Every row, every source.

JSON · row 0

{
  "supplementAName": "Vitamin D3",
  "supplementBName": "Vitamin K2",
  "interactionType": "synergy",
  "severity": "info",
  "evidenceLevel": "moderate",
  "description": "Vitamin D3 increases calcium absorption, while K2 activates
   osteocalcin and matrix GLA protein to direct calcium into bones and away
   from arteries.",
  "mechanism": "D3 upregulates calcium absorption; K2 activates calcium-binding
   proteins (osteocalcin, MGP) that shuttle calcium to bone matrix and prevent
   arterial calcification. Note: 2023 meta-analysis of 14 RCTs found vitamin K
   supplementation did NOT significantly prevent vascular calcification.",
  "recommendation": "Take together. K2 (MK-7 100-200mcg) is recommended whenever
   supplementing D3 above 2000 IU.",
  "directionality": "bidirectional",
  "minimumTimeSeparation": null,
  "sources": [
    { "text": "Masterjohn C. Vitamin D toxicity redefined. Med Hypotheses. 2007",
      "pmid": "17141962", "doi": "10.1016/j.mehy.2006.09.051" },
    { "text": "Kidd PM. Vitamins D and K as pleiotropic nutrients. J Orthomol Med. 2010" },
    { "text": "Chang MC, Choo YJ. ... Sarcopenia: A Systematic Review and Meta-Analysis.
       Nutrients. 2023.", "pmid": "36771225", "doi": "10.3390/nu15030521" },
    { "text": "Kuang X, et al. The combination effect of vitamin K and vitamin D on
       human bone quality: a meta-analysis of RCTs. Food & Function. 2020.",
      "pmid": "32219282", "doi": "10.1039/c9fo03063h" }
  ]
}

Notice the mechanism field includes a caveat from a 2023 meta-analysis that partially undercuts the popular claim. We do not omit inconvenient evidence.

When sources disagree

The mechanism field captures the disagreement and the row's evidence tier drops. A "strong" claim that gets contradicted by a quality 2023 meta-analysis becomes "moderate" in the next release. The popular old framing does not survive simply because it is popular.

Synergy · Vitamin C + Iron · Strong evidence

Vitamin C converts dietary iron into the more absorbable ferrous form.

Mechanism. Ascorbate reduces ferric (Fe³⁺) iron to ferrous (Fe²⁺), increasing absorption by 2–3x in iron-deficient subjects.

Recommendation. Take 200–500mg Vitamin C with iron supplements, especially non-heme (plant-source) iron.

Sources: PMID 1831564, PMID 33237064, PMID 31860103.

Caution · Zinc + Copper · Strong evidence

Chronic high-dose zinc (>40mg/day) induces copper deficiency.

Mechanism. Zinc upregulates intestinal metallothionein, which binds copper preferentially and sequesters it in enterocytes until they are shed.

Recommendation. When zinc exceeds 40mg/day for more than two months, add 1–2mg copper. Below 40mg/day, no copper needed.

Sources: PMID 38690587.

Timing-sensitive · Calcium + Magnesium · Moderate evidence

High-dose calcium and magnesium compete for absorption when taken simultaneously.

Mechanism. Both divalent cations use overlapping transporters (TRPM6/7 for Mg, calcium-sensing receptor and others for Ca). Saturation reduces absorption of whichever is in lower local concentration.

Recommendation. Separate by 2–3 hours. Common pattern: calcium with breakfast, magnesium at bedtime.

Sources: PMID 1955626, PMID 33237064.

Conflict · Iron + Calcium · Strong evidence

Calcium materially inhibits iron absorption when taken at the same dose.

Mechanism. Calcium competes with iron at the divalent metal transporter DMT1 and forms insoluble complexes in the gut lumen. Reductions of 30–60% in non-heme iron absorption have been measured.

Recommendation. Separate iron and calcium by at least 2 hours. Multivitamins that pack both reduce the value of both.

Sources: PMID 1831564, PMID 33237064, PMID 31860103.

Contraindicated · 5-HTP + L-Tryptophan · Strong evidence

Both convert to serotonin. Stacking them risks serotonin syndrome.

Mechanism. L-Tryptophan is converted to 5-HTP by tryptophan hydroxylase, then to serotonin. Supplemental 5-HTP bypasses the rate-limiting step. Combining them produces additive serotonin synthesis with no biological brake.

Recommendation. Do not combine. Choose one. Avoid either while on SSRIs or MAOIs.

Sources: PMID 28839121.

Contraindicated · Warfarin + St. John's Wort · Strong evidence

St. John's Wort dramatically accelerates warfarin metabolism.

Mechanism. Hyperforin in St. John's Wort activates the pregnane X receptor (PXR), which upregulates CYP3A4, CYP2C9, and P-glycoprotein. Both S- and R-warfarin are cleared faster, INR can drop below the therapeutic window, and stroke or clot risk rises.

Recommendation. Do not combine. Inform your prescriber if you have taken both.

Sources: PMID 15089812, PMID 10683750, PMID 33769581.

Timing-sensitive · Levothyroxine + Calcium · Strong evidence

Calcium chelates levothyroxine in the gut, cutting absorption by 20–25%.

Mechanism. Calcium cations form insoluble chelate complexes with levothyroxine in the acidic stomach, blocking transit across the intestinal mucosa. Subtherapeutic thyroid levels and dose escalation follow.

Recommendation. Separate by at least 4 hours. Take levothyroxine 30 minutes before breakfast on an empty stomach; calcium with later meals.

Sources: PMID 10838651, PMID 21595514, PMID 34766382.

When your stack is scored, every pairing is canonicalized by alphabetical sort before lookup. "Calcium + Iron" and "Iron + Calcium" resolve to the same row. The penalty applies once, not twice.

The directionality field on each row records whether the biology is symmetric (bidirectional, e.g. mineral competition) or asymmetric (A−affects−B, e.g. St. John's Wort accelerates warfarin but not vice versa). Asymmetric rows trigger a more specific recommendation in the app, but they still match in both stack orders.

• Genotype-specific interactions are out of scope. CYP2D6 poor metabolizers will respond differently to several SSRIs and stimulants. We flag the population-level risk; we do not pretend to know your variants.
• Dose-response curves are simplified. Most rows assume "typical" supplemental doses. Where the threshold matters (e.g. zinc >40mg/day for copper), it is named in the recommendation.
• Three-way interactions are not modelled. Adding a third agent that modifies one half of an existing pair (e.g. an SSRI on top of 5-HTP plus L-Tryptophan) is flagged via the individual pairings but not as a triple. A known limitation.
• Tier "limited" is small on purpose. If the literature is too thin to support a useful claim, we mark four rows "limited" rather than promote them to "emerging" for completeness.

• Quarterly PubMed re-query. Every interaction pair gets a fresh literature search. Findings that change the tier or recommendation are merged into the next app release.
• Public downgrades. When a strong claim weakens, the tier drops in the release notes. The old number is not quietly retained.
• Reader corrections. Found a citation we missed or a row that is out of date? Email support@nutristackapp.com with the PMID. Verified corrections ship in the next release.