Meta-analysisindependent cross-checkfull text checked★ key study I-CHK-01 in-depth review
Less antibiotics in farm animals, less resistance
Restricting the use of antibiotics in food-producing animals and its associations with antibiotic resistance in food-producing animals and human beings: a systematic review and meta-analysis.
Overview
Tang KL et al.. Lancet Planet Health 2017 · PMID 29387833 · DOI 10.1016/S2542-5196(17)30141-9
Restricting antibiotics in livestock reduced resistant bacteria in animals and, in fewer studies, in people.
Does the study support the specific claim? Yes · Methodological quality: high
Plain-language summary
Study explained simply
- What did the researchers want to find out?
Effect of restricting antibiotic use in food animals on resistance in animals and humans
- How was the study done?
Systematic review and meta-analysis (WHO-commissioned)
- How many people, animals or samples were studied?
181 studies (179 animal, 21 human); Food-producing animals; humans
- What was actually found?
In animals resistance was usually 10-15 percentage points lower with restrictions (poultry 18 points); in studied humans 24 points lower, mainly in farm workers (29 points) and much less in non-farm people (9 points, 3 studies).
- How large was the observed effect?
Random-effects risk differences
- What do the results mean in everyday life?
Prudent farm antibiotic use protects human medicine.
- What are the study's limitations?
Mostly cross-sectional, poorly adjusted studies; very high heterogeneity; few human studies, mostly farm workers; possible publication bias in human studies; almost no data from low- and middle-income countries.
- What does it say about Elawady's specific claim?
Supports.
- Do the results apply to Germany and Europe?
High.
- What can explicitly not be concluded from this study?
Cannot quantify effect on all consumers.
Scientific in-depth review
1. Full reference
- Authors
- Tang KL et al.
- Journal
- Lancet Planet Health
- Year
- 2017
- Identifier
- PMID 29387833 DOI 10.1016/S2542-5196(17)30141-9
- Verification
- verified – PubMed efetch 2026-10-25
- Retracted
- no
- Corrections
Erratum: Lancet Planet Health 2017;1(9):e359
- Extent checked
- full text
- Conflicts of interest and funding
Funded by WHO, which helped develop the question, design and protocol and peer-reviewed the search but had no role in extraction or interpretation. One author (JDK) reports Pfizer Canada and other vaccine-trial grants unrelated to the topic.
2. Research question
Effect of restricting antibiotic use in food animals on resistance in animals and humans
3. Study design and population
- Study design
- Meta-analysis – Systematic review and random-effects (DerSimonian-Laird) meta-analysis of observational and non-randomised intervention studies, commissioned by WHO AGISAR to inform WHO guidelines
- Population
Food-producing animals; humans
- Country
Multinational
- Period
Databases searched from inception to 14 July 2016, updated 27 January 2017
- Sample size
181 studies (179 animal, 21 human; 19 overlap); meta-analysed: 81 animal and 13 human studies. Animal designs: 126 cross-sectional, 50 longitudinal, 3 non-randomised trials; poultry (broilers/turkeys) in 87 studies.
- Inclusion and exclusion criteria
Original studies of any intervention reducing antibiotic use in food-producing animals (bans, organic, antibiotic-free/free-range, voluntary reduction, stewardship) with a non-restricted comparator (historical allowed), reporting genotypic/phenotypic resistance in animals or humans; no language, date or quality limits. Pooling only if >=6 studies.
4. Methods and endpoints
- Exposure / intervention
Restriction of antibiotic use
- Comparator
No restriction
- Endpoints
Prevalence of resistant bacteria
- Statistical methods
Absolute risk differences by antibiotic class within five pre-specified animal groupings and one human meta-analysis; random effects; Q and I2; stratified analyses and meta-regression; quality per adapted Downs and Black checklist (duplicate); Begg's test and trim-and-fill for publication bias.
5. Actual results
- Results
Animals: most pooled risk differences -0.10 to -0.15 (range -0.01 for cephalosporins in Enterobacteriaceae faecal samples, 95% CI -0.04 to 0.01, to -0.39 for macrolides in Enterococcus faecal samples, -0.56 to -0.23); e.g. tetracyclines -0.16 (-0.27 to -0.05), quinolones -0.01 (-0.02 to 0.00); poultry stratum -0.18 (-0.22 to -0.14); I2 51.3-99.4%. Humans (13 studies): pooled RD -0.24 (-0.42 to -0.06), I2 97.4%; farm workers and household members -0.29 (-0.54 to -0.04) vs non-farm workers -0.09 (-0.13 to -0.05; 3 studies). Lower-quality studies tended to show larger effects. Begg's test p=0.16 (animal), p=0.047 (human); trim-and-fill human RD -0.27 (-0.43 to -0.10).
