Rapid,Digital Dietary Assessment in Association with Cardiometabolic Biomarkers

Abstract

Purpose

To examine the associations between dietary intake as assessed by a rapid, image-based digital tool and biomarkers of cardiometabolic health.

Design

Retrospective analysis of adults with blood biomarkers performed by Boston Heart Diagnostics (BHD) between December 2020 and March 2022.

Setting

Outpatient centers serviced by BHD.

Subjects

546 adults, excluding those taking relevant medications and/or supplements known to affect blood test results.

Measures

Laboratory assays of blood specimens were performed by Boston Heart Diagnostics. Nutrient intake and diet quality data were obtained using Diet Quality Photo Navigation (DQPN®; US Patent #11,328,810 B2) technique via Diet ID™ tool.

Analysis

Pearson correlation coefficients (for continuous variables) and Spearman coefficients (for ordinal variables) were used to evaluate associations between nutrient intake data and laboratory data for the full study sample. Two-sided P-values < .05 were considered statistically significant.

Results

Both continuous and ordinal measures of diet quality correlated significantly with HDL-C and triglycerides (n = 485; P < .0 01); with hs-CRP (n = 441; P < .001); with HgbA1c (n = 345; P < .01); with fasting insulin (n = 372; P < .001); and with HOMA-IR (n = 319; P < .001).

Conclusion

Findings affirm that rapid, digital diet quality and composition assessment by pattern recognition rather than recall tracks significantly with key biomarkers of cardiometabolic health.

Keywords

cardiometabolic health cardiovascular nutrition diet quality nutrition

Purpose

Diet quality and composition exert profound effects on every aspect of health.¹ Impairment of overall diet quality is the single leading predictor of premature mortality and chronic morbidity in the United States,² and much of the world.¹ During the COVID pandemic, poor diet quality has been identified as a direct contributor to adverse outcomes of infection,³ as well as an indirect contributor via cardiometabolic disease.⁴ Along with the dire human costs of poor diet quality, the associated economic costs are extreme.⁵

Such considerations have prompted public health authorities to highlight the outsized influence of diet on population health, and call for relevant responses.² Among the actions espoused is the routine capture of nutrition in every electronic health record; in effect, the treatment of dietary intake as a vital sign. This perspective borrows an adage from the world of business - we manage what we measure- and evinces its comparable relevance to the world of medicine.

Medicine, in turn, has taken notice. In 2020, the American College of Cardiology published a position statement calling for the inclusion of nutrition as part of every patient assessment.⁶ That same paper called for the development of new approaches that would allow for efficient, accurate assessment of dietary intake at such scale, and within the standard clinical workflow.

Diet quality photo navigation is such an approach⁷ and is based on pattern recognition rather than recall or journaling, achieving a comprehensive assessment of dietary intake in as little as 60 seconds via any digital interface. The method has been compared favorably to the semi-quantitative food frequency questionnaire⁸ and the 24-hour recall.⁹ In this paper, we examine the associations between diet quality and composition assessed by this emerging method, and biomarkers of cardiometabolic health (ie, serum lipids, inflammatory cytokines, markers of insulin/glucose metabolism).

Methods

Design and Sample and Measures

This study used clinical care data that were not collected specifically for this study and the patient identifiers were not available to the researchers, thus fulfilling the criteria outlined by the department of Health and Human Services Office for Human Research Protections 45 CRF 46.104(d)4 as exempt research.¹⁰ This was confirmed in writing by Advarra Institutional Review Board. The study data set consisted of anonymized laboratory and clinical data from a convenience sample of Boston Heart Diagnostics’ patients who agreed that their anonymized data could be used in research, and who completed the Diet ID tool and relevant blood work for purposes other than this research study. Laboratory assays of blood specimens were performed by Boston Heart Diagnostics (Framingham, MA) according to standardized validated methods reported elsewhere.^11-13 Clinical and medication data were provided by ordering clinicians and patients at the time of sample submission between December 2020 and March 2022. Concurrently, nutrient intake data were obtained via the validated Diet Quality Photo Navigation (DQPN®) technique and Diet ID™ tool as reported in detail elsewhere.^7,8 Briefly, DQPN is a patented, image-based dietary assessment and powers Diet ID.⁷ A series of food images, each representing a unique dietary pattern type and quality level are viewed with a final selection.

Analysis

Statistical analyses were performed in Microsoft Excel (version 365 MSO 16.0.14326.20850). Pearson correlation coefficients (for continuous variables) and Spearman coefficients (for ordinal variables) were used to evaluate associations between nutrient intake data and laboratory data for the full study sample, and after excluding data from individuals taking relevant medications and/or supplements known to affect blood test results. Two-sided P-values <.05 were considered statistically significant. As this was a brief, exploratory analysis rather than an intervention trial with a single primary outcome, adjustment was not made for multiple comparisons.

