Associations between types of dietary fat and fish intake and risk of stroke in the Caerphilly Prospective Study (CaPS)

Background: Few population-based longitudinal studies on diet and stroke have been conducted, and associations between dietary fat and fish intake and risk of stroke are unclear.

Objectives: To prospectively examine relationships between intakes of total fat, saturated fat, unsaturated fat, white fish and oily fish and risk of stroke in a well-defined population of 2710 middle-aged men.

Study design: Prospective cohort study.

Methods: Detailed information on health and lifestyle factors was collected via interview, and diet was assessed on three occasions using a food frequency questionnaire. Stroke ascertainment was by self-report and inspection of clinical records. Extracted data were assessed by two independent experts.

Results: During a median follow-up of 18 years, 225 strokes (209 ischaemic and 19 haemorrhagic) were eligible for inclusion in the analyses. For most recent diet (i.e. food frequency questionnaire data collected immediately prior to the stroke event), there was a slightly lower risk of stroke with higher intakes of unsaturated fat and oily fish. Multiple adjusted hazard ratios (HRs) for the lowest vs highest quintiles of unsaturated fat and oily fish intakes were 0.66 [95% confidence interval (CI) 0.41-1.05, P trend = 0.13] and 0.66 (95% CI 0.41-1.05, P trend = 0.09), respectively. Baseline and cumulative diets showed a slightly higher risk of stroke with higher intake of white fish; HRs for the lowest vs highest quintiles were 1.16 (95% CI 0.76-1.77, P trend = 0.22) and 1.28 (95% CI 0.77-2.13, P trend = 0.48), respectively.

Conclusions: Overall, strong associations were not found between intakes of different types of fat and fish and risk of stroke in middle-aged men. The inverse associations between unsaturated fat and oily fish intakes and risk of stroke were weak, but the direction of association was broadly consistent with other studies; however, these relatively weak associations were not conventionally statistically significant.