Dietary Records and Self-Report Error as Inputs

The common ways intake is recorded

• Food frequency questionnaire (FFQ): a list of foods asking how often each is usually eaten.
• 24-hour dietary recall (24HR): a person describes everything eaten in the past day.
• Food record or app log: the person writes down or enters foods as they are eaten. For example, the prediabetes trial of a glucose-prediction diet relied on a smartphone app for self-reported intake. [1]

Food4Me, a large European trial, measured diet with an FFQ. [2]

How error is measured

Researchers compare self-reports with objective biomarkers. In the OPEN study, 484 healthy volunteers completed a food frequency questionnaire and 24-hour recalls, and their energy and protein intake were also measured using doubly labeled water and urinary nitrogen. [3] Findings:

• The FFQ led to severe attenuation of estimated disease risks for absolute protein or energy intake: a true relative risk of 2 would appear as 1.1 or smaller. [3]
• Using the 24-hour recall as the reference instrument can seriously underestimate how much error there is. [3]
• A follow-up that used a sugars biomarker found both the FFQ and the 24-hour recall biased, and the attenuation was greater in women than in men. [4]

Attenuation means the true relationship is blurred toward no effect. It is a research issue, but it shows that self-report is a rough measure.

Who is under-represented

High levels of misreporting, particularly among adults with obesity, have been observed when self-reported energy intake is compared with doubly labeled water. [5] A systematic review of 15 studies that included people with class III obesity found that people with class III obesity were underrepresented, and that five of six studies with usable data showed under-reporting. [5] Findings from validation studies may therefore not apply equally to everyone.

Why it matters for personalization

• A model that learns from logs reflects what was logged, including what was left out or misjudged.
• An app that shows nutrient numbers to the decimal still depends on what the user enters. See Nutrition Apps and Database Quality.
• If you have a history of disordered eating, ask a clinician before you start detailed tracking. See Eating Disorders and Disordered Eating.

How to read a claim based on self-reported data

1. Ask how intake was measured, and whether it was checked against a biomarker.
2. Treat small differences in reported intake with caution.
3. Remember that errors are not the same for everyone.
4. Look for objective outcome measures, not only reported ones.

For consumer tracking tools, see Nutrition Tracking and Continuous Glucose Monitors for People Without Diabetes. For the wider picture, see What Is Precision Nutrition?.