a Rhizopus species traditionally carried on hibiscus or banana-leaf wrappers Astuti et al. 2000. Tempeh spread to the Netherlands with returning colonials in the early 20th century, reached the broader West through the 1960s–70s vegetarian and natural-foods movements, and became a globally distributed product through the 1990s. Tempeh remains a daily protein source for tens of millions of Indonesians and the country's cheapest source of dietary protein per gram.

Stakes

The stakes shape, framed as the typical reader who skips this: ordinary Western default of animal-protein-dominant meals 7 days a week. The Huang 2020 NIH-AARP analysis is the lived-experience number — every 3% of daily calories shifted from animal to plant protein associated with a 10% reduction in all-cause mortality and 11–12% drop in cardiovascular mortality over 16 years Huang et al. 2020. The reader who never makes the swap is selecting a slightly worse trajectory on LDL, fibre intake (median Western adult eats ~half the recommended 25–30g/day), and saturated-fat displacement; none of these is a single dramatic event, all of them compound. Symmetric framing for the reader who already eats lentils-and-rice: tempeh is the incremental upgrade — higher protein density, complete amino-acid profile, fermented-mineral-bioavailability — not a categorical reframe.

Payoff

What the cascade looks like for the reader who installs one or two tempeh dinners a week, hinged to the evidence:

• Weeks 1–4. Direct meal-level felt experience: heavier satiety post-dinner than the equivalent-calorie pasta or chicken-and-rice meal — the protein + fibre combination is the lever. Postprandial glucose curves measurably flatter on CGMs for readers who track them.
• Months 2–6. If swap displaces ~50g/day of red or processed meat, modest LDL improvement of a few mg/dL — visible on the next lipid panel for readers in the borderline-elevated range; smaller for already-low-LDL readers Tokede et al. 2015. Iron and zinc status nudges up (fermented form is more bioavailable than the food-table number suggests).
• Year+. The compound substitution story — fewer high-SFA animal meals, more fibre, more plant protein — places the reader in the cohort whose long-term cardiovascular and all-cause mortality curves bend down. The size of the bend at one or two meals/week is a fraction of the full effect; the order of magnitude is consistent Huang et al. 2020.

Out of scope

Forward pointers: tofu as the lower-fibre, higher-versatility soy alternative; dietary fibre generally as the broader nutrient story tempeh is one delivery vehicle for; plant-forward eating patterns (Mediterranean, EAT-Lancet) as the macro story tempeh slots into; continuous glucose monitoring as the tool for personalising the postprandial-glucose claim; fermented foods as a broader microbiome story (yoghurt, kefir, kimchi, natto, sauerkraut).

Credibility range

Optimist case

Tempeh is a near-ideal protein vehicle. It is one of the few plant proteins approaching animal protein on PDCAAS, with the fibre, isoflavone, and fermented-mineral-bioavailability layers that animal protein can't match; the bioactive-peptide story is genuinely interesting (ACE-inhibitor, antioxidant sequences released by Rhizopus proteolysis); the Asian-cohort soy-and-T2D and soy-and-cardiovascular data are consistent and large; the cost is the cheapest of any complete-protein source; the cooking profile is broad; the safety record over generations of high-intake populations is excellent. The "substitute soy protein for animal protein" trial literature consistently lowers LDL even after the magnitude was revised down. As a once-or-twice-weekly default, the case is hard to argue against.

Skeptic case

The tempeh-specific clinical literature is thin. Almost all the cardiovascular and metabolic trial evidence is on isolated soy protein or soy milk, not whole tempeh; the read-across is mechanistic, not direct. The AHA 2006 downgrade is the honest read on soy's solo effect — small (~3% LDL) once the trial quality improved; the larger effect is the substitution story, which any plant protein delivers. Cohort studies showing strong soy-and-disease associations are concentrated in Asian populations whose total background intake (and life-long exposure) is hard to map onto a Western adult adding one block a week. Tempeh's B₁₂ reputation is folklore. Isoflavone safety is settled in iodine-sufficient adults but the population-level allergy and intolerance prevalence is non-trivial. The risk is overstating a worthy-but-modest food into "superfood" framing.

