Total sugar showed no link to breast cancer in 71,264 Korean women. Sweet foods as a group did, yet cakes, sweets and fizzy drinks on their own didn't

Ask how much sugar someone eats and you may be asking the wrong question. That is the idea behind a new analysis from the Health Examinees study, a large Korean cohort, published in the European Journal of Nutrition on 31 October. The researchers followed 71,264 women, aged 40 to 69 when they joined between 2004 and 2013, with an average age of 52. Cancer diagnoses came from the national registry up to the end of 2018. Over an average of 9.2 years, 713 women developed breast cancer.

Each woman answered a 106-item food questionnaire with an interviewer when she joined. The team built a sugar database for those foods, then sorted out seven groups they counted as high in sugar: sugary cereals, jam and butter, cakes and cookies, ice cream, sweets, carbonated drinks, and sugar-based Korean traditional teas. Women averaged 44 g of total sugar a day, and the largest single source by far was fruit, at 43%. Going by the shares the paper gives, those seven groups together came to only about an eighth of all the sugar eaten. That sum is ours.

Total sugar didn't predict anything: the top quarter of intake had a hazard ratio of 0.93 against the bottom quarter, with a confidence interval comfortably spanning 1. Fruit didn't either, at 0.89. The seven sweet groups taken together did. Women having them more than twice a week had a hazard ratio of 1.28 (95% CI 1.02 to 1.61) compared with women having them once a month or less, after adjusting for BMI, energy intake, education, alcohol, exercise, family history and reproductive history. Sugar from those foods gave the same figure, 1.28, comparing the top and bottom quarters.

In plain counts, before any adjustment, 206 of 17,814 women in the highest group developed breast cancer, against 147 of 17,626 in the lowest. That works out at roughly 12 and 8 per 1,000 women over about nine years (our division). So a relative figure of 28% sits on small absolute numbers, and breast cancer stayed uncommon in every group.

The surprise comes when the group is opened up. Cakes and cookies had a hazard ratio of 0.90, sweets 0.98, carbonated drinks 0.97 and traditional teas 1.07, all compatible with no link at all, even though the teas alone made up 45% of the sugar from these foods. The two that stood out were ice cream at least weekly (1.26, CI 1.00 to 1.58, which only just touches the line) and sugary cereal at least once every two months (1.34, CI 1.08 to 1.65). Sugary cereal supplied about 1% of the sugar from these foods, and the top category is a very low bar, so the authors themselves urge caution. They offer explanations involving insulin response at breakfast and saturated fat in ice cream, but the study didn't test either. The methods also describe no correction for running this many comparisons, and with seven foods plus subgroups, a chance finding or two is easy to come by.

The other limits matter as much. Diet was asked about once, at the start, then assumed to hold for around nine years. Sugar in the database wasn't split into added and natural, so the sweet foods stand in for added sugar. Women in the high group were younger, better educated and ate more calories (about 1,983 a day against 1,500); 49% were premenopausal against 31%. Some of that was adjusted for, but menopausal status isn't on the adjustment list and hormone therapy wasn't accounted for, which the authors flag. Subgroups hint that the link was stronger in women under 50 (1.52, against 1.15 in those 50 and over) and in those who had ever drunk alcohol (1.49, against 1.14 in never-drinkers). Of those two, only the alcohol difference passed the formal test for interaction (P = .02; P = .48 for age). The paper calls itself the first of its kind in an Asian population, so it needs repeating before anyone rewrites advice. The study was funded by the Korea Centers for Disease Control and Prevention, which had a role in design and data collection, and the authors declared no competing interests.