Longitudinal Associations between Monetary Value of the Diet, DASH Diet Score and the Allostatic Load among Middle-Aged Urban Adults

Lower cost can lead to poorer-quality diets, potentially worsening metabolic profiles. We explored these pathways among urban adults. Longitudinal data were extracted from 1224–1479 participants in the Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS) study. DASH(mean) (Dietary Approaches to Stop Hypertension) score was computed using four 24 h recalls (v1/v2: 2004–2013) linked with a national food price database to estimate monetary value of the diet [MVD(mean)]. Allostatic load (AL) was measured at visits 2 (v2) and 3 (v3) in 2009–2018. Mixed-effects regression and structural equation modeling (SEM) were conducted, linking MVD(mean)/DASH(mean) to AL [v2 and annual change(v3–v2)] and exploring mediating pathways between MVD(mean) and AL(v3) through DASH(mean), stratifying by sex, race and poverty status. MVD(mean) tertiles were linearly associated with contemporaneous DASH(mean), after energy adjustment. In mixed-effects regression models, DASH(mean) was consistently linked to lower AL(v2). DASH(mean) and MVD(mean) were positively associated with higher serum albumin(v2). In SEM, MVD(mean) was linked to AL(v3) through DASH(mean), mainly among Whites and specifically for the cholesterol and Waist-Hip-Ratio AL components. In summary, energy and other covariate-adjusted increase in MVD may have a sizeable impact on DASH which can reduce follow-up AL among urban White middle-aged adults. More studies are needed to replicate findings in comparable samples of urban adults.


Supplemental method S5: Stata do file for main analysis
Supplemental methods S1: Allostatic load A total AL score was computed using a method described in a previous study. [62] AL total score sums up cardiovascular (systolic and diastolic blood pressure, pulse rate), metabolic (total cholesterol, HDLcholesterol, glycosylated Hb, sex-specific waist-to-hip ratio) and inflammatory (albumin and C-reactive protein (CRP)) risk indicators. Clinical criteria summarized in Table 1 were used to obtain risk indicators which were summed with equal weighting to compute total AL score (range: 0-9).
[63] Using standard protocols, waist-to-hip ratio, radial pulse (beats/min), and systolic and diastolic blood pressure (mmHg) were measured by trained examiners. Specifically, blood pressure was measured using a mercury sphygmomanometer [63] The arithmetic mean of three systolic and diastolic pressures was used in analysis.

Supplemental Method S2: HomeScan data description
The Homescan panel is a nationwide sample of US households that record all packaged foods and beverages purchased from grocery stores, supermarkets, and other retail food stores continuously throughout the year. Households are followed prospectively and must report purchases for at least 10 months per year. The sample includes approximately 40,000-60,000 US households each year from 76 geographic markets, and Nielsen provides projection factor weights to generate nationally representative estimates. [129] Household members scan the Universal Product Code barcode on each purchased item after each shopping trip using a handheld scanner and report the quantity purchased. Methods for reporting price paid depend on the store where the purchase takes place. For most products, Nielsen imputes the price paid from store-level point-of-sales data ("ScanTrack") as the average price paid for the product from that store for the given week and market.
[130] However, for items purchased from stores not covered by ScanTrack, households must manually record the price paid; if the reported price is outside of the typical range, Nielsen replaces the reported value with the median regional price. [130]  (1) Refined breads, (3) Multigrain breads, (5) Low sodium breads, (6) Refined Quick breads 2.

Supplemental Method S4: Description of mixed-effects regression models
The main multiple mixed-effects regression models can be summarized as follows: Multi-level models vs. Composite models Eq.