Author's call

Lands optimist on the food itself, conservative on magnitude. Tempeh is genuinely better than its substitution counterparts (refined-carb meals, processed-meat-anchored meals) and modestly better than unfermented soy preparations on mineral bioavailability and protein digestibility; the LDL and glucose effects are real but small enough that nobody should treat it as a substitute for statins, exercise, or weight loss where any of those is indicated. Action class: do, at one to several servings per week, framed as a default protein swap rather than an intervention. Evidence score moderate (3): the soy-cardiovascular base is strong, the tempeh-specific clinical layer is thin. Controversy score modest (2): the phytoestrogen and thyroid misreads persist publicly even though the literature has settled them.

Stakeholder and incentive map

• Soy-food industry. Modest commercial push, fragmented; no single dominant tempeh brand. Soy-protein-isolate makers (powders, meat-replacement products) have larger budgets and rode the FDA 1999 health claim hardest, with the AHA 2006 revision a real headwind.
• Plant-based and vegan advocacy. Cultural endorsement: tempeh is a frequent menu item in plant-based positioning. Real signal, occasional overclaiming on B₁₂ and on "fermented = magic."
• Carnivore-diet and animal-protein-maximalist communities. Active counter-positioning, leaning on the phytoestrogen and thyroid myths and on the legitimate observation that animal protein has higher PDCAAS. Magnifies edge-of-evidence findings into categorical bans.
• Mainstream nutrition bodies (AHA, USDA, EAT-Lancet, EFSA). Consistent endorsement of soy foods in moderate amounts as part of plant-forward patterns; tempeh specifically rarely singled out from the soy category.
• Indonesian government and culinary heritage organisations. Soft promotion as cultural patrimony; not a meaningful Western-market force.

Population variability

• Equol producers. Roughly 30–50% of Western adults and 50–60% of Asian adults harbour gut bacteria that convert daidzein to equol, a more potent ER-β ligand than the parent isoflavone Setchell & Cole 2006. Equol-producer status modulates the isoflavone-specific fraction of soy's effects (bone, vasomotor symptoms, possibly LDL); the protein and fibre fractions don't depend on it.
• Baseline LDL. The Tokede 2015 subgroup data show larger LDL drops in hypercholesterolaemic baselines than in already-low-LDL adults Tokede et al. 2015. The reader with LDL of 180 has more to gain from the swap than the reader with LDL of 90.
• Background diet. Substituting tempeh for steak gives larger effects than substituting for chickpeas. The "swap-from-what" matters more than the absolute tempeh intake.
• Iodine intake. Adequate iodine (iodised salt, seafood, dairy) is the condition under which the thyroid-safety meta-analyses landed neutral Otun et al. 2019. Strict-plant + no-iodised-salt + no-seafood is the population corner where soy's TPO-inhibition signal could matter.
• Lifelong vs adult-onset intake. Asian cohorts with consistent inverse breast-cancer associations all have lifelong exposure starting in childhood; Western adult-onset intake does not have equivalent supporting data and should not be assumed to confer the same risk profile.
• Soy allergy. Roughly 0.4% of Western adults; absolute exclusion.
• Levothyroxine users. Real interaction; spacing is the answer.

