Associations of intakes of total protein, protein from dairy sources, and dietary calcium with risks of colorectal, breast, and prostate cancer: a prospective analysis in UK Biobank

Abstract

Background: Evidence concerning intakes of protein or sources of dairy protein and risks of colorectal, breast, and prostate cancers is inconclusive.

Methods: Using a subsample of UK Biobank participants who completed ≥2 (maximum of 5) 24-h dietary assessments, we estimated intakes of total protein, protein from total dairy products, milk, and cheese, and dietary calcium in 114,217 participants. Hazard ratios (HRs) and 95% confidence intervals (CI) were estimated using multivariable-adjusted Cox regression.

Results: After a median of 9.4 years of follow-up, 1193 colorectal, 2024 female breast, and 2422 prostate cancer cases were identified. There were inverse associations of total dairy protein, protein from milk, and dietary calcium intakes with colorectal cancer incidence (HR_{Q4 vs Q1}:0.80, 95% CI: 0.67-0.94; 0.79, 0.67-0.94; 0.71, 0.58-0.86, respectively). We also observed positive associations of milk protein and dietary calcium with prostate cancer risk (HR_{Q4 vs Q1}:1.12, 1.00-1.26 and 1.16, 1.01-1.33, respectively). No significant associations were observed between intake of dairy protein and breast cancer risk. When insulin-like growth factor-I concentrations measured at recruitment were added to the multivariable-adjusted models, associations remained largely unchanged. Analyses were also similar when looking at total grams of dairy products, milk, and cheese.

Conclusion: Further research is needed to understand the mechanisms underlying the relationships of dairy products with cancer risk and the potential roles of dietary protein and calcium.

Fig. 1. Multivariable-adjusted hazard ratios (95% CI) for intakes of total protein, protein from all dairy products and dairy sources, and dietary calcium and colorectal, breast, and prostate cancer risk.

All models used age as the underlying time variable, were stratified by age groups at recruitment, and were further adjusted for height, physical activity, Townsend deprivation index, education, employment status, smoking status, alcohol intake, ethnicity, diagnosis of diabetes, BMI, energy intake. For colorectal cancer analyses: all models were stratified by sex and further adjusted for menopausal status (women only), menopausal hormone therapy use (women only), red and processed meat intake, and non-steroidal anti-inflammatory drug use. For breast cancer analyses: models were further adjusted for menopausal hormone therapy use, oral contraceptive use, parity and age at first birth, age at menarche, BMI and menopausal status. For prostate cancer analyses: models were further adjusted for marital status. Abbreviations: BMI body mass index, CI confidence interval, HR hazard ratio, N number of participants, Q quantile, ref reference group.