A reflection-spectroscopy measured skin carotenoid score strongly correlates with plasma concentrations of all major dietary carotenoid species except for lycopene

Abstract

Skin carotenoids can be measured non-invasively using spectroscopy methods to provide a biomarker of total dietary carotenoid and carotenoid-rich fruit and vegetable intake. However, the degree to which skin carotenoid biomarkers reflect intakes of specific carotenoids must be determined for specific devices. Previously, findings were mixed regarding the correlation between reflection spectroscopy (RS)-assessed skin carotenoids and individual plasma carotenoid concentrations. The current study expands on prior analyses to examine the cross-sectional associations between adult RS-assessed skin carotenoids and individual carotenoid species intakes and plasma concentrations, controlling for potential covariates. We hypothesized that RS-assessed skin carotenoid scores would strongly correlate with all major plasma carotenoid species other than lycopene. Cross-sectional data from two prior studies (n = 213 and n = 162) examining the validity and sensitivity of RS-assessed skin carotenoids as a biomarker of fruit and vegetable intake were used. Skin carotenoids were assessed using the Veggie Meter ^®, which quantifies combined skin carotenoid concentrations. Plasma concentrations of α-carotene, β-carotene, β-cryptoxanthin, lycopene, and lutein and zeaxanthin were determined using high-performance liquid chromatography. Self-reported carotenoid intake was estimated from validated food frequency questionnaires. Skin carotenoid scores correlated moderately to strongly with individual plasma carotenoid species (Pearson’s r=0.52 – r=0.78) except for lycopene (r=0.04 to r=0.07). Low correlations between skin carotenoid score and lycopene plasma concentrations and intake could be due to differential deposition, preferential oxidation/degradation, and/or device measurement bias. Validating skin carotenoid measurement techniques relative to other concentration biomarkers informs the interpretation of skin carotenoid biomarkers.

Graphical Abstract

Using data from two prior studies, we examined correlations between reflection spectroscopy - assessed skin carotenoid scores and individual plasma carotenoids. Skin carotenoid scores were strongly correlated with all individual plasma carotenoids (r=0.52 – r=0.78) except for lycopene, where the correlations ranged from r=0.04 to r=0.07. Lycopene-rich food intake may not be detected well if using the reflection spectroscopy-device to assess dietary interventions.

1. Introduction

Consuming recommended amounts of fruits and vegetables (FV) is critical for optimal health.[1-3] Eating adequate portions of FV daily decreases the risks associated with developing preventable chronic diseases such as type 2 diabetes, cardiovascular disease, obesity, and certain cancers.[1-3] However, adults in the United States (US) do not consume the daily recommended FV intake,[4] with even less consumption of recommended FV among individuals from minority and marginalized communities.[5]

In light of these health concerns and disparities in minority and marginalized communities, it is imperative that appropriate and effective interventions are implemented to increase FV intake. To develop effective interventions, valid and reliable assessment measures of FV intake are needed.[6,7] Currently, the standard criterion to objectively assess FV intake is blood plasma carotenoids.[9] Carotenoids are plant pigments and antioxidants that play a vital role in protecting the human body from various diseases, and the major dietary sources of carotenoids are from various FV.[9,10] The six common carotenoids found in the US diet are lycopene, β-carotene, β-cryptoxanthin, α-carotene, lutein, and zeaxanthin.[9] These six carotenoids are also found in the skin, with the most abundant skin carotenoids in US adults being lycopene and β-carotene.[11] Although assessing FV intake by plasma carotenoid measurements is considered the criterion standard biomarker for FV intake,[7,8] measures of total or individual plasma carotenoids requires blood collection. Phlebotomy carries logistical and participation barriers in public health nutrition settings.

