In the context of an ageing global population, dementia poses a significant public health challenge. While there is no cure, understanding the risks for dementia and how these may be minimised is key to reducing the impact of the disease. As life expectancy improves, increasing proportions of the population are expected to survive into advanced old age. As such, understanding the risks for dementia in the oldest-old and how these may differ from earlier old age is of increasing importance. The existing literature specific to the oldest-old is lower in volume and many of the findings are inconsistent. The first two chapters provide a background to the thesis such that the reader may understand the context for the subsequent studies. The first of these chapters provides an overview of dementia, focussing on the impact of the disease and the requirement for further research. The concept of the oldest-old age group is described, along with a discussion regarding the complexities associated with studying those in advanced old age. The potential impact of diverse and complex health and disease profiles in this sector of the population are introduced. The thesis objectives are introduced within the text and summarised at the close of the chapter. The second chapter introduces the study cohort on which all of the studies included in the thesis are based - the Lothian Birth Cohort 1921 (LBC1921). The present thesis had three primary objectives. The first was to determine incident cases of dementia in a study cohort of oldest-old participants: the LBC1921. Dementia cases in this cohort were ascertained using existing data primarily and the dementia ascertainment method was developed following a systematic review of such methodology within the literature. While no 'gold standard' method was found, the evidence on which the methodology for this thesis was developed is presented and discussed. Using this method, 22.5% of the n=489 eligible participants were found to have developed dementia during the follow-up period. Comparing these results with 'expected' dementia rates in the cohort, the ascertainment method was determined to be relatively effective. The second objective of this thesis was to investigate potential risk factors for dementia in oldest age, with a focus on those that would be considered modifiable. The first study of risk factors considered a range of potentially modifiable health and lifestyle factors including hypertension, diabetes, obesity, smoking, hypercholesterolaemia and physical activity. The most well documented genetic risk factor for Alzheimer's disease - APOE ɛ4 - was also included in the analyses. Contrary to other studies of dementia in the oldest-old, the presented study found that carrying at least one APOE ɛ4 allele continued to be a statistically significant risk factor for dementia in those aged over 79 years (OR: 2.23; 95% CI: 1.29, 3.86). A history of hypertension was shown to decrease the risk for incident dementia after age 79 years (OR: 0.47, 95% CI: 0.23, 0.98). This is a similar pattern to that described within the literature on the oldest-old but differs in direction from the association observed in earlier old age. The results also indicated an increased risk for dementia with greater lifetime leisure-based physical activity (OR: 1.17, 95% CI: 1.04, 1.32). This finding was again contradictory to the findings of studies of dementia in earlier old age. A history of statin-use was also observed to increase risk for dementia (OR: 3.39, 95% CI: 1.04, 11.02), while increased height reduced the risk for dementia (OR: 0.72, 95% CI: 0.55, 0.95). Overall, the findings suggested that the risk factor profile for dementia in the oldest-old, as observed in the LBC1921, differs from the risk factor profile in earlier old age. The second study of risk factors examined the association between physical fitness and dementia. The published study presented within the chapter considered three specific measures of fitness in oldest age: grip strength, walking speed and lung function (FEV1). These analyses did not demonstrate an association between any of the fitness measures at age 79 years and subsequent dementia; FEV1 (HR per unit increase 1.30, p = 0.37), grip strength (HR 0.98, p = 0.35), walking speed (HR 0.99, p = 0.90). The findings were again different to those described in studies of younger participants and supported the possibility of a changed risk factor profile for dementia in oldest-age. The final study of risk factors considered whether DNA methylation-based measures of accelerated ageing may be associated with dementia risk. Such measures of accelerated ageing may be considered, in simplest terms, as whether someone's 'biological age' is more advanced than their chronological age. The results did not demonstrate any consistent association between recognised age acceleration measures and dementia. The third objective of the thesis was to revisit previous studies of non-pathological cognitive ageing in the LBC1921 and determine whether previously unidentified cases of dementia had influenced the findings. The study looked at five previous studies, and four factors reported to be associated with poorer cognitive ageing: smoking, APOE ɛ4, reduced fitness and lower vitamin B12. After excluding those participants who had gone on to develop dementia, the analyses were repeated. The overall findings were unchanged from the original studies, with all four factors continuing to be associated with poorer cognitive ageing (p<0.05). The final chapter of the thesis provides an overview and summary of the findings from the included studies. The general limitations, with regard to methodology and the study cohort, are outlined. The chapter closes with suggestions for further research.