The common belief vitamin C prevents colds is refuted for the general population and true only in a narrow form. On the current gold systematic review (29-31 trial comparisons, 11,306 participants), regular high-dose (>= 0.2 g/day) vitamin C does not reduce how many colds a normally-nourished person catches; it does modestly shorten a cold once it starts; and it halves cold incidence in one identified stratum — people under brief, heavy physical stress. Vitamin C started at the onset of symptoms (therapeutic use — the way most people actually reach for it) has shown no consistent benefit. The exposure tested is large doses added to the already-replete, not correction of a deficiency (Hemilä & Chalker, 2013).
Incidence: no effect in the general community (the belief, refuted)
Across 24 general-community comparisons (10,708 participants) regular vitamin C did not reduce the proportion of people catching at least one cold:
«Based on 24 entries with 10,708 participants from the general community who had no heavy short-term physical stress, the narrow CI, which is located close to the zero effect, refutes the possibility that regular vitamin C supplementation could reduce the average incidence of colds in the general community: RR 0.97 (95% CI 0.94 to 1.00).» (Hemilä & Chalker, 2013)
This is a precise null, not an unstudied gap: the CI is narrow and sits on 1.0. It is not an artifact of low-dose trials — restricting to >= 1 g/day gives RR 0.98 (0.95-1.01, 6661 participants), and the single 3 g/day trial gave RR 0.93 (0.73-1.20). The full 29-entry pool (RR 0.95, 0.92-0.98) is statistically significant, but the review reads the narrow CI as precluding any clinically relevant effect across broad population groups — a case where statistical significance and decision-relevance part company (Hemilä & Chalker, 2013).
The real exception: heavy acute physical stress (route-(b) effect modification)
The pooled incidence effect is heterogeneous (I2 = 38%), and the heterogeneity resolves into two subgroups that are homogeneous within (I2 = 0) and significantly different from each other — positive interaction evidence, not a mechanistic guess. In the physical-stress subgroup:
«In five trials with participants undergoing heavy acute physical activity in the subgroup at the bottom of Analysis 1.1, vitamin C halved the incidence of colds: RR 0.48 (95% CI 0.35 to 0.64) (Figure 2, Analysis 1.1.2). All of these five studies were randomised and double-blind.» (Hemilä & Chalker, 2013)
This is route-(b) (effect modification) in the five-routes scheme: the relative effect itself changes by stratum, backed by a positive, homogeneous, significant subgroup contrast — not merely a higher absolute benefit at a higher baseline risk. Two facts make the subgroup credible rather than a fishing artifact: doses were 0.25-1.0 g/day (so the benefit is not explained by high dose — similar doses did nothing in the general community), and all five trials were randomised and double-blind (Hemilä & Chalker, 2013).
The stratum is specifically short, acute exertion (marathon runners, skiers, subarctic soldiers). Two trials of chronic physical stress (marine recruits over 2 months; competitive swimmers over 3 months) found no incidence effect — the modifier is acute exertion, not athletic training in general (Hemilä & Chalker, 2013). Caveat on the outcome: some of these post-exertion colds may be exercise-induced bronchoconstriction (airway injury) rather than viral infection; the review holds the effect on acute respiratory symptoms as firm while noting the aetiology is not clean (Hemilä & Chalker, 2013).
Duration and severity: a real but modest effect of regular supplementation
Regular (prophylactic) vitamin C shortens colds that do occur — a consistent, statistically robust, but small effect:
«In adults the duration of colds was reduced by 8% (3% to 12%) and in children by 14% (7% to 21%). In children, 1 to 2 g/day vitamin C shortened colds by 18%.» (Hemilä & Chalker, 2013)
Read these as relative reductions on a short baseline: 8% of an adult cold of a few days is a fraction of a day per episode, for a handful of episodes a year, bought by taking vitamin C every day year-round. The review’s own verdict is that this «does not justify long-term supplementation in its own right» (Hemilä & Chalker, 2013). Severity is also reduced (combined P < 0.00001) but «modest in absolute terms» (Hemilä & Chalker, 2013). The child-vs-adult gap is the more robust contrast; within-trial dose hints (2 vs 1 g/day in children; 0/3/6 g/day linear trend in adults) suggest but do not establish dose-dependency (Hemilä & Chalker, 2013).
Therapeutic use (at onset): no consistent benefit shown — but the null is fragile
The intuitive move — regular use shortens colds, so take it when a cold starts — is not supported by the trials:
«No consistent effect of vitamin C was seen on the duration or severity of colds in the therapeutic trials.» (Hemilä & Chalker, 2013)
But this is a weaker null than the incidence null — insufficient/inconclusive rather than a clean no-effect verdict. Only 7 therapeutic comparisons (3249 episodes) exist, none in children, and the review argues the null may be a protocol artifact: supplementation often too short (< 5 days; 5-day trials hinted benefit), started too late (> 24 h after onset; “the sooner the better”), or too low a dose (an 8 g first-day dose beat 4 g). The authors call for further therapeutic RCTs and — given vitamin C’s low cost and safety — say it is not unreasonable for an individual patient to test therapeutic use on themselves. That is an individual-trial suggestion, not an evidenced population recommendation (Hemilä & Chalker, 2013).
Safety and the layer-1 verdict
No harm signal: 5.8% of vitamin-C recipients (>= 1 g/day) vs 6.0% of placebo reported adverse symptoms, none serious; vitamin C is considered safe to several g/day (Hemilä & Chalker, 2013). So the decision is not harm-vs-benefit but benefit-vs-nothing: for the general population the prophylactic lever is effectively null on incidence and small on duration, ranking low — the honest result is this does not move the outcome enough to be worth a daily habit, which is itself a decision-change (permission to stop). The lever is genuine only in the acute-physical-stress stratum.
Two cross-cutting lessons
Repletion is a different exposure than enhancement. The review deliberately tests large doses in the already-nourished (low-dose placebos were used to keep participants non-deficient). Separately, excluded UK trials from a low-dietary-intake era found incidence reductions in males (pooled RR 0.54, 0.40-0.74), attributed to low baseline dietary vitamin C rather than the supplement dose. Correcting a marginal deficiency and enhancing an already-replete person are not the same question -> Deficiency Repletion vs Enhancement (Hemilä & Chalker, 2013).
Pauling’s error is a transportability failure, not a data error. Pauling based mega-dose advocacy on the Ritzel ski-school trial — which the review places in the physical-stress stratum. Putting weight on that trial was reasonable; extrapolating its stratum-specific effect to the whole population was the mistake. A real effect in one stratum does not transfer to another by evidence grade alone (Hemilä & Chalker, 2013).
Open threads / gaps
- Therapeutic use is insufficient-evidence, not disproven — an adequately-dosed (>= 8 g adults,
= 2 g children), early (< 24 h), >= 5-day therapeutic RCT, including children, is the named gap the review itself flags. a therapeutic-protocol SR would upgrade it.
- Dietary-status as an effect modifier (the UK-male / fruit-juice-interaction signal) is a candidate route-(b) modifier not resolved by the included (>= 0.2 g/day) trials.
- The pneumonia signal (Pitt 1979: 8 pneumonias, 1 on vitamin C, P = 0.044) points beyond the cold to other respiratory infections — out of this page’s scope, a separate question.