Creme brulee: a dessert nobody has studied

Nobody has studied creme brulee. The name returns no composition record in either reference database, no trial, no retention factor for a baked custard, and no work at all on the torched sugar crust. What follows stands next to the dish instead of on it, because the research that exists is about custard and dairy desserts as a class. The closest measurement is about texture. Twelve healthy adults in a randomised crossover ate four model foods built from matched ingredients. The triglyceride response after the custard ran up to 122.7 percent higher than after the pudding or the biscuit (p < 0.0001). That was one test meal each in twelve people, with a laboratory custard standing in for the dessert, so it carries only as far as the soft egg and dairy matrix does. Soft food is also easy to keep eating. In the second of two randomised trials, 20 young adults eating chocolate custard until they felt full stopped 66 plus or minus 21 g earlier when the dessert was made to linger in the mouth.

How much is actually in it

• The nutritional quality of shop-bought dairy desserts sat still for five years, and only the products that manufacturers actually reformulated improved their score. — Yogurts and dairy desserts sold in Quebec, 468 products in 2018 to 2019 against 279 in 2023, scored with the 2023 Nutri-Score algorithm and weighted by sales data (n = 747 products across the two collections) Evidence C source
  Overall quality stable in all four categories (p at least 0.11); reformulated yogurts and dairy desserts improved from 3.95 ± 3.84 to 2.80 ± 2.13 (p = 0.03)

What a real portion looks like

• People eating chocolate custard until full stopped 66 g earlier when oral exposure to the dessert was high rather than low, and, only within that high-exposure condition, 42 g earlier when they ate slowly. — 60 young adults across two randomised 2 by 2 trials, eating chocolate custard with caramel sauce or with caramel fudge, fast or slow, until fullness (n = 40 in study 1 and 20 in study 2) Evidence B source
  42 ± 15 g less eaten with a slower eating rate within the high oral exposure condition (p = 0.04); 66 ± 21 g less with high versus low oral exposure; insulin 81% higher at meal end in the high exposure slow condition
• How big a spoonful someone takes of a custard is decided before it reaches the mouth: people took a large bite when the dessert looked and smelled creamy, whatever was actually in the cup. — Adults taking a first bite through a straw from a two-compartment cup, where the custard they saw and smelled differed from the custard they swallowed (n = three linked experiments, participant counts not given in the abstract) Evidence C source
  A favourable expectation produced a larger first bite irrespective of the custard actually ingested; blocking recognition by changing colour, odour or visual texture handed control back to the oral qualities

What your body absorbs

• Custard delivers its fat into the blood far faster than the same ingredients baked into a stiffer form: the triglyceride response after custard ran up to 123% higher than after pudding or biscuit. — 12 healthy adults in a randomised open label crossover, each eating four model foods of matched composition but different structure: custard, pudding, sponge cake and biscuit (n = 12) Evidence B source
  Triglyceride chylomicron response after custard up to 122.7% higher than after pudding and biscuit (p < 0.0001); vitamin D3 response from sponge cake 26.6% above biscuit; no difference for lutein

What it does to your health

• A single 192 g serving of custard cake shifted gut bacteria within 24 hours differently from a high glycaemic dessert, and the authors called the shift temporary. — 30 healthy adults aged 18 to 45 randomised to one of three Thai desserts alongside breakfast, the low glycaemic arm being 192 g of custard cake, with faecal sampling at baseline and 24 hours (n = 30, ten per arm) Evidence B source
  Collinsella and Bifidobacterium fell in the high glycaemic group relative to the low and medium groups; sugar intake correlated negatively with Lactobacillus in the medium and high groups but not in the low group

Safety, limits and interactions

• Plain caramel colouring E150a was the one additive linked to higher overall cancer incidence in a French cohort of 105 260 adults, and that industrial colouring is not the same thing as sugar burnt under a blowtorch. — 105 260 French adults without cancer at baseline (78.3% women, mean age 42.0), followed more than 7 years with repeated brand-specific 24 hour dietary records (n = 105 260, with 4226 incident cancers) Evidence C source
  Plain caramel E150a hazard ratio 1.15 (95% CI 1.07 to 1.25) for overall cancer in higher versus non or lower consumers, absolute risk at age 60 of 14.0% against 12.1%
• Across 5241 ready-to-eat samples from caterers and restaurants, including cooked custards and creams, exactly one carried staphylococcal enterotoxin, and it was a meat and vegetable dish. — 5241 ready-to-eat samples collected from caterers, hotels and restaurants in Istanbul between 2014 and 2016, one of the four sample groups being cooked sweets and desserts such as pudding, custard and cream (n = 5241) Evidence C source
  1 positive sample (0.019%), and it was not a dessert

Who this works differently for

• In a review of dairy and ageing, older women with high intakes of dairy desserts and ice cream were the one group where the association ran towards cognitive decline rather than away from it. — 6 studies screened from 661 records, 5 prospective cohorts and 1 randomised trial, in community-dwelling older people (n = 6 studies) Evidence C source
  High intake of dairy desserts and ice cream associated with cognitive decline in older women, while dairy overall was associated with lower risk of frailty and sarcopenia