Multivariate linear regression was used to assess associations between blood test results (dependent variable) and dietary intake measures after adjusting for age, sex, body mass index, and use of relevant medications and/or supplements. Statins, ezetimibe, niacin, red-yeast rice, and fish-oil were considered relevant to lipid levels; statins and fish oil were considered relevant to high-sensitivity c-reactive protein; diabetes medications were considered relevant to metabolic markers (glycosylated hemoglobin, glucose, insulin, and adiponectin levels). Data regarding vitamin and mineral supplements were not available.

Results

Laboratory assays and dietary intake assessment were available for a total sample of 546 adults (see Table 1). Diet quality was captured using the Healthy Eating Index 2015¹⁴ both as a continuous variable, and stratified into deciles (Diet ID quality scores). Both continuous and ordinal measures of diet quality correlated significantly with HDL-C and triglycerides (n = 485; P < .001); with hs-CRP (n = 441; P < .001); with HgbA1c (n = 345; P < .01); with fasting insulin (n = 372; P < .001); and with HOMA-IR (n = 319; P < .001). Adjustment for relevant medication use-lipid-lowering agents and pharmacotherapy for type 2 diabetes-generally strengthened these associations (data on file). The primary hypothesis that rapid assessment of diet quality would vary significantly with key measures of lipid metabolism, glucose and insulin metabolism, and inflammation was thus affirmed.

Table 1.

Demographic, Dietary, and Biomarker Characteristics.

Characteristics	Median	IQR	%	N
Age (years)	57	18.25		546
Sex (% female)			69.2	546
Weight (kg)	77.11	25.4		484
BMI (kg/m²)	27.1	7.72		483
Dietary intake
Energy intake (cal/day)	2082.62	653.01		546
Diet ID HEI score (1 to 100)	88	36		546
Diet ID quality (1 to 10)	8	4		546
Dietary fiber (g/d)	34.65	28.96		546
Dietary Omega-3 (g/d)	2.8	3.31		546
Dietary saturated fat (g/d)	17.28	13.6		546
Dietary sodium (mg/d)	2174.14	1594.33		546
Dietary added sugar (g/d)	9.02	35.86		546
Biomarkers
HDL cholesterol (mg/dL)	57	24		486
Triglycerides (mg/dL)	96	62.5		485
hsCRP (mg/L)	1.2	2.1		441
HbA1c (%)	5.4	0.5		345
Insulin (uU/mL)	9	8		372
HOMA-IR	2.2	2.4		319
Adiponectin (ug/mL)	12.3	9.6		209

The estimated intake of dietary fiber correlated inversely with triglycerides (n = 485; P < .001); with hs-CRP (n = 441; P < .001); with insulin (n = 372; P < .01); and with HOMA-IR (n = 319; P < .001). Estimated dietary intake of omega-3 fatty acids correlated inversely with triglycerides (n = 485; P < .01) and hs-CRP (n = 441; P < .001). Estimated intake of total added sugars correlated significantly and positively with fasting insulin (n = 372; P < .01) and with HOMA-IR (n = 319; P < .001). Estimated intake measures for saturated fat, sodium, and added sugars were all significantly, inversely correlated with adiponectin (n = 209; P < .001) (see Table 2). Significant covariance of diet quality measures and biomarkers was preserved in multivariate regression analysis with age, sex, BMI, and medication use entered as control variables (See Table 3).

Table 2.

Correlation of Dietary Intake Measures with Biomarkers of Cardiometabolic Risk.

Pearson R	HDL-C^a	Triglycerides	Hs-CRP^b	HbA1c^c	Glucose	Insulin	HOMA-IR^d	Adiponectin
Diet ID HEI^e	.25***	−.21***	−.16***	−.17**	−.1*	.23***	−.27***	.27***
Diet ID quality	.25***	−.21***	−.18***	−.18**	−.11*	−.23***	−.28***	.25***
Fiber	.05	−.15***	−.2***	−.13*	−.08	−.17**	−.21***	.14*
Omega-3	.04	−.13**	−.18***	−.08	−.08	−.14**	−.18**	0
Saturated fat	−.3***	.25***	.07	.18***	.14**	.27***	.28***	−.28***
Sodium	−.31***	.2***	.13**	.2***	.12*	.24***	.31***	−.33***
Added sugars	−.24***	.15**	.14**	.2***	.1*	.17**	.25***	−.24***
N	486	485	441	345	405	372	319	209

Pearson P-value <.05 = *; <.01 = **; <.001 = ***.

