Food diary underreporting and why calorie totals can mislead

Food diaries can miss energy intake even when completed carefully. Learn what independent measurements reveal and why average errors cannot correct your record.

日本語で読む（Japanese version）

Dinner is cleared away, and you open your food diary. Bread at breakfast, a lunch set, fish and rice in the evening: the main meals are there. But where should you put the food you tasted while cooking, or the drink you bought on the way home? For a portion you cannot remember, you enter the usual amount. The screen looks complete. Whether its numbers describe your meals accurately is another question.

A systematic review comparing adult dietary assessments with doubly labeled water measurements brought together 59 studies and 6,298 people. Most studies found that reported energy intake was too low. Methods using technology were not exempt.[5] A tidy calorie total needs to be read alongside the method that produced it and the independent measure used to check it.

The central question is whether the numbers you write down can be treated as accurate measurements of intake. Whether keeping a diary helps someone lose weight is a separate question. When weight does not change as expected, studies that look beyond the diary offer a way to examine the gap without turning a measurement problem into a judgment about the person doing the recording.

Doubly labeled water measures expenditure, not your meals

Doubly labeled water measures energy expenditure during everyday life. Researchers use it as a yardstick independent of the intake people report. Under conditions of stable body weight, energy intake equals energy expenditure, allowing researchers to compare the two and assess the validity of reported intake.[20] The comparison is with energy used by the body, rather than an independently observed list of everything someone ate.

That condition of stable weight matters. During weight loss, all the energy a person uses need not come from food eaten during the measurement period. Changes in the energy stored in the body also need consideration. A 2025 methods comparison in older adults with overweight or obesity examined this distinction directly. It compared a method using measured expenditure alone with a method calculating intake from expenditure plus changes in body energy stores.[2]

Both methods classified 50% of dietary recalls as underreported. However, the percentage classified as overreported changed from 10.2% with the expenditure comparison to 23.7% with the intake calculation. The percentages classified as plausible were 40.3% and 26.3%, respectively.[2] The reported meals had not changed between these analyses. The standard against which the reports were judged had changed, and so had some of the classifications.

These percentages describe the share of reports placed in a category. The 50% figure does not mean each person omitted half their food. Dietary assessment research can use percentages to describe both how often an error occurs and how large an error is. Those are different quantities. A count of reports classified as underreported cannot tell you the amount missing from any particular report.[2]

At home, yesterday’s diary and this morning’s scale reading may seem to provide the same kind of check. The research comparison, however, combines reported intake with independently measured expenditure and, in the newer method, changes in stored energy.[2] A household diary does not supply that independent expenditure measurement. Noticing a mismatch is a reasonable starting point; diagnosing research-defined underreporting from that mismatch alone goes beyond the information available.

Biomarker checks found 18–21% underreporting in 4-day food records

Dietary assessments take several forms. A recall asks someone to reconstruct recent eating. A food record captures eating during the recording period. A food-frequency questionnaire asks about the foods and amounts usually consumed. A 2020 review found that traditional methods can agree well with one another while all showing systematic misreporting when checked against objective indicators of nutrient intake.[15] Agreement between questionnaires does not establish that their shared answer is correct.

Systematic error means a tendency to be wrong in a particular direction, rather than simply scattered around the correct value. In the adult systematic review, the degree of underreporting varied considerably between studies using the same assessment method. Twenty-four-hour recalls generally showed less underreporting and less variation than other methods, but their errors still differed between studies.[5] A method with a smaller average error is not a guarantee of accurate intake for every user.

A 2018 validation study gives a concrete comparison. It involved 530 men and 545 women aged 50–74, followed through assessments over 12 months. Researchers compared repeated 24-hour recalls, unweighed 4-day food records, and food-frequency questionnaires against objective recovery biomarkers, including doubly labeled water.[9] This was a study of measurement accuracy. Participants were not assigned competing weight-loss treatments, so its results do not establish which recording method produces greater weight loss.

The schedule was specific: participants were asked to complete 6 recalls, 2 sets of 4-day food records, 2 food-frequency questionnaires, two 24-hour urine collections, and 1 doubly labeled water assessment during the 12 months. Urine measurements supplied objective indicators for protein, potassium, and sodium intake. At least 3 recalls were completed by 92% of men and 87% of women; the average numbers completed were 5.4 and 5.1, respectively.[9] The validation therefore concerned repeated assessments with documented completion, rather than an unspecified one-off questionnaire.