An alternative to plasma carotenoid measurement is pressure-mediated reflection spectroscopy (RS)-measured skin carotenoid score (SCS), generated by a portable device called the Veggie Meter^®,[12-14]. The Veggie Meter ^® is a non-invasive device, which quantifies multiple carotenoids in the skin, and does not differentiate the concentrations of specific carotenoid species. This technique is gaining interest for community-based nutrition intervention studies.[13,15,16]

With regards to interpreting a skin carotenoid biomarker of carotenoid and food intake, it is important to understand which carotenoids correlate with and contribute to the skin carotenoid measurement. Different carotenoids occur in different fruits and vegetables with some having predominantly β-carotene (e.g. carrots and sweet potatoes), while others may have predominantly lycopene (e.g. watermelon and tomatoes). Further, there are cultural and geographic differences in FV intake patterns, contributing to differences in carotenoid intake patterns.[17] Therefore, to effectively use the Veggie Meter ^® (VM) to evaluate FV interventions in multiple geographic regions and cultures, the degree to which different carotenoid species influence SCS must be determined. Whether the RS-assessed SCS are associated with each of the individual carotenoid species is unclear. For example, a cross sectional study of US adults found a correlation between non-log transformed plasma lycopene and RS-assessed SCS of r = 0.07, whereas the correlation was r>0.59 for all other carotenoid species.[13] In US toddlers, intake of β-carotene and lutein & zeaxanthin (combined) were predictors of SCS, while intakes of alpha-carotene, β-cryptoxanthin, and lycopene were not.[18] Casperson, et al. reported a correlation of 0.42, p < 0.001 between plasma lycopene and RS-assessed skin carotenoids,[19] similar to Faraji, et al. who reported a correlation of 0.41, p < 0.05 between plasma lycopene and RS-assessed skin carotenoids.[20] Further, in a randomized controlled trial of a lycopene-rich juice, participants consumed 28 mg/day of lycopene and 31 mg/day of total carotenoids (which included β-carotene), producing a 114 unit increase in SCS over six weeks;[19] however, in a similar randomized controlled trial of a β-carotene rich juice, participants consumed a much smaller amount of total carotenoids (8 mg/day), the majority of which was β-carotene, producing a 123 unit increase in RS-assessed SCS over six weeks.[21] Moreover, studies examining correlations between skin and plasma carotenoids using Resonance Raman Spectroscopy (RRS), a similar measurement method, did not find differences between the strengths of correlation between plasma concentrations and skin carotenoid measures for lycopene versus other carotenoids.[11] RRS has been validated by measurements of carotenoids and lycopene from skin biopsies and matching plasma carotenoids.[11] Thus, studies are mixed regarding the associations between skin carotenoids and individual plasma carotenoids species. These mixed results could be due to collinear intakes of multiple carotenoids, differential rates of skin carotenoid deposition from plasma, preferential lycopene oxidation/degradation in the skin, or device bias to detect β-carotene and other carotenoids with a similar absorbance profile.[22,23] Furthermore, in the prior studies [19,20], the participants were not diverse in terms of race/ethnicity, age, and health status: more studies are needed to examine correlations between SCS and individual plasma carotenoids by race and ethnicity, age, and weight status in order to evaluate interventions that occur in diverse populations. Therefore, the purpose of this study is to examine the cross-sectional associations between RS-assessed SCS and individual carotenoid species (measured in the plasma and via self-report) among adults from racially and ethnically diverse backgrounds. We hypothesized that RS-assessed SCS would be strongly correlated with all individual plasma carotenoid concentrations other than lycopene. We conducted correlation analyses to examine the cross-sectional associations between RS-assessed SCS and individual carotenoid species and controlled for covariates including age, sex, race and BMI.

2. Materials and Methods

2.1. Study Design, Recruitment, and Enrollment

In this study, we utilized samples from two research studies: a cross-sectional validation study[13] and a randomized controlled trial of a β-carotene-rich juice,[21] both described in detail elsewhere. The cross-sectional observational validation study (hereafter deemed “Study 1”) recruited participants from two diverse locations within the United States: Greenville, North Carolina (NC), and the St. Paul/Minneapolis, Minnesota (MN) metropolitan areas, with a total of 213 participants completing the study.[13] The randomized controlled feeding trial (n = 162, hereafter deemed “Study 2”) had recruitment sites in Greenville, NC, St. Paul /Minneapolis, MN, and Baylor College of Medicine in the Houston, Texas area.[21] In the current paper, only the baseline data from Study 2 were included in the analysis. Both Study 1 and Study 2 were approved by the East Carolina University Medical Center Institutional Review Board, #UMCIRB 17-001242. Study 2 was registered at clinicaltrials.gov as ID: NCT04056624.

