Among 9,567 Italian adults aged 50 and over in the Moli-sani cohort, cluster analysis split participants into Healthy (42.6%), Metabolic Risk (28.6%), and Functional and Psycho-social Vulnerability (28.8%) profiles. Compared with the Healthy group, all-cause mortality hazard was 1.35 for Metabolic Risk and 1.89 for Functional and Psycho-social Vulnerability. Those hazards rose to 1.67 and 2.01 among people with higher ultra-processed food (UPF) share of food weight (interaction p = 0.037). Mediterranean and pro-vegetarian adherence tracked with lower vulnerability.

The headline interaction is fragile. It vanished when UPF was measured as a percentage of energy rather than weight, and weight-based metrics are partly driven by beverages and heavy, water-rich foods, so the signal may reflect how UPF is counted rather than a true biological effect. An interaction p of 0.037 without correction for multiple subgroup tests, plus no interaction for overall diet quality, argues for caution. Dietary data come from a single baseline questionnaire, and vulnerable participants may eat differently for reasons tied to illness, income, or disability, leaving residual confounding likely. The clustering also mixes functional and psychosocial traits with clinical ones, making the profiles hard to act on directly.

The sturdier takeaway is confirmatory: frailty, poor function, and psychosocial strain carry roughly double the mortality risk, and better-quality, less processed eating patterns accompany lower odds of that vulnerable state. For older adults, diet quality, mobility, and social wellbeing deserve attention together. The UPF modification claim is hypothesis-generating at best.