- Absolute risks
Absolute reductions in the proportion of resistant isolates mostly 10-15 percentage points in animals; 24 percentage points lower prevalence of resistant bacteria in studied humans (9 in non-farm populations, 3 studies).
- Authors' conclusion
Restrictions associated with reduced resistance in animals; smaller evidence in humans; general population implications less clear.
6. Strengths
- Very comprehensive search of 9 databases plus grey literature, experts and WHO GDG; no language restriction
- Duplicate screening (kappa 0.74), extraction and quality assessment
- Pre-specified analytic groupings, stratified analyses and publication-bias assessment
- Absolute risk differences reported
7. Weaknesses
- Underlying studies mostly cross-sectional with few adjusting for confounders
- Extreme heterogeneity (I2 up to 99.4%; human 97.4%)
- Human evidence limited (13 pooled studies), mostly farm workers; only 3 studies in non-farm populations
- Possible publication bias in human studies (Begg p=0.047)
- Lower-quality studies overestimate effects
- Almost no data from major antibiotic-using countries (one study from India, China or Brazil); none from Egypt
- Protocol pre-determined but no registration number given in retrieved text
8. Risk of bias
- Orienting tool
- AMSTAR 2
- Type of assessment
- structured assessment based on the full text by a single reviewer (no formal double review)
- Full formal assessment
- no
- Overall risk of bias
- some concerns
- Rationale
AMSTAR 2 overall confidence: moderate to low (no critical flaw clearly present; non-critical weaknesses and unverifiable excluded-study list). Overall unchanged (some_concerns). Single-reviewer structured judgement; a formal assessment needs two independent reviewers and the protocol.
| Domain | Judgement | Rationale |
|---|---|---|
| 1 PICO components | low | Population, intervention, comparator and outcome clearly specified. |
| 2 Protocol established a priori (critical) | some concerns | 'Predetermined protocol' developed with WHO, but no registration reported in the retrieved text. |
| 3 Explanation of included designs | low | Inclusive approach justified to be comprehensive; quality handled by sensitivity analyses. |
| 4 Comprehensive search (critical) | low | Nine databases, grey literature, reference lists, conference proceedings, experts; no language limits. |
| 5 Duplicate selection | low | Two reviewers independently (kappa 0.74). |
| 6 Duplicate extraction | low | Two reviewers extracted data. |
| 7 List of excluded studies with reasons (critical) | some concerns | Flow diagram with exclusion categories; full excluded list not visible in retrieved text (appendix not available). |
| 8 Description of included studies | low | Study characteristics tabulated and in appendix. |
| 9 Risk of bias of individual studies (critical) | low | Adapted Downs and Black criteria applied in duplicate. |
| 10 Funding of included studies | high | Not reported. |
| 11 Appropriate meta-analytic methods (critical) | some concerns | Random effects and minimum of 6 studies, but pooling of very heterogeneous, mostly unadjusted cross-sectional data (I2 up to 99.4%). |
| 12 Impact of RoB on results | low | Stratified analyses by quality criteria and meta-regression. |
| 13 RoB considered in interpretation (critical) | low | Authors state human evidence less robust and lower-quality studies overestimate effects. |
| 14 Heterogeneity explained | some concerns | Explored by stratification/meta-regression; largely unexplained. |
| 15 Publication bias (critical) | low | Funnel plots, Begg's test, trim-and-fill reported. |
| 16 Conflicts of interest of review | low | Funding and author interests declared; WHO role described. |
Full-text review (25 October 2026): Added search dates/databases, design mix (126 cross-sectional), poultry stratum RD -0.18, class-specific RDs, I2 values, human subgroups (farm workers -0.29 vs non-farm -0.09), Begg's tests, WHO role and COI; AMSTAR 2 itemised. Bias unchanged.
9. Assessment of the statistical interpretation
Consistent direction across species, bacteria and quality strata supports a real association, but pooled magnitudes are uncertain given heterogeneity and confounding. Evidence for transmission to the general (non-farm) population via food is weak and indirect.
10. Relevance to the specific claim
11. Independent contrary findings
Not applicable.
12. Transferability to Germany and Europe
High: most studies from Europe and North America; EU has banned antibiotic growth promoters since 2006.
13. Overall methodological judgement
High-quality, WHO-commissioned review showing that restricting antibiotic use in food animals is associated with less resistance in animals (including poultry) and in people with direct animal contact; the effect on consumers in general is plausible but poorly quantified.
| Multidimensional assessment | |
|---|---|
| Authenticity of the source | verified |
| Methodological quality | high |
| Risk of bias | some concerns |
| Precision of results | moderate |
| Directness | direct |
| Reproducibility | replicated |
| Consistency with the overall evidence | consistent |
| Regional transferability | high |
| Adequacy of the interpretation (by the archive) | not applicable |
14. Open questions
Effect on general consumers via food; data from low- and middle-income countries including Egypt; resistome-level effects.