The decision. For someone eating a vegetarian or vegan diet, should they supplement vitamin B12, and how is their B12 status actually measured? This is the repletion mirror of the general supplements question: whereas supplementing the already-replete adult to prevent disease is null-to-harm (the enhancement arm -> Vitamin and Mineral Supplements for Disease Prevention), a vegetarian/vegan is at genuine risk of a real dietary shortfall of a nutrient with essentially no plant source — so this sits on the steep lower (repletion) arm of Deficiency Repletion vs Enhancement, where correcting the deficit can matter. The single source held is a descriptive prevalence review (Pawlak et al., 2013); its findings are prevalence and measurement facts, not an effect estimate on a patient-important outcome.
The actionable core — a genuinely non-optional supplement for this stratum
B12 is synthesized only by microorganisms, so it is absent from plant foods except via soil contamination or fortification (Pawlak et al., 2013); a vegan’s diet contains none unless fortified, and a vegetarian’s few natural sources (dairy, eggs) empirically fail to keep many replete. This is not the discretionary micronutrient supplementation the fabric ranks at the bottom of the hierarchy — it is a structural gap created by the diet itself, so B12 is one of the few supplements that is close to non-optional in an identified stratum, and Pawlak’s own recommendation is unambiguous: «Supplements, when taken in an adequate dose, are very effective in both the prevention and treatment of B12 deficiency and are very cost effective» (Pawlak et al., 2013). Large, cheap, well-tolerated, and correcting a genuine shortfall — the classic profile of a high-value lever for the stratum that has the deficit.
The important restraint (see Surrogate section below): warranted rests on the repletion logic and the low cost/harm of the lever, NOT on a held demonstration that biochemical deficiency in this stratum translates into a patient-important harm. The recommendation is well-founded as cheap insurance against a real shortfall; it is not licensed to claim a quantified clinical benefit the source does not carry.
Dosing note (out of scope to prescribe; carried for magnitude). Because fractional absorption is
low, the maintenance dose is far above the 2.4 ug/day RDA — cited secondhand as ~100x RDA for
maintenance and ~200x in established deficiency, and >=250 ug for best results
(Pawlak et al., 2013). Selection/titration is a
prescriber act (needs this person’s labs) and is out of the wiki’s scope; the number is here only to
size the lever, not to prescribe.
Prevalence — high, but cutoff-driven and one-armed
Across 18 studies using accurate biomarkers, the prevalence of B12 depletion/deficiency among vegetarians ranged «from about 11 to 90%» (Pawlak et al., 2013) — 62% in pregnant women, 25-86% in children, 21-41% in adolescents, 11-90% in adults/elderly. Two load-bearing caveats on that band:
- It is prevalence, not risk difference. Every rate is a one-arm proportion; the review holds no omnivore comparator anywhere, so nothing here quantifies how much the diet (versus the marker cutoff, life-stage, or absorption problems) drives the deficiency .
- Cutoff heterogeneity is a major driver of the band, not just the population. The serum-MMA
threshold «varied widely» (source’s words) — roughly 3-fold, from
>0.260to>0.75 umol/L; the lowest rate (Gilsing’s 11% in UK vegans) used the least stringent criterion, and rates fell further where studies required both low holo-TCII and elevated MMA (Pawlak et al., 2013). So the ~8-fold11-90%span is partly an artifact of which cutoff each study chose (compounded by real life-stage/duration differences) — read it as cutoff-and-population-dependent, not as a stable population parameter.
The measurement finding — serum B12 is the wrong marker
The decision-relevant measurement point: the marker changes the answer. «MMA and holo-TCII are the most accurate» while serum/plasma B12 and MCV are unreliable (Pawlak et al., 2013); studies resting on serum B12 «likely underestimated the rate of deficiency due to the inherent inaccuracy of this assessment method» (Pawlak et al., 2013). Holo-TCII marks the earlier depletion stage; MMA marks deficiency once stores are exhausted — and in 3 of 4 studies reporting both, the holo-TCII (depletion) rate exceeded the MMA (deficiency) rate, consistent with depletion preceding deficiency. Decision consequence: a normal serum-B12 result does not rule out a depleting vegetarian; assessment should use MMA and/or holo-TCII. This is a domain-specific instance of the wiki’s broader point that the instrument that produced a status estimate must be interrogated before the estimate is trusted -> Measurement Error in Dietary Assessment.
