You're not average

TL;DR. Nutrition guidelines are written for an average person who doesn't exist. The PREDICT study put continuous glucose monitors and standardized muffins onto more than a thousand subjects, including hundreds of twin pairs, and found that less than 1% sat at the mean response for glucose, insulin, and triglycerides at the same time. Identical twins eating identical meals can get tenfold different metabolic responses. Genes explain some of this, the microbiome explains more, and life stage and context fill in the rest. The right takeaway is not to chase a personalized diet from a $300 saliva kit — it's to know which decisions actually shift with your biology, and to use the new tools (a two-week CGM, a quarterly biomarker panel, your own felt sense) to test what works for you. Real food, sleep, and movement are still the foundation. Personalization is the upper layer.

What you'll learn

• Why population-level dietary averages are a statistical artifact that fits almost no individual.
• What three forces actually drive the variation: genes, microbiome, and context.
• How life stage rewrites the priority list — what changes for pregnancy, childhood, midlife, and older adults.
• What the new bio-observability tools (CGMs, direct-to-consumer biomarker panels) actually show, and what they don't.
• Where personalization clearly matters — lactose, gluten sensitivity, caffeine, APOE4, iron status.
• The honest limit: n=1 data is noisy and shouldn't be allowed to crowd out the basics.

1. The average person doesn't exist

Tim Spector's group at King's College London, with Mass General, Stanford, and the spinout ZOE, ran the PREDICT 1 study: more than 1,000 people, hundreds of them twin pairs, eating standardized muffins under continuous glucose, insulin, and triglyceride monitoring — roughly 130,000 meals and two million glucose readings. Published in Nature Medicine in 2020, the headline was that less than 1% of subjects sat near the population average for all three markers at once. Identical twins eating the same muffin could show up to a tenfold difference in glucose response. The same orange juice that spiked Spector himself barely moved his wife's blood sugar. In the Stanford DIETFITS trial of 609 adults, some on the low-fat arm lost 27 kilograms while others on the same arm gained 9.

This is not about outliers. The bell curve is wide enough that the mean is a fiction. The Dietary Reference Intakes anchoring nearly every food label were designed, as Chapter 109 of Modern Nutrition in Health and Disease admits, to cover nearly all healthy individuals in a defined life-stage and sex group — not to predict what any one person needs. The textbook itself flags that DRIs handle chronic-disease endpoints poorly. "Eat 2,000 calories a day," "fat should be 30% of intake," and "everyone needs three servings of dairy" are population statements being misused at the individual level.

2. What actually drives the variation

PREDICT and the parallel work coming out of the Weizmann Institute (Zeevi et al., Cell, 2015) pin the variation in glycemic response on three contributors, in roughly this order.

Genetics — modest but real. Genes explain about 30% of glucose response and under 5% of post-meal fat response. Within that 30%, a few common variants do meaningful work. FTO tracks hunger and body-weight regulation. MTHFR C677T and A1298C affect folate metabolism and homocysteine. APOE — particularly the E4 allele — modifies how strongly saturated fat raises LDL and roughly doubles Alzheimer's risk per copy. FADS1/FADS2 variants govern conversion of plant omega-3 (ALA) to long-chain EPA and DHA; slow converters get less benefit from flaxseed and more from fatty fish. LCT (lactase persistence) determines adult milk-sugar digestion. Chapter 38 of Modern Nutrition runs these as the canonical nutrigenomics examples; Chapter 123 is careful to mark which have demonstrated diet-by-gene interactions and which are marketing.

Microbiome — bigger than genes for glucose. Spector's twin work showed identical twins, sharing 100% of their DNA, share only 37% of gut microbial species — barely more than unrelated strangers. The Weizmann study found microbiome features were the single largest predictor of post-meal glucose response. ISAPP-defined probiotics, prebiotics, synbiotics, postbiotics, and fermented foods (Chapter 37 of Modern Nutrition, Holscher & Donovan) act on this layer. Two people can eat the same banana and one spikes harder than soda while the other barely moves, because their microbial communities ferment it differently.

Context — sleep, stress, exercise, time of day. The same person can show different responses morning versus evening. Six days of four-hour sleep induces measurable prediabetes in healthy adults. Stress raises cortisol and shifts glucose handling. A 15-minute walk after a meal flattens the spike. Means's framing in Good Energy is that this bucket is the most actionable: you can't change your APOE genotype, and you can only slowly reshape your microbiome, but you can decide whether to eat a heavy carbohydrate meal at 10 p.m. tonight.

3. Life stage rewrites the priority list

The most reliable form of "personalization" is the one nobody markets: matching the advice to your life stage. Willett's tailoring chapter in Eat, Drink, and Be Healthy and Part III of Modern Nutrition (Chs 52–56) lay it out.

Pregnancy. 400 mcg folic acid daily before conception prevents most neural tube defects. Iron needs roughly double. Iodine is critical for fetal neurodevelopment. Choline (~450 mg/day) supports neural tube closure and is under-consumed in most American diets. Two to three weekly servings of low-mercury fish (salmon, sardines, cod, anchovy, tilapia) deliver EPA and DHA without the mercury load of larger predators. Caloric needs rise only in the third trimester, by about 200 calories — "eating for two" is wrong by an order of magnitude.

