Observational Dose-Response Meta-Analysis Methods May Bias Risk Estimates at Low Consumption Levels: The Case of Meat...

Area of Science:

• Nutritional Epidemiology
• Cancer Research
• Biostatistics

Background:

• Observational studies on diet and health are prone to bias, especially confounding factors.
• Standard dose-response meta-analysis (DRMA) methods may yield biased or overly certain risk estimates for food-health associations.
• Investigating colorectal cancer (CRC) risk associated with red meat (RM) and processed meat (PM) consumption requires careful methodological consideration.

Purpose of the Study:

• To evaluate the empirical evidence for CRC association with unprocessed red meat (RM) and processed meats (PM) using DRMA models.
• To assess the consistency of CRC association for low and high consumers under different modeling assumptions.
• To compare classical DRMA models against an empirical model and evaluate the impact of reference consumer selection.

Main Methods:

• Utilized data from Global Burden of Disease systematic reviews, compiling studies on PM (29 cohorts) and RM (23 cohorts).
• Fitted DRMA models using only lower consumers (below US median intake) and compared with models using all consumption levels.
• Compared classical DRMA models against an empirical model, assessing nonlinear, nonmonotonic relationships and the influence of reference groups.

Main Results:

• No significant association was found between 50 g/d RM consumption and CRC using empirical models with lower consumption (RR 0.93) or all consumption levels (RR 1.04).
• Classical models indicated a significant association for RM (RR 1.09 at 50 g/d).
• No significant association was found between 20 g/d PM consumption and CRC using lower consumer data (RR 1.01), regardless of model choice.
• Using all consumption data for PM showed an association with CRC at 20 g/d for empirical models (RR 1.07) and as little as 1 g/d for classical models.
• Empirical DRMA revealed nonlinear, nonmonotonic relationships for both PM and RM.
• Reference group selection (nonconsumer vs. mixed) did not significantly affect CRC association in lowest consumption arms for RM or PM.

Conclusions:

• Classical DRMA model assumptions and the inclusion of higher consumption levels significantly influence the observed association between CRC and low RM/PM consumption.
• The empirical DRMA suggests that a no-risk limit of 0 g/d for RM and PM consumption is inconsistent with the current evidence.
• Accurate risk assessment for meat consumption and CRC requires careful consideration of modeling techniques and consumption levels.