Red and processed meat intakes and cardiovascular disease and type 2 diabetes mellitus: An umbrella systematic review and assessment of causal relations using Bradford Hill’s criteria

Figures & data

Table 1. Description of PICOS criteria for an umbrella review assessing causal relations between red and processed meat intake and cardiometabolic disease.

Variable	Description
Population	Human adults with a mean age ≥ 19 years who are free of the cardiometabolic disease outcome of interest at enrollment
Intervention/Exposure	Red or processed meat consumption listed as an a priori^a independent exposure or variable of interest
Comparator	No or lower red meat intake with no restrictions on alternative food sources
Outcome	Report of any CVD and/or T2DM risk factors, incidence or mortality outcomes^b
Study Design	Systematic reviews or systematically searched meta-analyses of observational and experimental human studies

PICOS (Population, Intervention/Exposure, Comparator, Outcome, Study Design); CVD (Cardiovascular Disease; T2DM (Type 2 Diabetes Mellitus).

a Red or processed meat consumption was determined to be an a priori independent exposure or variable if listed in the search terms, objectives, or methods sections of the review.

b See Supplemental Table S3 for list of outcomes.

Table 2. The Bradford Hill causality criteria descriptions (Hill 1965) for an umbrella review assessing red and processed meat intake and cardiometabolic disease.

Bradford Hill criteria	General description
Strength	A large association is more likely to be causal but a small association does not mean that there is no causal effect
Consistency	Consistent findings observed for different populations, in different places, at different times, under different circumstances strengthens the likelihood of causality
Specificity	If the exposure is not related to other diseases or causes of death, then this strengthens the likelihood of causality
Temporality	The exposure must occur before the outcome
Biological Gradient	Greater exposure leads to either increased or decreased incidence of the outcome
Plausibility	A plausible biological mechanism exists to explain the relation between the exposure and outcome
Experiment	When available, experimental evidence can provide insights into direct cause and effect relationships
Coherence	Agreement between the observational and experimental findings increase the likelihood for causality
Analogy	The effect of similar exposures on the outcomes of interest may be considered to infer causality

Table 3. Bradford Hill causality criteria summary for all meat type categories and CVD and T2DM outcomes.

Bradford Hill causality criteria	Cardiovascular disease outcomes			Type 2 diabetes mellitus outcomes
	Mixed URM + PM	PM	URM	Mixed URM + PM	PM	URM
Strength	X	X	X	✔	✔	Could not assess
Consistency	X	X	X	✔	✔	Could not assess
Specificity	X	X	X	X	X	X
Temporality	✔	✔	✔	✔	✔	✔
Biological Gradient	✔ ^a	✔	✔^a	✔^a	✔	X
Plausibility	✔	✔	✔	✔	✔	✔
Experiment	X	Could not assess	X^b	X	Could not assess	X
Coherence	X	Could not assess	X^b	X	Could not assess	Could not assess
Analogy	X	X	X	X	X	X

CVD (Cardiovascular Disease); T2DM (Type 2 Diabetes Mellitus); URM (Unprocessed Red Meat); PM (Processed Meat). ✔ = criteria are supported; X = criteria are not supported; Could not assess = criteria could not be assessed due to lack of reviews.

This summary table is representative of all reviews included into this umbrella review and is supported by the secondary analysis of High and Moderate quality reviews unless otherwise noted.

a There was not support for the causal criterion in the post hoc secondary analysis of High and Moderate quality reviews.

b This criterion could not be assessed in the post hoc secondary analysis because there was no experimental evidence from High and Moderate quality reviews.

Figure 1. Article screening process flow chart for an umbrella review assessing red and processed meat intake and cardiometabolic disease.

Table 4. Mixed unprocessed red meat + processed meat and CVD outcomes: Bradford Hill causality criteria and AMSTAR 2 quality findings.

Bradford Hill causality criteria*
		Outcome	Total reviews	Total assessments	Statistically significant assessments	ES ≥1.2 assessments
Strength and Consistency	Not Supported	CVD RF (OBS)	3	3	2	1^†
		CVD Incidence	5	11	7	1^‡
		CVD Mortality	4	6	2	0
Biological Gradient	Supported	CVD RF (OBS)	2	2	2	–
		CVD Incidence	6	9	8	–
		CVD Mortality	4	6	4	–
Experiment	Not Supported	CVD RF (RCT)	4	31	8	–
Coherence	Not Supported	Lack of coherence between observational findings and experimental evidence
AMSTAR 2 quality of assessments
Outcome	Total reviews	Total assessments	Quality of assessments:
			High	Moderate	Low	Critically low
CVD RF (OBS)	3	3	0	0	1	2
CVD Incidence	8	15	2	1	10	2
CVD Mortality	6	9	1	1	5	2
CVD RF (RCT)	4	31	7	6	0	18

CVD (Cardiovascular Disease); RF (Risk Factors); OBS (outcomes from Observational studies); RCT (outcomes from Randomized experimental studies); ES (Effect Size); AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews).