2.2. Skin carotenoids

Both studies utilized the Veggie Meter^®, a pressure-mediated reflection spectroscopy device (Longevity Link, LLC), to measure skin carotenoids.[13,21] SCS is assessed on a scale ranging from 0 to 800. Participants were asked to wash and dry their hands first and then proceeded to place their right index finger into the Veggie Meter^® for readings. Participants inserted and removed their fingers three times for a total of 3 consecutive scans. The mean SCS of the three Veggie Meter^® scans was computed and used in the analyses.[13,21]

2.3. Plasma carotenoids

In both studies, participants' blood was drawn after a minimum of 7 hours fasting, and their plasma carotenoids, glucose, and cholesterol concentrations were evaluated. Plasma carotenoid concentrations were measured using high performance liquid chromatography (HPLC)-photodiode array detection as previously described.[13,21] The following carotenoids were quantified: α-carotene, β-carotene, β-cryptoxanthin, lycopene, lutein and zeaxanthin. Values for these carotenoids were used in analyses involving the individual plasma carotenoid species.

2.4. Dietary assessment measures

In Study 1, the 152-item Harvard semi-quantitative, validated food frequency questionnaire (FFQ) was used.[24] The instrument is designed to assess FV intake and carotenoid intake from foods and dietary supplements.[25] A “times/day” variable was created for total fruit, total vegetables, and total FV (including white potatoes), calculated as the sum of the times per day estimates for the FV items. Total dietary carotenoid intake was the sum of participants’ daily intakes for the following: α-carotene, β-carotene, β-cryptoxanthin, lycopene, lutein and zeaxanthin. In Study 2, the validated National Cancer Institute (NCI) Diet History Questionnaire (DHQ)-III was used.[26] There are 135 food/beverage items and 26 dietary supplement questions. Participants reported on intakes from the past month, with portion size, version of the questionnaire online. FV intake was estimated by summing the total fruit and total vegetable variables created by the DHQ. Total carotenoid intake was calculated by summing the amounts of individual carotenoids available in the DHQ-III output.[26]

2.5. Covariates

In both Study 1 and Study 2, self-administered questionnaires and surveys were employed to gather participants' demographic information such as marital status, educational attainment, annual household income, ultraviolet (UV) exposure, and vitamin/supplement usage.[13,21] The covariates that were significant covariates in field-based predictive models in the prior studies were used in the current study: Age, sex, race and body mass index (BMI).

2.6. Statistical analyses

All analyses were completed separately by study. There was some overlap in the participants, as 48 (29.6%) people who participated in Study 1 also later participated in Study 2. These participants were included in the analyses for both studies. For all plasma carotenoid (species) concentration and carotenoid species intakes, a log2-transformation was applied to improve linearity and stabilize the variance in correlation and regression analyses. Percent of each carotenoid species in plasma was calculated by dividing the amount of each carotenoid species by the total plasma carotenoid concentration (including α-carotene, β-carotene, β-cryptoxanthin, lycopene, and lutein and zeaxanthin). In both samples, we first examined Pearson’s correlations among individual carotenoid species in plasma, among individual carotenoid species intake as measured by the FFQ/DHQ, and across plasma carotenoid species and carotenoid species intake. We then examined Pearson’s correlations between plasma carotenoid species concentrations and SCS, between percent plasma carotenoid species and SCS, as well as between carotenoid (species) intake and SCS. As an effort to minimize confounding effects from covariates, partial correlations were obtained while adjusting for age, sex, race, and BMI. The covariates were selected based on the results in previous analyses of the cross-sectional and the randomized controlled studies.[13,21] Lastly, Pearson’s correlations among subgroups of participants by age (≤30 vs. >30), sex (Male vs. Female), race (Non-Hispanic Black vs. Asian vs. Non-Hispanic White vs. Hispanic), BMI (<25 vs. ≥25) were compared for possible modifying effects of the covariates.

Ermakov and Gellermann (2012) described how the pressure-mediated reflection spectroscopy method utilized in the Veggie Meter ^® was developed, with a focus on the spectra of β-carotene.[22] As β-carotene is theoretically the major carotenoid detected by this method, and is also a major dietary and skin carotenoid, we tested whether other carotenoids in addition to β-carotene are determinants of SCS. Therefore, we also examined partial correlations between SCS and carotenoid species when adjusted for β-carotene, age, sex, race and BMI. This was to determine the additional contribution that carotenoid species would make to the SCS if conditioned on the β-carotene measurement. The data was analyzed using SAS Version 9.4 (SAS Institute Inc., Cary, NC). A familywise significance level of 0.05 was used throughout the paper. Since multiple plasma and intake carotenoids species were tested in both studies, a Bonferroni correction factor was applied to account for multiple comparisons.

