All topics › Pregnancy preparing for birth
Is any amount of alcohol safe in pregnancy, and what do we actually know about light drinking, miscarriage, birthweight, and fetal alcohol spectrum disorders?
Heavy drinking in pregnancy causes well-established harm. But the question most parents actually ask is narrower: is any amount safe? A glass of wine at a wedding, a drink before you knew you were pregnant — what does the evidence say about light drinking?
The UK Chief Medical Officers' guidance (2016) is unambiguous: if you are pregnant or think you could become pregnant, the safest approach is not to drink alcohol at all [1]. That same guidance also says the risk is likely low after only small amounts of alcohol before the pregnancy was known, and that most such pregnancies are unlikely to have been affected [1] — which is both a medical statement and a mercy.
Here's the honest shape of the evidence: heavy alcohol use is an established cause of fetal alcohol spectrum disorders (FASD), with lifelong effects on the child [1]. Light drinking (a few units a week) has not been shown to cause consistent harm in the available studies — but those studies are observational, light drinkers tend to be more socioeconomically advantaged than abstainers, and the evidence is too thin to declare any amount safe [2][3]. The "don't drink at all" advice is precautionary guidance built on that uncertainty, not a threshold established by trials.
The UK CMO guideline is a precautionary one: because no safe threshold has been established, the safest approach is not to drink at all [1]. It is not saying that a single small drink is proven to cause harm — it is saying the evidence can't rule harm out. The same document deliberately includes the counterweight: after small amounts drunk before the pregnancy was known, "most pregnancies are unlikely to have been affected" [1]. That sentence exists because the "before I knew" scenario is the most common source of anxiety, and the guidance intends to defuse it rather than feed it.
Fetal alcohol spectrum disorders — the umbrella term for the growth, facial, and neurodevelopmental effects of prenatal alcohol exposure — are the one outcome where causation is settled: the CMO report describes alcohol as causing "a range of lifelong conditions" under the FASD umbrella, with severity linked to the amount drunk and the developmental stage at the time [1]. The dose and pattern that produces FASD is sustained heavy drinking — not an occasional glass — but nobody has identified a threshold below which FASD risk is zero, which is why the guidance doesn't try to. How common FASD is in the UK isn't directly measured, and the modelled estimates are too uncertain to quote usefully, so this topic doesn't.
The best synthesis of prospective evidence (Mamluk et al., 2017: a systematic review and meta-analysis of prospective studies of light drinking, defined as up to 32 g of alcohol per week — roughly 4 UK units) found the evidence sparse beyond birth size and gestational age [2]. For being born small for gestational age, seven adjusted studies gave a summary odds ratio of 1.08 (95% CI 1.02–1.14) — a small association; for preterm birth, nine studies gave 1.10 (0.95–1.28), compatible with no association at all [2]. The authors' conclusion was blunt: residual confounding means no robust conclusions can be drawn [2].
Illustratively: by definition, about 100 in 1,000 babies are small for gestational age (it's the bottom 10th percentile). An odds ratio of 1.08 would move that to roughly 108 per 1,000 — about 8 extra per 1,000 — if the association were causal, which the review itself does not claim [2].
An earlier synthesis (Patra et al., 2011: 36 observational studies) found no detectable increase in low birthweight or small-for-gestational-age up to 10 g of alcohol a day (a little over one UK unit), and no detectable increase in preterm birth up to 18 g a day — with risk rising above those levels [3]. That is sometimes quoted as evidence of a safe zone. It isn't one: the exposure data were self-reported, confounder adjustment varied, and "healthy drinker" confounding can make small amounts look protective when they aren't [3]. The CMO expert group drew the same line from its own evidence review: the risks of low birthweight, preterm birth, and being small for gestational age "may all be increased" above 1–2 units a day [1].
The studies don't share a definition, which is part of why they can't be stacked neatly. Mamluk's review defined light drinking as up to 32 g of alcohol a week — roughly 4 UK units, or about two and a half small glasses of wine [2]. Kelly's Millennium Cohort analysis used up to 1–2 units a week or per occasion [4]. Patra's dose-response work went as low as 10 g a day (a little over one unit) [3]. All of these rely on mothers' own reports of what they drank — and pregnancy is exactly the context where drinking is most likely to be under-reported [2][3]. So "no harm detected at light levels" means "no harm detected at reported light levels, in studies that couldn't fully adjust for who the light drinkers were."
The largest long-term follow-up (Kelly et al., 2009: ~12,500 mothers in the UK Millennium Cohort) asked about drinking at 9 months postpartum and tested the children at age 3. It found no consistent evidence that light drinking (up to 1–2 units a week or per occasion) harmed behaviour or cognition — and in unadjusted analyses, children of light drinkers actually scored slightly higher on cognitive tests [4]. That advantage shrank after adjusting for the mothers' education, income, and home environment — which is the point: in this cohort, light drinkers were systematically more advantaged than abstainers, and no statistical adjustment fully removes that [4]. Later follow-ups in other cohorts tell the same story: no consistent harm detected, no proof of safety [2].
The "before I knew" drink. This is the scenario the CMO guidance explicitly addresses: a small amount of alcohol consumed before the pregnancy was known carries likely low risk, and most such pregnancies are unlikely to have been affected [1]. That is not a guarantee — no study can give you one — but it is the closest thing to official reassurance that exists. Anxiety about it is normal and common; the guidance is written with that in mind.
Abstinence is a social decision as much as a medical one. Nine months without alcohol affects evenings out, celebrations, work events, and how you explain yourself to people who don't know you're pregnant. It helps to have a plan before you need one: a non-alcoholic drink you actually enjoy, a ready answer for curious colleagues ("I'm on a health kick" works until you're ready to share the news), and — biggest of all — a partner who joins in. Partners matter here: there is no trial evidence on a partner's drinking and pregnancy outcomes, but a partner who drinks alongside an abstinent pregnant person is making her life harder for no medical reason — and a partner who joins in makes abstinence easier socially. None of the studies measured the parental side of abstinence — its effect on mood, social life, or relationship dynamics is an evidence gap, not a finding.
If cutting down is hard, that's a medical issue, not a moral one. Difficulty stopping drinking in pregnancy deserves the same non-judgmental clinical response as any substance-use concern — tell your midwife or GP, who can offer support and referral. The stigma around drinking in pregnancy keeps some people from asking for help, which helps nobody.
| Exposure / comparison | Outcome | Effect in absolute terms | Evidence rating |
|---|---|---|---|
| Any drinking vs none | — | UK CMO: safest is not to drink at all (precautionary guidance; no established safe threshold) [1] | C (guidance built on uncertain evidence) |
| Light drinking ≤32 g/week (~4 units) vs none | Small for gestational age | Summary OR 1.08 (1.02–1.14); illustratively ~100 → ~108 per 1,000 (observational; causality unproven) [2] | C |
| Light drinking ≤32 g/week vs none | Preterm birth | Summary OR 1.10 (0.95–1.28) — compatible with no association [2] | C (for the null) |
| ≤10 g/day vs none | Low birthweight / SGA | No detectable increase in 36-study meta-analysis (observational; self-reported exposure) [3] | C |
| ≤18 g/day vs none | Preterm birth | No detectable increase [3] | C |
| Heavy sustained drinking | FASD (growth, facial, neurodevelopmental effects) | Established cause; severity linked to amount drunk; no safe threshold identified [1] | B (causation) |