Introduction Analysis of dietary patterns offers an alternative approach to the investigation of diet and health. The aim of this study was to compare different approaches to assessing dietary patterns in the same cohort, and their associations with all-cause mortality.

Methods Analysis was based on 972 participants of the British Diet and Nutrition Survey of adults aged >65 years in 1994/1995 who were followed-up for mortality status until 2008. We examined the generalisability of reduced rank regression (RRR) methods from a previous study of all-cause mortality using: (1) an RRR-derived dietary pattern (known as the “Unhealthy Eating Index—UHI”; high in red meat, added fat, potatoes, refined grains, processed meat; low in fruit) and (2) dietary patterns derived from exploratory RRR using BMI, total cholesterol, HDL cholesterol, triglycerides and blood pressure as the intermediate markers. Three measures of diet quality, the Healthy Diet Score, the Recommended Food Score (RFS) and the Mediterranean Diet Score (MDS) were also investigated. Cox proportional hazards regression was conducted using follow-up time as the time variable.

Results After adjustment for potential confounders, the MDS, the RFS and the UHI remained associated with mortality (highest vs lowest quartile; MDS HR 0.77, 95% CI, 0.61 to 0.97; RFS HR 0.67, 95% CI 0.52 to 0.86; UHI HR 1.14, 95% CI 1.09 to 1.82). No significant associations were shown for the Healthy Diet Score or any of the exploratory RRR dietary patterns.

Conclusion Not all dietary patterns were associated with all-cause mortality. Further work is needed to test the generalisabilty of dietary patterns across cohorts.