3. Results

3.1. Participant Characteristics

Characteristics of the samples from Study 1 and Study 2 are summarized in Table 1 and were previously published.(13,21) In brief, of the 213 eligible and enrolled Study 1 participants, over half (n = 151, 70.9%) of the 213 participants were female. Sixty-one participants (28.6%) were African American, 53 (24.9%) were Asian, 70 (32.9%) were White, and 29 (13.6%) were Hispanic/Latinx. Additionally, mean age was 33.9 years, mean BMI was 25.6 kg/m², mean body fat was 27.1 percent, and mean SCS was 318.9. Among those enrolled in Study 2, approximately half (n = 83, 51.2%) in the randomized study were female. Forty-one (25.3%) self-categorized as African American, 40 (24.7%) Asian, 44 (27.2%) white, and 37 (22.8%) Hispanic. Mean age was 32.3 years, with mean BMI and mean body fat percentage of 25.5 kg/m² and 24.6 percent, respectively. Mean SCS was 298.

Table 1.

Baseline Characteristics of racially and ethnically diverse adults included in the two study samples from Study 1 (n = 213) and Study 2 (n =162).

		Study 1 (n=213)		Study 2 (n=162)
Variable	Category	n	%	n	%
Site	East Carolina University	121	56.8	61	37.6
	University of Minnesota	92	43.2	41	25.3
	Baylor College of Medicine	NA	NA	60	37.0
Sex	Male	62	29.1	79	48.8
	Female	151	70.9	83	51.2
Race/ethnicity	Non-Hispanic African American/Black	61	28.6	41	25.3
	Asian	53	24.9	40	24.7
	Non-Hispanic White	70	32.9	44	27.2
	Hispanic/Latinx	29	13.6	37	22.8
Marital status	Single/ Separated/ Divorced/ Widowed	121	56.8	101	62.3
	Married/ Living with a partner	92	43.2	61	37.6
Education	Some college or less	60	28.2	41	25.3
	College graduate	153	71.8	121	74.7
Income	$39,999 or less	65	30.5	43	26.5
	$40,000-$79,999	53	24.9	58	35.8
	$80,000 or more	73	34.3	48	29.6
	Missing	22	10.3	13	8.02
		Mean	SD	Mean	SD
Age, years		33.9	12.5	32.3	11.1
BMI, kg/m2		25.6	3.98	25.4	3.81
Body fat, %		27.1	7.78	24.6	8.00
Skin carotenoid status		319	114	298	97.1
Sun exposure, hrs/day		1.30	1.05	2.31	1.03

Dietary and carotenoid intake of the two participant samples is displayed in Table 2. Study 1 participants self-reported a mean (SD) total FV of 5.1 (3.4) cups/d, total self-reported carotenoid intake of 14,805 (10,261) μg/d, and plasma total carotenoid concentration of 140 (58.9) mcg/dL. Study 2 participants reported a mean (SD) total FV of 3.4 (2.9) cups/d, total carotenoid intake of 15,887 (15,493) μg/d and had plasma total carotenoid concentrations of 161 (70) mcg/dL.

Table 2.

Dietary and carotenoid intake descriptive statistics of the two study samples of racially and ethnically diverse adults. *Study 1 did not obtain self-reported dietary data for two participants. ** Percent plasma carotenoids concentrations were calculated as each species concentrations divided by the total plasma carotenoids concentration x 100%.

