Why test on yourself

01The same meal, different responses

In one study, 800 people wore a glucose monitor for a week, and the rise in their blood sugar after the same meals varied widely from person to person (Zeevi et al., 2015). A larger study, of 1,002 adults in the UK, found the same for blood fats and insulin as well as blood sugar (Berry et al., 2020).

That is the usual case for testing on yourself: a group’s average can hide large differences between the people in it. Both studies measured body chemistry such as blood sugar, though, not how people felt, so they say nothing directly about symptoms.

02Much of the difference can be noise

There is a catch: one person’s responses vary too. In the study of 1,002 adults, people were given the same test meals twice under similar conditions, and their two responses agreed only partly (Berry et al., 2020). The authors noted that when each person is tested just once, much of the apparent difference between people could be chance swings and other error.

The statistician Stephen Senn makes the same point in general terms. When each person is tested only once, their own response is mixed up with their everyday ups and downs, and telling the two apart takes testing the same person more than once (Senn, 2016). He concluded that the widespread belief that people respond to medicines in strongly personal ways lacks sound statistical support (Senn, 2016). He has also argued that figures claiming most medicines don’t work for most people are being misread (Senn, 2018).

This cuts both ways. It is a warning against assuming that everyone reacts differently to everything. It is also the best reason to repeat a test: researchers tell a real personal difference from noise by comparing the same person with and without a change, more than once (Senn, 2016).

03What one-person food studies have managed

Researchers have argued that judging nutrition one person at a time needs new kinds of study, and that one-person (n-of-1) trials can be one of them (Schork and Goetz, 2017). Group trials, one perspective notes, measure each person too rarely to show how that person responds (Potter et al., 2021).

So far there are few such studies. A 2023 review found only 7 that met its criteria, with 83 adults in all (Allman-Farinelli et al., 2023). They can also be hard to keep going. In one series of 54 one-person trials, children and teenagers switched between two versions of a restricted diet every 8 weeks, and fewer than half finished (Kaplan et al., 2022).

04Why a test on one person is still hard

Formal one-person trials borrow the tools of group trials: a random order, gaps between periods so one doesn’t spill into the next, several switches back and forth, and sometimes a dummy version of the thing being tested (Lillie et al., 2011). Even so, their answers are often unclear. In a review of 108 published trials, 38% of the people whose later decisions were reported had results that didn’t clearly point either way (Gabler et al., 2011).

A test at home usually has fewer of these tools, so one run on one day says little on its own. The rest of this guide takes the parts of a careful test one at a time, starting with a note made on the day.

05What the studies found

• In a 2015 study, 800 people wore a glucose monitor for a week while their responses to 46,898 meals were measured (Zeevi et al., 2015). Blood sugar after identical meals varied widely between people, and a computer model using blood tests, diet, body measurements, activity and gut microbes predicted each person’s response.
• In a 2020 study of 1,002 twins and unrelated healthy adults in the UK, rises in blood fats, blood sugar and insulin after identical meals varied widely between people (Berry et al., 2020). When people were given the same test meals twice, the two responses agreed only partly: on a repeatability scale where 1 is a perfect match, scores ran from 0.46 for blood fats to 0.74 for blood sugar. The company Zoe Global supported the study, alongside research grants, and 17 of the 25 authors were Zoe consultants or current or former staff.
• In a 2016 methods paper, Senn explained that separating a person’s own response from their day-to-day variation needs repeated tests in the same person, as in repeated crossover trials or a series of one-person trials (Senn, 2016). He concluded that the widespread belief in strongly personal responses to medicines lacks sound statistical support.
• In a 2018 comment in Nature, Senn argued that figures claiming most medicines don’t work for most people are being misread (Senn, 2018).
• In a 2017 review, two researchers argued that judging nutrition plans tailored to each person needs new study designs, and that one-person trials can be used to test such plans (Schork and Goetz, 2017).
• In a 2021 perspective, three researchers noted that group nutrition trials measure too rarely to show how each person responds (Potter et al., 2021). They described one-person designs that follow ordinary days with repeated real-time reports, and designs that switch between periods with and without a change.
• In a 2023 review, only 7 of the 211 articles screened were one-person nutrition studies that qualified, with 83 adults in 5 countries; they covered prediabetes, diabetes, irritable bowel syndrome, weight management and healthy people (Allman-Farinelli et al., 2023). Reporting was mostly adequate, but the way diet was measured ranged from poor to good, and the authors called the evidence base small.
• In a 2022 series of 54 one-person trials at 19 sites, children and teenagers aged 7 to 18 with active inflammatory bowel disease compared two versions of a restricted diet (Kaplan et al., 2022). After 2 weeks on their usual diet, each was randomly given one of two orders of four alternating 8-week periods. Only 21 (39%) completed all four, 9 (17%) completed only the first two, and 24 (44%) withdrew; the authors wrote that diet changes are hard to study this way.
• In a 2011 review, researchers described how one-person trials can use the tools of group trials, including a random order, washout and crossover periods, and dummy (placebo) comparisons (Lillie et al., 2011). They noted that such trials were widely used in education but rarely in medicine.
• In a 2011 review of 108 one-person trials published from 1985 to 2010, with 2,154 people in all, about half picked the better option by a statistical cut-off (Gabler et al., 2011). Among 488 people whose later decisions were reported, 54% made decisions that matched their trial’s result, 8% went against it, and 38% had unclear results.

