Skinny women who get pregnant are at a far greater risk of miscarriage than those of normal weight, a study has found. But they may be able to reduce the risk significantly by eating the right food or taking vitamins, say researchers.

Scientists from the London School of Hygiene and Tropical Medicine questioned 603 British women aged 18 to 55 who had miscarried within three months of getting pregnant.

Their social and lifestyle information was compared with that from more than 6,000 women whose pregnancies had progressed beyond 12 weeks.

As well as confirming previously known miscarriage risk factors, such as age and alcohol consumption, the findings revealed a number of surprises.

One was that underweight women had a 72 per cent higher risk of miscarrying in the first trimester. On the other hand, women who took vitamin supplements during early pregnancy reduced the risk by around 50 per cent.

This was especially true if they took folic acid or iron and multivitamins. Daily consumption of fruit and vegetables also halved the chances of having a miscarriage, and chocolate also gave some benefit.

Women who were not married or living with a partner were found to be at greater risk, while those whose pregnancy was "planned" were 40 per cent less likely to miscarry than those for whom conception was an accident.

A woman who changed partner after becoming pregnant increased her odds of a miscarriage by 60 per cent.Having had a previous abortion raised the chances of a subsequent miscarriage by more than 60 per cent, while fertility problems were associated with 41 per cent higher odds.

The study also confirmed that morning sickness is a good sign. Women who suffered from morning sickness in the first 12 weeks of pregnancy were almost 70 per cent less likely to miscarry. The more severe the sickness, the better the odds of the pregnancy continuing.

The research leader, Noreen Maconochie, a senior lecturer at the London School of Hygiene and Tropical Medicine, said: "The findings related to low pre-pregnancy weight, previous termination, stress and change of partner are noteworthy, and we suggest further work be initiated to confirm these findings in other study populations."