All topics › Pregnancy preparing for birth
Which pregnancy food rules are backed by real evidence, and which rest on thin data?
The NHS pregnancy food list is long and forbidding: no soft cheese, no smoked salmon, no cured ham, no liver, no runny eggs (unless they're the right eggs), no liquorice tea, careful with the tuna. But the items on the list are not there for the same reasons, and they are not equally well supported.
The honest split is this: a few restrictions rest on real evidence of harm, especially around undercooked meat and high-mercury fish. A larger number are precautionary, based on thin, one-group, or disputed evidence. And the "what to eat" side of the story is thinner than the "what to avoid" side — with one genuine exception, fish.
This topic follows the standard guidance where it exists and says plainly where the evidence stops.
Listeria monocytogenes is a bacterium that thrives in ready-to-eat chilled foods — soft mould-ripened cheese, pâté, cold smoked or cured fish, deli salads and prepacked sandwiches. In most adults it causes mild illness or nothing at all. In pregnancy it can cross the placenta and cause miscarriage, stillbirth, or serious infection of the newborn.
How common is it? In England and Wales in 2024, surveillance recorded 179 confirmed listeriosis cases, of which 37 were pregnancy-associated — roughly one in five. Among pregnancy-associated cases where the outcome was known, 43.7% ended in miscarriage or stillbirth. That sounds terrifying, and it is a grim number, but remember the denominator: 37 cases against roughly 600,000 births a year in England and Wales is about 6 in 100,000 births. Rare, but with devastating outcomes when it happens. [B — national surveillance]
Pregnancy is a genuine risk factor: one French surveillance-based analysis estimated that pregnant women face roughly 100 to 1,000 times the listeriosis risk of healthy under-65s. [B — surveillance-based estimate]
The practical upshot: the rule is not really "avoid cheese" — it is "avoid ready-to-eat chilled foods unless they are cooked until steaming hot." That is why cold cured meats and cold smoked fish are on the list even though they don't fit any single theme. The investigated outbreaks in 2024 included smoked fish, garlic sausage, prepacked sandwiches, and a chocolate-and-strawberry mousse — which tells you the risk is about cold, stored, ready-to-eat food generally, not a magic property of any one item.
Toxoplasma gondii is a parasite whose congenital form — infection that passes to the baby — can cause miscarriage, stillbirth, or lasting damage. In Europe, congenital toxoplasmosis affects between 1 and 10 in 10,000 newborns. [B — epidemiology]
The strongest European case-control study of how pregnant women actually get infected (252 infected women, 858 controls, six centres) found that undercooked and cured meat accounted for an estimated 30 to 63% of infections, depending on the centre. Contact with cats was not associated with infection in that study. [B — multicentre case-control]
That flips the popular wisdom: the cat litter story is the memorable one, but the evidence points at the steak. Meat that is rare, blue, or cured-but-not-cooked (parma ham, salami) is the main food route, and this is one restriction with a clear evidence base behind it. Cook meat — including cold cured meats — until steaming hot.
Fish is where the list stops being a simple story of avoidance, because the evidence genuinely pulls in two directions.
On the harm side: methylmercury, which accumulates in long-lived predator fish, is a known neurotoxin. The Faroe Islands cohort, where exposure came largely from whale meat and blubber, found deficits in motor function, attention and verbal tests in children exposed in the womb, persisting to ages 14 and 22. Regulators responded: EFSA set a tolerable weekly intake of 1.3 micrograms of methylmercury per kilogram of body weight, and guidance caps tuna and bans the big predators. [B — cohort + regulatory]
On the benefit side: the Seychelles cohort, where mercury exposure came mainly from ocean fish in a high-fish diet, found no association with adverse neurodevelopmental outcomes at all. And a large UK cohort (ALSPAC, 11,875 women) found that women who ate no seafood had children with 48% higher odds of being in the lowest quartile for verbal IQ compared with women who ate more than 340 grams of seafood a week, after adjusting for 28 confounders. The authors concluded that advice to limit seafood could actually be detrimental. [C — observational cohort]
Whether the difference between the cohorts is about the exposure vehicle (whale blubber versus ocean fish), the protective nutrients in fish, or study differences is unresolved. The guidance handles it pragmatically: avoid the seven highest-mercury species, limit tuna to four cans or two steaks a week and oily fish to two portions, and eat plenty of low-mercury fish. The strongest genuine "what to eat" evidence in this whole topic is the fish benefit story — and it is observational.
