Closer relationships between caregivers and care recipients with dementia are associated with positive outcomes for care recipients, but it is unclear if closeness is a risk or protective factor for the health and psychological well-being of caregivers. We examined 234 care dyads from the population-based Cache County Dementia Progression Study. Caregivers included spouses (49%) and adult offspring (51%). Care recipients mostly had dementia of the Alzheimer's type (62%). Linear mixed models tested associations between relationship closeness at baseline or changes in closeness prior to versus after dementia onset, with baseline levels and changes over time in caregiver affect (Affect Balance Scale, ABS), depression (Beck Depression Inventory, BDI), and mental and physical health (components of the Short-Form Health Survey, SF-12). After controlling for demographic characteristics of the caregiver, number of caregiver health conditions, and characteristics of the care recipient (type of dementia, functional ability, and behavioral disturbances), we found that higher baseline closeness predicted higher baseline SF-12 mental health scores (better mental health) and lower depression. Higher baseline closeness also predicted greater worsening over time in ABS and SF-12 mental health. In addition, caregivers who reported a loss of closeness in their relationship with the care recipient from pre- to post-dementia displayed improved scores on ABS and SF-12 mental health, but worse SF-12 physical health over the course of the study. These results suggest that closeness and loss of closeness in the care dyad may be associated with both positive and adverse outcomes for caregivers, both cross-sectionally and over time.