The stratum finding — every vegetarian subtype, with two gradients
The review’s main claim: «vegetarians develop B12 depletion or deficiency regardless of demographic characteristics, place of residency, age, or type of vegetarian diet» (Pawlak et al., 2013) — i.e. it is not confined to vegans. Two gradients ride on top of the shared risk:
- Strictness: vegan > lacto-/lacto-ovo-vegetarian (Herrmann 2001: 43% vegan vs 32% LV/LOV) (Pawlak et al., 2013).
- Duration / since-birth: those vegetarian from birth fared worse than later adopters (US macrobiotic children: 67% lifelong vs 25% later), plausibly via low maternal stores passing to offspring (Pawlak et al., 2013). The common belief that deficiency takes 20-30 years to develop is refuted — Donaldson’s raw vegans reached 47% deficient after only ~2-4 years (Pawlak et al., 2013).
This makes B12 a diet-defined monitoring flag for the whole vegetarian stratum (a route-(c)-flavored the exposure — the diet — creates the shortfall) rather than an effect-modification subgroup claim . Note supplement use did not guarantee repletion in these observational samples — Herrmann 2003 found 31% of LV/LOV supplement users still deficient (vs 68% of non-users), and, anomalously, vegan supplement users were more often deficient (88%) than non-users (83%), which Pawlak reads as too-infrequent or too-low dosing rather than supplements failing (Pawlak et al., 2013).
CRITICAL — biochemical deficiency is a SURROGATE, not a patient-important outcome
The prevalence numbers measure a biomarker, not harm, and the source itself supplies the discipline. In the study that looked, «none of the vegetarians included in their study had clinical symptoms despite the fact that about two-thirds of the sample had B12 depletion or deficiency, as indicated by both low holo-TCII and elevated MMA» (Pawlak et al., 2013). So an X% biochemically deficient rate must NOT be read as X% harmed — the transmission from marker to patient-important outcome is exactly the surrogate-validity question the wiki holds open by default -> Surrogate Outcomes.
Both arms must be stated, because Pawlak states both. He immediately counter-caveats that biochemical deficiency should not be dismissed as harmless: «one has to be careful with concluding that biochemical indicators of B12 defi- ciency in vegetarians are not associated with any adverse symptoms» (Pawlak et al., 2013). His reasons: mild symptoms (lethargy, forgetfulness) may go undetected or be blamed on aging; the hematological sign (elevated MCV) can be masked by concurrent iron deficiency or high folate — both more common in vegetarians; and B12-dependent enzyme function is already impaired at the biochemical stage, plausibly raising homocysteine. Neither side is an outcome finding. The associations Pawlak cites between deficiency and vascular disease / dementia / bone (his refs 17, 35, 36) are cited, not evidenced in this review, and the one within-study clinical read found zero symptoms. The honest state is insufficient evidence on marker-to-outcome transmission in this stratum, held apart from the well-supported prevalence and measurement facts.
Confidence — graded per claim, not as a blob
- Prevalence + measurement facts: well-supported by an appropriately-designed (biomarker-restricted) descriptive review — held with reasonable confidence, subject to the cutoff-dependence and one-arm/no-comparator caveats above.
- The supplement-is-warranted recommendation:
medium— it rests on sound repletion logic plus the lever’s low cost/harm, not on a held demonstration that correcting biochemical deficiency improves a patient-important outcome in this stratum. - Clinical-harm transmission: insufficient evidence — do not upgrade the prevalence numbers into a harm claim.
No [E-independent] is claimed: a single descriptive review is one route, not independent convergence.
Gaps (G)
- No patient-important-outcome effect for this stratum. Does correcting B12 deficiency in vegetarians/vegans measurably reduce a hard endpoint (neurological, cognitive, vascular, or pregnancy/offspring outcome)? Not held. — an outcome SR/RCT on B12 repletion -> patient-important endpoints in this stratum`
- No omnivore comparator / no relative effect. Prevalence is one-armed; the diet-attributable fraction is unquantified.
- Dose-response for repletion is unheld — the ~100x/200x-RDA and
>=250 ugfigures are cited secondhand, not pooled; a same-outcome dose-response is absent. - Maternal/infant B12 and offspring neurodevelopment is flagged by the since-birth gradient but not quantified on outcomes here — a candidate high-value cell (a real developmental window, cf. the folate and iodine window-gated repletion cases on Deficiency Repletion vs Enhancement).