What is still unknown

• Pooling 15 cohorts and 930 677 people, dairy desserts showed a protective looking link with high blood pressure that vanished once dose was taken into account, and the authors graded that evidence very low. — 930 677 participants with 363 459 incident hypertension cases across 15 prospective cohorts, searched to December 2018 and graded with GRADE (n = 930 677) Evidence C source
  The pairwise protective association for dairy desserts was not supported by linear dose-response analysis; sugar-sweetened beverages meanwhile showed a risk ratio of 1.10 (95% CI 1.08 to 1.12) per 355 mL

The bottom line

This is a dessert, and nothing measured here speaks to whether to eat one. What the two trials measure is how fast a serving disappears. In the first of those two trials, 40 adults ate 42 plus or minus 15 g less when they paused between spoonfuls, and only when each spoonful stayed in the mouth (p = 0.04). Expectation sets the size of the bite before the spoon lands. In three linked laboratory experiments using a straw and a two-compartment cup, people took a large first bite whenever the dessert looked and smelled creamy, whatever they actually swallowed. The abstract never says how many people took part. The only measured statement about composition here is Canadian. Across 747 packaged yogurts and dairy desserts scored in 2018 to 2019 and again in 2023, overall quality sat still in all four categories (p at least 0.11). Only the products manufacturers actually reformulated improved, from 3.95 plus or minus 3.84 to 2.80 plus or minus 2.13 (p = 0.03). Those are pots on a shelf, and this dessert is usually made in a kitchen. The health findings are thin too. Thirty healthy adults, ten per arm, ate one Thai dessert with breakfast, and the 192 g custard cake arm moved gut bacteria differently from the high glycaemic arm within 24 hours, a shift the authors themselves call temporary. Across 15 prospective cohorts and 930 677 people, dairy desserts looked protective against high blood pressure until dose was taken into account. The association did not survive that step, and the authors graded the evidence very low. A review of six studies in community-dwelling older people found high intakes of dairy desserts and ice cream associated with cognitive decline in older women, which is observational and may be marking who eats them. On caution the page is thinner still. In a French cohort of 105 260 adults followed more than 7 years, plain caramel colouring E150a carried a hazard ratio of 1.15 for overall cancer (95% CI 1.07 to 1.25). That is an absolute risk at age 60 of 14.0 percent against 12.1, in an analysis its own authors call novel and exploratory. E150a is a manufactured colourant for cola and sauces. Burning sugar under a blowtorch is a different process, and that study never measured it. Of 5241 ready-to-eat samples taken from caterers, hotels and restaurants in Istanbul, one carried staphylococcal enterotoxin, and it was a meat and vegetable dish. We have no card about egg or dairy allergy, none about pregnancy, and none about medicines, so this page cannot tell you who should skip it.

Sources

1. crbr-r1-09 · cross-sectional
   Only Yogurts and dairy desserts that specifically underwent reformulation between both data collections showed an improved overall nutritional quality (i.e., lower score; 3.95 ± 3.84 vs. 2.80 ± 2.13, p = 0.03).
   Appl Physiol Nutr Metab, 2026 · checked 2026-09-24
2. crbr-r1-02 · randomised trial
   In study 1, participants ate (mean ± SEM) 42 ± 15 g less in the slow- than in the fast-ER condition, only within the high-OSE condition (P = 0.04). In study 2, participants ate 66 ± 21 g less in the high- than in the low-OSE condition and there were no intake differences between slow and fast ER (P = 0.35).
   Am J Clin Nutr, 2020 · checked 2026-09-24
3. crbr-r1-03 · controlled sensory experiment
   When this recognition was favorable, e.g., when the upper custard was known to be creamy, a relatively large bite was taken, irrespective of the custard that was actually ingested.
   Physiol Behav, 2004 · checked 2026-09-24
4. crbr-r1-01 · randomised crossover
   Custard triglyceride response is higher than pudding and biscuit responses (up to +122.7%, p < 0.0001).
   Mol Nutr Food Res, 2020 · checked 2026-09-24
5. crbr-r1-04 · randomised trial
   In the high-GI dessert group, the abundance of Collinsella and Bifidobacterium decreased compared to the low- and medium-GI groups, while Roseburia and Ruminococcus showed slight increases.
   Nutrients, 2024 · checked 2026-09-24
6. crbr-r1-07 · prospective cohort
   After False Discovery Rate correction, only plain caramel (European code: E150a) was associated with overall cancer [1.15 (1.07-1.25); 14.0%, 12.1%] and beta-carotene (E160a) with overall [1.16 (1.07-1.25); 13.7%, 11.9%] and breast cancer [1.41 (1.23-1.62); 6.2%, 4.4%].
   Eur J Epidemiol, 2026 · checked 2026-09-24
7. crbr-r1-08 · prevalence survey
   Among all samples, only 1 (0.019%) RTE meal (vegetable meal with meat) was found to be contaminated with SEs, a good result in terms of staphylococcal food poisoning risk and public health.
   J Food Prot, 2017 · checked 2026-09-24
8. crbr-r1-06 · systematic review of observational studies
   In older women, high intakes of dairy desserts and ice cream were associated with cognitive decline.
   Adv Nutr, 2019 · checked 2026-09-24
9. crbr-r1-05 · meta-analysis of cohorts
   The pair-wise protective association of dairy desserts was not supported by linear dose-response analysis.
   J Am Heart Assoc, 2019 · checked 2026-09-24