	Study 1 (n=211*)		Study 2 (n=162)
	Mean	SD	Mean	SD
Calorie intake, kilocalories/day	1,819	866	1,801	1,056
Protein intake, g/d	73.9	33.1	75.0	51.1
Total fat intake, g/d	70.2	37.4	70.5	41.5
Carbohydrate intake, g/d	220	115	211	131
Total fruits intake, cups/d	1.62	1.48	1.46	2.00
Total vegetables intake, cups/d	3.44	2.60	1.91	1.50
Total fruits and vegetables intake, cups/d	5.05	3.39	3.37	2.89
Total carotenoids intake, μg/d	14,805	10,261	15,887	15,493
Total carotenoids intake w/o lycopene, μg/d	10,665	9,195	10,936	12,756
α-carotene intake, μg/d	793	1,158	585	810
β-carotene intake, μg/d	5,565	5,043	4,963	5,567
β-cryptoxanthin intake, μg/d	146	180	120	158
Lutein and zeaxanthin intake, μg/d	4,161	3,541	5,268	7,074
Lycopene intake, μg/d	4,140	2,940	4,951	4,621
Plasma Concentrations
Total plasma carotenoids, μg/dL	140	58.9	161	70
Total plasma carotenoids w/o lycopene, μg/dL	94.7	53.7	107	63.9
α-carotene in plasma, μg/dL	11.5	9.0	11.3	14.8
β- carotene in plasma, μg/dL	35.7	29.4	36.9	33.5
α- and β-cryptoxanthin in plasma, μg/dL	19.2	13.6	26.3	16.7
Lutein and zeaxanthin in plasma, μg/dL	28.4	11.6	32.1	13.9
Total Lycopene in plasma, μg/dL	45.3	16.7	54.8	21.9
Total cholesterol in plasma, mg/dL	146	41	152	40.3
Percent Plasma Carotenoids Concentrations**
Percent lycopene in plasma	34.7%	11.3%	36.6%	13.0%
Percent β-carotene in plasma	23.3%	8.71%	20.4%	9.40%
Percent lutein and zeaxanthin in plasma	20.9%	5.76%	20.8%	6.38%
Percent α- and β-cryptoxanthin in plasma	13.3%	5.38%	16.1%	4.99%
Percent α-carotene in plasma	7.86%	3.17%	6.12%	4.09%

Table 2 also shows the percentages of individual carotenoid species in plasma. Lycopene was the most prevalent carotenoid in the plasma (35% in Study 1, 37% in Study 2), followed by β-carotene (23%, 20%, Study 1 and 2, respectively) and lutein/zeaxanthin (21% in both studies).

3.2. Correlations between SCS and individual carotenoid species

Bivariate Pearson’s correlations tested how carotenoid species intake and plasma concentrations were associated with SCS (Table 3). In both Study 1 and 2, consistently strong Pearson’s correlations were observed between SCS and individual plasma carotenoid species (r = 0.52 – 0.72, p < 0.001) other than lycopene (r = 0.04 – 0.07, p > 0.05). Pearson’s correlations between SCS and self-reported intake of carotenoid species other than lycopene were low to moderate (Study 1: r = 0.28 – 0.39, p < 0.001; Study 2: r=0.20 - 0.21, p < 0.05). SCS had low, non-significant Pearson’s correlations with lycopene intake (r = 0.05 – 0.08, p > 0.05) in both samples. With a Bonferroni correction for 38 correlation tests, any p-value < 0.001 is considered statistically significant for a familywise 0.05 significance level.

Table 3.

Pearson’s correlations and partial correlations, adjusted for age, sex, race, and BMI (r-1) and adjusted for covariates and β-carotene (r-2), of self-reported carotenoids and plasma carotenoids with skin carotenoid status among the two study samples of racially and ethnically diverse adults.