06Sources on this page

Each source is listed once, with links to its record. The reference list says what was read for each, and when it was checked.

1. Allman-Farinelli et al., 2023Allman-Farinelli M, Boljevac B, Vuong T, Hekler E. 2023. Nutrition-Related N-of-1 Studies Warrant Further Research to Provide Evidence for Dietitians to Practice Personalized (Precision) Medical Nutrition Therapy: A Systematic Review. Nutrients 15(7):1756.DOI 10.3390/nu15071756 · PubMed 37049595 · PMC10097352
2. Berry et al., 2020Berry SE, Valdes AM, Drew DA, Asnicar F, Mazidi M, Wolf J, et al. 2020. Human postprandial responses to food and potential for precision nutrition. Nature Medicine 26(6):964–973.DOI 10.1038/s41591-020-0934-0 · PubMed 32528151 · PMC8265154
3. Gabler et al., 2011Gabler NB, Duan N, Vohra S, Kravitz RL. 2011. N-of-1 trials in the medical literature: a systematic review. Medical Care 49(8):761–768.DOI 10.1097/MLR.0b013e318215d90d · PubMed 21478771
4. Kaplan et al., 2022Kaplan HC, Opipari-Arrigan L, Yang J, Schmid CH, Schuler CL, Saeed SA, et al.; ImproveCareNow Pediatric IBD Learning Health System. 2022. Personalized Research on Diet in Ulcerative Colitis and Crohn's Disease: A Series of N-of-1 Diet Trials. American Journal of Gastroenterology 117(6):902–917.DOI 10.14309/ajg.0000000000001800 · PubMed 35442220
5. Lillie et al., 2011Lillie EO, Patay B, Diamant J, Issell B, Topol EJ, Schork NJ. 2011. The n-of-1 clinical trial: the ultimate strategy for individualizing medicine? Personalized Medicine 8(2):161–173.DOI 10.2217/pme.11.7 · PubMed 21695041 · PMC3118090
6. Potter et al., 2021Potter T, Vieira R, de Roos B. 2021. Perspective: Application of N-of-1 Methods in Personalized Nutrition Research. Advances in Nutrition 12(3):579–589.DOI 10.1093/advances/nmaa173 · PubMed 33460438 · PMC8166550
7. Schork and Goetz, 2017Schork NJ, Goetz LH. 2017. Single-Subject Studies in Translational Nutrition Research. Annual Review of Nutrition 37:395–422.DOI 10.1146/annurev-nutr-071816-064717 · PubMed 28715990 · PMC6383767
8. Senn, 2016Senn S. 2016. Mastering variation: variance components and personalised medicine. Statistics in Medicine 35(7):966–977.DOI 10.1002/sim.6739 · PubMed 26415869 · PMC5054923
9. Senn, 2018Senn S. 2018. Statistical pitfalls of personalized medicine. Nature 563(7733):619–621.DOI 10.1038/d41586-018-07535-2 · PubMed 30482931
10. Zeevi et al., 2015Zeevi D, Korem T, Zmora N, Israeli D, Rothschild D, Weinberger A, et al. 2015. Personalized Nutrition by Prediction of Glycemic Responses. Cell 163(5):1079–1094.DOI 10.1016/j.cell.2015.11.001 · PubMed 26590418