^aHDL-C = high-density lipoprotein cholesterol.

^bhs-CRP = high-sensitivity C-reactive protein.

^cHbA1c = hemoglobin A1c.

^dHOMA-IR = homeostatic model assessment for insulin resistance.

^eHEI = healthy eating index.

Table 3.

Multivariate Regression Analysis of Dietary Intake Measures with Biomarkers.

	HDL-C^a		Triglycerides		Hs-CRP^b		HbA1c^c		Glucose		Insulin		HOMA-IR^d		Adiponectin
	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value	Coefficient	P-value
Diet ID HEI^e	.084	.049	−.665	.01	−.003	.611	−.004	.005	−.059	.171	−.116	.022	−.024	.001	.051	.081
Diet ID quality	.797	.037	−5.822	.011	−.029	.638	−.034	.007	−.499	.193	−.967	.033	−.212	.001	.347	.177
Fiber	.050	.32	−.672	.025	−.012	.121	−.004	.012	−.068	.173	−.137	.022	−.025	.004	.025	.472
Omega-3	.469	.202	−4.299	.051	−.101	.083	−.015	.211	−.405	.275	−.818	.058	−.111	.079	−.175	.474
Saturated fat	−.137	.108	1.759	.001	−.014	.313	.004	.098	.090	.271	.263	.006	.042	.003	−.041	.428
Sodium	−.001	.237	.008	.138	.000	.845	.000	.045	.000	.622	.002	.127	.001	.001	−.001	.073
Added sugars	−.041	.116	.089	.566	.002	.599	.003	.0003	.034	.185	.030	.349	.014	.0022	−.024	.178
N	369		368		342		266		312		290		253		164

^aHDL-C = high-density lipoprotein cholesterol

^bhs-CRP = high-sensitivity C-reactive protein

^cHbA1c = hemoglobin A1c

^dHOMA-IR = homeostatic model assessment for insulin resistance

^eHEI = healthy eating index.

Discussion

Summary

This study demonstrates that a rapid, digital dietary assessment tool based on pattern recognition correlates significantly with biomarkers of cardiovascular risk, insulin and glucose metabolism, and inflammation, as well as adiponectin. These results compare favorably to the strength and consistency of correlations of diet quality measured by traditional recall methods with cardiometabolic biomarkers.^15,16 This study serves as a proof of principle: a rapid, scalable dietary assessment method with data generation in as little as one minute reliably predicts variation in key biomarkers of cardiometabolic health.

Significance

Associations remained significant in multivariate analysis and were generally strengthened with adjustment for intake of pertinent classes of medication. For most of these associations, the overall diet quality was the principal correlate. These findings all affirmed the primary study hypothesis that diet quality, derived from dietary composition as assessed using diet quality photo-navigation (8; and see https://www.dietid.com/the-science) would track significantly with key biomarkers of cardiometabolic health.

The correlations observed, while highly significant, were generally of low to moderate strength, with Pearson correlation coefficients ranging from roughly .2 to .5. This is to be expected, since even perfect measurement of dietary intake on a metabolic ward correlates far from perfectly with blood markers.¹⁷ Factors other than intake affect most biomarkers influenced by diet, including but not limited to genetic variation, enzymatic activity, hormonal factors, microbiome composition, physical activity, body composition, medication use, and supplement use. This analysis in a convenience sample was insensitive to most of these factors, all of which potentially bias correlates with dietary intake toward the null. The preservation of significant associations in the hypothesized pattern despite these influences is noteworthy.

Limitations

The study also has important limitations. The sample was one of convenience, and all participants were patients of a single national laboratory. Several important potential cofactors were not assessed.

So What? (Implications for Health Promotion Practitioners and Researchers)

What is already known on this topic?

Impairment of overall diet quality is the single leading predictor of premature mortality and chronic morbidity in the United States.

What does this article add?

This study serves as a proof of principle: a rapid, scalable dietary assessment method with data generation in as little as one minute reliably predicts variation in key biomarkers of cardiometabolic health.

What are the implications for health promotion practice or research?

Diet quality and composition can be assessed using diet quality photo-navigation to track significantly with key biomarkers of cardiometabolic health.

Footnotes

Authors’ Contributions

Study conception and design - MLD, DLK, LQR; Data analysis and interpretation of results - GLB, JEJ, MLD; Draft and revise manuscript - DLK, MLD, LQR. All authors reviewed and approved the final version of the manuscript.