Lactation. Energy needs rise by ~500 calories/day. Calcium, iodine, and adequate B12 (especially for vegan or vegetarian mothers) matter most. The microbiome the infant inherits is shaped by maternal diet and delivery mode.

Childhood. Calorie restriction is contraindicated. Iron deficiency is the most common micronutrient deficiency in U.S. toddlers and can cause cognitive impairment that may not fully reverse. DHA matters for brain development. Restrictive labeling and parental pressure are independent risks: Birch and Satter's research shows mothers who restricted "bad" foods had daughters with the most overeating at age 9, and U.S. children are 242 times more likely to develop an eating disorder than type 2 diabetes.

Adolescence. Peak bone mass is largely set by the early 20s — calcium, vitamin D, and weight-bearing activity matter most here and almost not at all after age 30 in the same way. Menstruating adolescents need extra iron.

Adults 30–60. The demographic most nutrition advice is written for, and even here it splits by sex, body composition, and disease status. PCOS shifts the case toward a lower glycemic-load pattern. The standard one-hour OGTT misses an estimated 70% of gestational diabetes cases; a CGM during pregnancy can catch the rest.

Adults 60+. Sarcopenia, not obesity, becomes the limiting factor for healthspan. Protein needs rise toward 1.0–1.2 g/kg (versus the RDA of 0.8), with resistance training as the non-negotiable companion. B12 absorption falls with age — atrophic gastritis is common — so supplementation or fortified foods matter. Vitamin D, vitamin K, and adequate (not high) calcium support bone alongside loading exercise.

4. Bio-observability: tools that let you see your own response

The technical shift Means writes about in Good Energy is that the data layer is no longer locked behind a clinic.

Continuous glucose monitors. Levels, Stelo, and Lingo have made CGMs available to non-diabetics. A two-week run gives ~35,000 data points versus one annual fasting-glucose snapshot. You'll see your spike from white rice versus brown, fruit alone versus with protein, a 9 p.m. dessert versus a 7 p.m. one. Means and Lustig both push fasting insulin and HOMA-IR as the most valuable single blood tests — fasting glucose stays normal for years while insulin climbs.

Direct-to-consumer biomarker panels. Function Health, InsideTracker, and similar services offer 30–100+ biomarker panels (lipid fractionation, hsCRP, ApoB, fasting insulin, hormones, micronutrients) without a referral. Four times a year is plenty; daily is theater.

DEXA. Body composition over BMI. A "normal-weight" adult can be sarcopenic and metabolically unwell; the scale won't catch it.

HRV and sleep wearables. Recovery and sleep architecture as a proxy for the autonomic context that shapes everything else. Garmin, Whoop, Oura, and Apple Watch all do this adequately.

What they don't do is replace judgment. CGMs in non-diabetics will show spikes from foods that aren't actually causing harm; the right read is pattern over time, not panic per meal.

5. Where personalization clearly matters

A short list of cases where individual variation is large enough to act on.

Lactose. Lactase persistence (LCT) is the default in much of Northern Europe and rare in much of East Asia and Africa. Cheese (most lactose removed by fermentation and aging) and yogurt (lactose pre-digested by bacterial cultures) are tolerated by many people who can't drink a glass of milk.

Gluten. Celiac disease is about 1% of the population and requires strict lifelong avoidance. Non-celiac gluten sensitivity is real but rarer than the marketplace suggests: the Italian re-challenge studies Spector cites found that of 392 self-identified gluten-sensitive subjects, only 7% met NCGS criteria and 0.5% had wheat allergy after blinded testing. Most people who feel better gluten-free are responding to reduced ultra-processed intake, not gluten itself.

Caffeine. CYP1A2 variants produce ~twofold differences in caffeine clearance. Fast metabolizers can have an espresso at 5 p.m. and sleep; slow metabolizers can't have one at noon. Smokers metabolize caffeine twice as fast as nonsmokers; oral contraceptives slow it down.

APOE4 and saturated fat. Carriers of one or two E4 alleles show steeper LDL responses to saturated fat than non-carriers — one of the cleaner gene-by-diet interactions in the literature.

Iron. The only deficiency with a clear case for supplementation — but only with documented low ferritin or hemoglobin on labs. Excess iron is harmful: hemochromatosis is the most common single-gene disorder in people of Northern European ancestry, and free iron drives oxidative stress. Don't supplement iron on instinct.

6. The honest limit

Personal data is noisy. A single high glucose reading after a meal could be the meal, sleep debt, last night's wine, or the handful of almonds you forgot to log. Run an experiment for at least two weeks before drawing a conclusion. Hold one variable at a time.

The deeper trap is letting the optimization layer crowd out the foundation. Real food, sleep, movement, sunlight, social connection, and stress regulation account for the vast majority of metabolic and longevity outcomes. If you're sleeping six hours, sitting all day, and eating ultra-processed food twice a day, no genetic test will rescue you. If the basics are in place, the personalized refinements buy small percentages — real, but small.

