Otherindependent cross-checkabstract checked I-UPF-103 standard review
Can experts agree on what is ultra-processed?
Reliability and validity of assigning ultraprocessed food categories to 24-h dietary recall data.
Overview
Sneed NM et al.. Am J Clin Nutr 2023 · PMID 36789937 · DOI 10.1016/j.ajcnut.2022.10.016
With training and detailed food records, two coders agreed on the Nova category for 88% of foods.
Does the study support the specific claim? Partly · Methodological quality: moderate
Plain-language summary
Study explained simply
- What did the researchers want to find out?
To test reliability of Nova coding.
- How was the study done?
Agreement study of coded dietary recalls.
- How many people, animals or samples were studied?
3,099 foods from 610 children.
- What was actually found?
Substantial agreement (kappa 0.75).
- How large was the observed effect?
Kappa 0.75.
- What do the results mean in everyday life?
The UPF category can be used consistently by trained researchers.
- What are the study's limitations?
One setting; needs training.
- What does it say about Elawady's specific claim?
Partly supports C-UPF-01's 'scientific category'.
- Do the results apply to Germany and Europe?
Applicable.
- What can explicitly not be concluded from this study?
Does not show the label test is precise.
Scientific in-depth review
1. Full reference
- Authors
- Sneed NM et al.
- Journal
- Am J Clin Nutr
- Year
- 2023
- Identifier
- PMID 36789937 DOI 10.1016/j.ajcnut.2022.10.016 · PMC10196599
- Verification
- verified – PubMed efetch 2026-10-26; Crossref works API 2026-10-26
- Retracted
- no
- Corrections
- none
- Extent checked
- abstract only
- Conflicts of interest and funding
No COI statement in PubMed record; funding listed: AHRQ HHS, NCATS NIH HHS, NHLBI NIH HHS, NIDDK NIH HHS
2. Research question
Can trained coders reliably assign Nova categories to foods from 24-h recalls?
3. Study design and population
- Study design
- Other – Secondary analysis of 24-h dietary recalls from an RCT (reliability/validity study of Nova coding)
- Population
Children aged 3-5 years in an RCT
- Country
USA
- Period
Baseline and yearly for 3 years
- Sample size
610 children; 5,546 recall days; 3,099 unique foods
- Inclusion and exclusion criteria
Valid NDSR recalls
4. Methods and endpoints
- Exposure / intervention
Nova coding by trained coder pairs
- Comparator
Second independent coder
- Endpoints
Percent concordance; Cohen's kappa; construct validity
- Statistical methods
Interrater agreement statistics
5. Actual results
- Results
Coder concordance 88.3%; kappa 0.75. UPF were 67% of unique foods and on average 62% of daily calories; UPF supplied 94% of added sugar and 47% of protein.
- Absolute risks
Not applicable
- Authors' conclusion
This method of Nova-classifying NDSR-based recalls was reliable and valid for identifying intake of processed foods including UPF.
6. Strengths
- Large number of foods
- Trained, certified coders
7. Weaknesses
- Single dataset and protocol; US children
- Reliability depends on detailed food descriptions and training
8. Risk of bias
- Orienting tool
- none applicable
- Type of assessment
- simplified assessment (abstract only)
- Full formal assessment
- no
- Overall risk of bias
- not assessable
- Rationale
Measurement study.
9. Assessment of the statistical interpretation
Kappa 0.75 = substantial agreement; contrasts with 0.32-0.34 among untrained specialists (I-UPF-02).
10. Relevance to the specific claim
Supports that NOVA can be applied reliably under standardised conditions, partly countering I-UPF-02; does not make the category 'precise' in everyday use.
11. Independent contrary findings
I-UPF-02 Braesco 2022 found low agreement.
12. Transferability to Germany and Europe
Applicable to research use.
13. Overall methodological judgement
Reliability is achievable with training and detailed data.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | moderate |
| Risk of bias | not assessable |
| Precision of results | high |
| Directness | direct |
| Reproducibility | partly replicated |
| Consistency with the overall evidence | mixed |
| Regional transferability | moderate |
| Adequacy of the interpretation (by the archive) | not applicable |
14. Open questions
Reliability with product labels in Egypt.