Knowledge gaps

• Tempeh-specific RCTs. Almost every soy-cardiovascular trial uses isolated protein or soy milk; tempeh-specific clinical-endpoint trials are scarce. The mechanistic read-across is plausible — same isoflavone and protein content, mostly — but the trial layer is missing.
• The fermentation-bioactive question. Whether the Rhizopus-derived peptides and de-glycosylated aglycones deliver measurable clinical benefit beyond what unfermented soy delivers is open. Mechanism plausible, dose-response in humans not characterised.
• Microbiome effects. Tempeh's impact on stool microbiota in humans is documented in a handful of small studies; the magnitude, persistence, and clinical relevance are unclear. Pasteurised vs unpasteurised tempeh comparisons are essentially absent.
• Bioactive-peptide pharmacokinetics. ACE-inhibitor peptide sequences have been identified in tempeh hydrolysates; whether they survive gastric digestion in active form at meal-realistic doses is not characterised.
• Western-cohort soy + endpoint data. Most cohort data is from Asian populations with very different exposure ranges; Western prospective data at the relevant intake range remains underpowered.

Scope vs brief. The brief named complete protein, fibre, isoflavones, fermentation bioactives, LDL, postprandial glucose, satiety, mineral bioavailability, gut microbiome, and protein digestibility. The article covers all of these except gut microbiome, which is mentioned only in the mechanism section as part of the fermentation story. Reason: the tempeh-specific human microbiome literature is genuinely thin (a handful of small studies, no consistent quantification of stool-microbe shifts or downstream clinical endpoints), and a dedicated microbiome section would have either padded with mechanistic speculation or read as overclaiming. Flagged in the dossier's knowledge gaps; will earn its own treatment once the literature thickens.

The tempeh-specific vs soy-generally read-across. The hardest editorial call. Almost all the cardiovascular and metabolic trial evidence is on isolated soy protein or soy milk, not tempeh — the read-across to tempeh is mechanistic (same isoflavones, same protein, plus fermentation-specific advantages on phytate and aglycone form). The article uses soy-protein literature as the load-bearing evidence and flags the tempeh-specific gap honestly in the evidence section's framing. Alternative considered: restrict claims to the few tempeh-specific human trials and lose most of the LDL/glucose story; rejected as more misleading than the read-across.

Dream narrative written below threshold. Overall score landed around 22 (well below the 40+ where the narrative is obligatory). Chose to write one anyway because the substitution-cumulates-over-years story has a real, evidence-hinged aspirational cascade (NIH-AARP plant-for-animal mortality data); the dek and tagline lean lightly on it without lifting marketing voice all the way. A purely-clinical dek would have under-sold what the food honestly does over years.

Rating difficulties.

• longevity = 3. The soy-specific lipid effect alone wouldn't earn this; the substitution-with-animal-protein effect does, but only if you credit tempeh with the broader plant-substitution story. Defensible because that is the mechanism by which tempeh acts on mortality at population scale; the alternative score of 2 would have been defensible too.
• energy = 1 / beauty_cumulative = 1. Both borderline 0-vs-1 calls. Kept at 1 because the stable-glucose-prevents-afternoon-flatness effect is real and felt, and the long-run body-composition story rides on the broader dietary pattern. Either could have honestly been a 0.
• evidence = 3, not 4. The soy-protein-and-LDL trial base could support a 4 on its own; downgraded to 3 because the tempeh-specific clinical layer is thin and the entry's claims rest on read-across.
• applicability = 3, not 4. Borderline. The substitution decision applies to most omnivore adults; the food itself is a niche cultural product in Western markets. Kept at 3.

Contraindications token. Considered thyroid-condition for the levothyroxine interaction; rejected because the closed-vocab token signals "this is unsafe for you," whereas the levo case is a timing issue solved by four-hour spacing. Covered in prose + warning callout instead.

Future links once they exist. Tofu, tempeh-vs-tofu-vs-natto comparison, plant-forward eating patterns / Mediterranean, dietary fibre, continuous glucose monitoring, fermented foods generally. Pointed to in the closing addressing section as forward signposts; wire up when adjacent entries land.

Separate-entry candidates surfaced. Soy isoflavones (the SERM / equol-producer pharmacology is its own entry, not a tempeh sub-section). Phytate as anti-nutrient (cross-cutting across legumes and grains; tempeh is one fermentation answer among several). Plant vs animal protein substitution (the broader cardiometabolic story; tempeh is one tactic).