Loneliness has been reframed from a ‘social problem of old age’ into a major public health problem. This transformation has been generated by findings from observational studies of a relationship between loneliness and a range of negative health outcomes including dementia. From a public health perspective, key to evaluating the relationship between loneliness and dementia is examining how studies define and measure loneliness, the exposure variable, and dementia the outcome. If we are not consistently measuring these then building a body of evidence for the negative health outcomes of loneliness is problematic. Three key criteria had to meet for studies to be included in our analysis. To test the proposition that loneliness is a cause of dementia we only included longitudinal studies. For inclusion studies had to measure loneliness at baseline, have samples free of dementia and assess dementia at follow up (specified as a minimum of 12 months). We identified 11 papers published between 2000 and 2018 that meet these criteria. These studies included seven different countries and only one was specifically focused upon dementia: all other studies were cohort studies focused upon ageing and health and wellbeing. There was extensive heterogeneity in how studies measured loneliness and dementia and in the use of co-variates. Loneliness was measured by either self-rating scales (n = 8) or scales (n = 3). Dementia was assessed by clinical tests (n = 5), diagnostic/screening tools (n = 3), cognitive function tests (n = 1), and self-reported doctor diagnosis (n = 2). Substantial variation in loneliness prevalence (range 5–20%) and dementia incidence (5–30 per 1000 person years at risk). Six studies did not report a statistically significant relationship between loneliness and dementia. Significant excess risk of dementia among those who were lonely ranged from 15% to 64%. None of these studies are directly comparable as four different loneliness and dementia measures were used. We suggest that the evidence to support a relationship between loneliness and dementia is inconclusive largely because of methodological limitations of existing studies. If we wish to develop this evidence base, then using a consistent set of loneliness and dementia outcome measures in major longitudinal studies would be of benefit.