	Correlations with Skin Carotenoid Score	Pearson r	Partial r-1	Partial r-2
Study 1	Plasma total carotenoids	0.71***	0.68***	0.18*
	Plasma total carotenoids w/o lycopene	0.78***	0.77***	0.41***
	α-carotene in plasma	0.66***	0.66***	0.19**
	β-carotene in plasma	0.72***	0.71***
	α- and β-cryptoxanthin in plasma	0.63***	0.58***	0.29***
	Lutein and zeaxanthin in plasma	0.64***	0.59***	0.38***
	Lycopene in plasma	0.04	0.06	−0.16*
	Percent α-carotene in plasma	0.30***	0.29***
	Percent β-carotene in plasma	0.54***	0.52***
	Percent α- and β-cryptoxanthin in plasma	0.30***	0.24***
	Percent lutein and zeaxanthin in plasma	−0.02	−0.02
	Percent lycopene in plasma	−0.63***	−0.61***
	Total carotenoids intake	0.38***	0.37***	0.09
	Total carotenoids intake w/o lycopene	0.41***	0.39***	0.14
	α-carotene intake	0.37***	0.33***	0.04
	β-carotene intake	0.39***	0.37***
	β-cryptoxanthin intake	0.28***	0.28***	0.11
	Lutein and zeaxanthin intake	0.38***	0.37***	0.10
	Lycopene intake	0.05	0.09	−0.03
Study 2	Plasma total carotenoids	0.63***	0.59***	0.08
	Plasma total carotenoids w/o lycopene	0.71***	0.69***	0.24**
	α-carotene in plasma	0.63***	0.62***	0.21**
	β-carotene in plasma	0.68***	0.67***
	α- and β-cryptoxanthin in plasma	0.60***	0.54***	0.15
	Lutein and zeaxanthin in plasma	0.52***	0.46***	0.13
	Lycopene in plasma	0.07	0.07	−0.13
	Percent α-carotene in plasma	0.51***	0.51***
	Percent β-carotene in plasma	0.54***	0.55***
	Percent α- and β-cryptoxanthin in plasma	0.17*	0.08
	Percent lutein and zeaxanthin in plasma	−0.10	−0.12
	Percent lycopene in plasma	−0.57***	−0.55***
	Total carotenoids intake	0.19*	0.21**	−0.01
	Total carotenoids intake w/o lycopene	0.21**	0.22**	−0.00
	α-carotene intake	0.20*	0.21**	0.06
	β-carotene intake	0.21**	0.23**
	β-cryptoxanthin intake	0.20*	0.20*	0.07
	Lutein and zeaxanthin intake	0.20*	0.20*	−0.01
	Lycopene intake	0.08	0.10	0.00

^*:p < 0.05

^**:p < 0.01

^***:p < 0.001

Intake and plasma measures were log2-transformed.

Partial r-1: given age, sex, race, and BMI.

Partial r-2: given age, sex, race, BMI, and β-carotene in plasma or intake.

Table 3 also shows correlations between SCS and percent of each carotenoid species in plasma. There were statistically significant, positive correlations (r = 0.30 – 0.54, p < 0.001) between SCS and percent α- and β-carotene, and a weaker correlation between SCS and α- and β-cryptoxanthin (study 1: r = 0.30, p < 0.001; study 2: r = 0.17, p < 0.05). There was a weak negative, not statistically significant correlation (r = −0.10 - −0.02, p > 0.05) between SCS and percent lutein/zeaxanthin. There was a strong, negative, statistically significant correlation (r = −0.63 - −0.57, p < 0.001) between SCS and percent lycopene. This indicates that the SCS was lower among those with a greater percentage of lycopene in plasma.

Since ANOVA tests revealed some significant differences (p < 0.0003 with Bonferroni correction for 160 comparisons) of plasma carotenoid (species) concentration among subgroups of participants, particularly by race and BMI (< 25 vs. ≥ 25 kg/m²) (Supplemental Table 1), confounding effects of the covariates to the correlations were of concern. To minimize such confounding effects, Table 3 shows partial corrections (r-1) to the corresponding Pearson’s correlations while adjusting for age, sex, race, and BMI. The sizes and statistical significance of the partial correlations remained essentially the same as those of Pearson’s correlations. This indicates minimal confounding effects of the covariates.

To examine possible modifying effects of the covariates on Pearson’s correlations, we compared Pearson’s correlations among subgroups of participants defined by age (≤ 30 vs. > 30 years), sex, race, and BMI (< 25 vs. ≥ 25 kg/m²) (Supplemental Table 2). In Study 2, Pearson’s correlations between skin carotenoid score and plasma α- and β-Cryptoxanthin were found to be significantly different between the two age groups (p < 0.0003 with a Bonferroni correction for 160 tests), which leaded to a significant difference also in Pearson’s correlations between skin carotenoid score and plasma total carotenoids between the two age groups. However, the same result was not observed in Study 1. Thus, it is not clear whether this finding is spurious.

To examine the contribution of individual plasma carotenoids to SCS, independently of β-carotene, Table 3 also shows partial correlations between SCS and individual plasma carotenoids adjusting for plasma β-carotene and the covariates (r-2). SCS had a partial correlation of r = 0.19-0.21 (p < 0.01) with α-carotene in plasma in both studies, notably lower than the corresponding Pearson’s correlations of r = 0.63-0.66. This indicates shared variance between plasma α- and β-carotene, but that plasma α-carotene independently contributes to the SCS. Similar patterns were observed for other plasma carotenoid species, except for lycopene. The partial correlation between SCS and plasma lycopene concentration was negative (Study 1: r = −0.16, p < 0.05; Study 2: r = −0.13, p > 0.05), indicating that plasma lycopene concentrations are not predictive of SCS.

