Fig. 2

Association between MedDiet adherence and risk of dementia in sub-group and sensitivity analyses. MedDiet adherence level was split into tertiles, with the dashed line reflecting the low MedDiet adherence reference group for each MedDiet score. Analyses include: A) Primary analyses for the MEDAS, MEDAS continuous and PYRAMID scores (n = 60,298, including 882 dementia cases); B) Including participants with a minimum of 2 dietary reports (n = 38,794, including 479 dementia cases); C) After excluding participants with extreme energy intakes (n = 59,627, including 867 dementia cases); D) Excluding participants with less than 2 years (n = 59,594, including 843 dementia cases) and 5 years (n = 58,196, including 698 dementia cases) follow up; E) Adjusting for potential mediators, including BMI (n = 60,163, including 876 dementia cases), history of depression (n = 58,837, including 851 dementia cases), or stroke (n = 60,298, including 882 dementia cases); F) Stratified into low (n = 21,009, including 261 dementia cases), medium (n = 20,000, including 313 dementia cases) and high (n = 19,273, including 308 dementia cases) genetic risk categories; G) Stratified into APOE4 carriers (n = 16,644, including 467 dementia cases) and non-carriers (n = 43,651, including 415 dementia cases); H) With imputed missing data (n = 196,335, including 5001 dementia cases); I) Restricted to fatal (n = 59,627, including 260 dementia cases) and non-fatal (n = 60,038, including 622 dementia cases) dementia cases; and J) Stratified into higher (n = 33,281, including 430 dementia cases) and lower (n = 27,007, including 452 dementia cases) education status groups