Tribole and Resch's interoception thread closes the loop. The most powerful "personalized nutrition" instrument you own is the one humans evolved with: hunger, fullness, satisfaction, and how your body feels three hours after a meal. The CGM is a useful prosthetic when interoception has been worn out by dieting or distraction; it isn't a replacement for the felt sense it's trying to rebuild.

Frequently Asked Questions

Should I get genetic testing for nutrition?

For most people, no — not yet. The demonstrated gene-by-diet interactions (APOE × saturated fat, LCT and lactose, MTHFR and folate, FADS and omega-3 conversion) can usually be inferred from labs, family history, and ancestry, and the consumer testing market overpromises. A clinical genetics consult is reasonable with a strong family history of premature heart disease, hemochromatosis, or familial hypercholesterolemia.

Are CGMs useful for non-diabetics?

Yes, for a defined window. A two-to-four-week run will teach you more about your own glucose response than any book: which foods spike you, how late-night meals compare to mid-day, how a walk after dinner changes the trace. Wearing one indefinitely tips into theater and anxiety.

How do I find a registered dietitian who does personalization well?

Look for credentials (RD/RDN), a clinical specialty that matches your situation (sports, pediatrics, IBD, eating disorders, geriatrics), and willingness to discuss bio-observability data without selling a single-protocol cure. The Academy of Nutrition and Dietetics directory lists specialists by area.

What's the difference between nutrigenomics and precision nutrition?

Nutrigenomics is the science of how nutrients interact with genes. Precision nutrition is the broader applied agenda of using genomic, microbiome, metabolomic, and behavior data for individual recommendations. Chapter 123 of Modern Nutrition is candid that precision nutrition is still aspirational, not deployed at scale.

Is "metabolic typing" or the "blood type diet" real?

No. Neither has held up in trials. A 2014 PLOS ONE study of 1,455 people found "blood type diet" adherence did associate with some health markers — but the same way for every blood type, meaning the diet was working, not the matching.

How do I know if I'm gluten-sensitive?

First rule out celiac via blood tests (tissue transglutaminase IgA) and, if positive, endoscopy — done while still eating gluten. If celiac is negative and symptoms persist, a structured elimination-and-rechallenge protocol with a dietitian is the standard, ideally blinded.

When does personalization matter most?

At extremes and transitions. Pregnancy, lactation, childhood, adolescence, peri- and post-menopause, and 60+ each rewrite the priority list. A diagnosed condition — type 2 diabetes, PCOS, CKD, IBD, celiac, an eating disorder, a cardiovascular event — moves you from population guidance to individual care.

Sources

1. Berry, S. E., et al. "Human postprandial responses to food and potential for precision nutrition." Nature Medicine, 2020;26(6):964–973. DOI: 10.1038/s41591-020-0934-0. The PREDICT 1 paper.
2. Zeevi, D., et al. "Personalized Nutrition by Prediction of Glycemic Responses." Cell, 2015;163(5):1079–1094. DOI: 10.1016/j.cell.2015.11.001. Weizmann Institute glycemic prediction study.
3. Spector, T. Spoon-Fed: Why Almost Everything We've Been Told About Food Is Wrong (2020). Twin variation, 37% microbe sharing, PREDICT framing.
4. Means, C. Good Energy (2024). Bio-observability, fasting insulin, CGM interpretation, Function Health.
5. Tucker, K. L., Duggan, C. P., Jensen, G. L., et al. (eds). Modern Nutrition in Health and Disease, 12th edition (2026). Chapters 37 (biotics, ISAPP definitions), 38 (nutritional genomics — FTO, MTHFR, APOE, FADS, LCT), 52–56 (life stage), 109 (DRI limits), 123 (precision nutrition).
6. Willett, W. C. Eat, Drink, and Be Healthy, revised edition (2017). Life-stage tailoring; pregnancy, diabetes, older adults.
7. Tribole, E., Resch, E. Intuitive Eating, 4th edition (2020). Interoceptive awareness; the limits of externalized eating metrics.
8. Araújo, J., Cai, J., Stevens, J. "Prevalence of Optimal Metabolic Health in American Adults: NHANES 2009–2016." Metabolic Syndrome and Related Disorders, 2019;17(1):46–52. DOI: 10.1089/met.2018.0105.
9. Volpi, E., et al. "Is the optimal level of protein intake for older adults greater than the recommended dietary allowance?" Journals of Gerontology Series A, 2013;68(6):677–681. Sarcopenia and protein in older adults.

Related modules

• C5: The microbiome — your other genome
• C2: Inside your cells — mitochondria, ATP, and insulin
• C3: Macronutrients without the moralism

Related glossary terms

• Nutrigenomics
• Precision nutrition
• Interoception
• PREDICT study
• Continuous glucose monitor (CGM)
• MTHFR
• APOE
• FTO
• FADS
• LCT (lactase persistence)

---

This is the last module of the Core tier. From here, two paths: the Deep tier covers clinical conditions, evaluating health claims, and the political economy of food; or rerun the Beginner tier to let the habits sink in. The Core arc — what food is, how cells use it, what the macros and micros actually do, why sugar and ultra-processed food keep showing up as the consistent culprits, what the microbiome adds, why you're not average — is now complete.