Dairy intake and cancer risk: focusing on dairy fat content and consumption patterns
Lung, breast, prostate and colorectal cancers are the four major cancers in the US and account for over 40% of all cancer cases. Current evidence on total and specific dairy product intakes and cancer risk for these four organ sites is controversial. Recent meta-analyses of dairy intake and cancer risk reported positive associations of total dairy, cheese and low-fat milk intakes with prostate cancer risk, whereas inverse associations of total dairy intake with risk of colorectal, breast and lung cancers were reported. Besides dairy intake, both fat content of dairy products and dairy consumption patterns (e.g., types of milk added to breakfast cereals vs coffee) may affect cancer risk. However, to our knowledge, no previous study of cancer investigated these factors together. We hypothesize that dairy fat content and dairy consumption patterns, as well as total and specific dairy product intakes, are associated with cancer risk. The PLCO is uniquely suited to test this hypothesis, given that the Diet History Questionnaire (DHQ) collected detailed information on consumption of various dairy products by fat content (e.g., milk: whole, 2%, 1%, or skim/nonfat) and consumption patterns at which they were consumed with or without another food (e.g., type of milk consumed as a beverage, added to cold cereals, or added to coffee). Within the PLCO, in earlier analyses using Dietary Questionnaire, total dairy intake was not associated with colorectal adenoma, a known precursor to colorectal cancer (Dixon et al. 2007), whereas intake of low-fat dairy, not total or high-fat dairy, was positively associated with prostate cancer risk (Ahn et al. 2007). In this project, we will extend these previous analyses by: 1) including more prostate and colorectal cancer cases as well as lung and breast cancers as outcomes; and 2) evaluating associations of risk of each cancer site with dairy intake by fermentation and individual products, and by the preference in dairy fat content by products and consumption patterns that were assessed in DHQ. In our statistical analyses, we will run multivariate Cox proportional hazards regression models, separately for each cancer site, to estimate risk of each cancer by dairy product intake levels (Aim 1), or by the preference in fat content of dairy products and/or dairy consumption patterns (Aim 2). For the Aim 2, we will model variables on the preference in fat content of each dairy product and/or consumption patterns individually or by data-driven or pre-defined grouping.
Aim 1: to investigate whether dairy product intakes, as total (all dairy products combined), by fat content (full-fat or low/non-fat), by fermentation (fermented or unfermented), or by individual dairy product (also by fat content if available), are associated with risk of cancer in each of the lung, breast, prostate or colon/rectum.
Aim 2: to investigate whether the preference in fat content by dairy products and/or dairy consumption patterns, as assessed individually or by grouping, is associated with risk of lung, breast, prostate or colorectal cancer.
Gerd Bobe (Oregon State University)
Jackilen Shannon (Oregon Health & Science University)
Major PLCO investigators