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr. Dansinger has served as a compensated advisor to Diet ID, Inc, but holds no interest in the company; he is also an employee of Boston Heart Diagnostics. Gary Breton and Justin Joly are employees of Boston Heart Diagnostics. Lauren Rhee is an employee of Diet ID, Inc, and holds equity in the company. Dr Katz is the founder and principal owner of Diet ID, Inc, a for-profit company.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Author’s Note

Protocol was reviewed by the Advarra Institutional Review Board (Columbia, MD) and recognized as Exempt from IRB Oversight under 45 CFR 46.104(d)4.

ORCID iDs

Lauren Q. Rhee

David L. Katz

References

GBD 2017 Diet Collaborators . Health effects of dietary risks in 195 countries: A systematic analysis for the Global Burden of Disease Study 2017. Lancet. 1990;393:1958-1972.

Mozaffarian

Glickman

. Our Food Is Killing Too Many of Us, 2019: New York Times. https://www.nytimes.com/2019/08/26/opinion/food-nutrition-health-care.html

Merino

Joshi

Nguyen

, et al. Diet quality and risk and severity of COVID-19: a prospective cohort study. Gut. 2021;70(11):2096-2104. doi:10.1136/gutjnl-2021-325353

O’Hearn

Liu

Cudhea

, et al. Coronavirus Disease 2019 hospitalizations attributable to cardiometabolic conditions in the United States: A comparative risk assessment analysis. J Am Heart Assoc 2021;10(5):e019259. doi:10.1161/JAHA.120.019259.

Katz

Govani

Anderson

, et al. The financial case for food as medicine: Introduction of a ROI calculator. Am J Health Promot. 2022;36(5):768-771. doi:10.1177/08901171211070751

Vadiveloo

Lichtenstein

Anderson

, et al. American Heart Association Council on Lifestyle and Cardiometabolic Health; Council on Arteriosclerosis, Thrombosis and Vascular Biology; Council on Cardiovascular and Stroke Nursing; Council on Clinical Cardiology; and Stroke Council. rapid diet assessment screening tools for cardiovascular disease risk reduction across healthcare settings: A scientific statement from the American Heart Association. Circ Cardiovasc Qual Outcomes 2020;13(9):e000094.

Katz

. Diet Mapping Processes and Systems to Optimize Diet Quality And/or Minimize Environmental Impact. (U.S. Patent Number: 11328810). U.S. Patent and Trademark Office; 2018.

Katz

Rhee

Katz

, et al. Dietary assessment can be based on pattern recognition rather than recall. Med Hypotheses. 2020;26:109644. Epub ahead of print. PMID: 32131036. doi:10.1016/j.mehy.2020.109644

Turner-McGrievy

Hutto

Bernhart

Wilson

. Comparison of the Diet ID platform to the Automated Self-Administered 24-hour (ASA24) dietary assessment tool for assessment of dietary intake. J Am Nutr Assoc 2022;41(4):360-382.

10.

Office for Human Research Protections (OHRP) . Exemptions (2018 Requirements), 2021. https://www.hhs.gov/ohrp/regulations-and-policy/regulations/45-cfr-46/common-rule-subpart-a-46104/index.html

11.

Diffenderfer

Rajapakse

Pham

, et al. Plasma fatty acid profiles: Relationships with sex, age, and state-reported heart disease mortality rates in the United States. J Clin Lipidol. 2022;16(2):184-197.

12.

Dansinger

Williams

Superko

, et al. Effects of weight change on apolipoprotein B-containing emerging atherosclerotic cardiovascular disease (ASCVD) risk factors. Lipids Health Dis. 2019;18(1):154.

13.

Gill

Superko

McCarthy

, et al. Cardiovascular risk factor reduction in first responders resulting from an individualized lifestyle and blood test program: A randomized controlled trial. J Occup Environ Med. 2019;61(3):183-189.

14.

Krebs-Smith

Pannucci

Subar

, et al. Update of the healthy eating index: HEI-2015. J Acad Nutr Diet. 2018;118. doi:10.1016/j.jand.2018.05.021

15.

Park

Shin

, et al. Diet quality and diet patterns in relation to circulating cardiometabolic biomarkers. Clin Nutr. 2016;35(2):484-490.

16.

Phillips

Harrington

Perry

. Relationship between dietary quality, determined by DASH score, and cardiometabolic health biomarkers: A cross-sectional analysis in adults. Clin Nutr. 2019;38(4):1620-1628.

17.

Jenab

Slimani

Bictash

, et al. Biomarkers in nutritional epidemiology: applications, needs and new horizons. Hum Genet. 2009;125(5-6):507-525.