* Refer to Table 2 for Bradford Hill Causality Criteria descriptions.

† Assessment is not statistically significant (RR: 1.31 95% CI: 0.91, 1.89) (Kim and Je 2018).

‡ Assessment is not statistically significant (RR: 1.33 95% CI: 0.95, 1.89(Vernooij et al. 2019).

Table 5. Processed meat and CVD outcomes: Bradford Hill causality criteria and AMSTAR 2 quality findings.

Bradford Hill causality criteria*
		Outcome	Total articles	Total assessments	Statistically significant assessments	ES >1.2 assessments
Strength and Consistency	Not Supported	CVD RF (OBS)	1	1	1	0
		CVD Incidence	5	7	6	1
		CVD Mortality	2	3	2	1^†
Biological Gradient	Supported	CVD RF (OBS)	1	1	1	–
		CVD Incidence	7	11	9	–
		CVD Mortality	4	6	5	–
Experiment	Could Not Assess	CVD RF (RCT)	0	0	0	–
Coherence	Could Not Assess	Could not assess due to lack of experimental data
AMSTAR 2 quality of assessments
Outcome	Total articles	Total assessments	Quality of assessments:
			High	Moderate	Low	Critically low
CVD RF (OBS)	1	1	0	0	0	1
CVD Incidence	8	12	2	2	6	2
CVD Mortality	4	6	1	0	3	2
CVD RF (RCT)	0	0	0	0	0	0

CVD (Cardiovascular Disease); RF (Risk Factors); OBS (outcomes from Observational studies); RCT (outcomes from Randomized experimental studies); ES (Effect Size); AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews).

* Refer to Table 2 for Bradford Hill Causality Criteria descriptions.

† Assessment is not statistically significant (RR: 1.52 95% CI: 0.52, 4.66) (Abete et al. 2014).

Table 6. Unprocessed meat and CVD outcomes: Bradford Hill causality criteria and AMSTAR 2 quality findings.

Bradford Hill causality criteria*
		Outcome	Total articles	Total assessments	Statistically significant assessments	ES >1.2 assessments
Strength and Consistency		CVD RF (OBS)	0	0	0	0
	Not Supported	CVD Incidence	4	4	4	0
		CVD Mortality	2	2	0	0
Biological Gradient	Supported	CVD RF (OBS)	0	0	0	–
		CVD Incidence	6	8	5	–
		CVD Mortality	2	3	3	–
Experiment	Not Supported	CVD RF (RCT)	1	4	0	–
Coherence	Not Supported	Lack of coherence between observational findings and experimental evidence
AMSTAR 2 quality of assessments
Outcome	Total articles	Total assessments	Quality of assessments:
			High	Moderate	Low	Critically low
CVD RF (OBS)	0	0	0	0	0	0
CVD Incidence	7	9	2	2	3	2
CVD Mortality	3	4	0	1	3	0
CVD RF (RCT)	1	4	0	0	0	4

CVD (Cardiovascular Disease); RF (Risk Factors); OBS (outcomes from Observational studies); RCT (outcomes from Randomized experimental studies); ES (Effect Size); AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews).

* Refer to Table 2 for Bradford Hill Causality Criteria descriptions.

Table 7. Mixed unprocessed red meat + processed meat and T2DM outcomes: Bradford Hill causality criteria and AMSTAR 2 quality findings.

Bradford Hill Causality Criteria*
		Outcome	Total reviews	Total assessments	Statistically significant assessments	ES ≥1.2 assessments
Strength and Consistency	Supported	T2DM RF (OBS)	2	2	1	1^†
		T2DM Incidence	6	6	6	6
Biological Gradient	Supported	T2DM RF (OBS)	1	1	1	–
		T2DM Incidence	6	6	5	–
Experiment	Not Supported	T2DM RF (RCT)	3	12	6	–
Coherence	Not Supported	Lack of coherence between observational findings and experimental evidence
AMSTAR 2 quality of assessments
Outcome	Total reviews	Total assessments	Quality of assessments:
			High	Moderate	Low	Critically low
T2DM RF (OBS)	2	2	0	0	1	1
T2DM Incidence	8	8	1	1	4	2
T2DM RF (RCT)	3	12	6	6	0	0

T2DM (Type 2 Diabetes Mellitus); RF (Risk Factors); OBS (outcomes from Observational studies); RCT (outcomes from Randomized experimental studies); ES (Effect Size); AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews).

* Refer to Table 2 for Bradford Hill Causality Criteria descriptions.

† Assessment is not statistically significant (RR: 1.31 95% CI: 0.91, 1.89) (Kim and Je 2018).

Table 8. Processed meat and T2DM outcomes: quantitative Bradford Hill causality criteria and AMSTAR 2 quality findings.

