Food Frequency Questionnaire

A food frequency questionnaire presents a fixed list of foods and asks, for each, how often it was consumed over a defined period — commonly the past month, the past six months, or the past year — usually with a portion-size question attached.

It is the least precise of the major dietary assessment instruments per item, and it is the most widely used. Both facts follow from the same design decision.

What it is actually for

An FFQ is built to rank, not to quantify. Its intended output is a defensible sorting of a cohort into intake categories — lowest fifth to highest fifth of fibre intake, say — so that disease outcomes can be compared across those categories.

It is not built to tell any individual their calorie intake, and using it that way misreads the instrument. This distinction is the single most common source of confusion when FFQ-derived numbers appear outside the epidemiological literature: a questionnaire that reliably separates high from low consumers may still be substantially wrong about either one's absolute total.

Why the long reference period matters

Chronic disease relates to habitual intake over years, not to what someone ate on a Tuesday. A weighed record covers three to seven days; a 24-hour recall covers one day and needs repetition to approach usual intake. Neither reaches a year.

The FFQ reaches it by asking about frequency rather than events — which trades precision for coverage, deliberately. For an exposure that acts over decades, the trade is usually the right one.

Known biases

• Fruit and vegetable intake is over-reported, consistently and substantially. Socially desirable foods with many list entries accumulate reported frequency.
• Energy is under-reported, as with every self-report method.
• Portion assumptions dominate. Because portions are asked coarsely or assumed from population defaults, an error in the assumed portion propagates to every respondent.
• The food list is the ceiling. Anything not on the list is invisible, which makes an FFQ population-specific by construction: an instrument validated in one dietary culture cannot be transferred to another without revalidation, and this is a recurring failure when questionnaires are reused across populations.

Calibration

Because these biases are systematic rather than random, they can be partially corrected. Calibration sub-studies administer a more precise instrument — repeat recalls, weighed records, or doubly labelled water — to a subset of the cohort, estimate the relationship between FFQ-reported and reference-measured intake, and apply the correction to the full sample.

Calibration is what makes FFQ-based findings usable, and a report that presents uncalibrated FFQ energy figures as absolute intake should be read with that omission in mind.

Relationship to app-based logging

They sit at opposite ends of the same trade-off, and neither substitutes for the other.

	FFQ	App logging
Reference period	Months to a year	Whatever you logged
Per-item precision	Low	Higher
Coverage of unlogged eating	Partial, via frequency questions	None
Best at	Ranking a population	Tracking one person over time

The third row is the one worth dwelling on. An FFQ asks how often you eat crisps and gets an answer, however imprecise. An application records the crisps you entered and knows nothing about the rest. For an individual tracking their own trend that is acceptable, because the omission is roughly constant and the trend survives it. For characterising intake in absolute terms it is not.