Liver concentrates preformed vitamin A (retinol), and very high intakes in early pregnancy are teratogenic — they can cause malformations. A study of 22,748 pregnancies found an apparent threshold around 10,000 IU per day of supplemental preformed vitamin A, above which about 1 attributable malformation occurred per 57 exposed births, concentrated in exposure before the seventh week of gestation. [C — observational; the exact threshold has been disputed by later work]
The NHS advice is simple: avoid liver and liver products in pregnancy. You do not need a precise personal dose calculation to follow that — but note the limits of the evidence: the threshold is debated, β-carotene (the vitamin A in vegetables) is not considered teratogenic, and the risk is concentrated in the first weeks, often before a pregnancy is detected.
Liquorice is the restriction most people have never heard of. One Finnish cohort (the GLAKU study, 378 adolescents followed since birth) found that mothers with high glycyrrhizin exposure during pregnancy — about 500 mg a week or more, roughly 250 grams of liquorice — had children with roughly 7 IQ points lower scores, more ADHD-type problems, and, in girls, earlier puberty. [B/C — single cohort]
But the foundations are thin: the 500 mg/week cut-off was a distributional split in that cohort, not an established toxicity threshold; all the findings come from one research group; and Finland is the main country to have issued guidance against liquorice in pregnancy. This is a precautionary restriction on a genuine but fragile evidence base — not a proven dose problem. Occasional liquorice sweets are not what the study was about; the study was about very heavy consumption.
You may have seen advice to avoid chocolate in late pregnancy because its polyphenols could narrow the fetal ductus arteriosus (a blood vessel that should stay open until birth). This claim comes almost entirely from one unblinded research group in Brazil: in an open trial of 51 fetuses with ductal constriction, 96% showed reversal after three weeks of withdrawing polyphenol-rich foods. [C/D — unblinded, single group, unreplicated]
No pregnancy guideline endorses this restriction, and the study has not been independently replicated. There is also a separate, inconsistent literature suggesting chocolate might reduce preeclampsia: two observational studies found less preeclampsia with more chocolate, while a large cohort and two cocoa trials found no effect. [C/D — mutually inconsistent]
The bottom line: there is no reliable evidence that normal chocolate consumption in pregnancy harms the baby. This one is better treated as a debunked scare than as a warning.
The raw-egg rule changed in 2017, when the Food Standards Agency reinstated the advice that Lion-marked eggs could be eaten raw or runny in pregnancy. The current NHS guidance: raw or runny hen eggs are fine if they are stamped with the British Lion mark or are Laid in Britain; other eggs should be fully cooked until the white and yolk are solid.
If you are reading this outside the UK, check your own country's guidance — the egg situation depends on local production practices, and this is one area where the UK rules are genuinely more relaxed than the US ones.
| Risk | Best available estimate | Confidence |
|---|---|---|
| Listeriosis in pregnancy (England & Wales, 2024) | 37 cases; ~6 in 100,000 births (illustrative); 43.7% of known outcomes ended in miscarriage or stillbirth | B — surveillance |
| Congenital toxoplasmosis (Europe) | 1–10 per 10,000 newborns | B — epidemiology |
| Malformation from high supplemental preformed vitamin A | ~1 per 57 exposed births above 10,000 IU/day, before week 7 (disputed threshold) | C — observational |
| Mercury harm from high-mercury fish | No single absolute number; avoid the seven highest-mercury species, limit tuna | B/C — cohorts + guideline |
| IQ effects of heavy liquorice consumption | ~7 IQ points lower at ≥500 mg/week glycyrrhizin in one cohort (no population risk figure) | B/C — single cohort |
| Late-pregnancy chocolate/polyphenol risk | No confirmed risk figure; claim unreplicated | C/D — one unblinded group |