Bradford Hill causality criteria*
		Outcome	Total reviews	Total assessments	Statistically significant assessments	ES ≥1.2 assessments
Strength and Consistency	Supported	T2DM RF (OBS)	0	0	0	0
		T2DM Incidence	5	5	5	5
Biological Gradient	Supported	T2DM RF (OBS)	0	0	0	–
		T2DM Incidence	6	6	6	–
Experiment	Could Not Assess	T2DM RF (RCT)	0	0	0	–
Coherence	Could Not Assess	Could not assess due to lack of experimental data
AMSTAR 2 quality of assessments
Outcome	Total reviews	Total assessments	Quality of assessments:
			High	Moderate	Low	Critically low
T2DM RF (OBS)	0	0	0	0	0	0
T2DM Incidence	7	7	1	1	3	2
T2DM RF (RCT)	0	0	0	0	0	0

T2DM (Type 2 Diabetes Mellitus); RF (Risk Factors); OBS (outcomes from Observational studies); RCT (outcomes from Randomized experimental studies); ES (Effect Size); AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews).

* Refer to Table 2 for Bradford Hill Causality Criteria descriptions.

Table 9. Unprocessed red meat and T2DM outcomes: quantitative Bradford Hill causality criteria and AMSTAR 2 quality findings.

Bradford Hill causality criteria*
		Outcome	Total reviews	Total assessments	Statistically significant assessments	ES ≥1.2 assessments
Strength and Consistency	Could Not Assess	T2DM RF (OBS)	0	0	0	0
		T2DM Incidence	0	0	0	0
Biological Gradient	Not Supported	T2DM RF (OBS)	0	0	0	–
		T2DM Incidence	2	2	1	–
Experiment	Not Supported	T2DM RF (RCT)	1	3	0	–
Coherence	Could Not Assess	Could not assess due to lack of observational data
AMSTAR 2 quality of assessments
Outcome	Total reviews	Total assessments	Quality of assessments:
			High	Moderate	Low	Critically low
T2DM RF (OBS)	0	0	0	0	0	0
T2DM Incidence	2	2	1	0	1	0
T2DM RF (RCT)	1	3	3	0	0	0

T2DM (Type 2 Diabetes Mellitus); RF (Risk Factors); OBS (outcomes from Observational studies); RCT (outcomes from Randomized experimental studies); ES (Effect Size); AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews).

* Refer to Table 2 for Bradford Hill Causality Criteria descriptions.

Table 10. Bradford Hill causality inference summary table for an umbrella review assessing red and processed meat intake and cardiometabolic disease.

Outcome	Meat type	Causal inference summary
Cardiovascular Disease	Mixed Unprocessed Red Meat + Processed Meat	We infer that a causal relationship is not likely due to consistently weak associations among observational studies, which are coherent with non-significant experimental evidence, despite a temporal relationship and plausible biological mechanisms.
	Processed Meat	We infer that a causal relationship is not likely due to consistently weak associations among observational studies, despite the potential for a dose-response relationship and plausible biological mechanisms. No reviews of experimental evidence were found; thus, coherence could not be assessed and contribute to this inference.
	Unprocessed Red Meat	We infer that a causal relationship is not likely due to consistently weak associations among observational studies, which were coherent with non-significant experimental evidence, despite a temporal relationship and plausible biological mechanisms.
Type 2 Diabetes Mellitus	Mixed Unprocessed Red Meat + Processed Meat	We infer a potential for a causal relationship due to consistently strong associations among observational studies, plausible mechanisms, and a potential for a dose-response relationship. Yet, coherence is lacking with the experimental evidence reporting non-significant findings.
	Processed Meat	We infer a potential for a causal relationship due to consistently strong associations among observational studies, plausible mechanisms, and a potential for a dose-response relationship. No reviews of experimental evidence were reported; thus, coherence could not be assessed and contribute to this inference.
	Unprocessed Red Meat	Insufficient evidence precluded evaluation.

Hill, A. B. 1965. The environment and disease: association or causation? Proceedings of the Royal Society of Medicine 58 (5):295–300. doi: 10.1177/003591576505800503.

 PubMedGoogle Scholar

Kim, Y, and Y. Je. 2018. Meat consumption and risk of metabolic syndrome: results from the korean population and a meta-analysis of observational studies. Nutrients 10 (4):390. doi: 10.3390/nu10040390.

 PubMed Web of Science ®Google Scholar

Vernooij, R. W., M. D. Zeraatkar, M. A. Han, R. E. Dib, M. Zworth, K. Milio, D. Sit, Y. Lee, H. Gomaa, C. Valli, et al. 2019. Patterns of red and processed meat consumption and risk for cardiometabolic and cancer outcomes: a systematic review and meta-analysis of cohort studies. Annals of Internal Medicine 171 (10):732–41. doi: 10.7326/M19-1583.

 PubMed Web of Science ®Google Scholar

Abete, I., D. Romaguera, A. R. Vieira, A. Lopez de Munain, and T. Norat. 2014. Association between total, processed, red and white meat consumption and all-cause, CVD and IHD mortality: a meta-analysis of cohort studies. The British Journal of Nutrition 112 (5):762–75. doi: 10.1017/S000711451400124X.

 PubMed Web of Science ®Google Scholar

Data availability statement

All results described in the manuscript are provided in the Supplemental Material.