Average energy intake was underestimated by 15–17% in the recalls, 18–21% in the food records, and 29–34% in the food-frequency questionnaires.[9] The diary did not automatically come first for accuracy. Equally, these results do not mean that recalling meals is always better than recording them. The recalls were repeated, and the food records were unweighed. Those details define the comparison and matter when considering whether a personal tracking routine resembles the research method.

The tools also describe different periods of eating. A frequency questionnaire looks back at usual diet as a whole. A record describes the days on which it was kept. In this comparison, repeated recalls and the 4-day records gave better estimates of absolute intake for the nutrients examined than the frequency questionnaires.[9] Treating all these tools as interchangeable calorie ledgers obscures their different purposes and the different errors found in the study.

Correcting energy values does not necessarily correct every nutrient. In the same study, energy adjustment improved estimates of protein and sodium from the frequency questionnaires, but it did not produce the same improvement for potassium.[9] A calculation that brings one part of a dietary report closer to an objective indicator need not repair the rest. There is no universal adjustment that turns an uncertain food report into an accurate account of every nutrient.

The 2020 review similarly found that foods are not all underreported to the same degree, with protein relatively less underreported. It also noted that the specific foods most often underreported remain unclear.[15] Tastes while cooking and drinks between meals are sensible things to check in a personal diary, but the research here does not establish them as the missing foods in everyone’s record. An everyday example can guide a question without becoming a scientific finding.

Body size does not settle whose report is accurate

Body size and underreporting have been linked in many discussions of dietary measurement. The 2020 review described a tendency for energy underreporting to increase as BMI, an index of body size, increased.[15] The 2018 validation study also found underreporting more common among participants with obesity.[9] These findings deserve to be retained, rather than dismissed because they are uncomfortable or because another study reaches a different conclusion.

There is, however, an important conflicting result. A 2022 study compared reported intake with doubly labeled water expenditure in 221 adults. In absolute terms, underreporting was greater in participants with obesity, those who were more active, and younger people. After the researchers used a body-size adjustment called allometric scaling, the effect of body mass disappeared. For weight-stable adults, the authors concluded that people with obesity did not underreport to a greater extent than people without obesity.[6]

Allometric scaling is not simply dividing intake or error by body weight. It takes account of how quantities change with body size.[6] This matters because comparing absolute intake–expenditure gaps between groups with different energy needs may give a different impression from a comparison adjusted for body size. The result challenges a broad statement that one body-size group is always less accurate, without denying that differences appeared in the unadjusted comparison.

The studies therefore need to be read with their scales of comparison visible. “Larger absolute gaps were observed” and “the difference remained after accounting for body size” are not equivalent conclusions. In the 2022 study, they led to different answers.[6] When someone turns a finding about BMI into a verdict about an individual’s honesty, both the statistical comparison and the limits of self-report measurement have been lost.

Representation is another concern. A 2024 systematic review examined studies involving adults with class III obesity, defined as BMI of 40 or above. It identified 15 eligible studies. Among the 8 studies that gave relevant participant counts, only 63 of 1,784 participants, or 3.5%, belonged to this group. Of the 6 studies providing validity data for this group, 5 found underreporting.[1] There was evidence of measurement error, alongside very limited representation.

A broad label such as “adults with obesity” can conceal who actually took part. The review’s authors identified underrepresentation as a problem for understanding validity in adults with class III obesity.[1] It is reasonable to report the observed underreporting and also explain why the evidence cannot characterize every person in that category. Participant numbers help define the reach of the conclusion; they do not merely decorate it.

For a reader using a food diary, these studies favor checking the measurement conditions rather than using body size as a shortcut to credibility. Ask how the discrepancy was expressed, whether body size was accounted for, and which participants were studied.[1][6] A diary remains a report whose accuracy needs assessment. Neither a larger nor a smaller body makes its calorie total self-validating.

Numbers: the gap between reports and measurements

59 studies, 6,298 adults: the systematic review found underreporting in most studies.[5]

15–17%: average energy underestimation with repeated 24-hour recalls in the validation study of adults aged 50–74.[9]

18–21%: average underestimation with unweighed 4-day food records in that same study.[9]

29–34%: average underestimation with food-frequency questionnaires in the same comparison, not a personal correction rate.[9]

6,497 measurements: doubly labeled water observations used to develop an expenditure prediction equation; measurements are not synonymous with people.[3]