Are Reproducible Dietary Patterns Consistently Associated With Disease Outcomes or Their Drivers in Italy? A Systematic Review

The strength, direction, and trend of associations between specific diseases and reproducible a posteriori dietary patterns (DPs) based on principal component analysis (PCA) or exploratory factor analysis (EFA) have rarely been investigated across populations. We conducted a systematic review of PCA/EFA-based DPs identified in Italy to explore 2 methodological issues: 1) cross-study reproducibility of Italian DPs; 2) consistency of associations between reproducible DPs and the same/similar disease outcomes/DP drivers/correlates. The systematic review process and findings on DP cross-study reproducibility were published separately. This paper focuses on associations, summarizing the data in figures and tables, with post-hoc criteria for similarity among target variables, statistical methods, and adjustment for confounding. Predefined rules of inference were used to evaluate selected Hill’s causal criteria (consistency, strength, and dose–response effects) and draw valid scientific conclusions on the association between PCA/EFA-based DPs and similar/the same target variables. Fifty-two articles, primarily on EFA-based DPs derived from food frequency questionnaires, were included. Regression models were used to explore the relationships between DPs and disease outcomes/DP drivers, aligning with original research questions, study designs, and literature on confounding. When considering similar target variables, 9 groups of reproducible DPs showed >50% statistically significant associations in the same direction across 1–3 groups of target variables, such as socioeconomic characteristics, incidence of chronic diseases, overall/cause-specific mortality, cardiovascular disease risk factors, pregnancy/breastfeeding-related and elderly-related outcomes. Groups targeting dairies/sweets and vegetable sources of fats showed >50% nonsignificant findings across all similar target variables. Overall, 54% of findings were nonsignificant. When considering the same target variable, the median number of DPs per group was equal to 2 (interquartile range: 2–2.5). Together with population comparability issues, this prevented us from reliably performing any meta-analyses. At this stage, valid scientific conclusions cannot be drawn to inform Italian nutritional recommendations.