The oldest-old (people aged 80 or older) constituted 5 % of the population in Sweden in 2000, an increase from 1.5 % fifty years earlier. The immune system undergoes dramatic changes at high age, sometimes referred to as “immunosenescence”. However, the natures of these changes, and in particular, their clinical consequences are incompletely understood. In a previous longitudinal study, a set of immune parameters were identified and termed immune risk phenotype (IRP) because of an association with increased mortality. The IRP consists of changes in the T lymphocyte compartment, in particular an inverted CD4/CD8 ratio. The IRP was found to be associated with cytomegalovirus (CMV) infection, which through expansions of cytolytic anti-viral CD8 cell responses was ascribed a role in the development of IRP. The general aim of this thesis was to increase the knowledge of changes in the immune system and their clinical consequences in the oldest-old. The population-based random sample of the longitudinal NONA-Immune Study (n = 138, mean age 90 years at baseline) was used for all investigations. In paper I, the effects on sample size of various exclusion protocols for immune studies of the elderly was examined. The commonly used SENIEUR protocol, selecting individuals representing ‘normal ageing’, excluded 90 % of nonagenarians. Based on different protocol criteria, individuals were grouped into ‘very healthy’, ‘moderately healthy’ or ‘frail’. The prevalence of CMV was similar across the groups. Further, differentiated CD8 populations associated with CMV, i.e. those expressing CD56, CD57 and CD45RA while lacking expression of CD27 and CD28, were equally distributed across the groups of the oldest-old, but were, as expected, significantly increased in the elderly compared to a middle aged control group. The findings showed that lymphocyte subsets associated with IRP might serve as significant biomarkers of ageing independent of the overall health status, also supporting the notion that immunological studies of the oldest-old should be done in population-based non-selected populations. The IRP and the presence of low-grade inflammation, for example increase of   IL-6 in plasma, constitute major predictors of 2-year mortality in the oldest-old. In paper II, the CD4/CD8 ratio and IL-6 were found to predict 97 % of observed survival and 57 % of deaths over 2 years. The impact of IRP and IL-6 on 2-year survival was independent of age, sex and several diseases. The longitudinal design allowed temporal evaluations, suggesting a sequence of events starting with IRP and leading to inflammation in the decline state.                                      Four-year mortality in the oldest-old (paper III) was found to be mainly related to markers of inflammation and IRP. Individuals with both inverted CD4/CD8 ratio and high IL-6 level had significantly higher 4 year mortality (82 %) compared to individuals with CD4/CD8 ratio ³ 1 and low IL-6 level (29 %) at baseline. The presence of IRP and increased IL-6 level showed some associations with presence of diseases; in particular, IL6 was associated with the presence of cognitive impairment. However, despite being strong predictors of mortality, IRP and IL-6 could not be linked to any specific cause of death, probably due to the multi-factorial nature of these factors.                                                                                                                             The prevalence of antinuclear antibodies (ANA) in the oldest-old was higher compared to younger controls (paper IV). The difference across age was most pronounced in men, showing low levels at younger age, whereas the prevalence among the oldest-old men reached a similar level as in women. There was no association between the presence of ANA and IRP, CMV status or health status in the oldest-old.