Food price and diet and health outcomes: 20 years of the CARDIA Study

Abstract

Background: Despite surging interest in taxation as a policy to address poor food choice, US research directly examining the association of food prices with individual intake is scarce.

Methods: This 20-year longitudinal study included 12 123 respondent days from 5115 participants in the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Associations between food price, dietary intake, overall energy intake, weight, and homeostatic model assessment insulin resistance (HOMA-IR) scores were assessed using conditional log-log and linear regression models.

Results: The real price (inflated to 2006 US dollars) of soda and pizza decreased over time; the price of whole milk increased. A 10% increase in the price of soda or pizza was associated with a -7.12% (95% confidence interval [CI], -63.50 to -10.71) or -11.5% (95% CI, -17.50 to -5.50) change in energy from these foods, respectively. A $1.00 increase in soda price was also associated with lower daily energy intake (-124 [95% CI, -198 to -50] kcal), lower weight (-1.05 [95% CI, -1.80 to -0.31] kg), and lower HOMA-IR score (0.42 [95% CI, -0.60 to -0.23]); similar trends were observed for pizza. A $1.00 increase in the price of both soda and pizza was associated with greater changes in total energy intake (-181.49 [95% CI, -247.79 to -115.18] kcal), body weight (-1.65 [95% CI, -2.34 to 0.96] kg), and HOMA-IR (-0.45 [95% CI, -0.59 to -0.31]).

Conclusion: Policies aimed at altering the price of soda or away-from-home pizza may be effective mechanisms to steer US adults toward a more healthful diet and help reduce long-term weight gain or insulin levels over time.

Figure 1

Association between a one dollar increase in the price of selected foods and beverages with change in total energy intake (A), body weight (B), and HOMA-IR (C) ^{a a} Each food/beverage and outcome variable were modeled independently (n=12 models) as linear regression models of outcome (total energy intake (kcal, n observations = 12,007), weight (lbs, n observations = 11,972), and HOMA-IR (n observations = 10,218)) on the price (in dollars) of soda, whole milk, hamburgers, and pizza. All models adjusted for the following covariates: age (continuous), race, gender, income (low (<$25,000), middle ($25,000-<$50,000), high (≥$50,000) [referent], missing income), education (< high school (HS), completed HS [referent], 3 years college, ≥ 4 years college), family structure (single, married [referent], single with children, married with children), logged cost of living, imputed price (indicator variable, yes/no), and CARDIA study center. Models with weight as the dependent variable also adjusted for participants' height. Models adjust for clustering at the individual level. Specific food and beverage models also adjust for the following covariates: Soda, logged price of wine; Whole milk, logged price of coffee; Burger, logged price of fried chicken, steak, and parmesan cheese; Pizza,logged price of fried chicken. ^b Estimate is significantly different from zero, p<0.05.

Figure 2

Association between a one dollar increase in the price of soda alone, pizza alone, or both soda and pizza with change in total energy intake (a), body weight (b), and HOMA-IR score (c)^{a a} Estimates derived from linear regression model of outcome (total energy intake (kcal, n observations = 12,007), body weight (lbs, n observations = 11,972), and HOMA-IR (n observations = 10,218)) on the prices (in dollars) of soda, whole milk, hamburgers, and pizza. All models adjusted for age (continuous), race, gender, income (low (<$25,000), middle ($25,000-<$50,000), high (≥$50,000) [referent], missing income), education (< high school (HS), completed HS [referent], 3 years college, ≥ 4 years college), family structure (single, married [referent], single with children, married with children), logged price of the replacement beverage wine and orange juice, the logged cost of living index, having imputed prices (indicator variable, yes/no), and CARDIA study center. Models adjust for clustering at the individual level. Models with weight as the dependent variable also adjusted for participants' height. ^b Estimate is significantly different from zero, p<0.05.