For self-reported intake of carotenoid species other than lycopene, the partial correlations with SCS dropped to r = −0.01 – 0.11 (p > 0.05), indicating minimal contributions of these carotenoid species to the skin carotenoid score when adjusted for β-carotene intake and the covariates. For lycopene intake, however, the partial correlation with SCS dropped to r = −0.13 – 0.00 (p > 0.05), indicating a null contribution of lycopene intake to the SCS when adjusted for β-carotene intake and the covariates.

To determine if intake of and plasma concentrations of individual carotenoids were associated, and to determine if specific carotenoids are related to intake of and plasma concentrations of other carotenoid species, correlations for plasma and intake were examined. Table 4 displays Pearson’s correlations between individual carotenoid species in plasma and between individual carotenoid species as measured by the food frequency questionnaires from each study. We found that lycopene intake had a low correlation with plasma lycopene (Study 1: r = 0.17, p < 0.05;Study 2: r = 0.14, p > 0.05) when compared to the moderate to large correlations between intake of all other carotenoids and plasma measures of the same (r = 0.30 – 0.53, p < 0.001).

Table 4.

Pearson’s correlations between individual carotenoid species in plasma, between individual carotenoid species intake as measured by the food frequency questionnaires, and across individual plasma carotenoid species and individual carotenoid species intake in the two study samples of racially and ethnically diverse adults.

Variable 1	Variable 2	Study 1	Study 2
Across Intake and Plasma Concentration Correlations
Intake species	Plasma species
Lycopene	Lycopene	0.17*	0.14
β-carotene	β-carotene	0.44***	0.34***
α-carotene	α-carotene	0.53***	0.39***
β-cryptoxanthin	α- and β-cryptoxanthin	0.34***	0.31***
Lutein and zeaxanthin	Lutein and zeaxanthin	0.37***	0.30***
Between Plasma Concentration Correlations
Plasma species 1	Plasma species 2
Lycopene	α-carotene	0.09	0.14
Lycopene	β-carotene	0.16*	0.18*
Lycopene	α- and β-cryptoxanthin	0.13	0.29***
Lycopene	Lutein and zeaxanthin	0.22**	0.17*
β-carotene	α- and β-cryptoxanthin	0.63***	0.72***
β-carotene	Lutein and zeaxanthin	0.57***	0.63***
β-carotene	α-carotene	0.83***	0.80***
α-carotene	α- and β-cryptoxanthin	0.52***	0.61***
α-carotene	Lutein and zeaxanthin	0.50***	0.60***
α- and β-cryptoxanthin	Lutein and zeaxanthin	0.65***	0.69***
Between Intake Correlations
Intake species 1	Intake species 2
Lycopene	α-carotene	0.39***	0.46***
Lycopene	β-carotene	0.31***	0.44***
Lycopene	β-cryptoxanthin	0.27***	0.53***
Lycopene	Lutein and zeaxanthin	0.27***	0.39***
α-carotene	β-carotene	0.81***	0.75***
α-carotene	β-cryptoxanthin	0.45***	0.54***
α-carotene	Lutein and zeaxanthin	0.61***	0.53***
β-carotene	β-cryptoxanthin	0.53***	0.67***
β-carotene	Lutein and zeaxanthin	0.88***	0.90***
β-cryptoxanthin	Lutein and zeaxanthin	0.49***	0.65***

^*:p < 0.05

^**:p < 0.01

^***:p < 0.001

Plasma and intake measures were log2-transformed.

For both studies, there were consistently high, positive correlations among plasma β-carotene, α-carotene, β-cryptoxanthin, and lutein/zeaxanthin concentrations (r=0.50 - r=0.83, p<0.001). This pattern was also reflected in the self-reported β-carotene, α-carotene, β-cryptoxanthin, and lutein/zeaxanthin intakes (r=0.45 – r=0.90, p < 0.001). There was not a consistently, strongly positive correlation between lycopene intake or plasma lycopene with intakes and plasma concentrations of other carotenoid species, respectively. Correlations between lycopene and the other carotenoid species were higher in self-reported intake (r=0.27 – r=0.53, p < 0.001) than in plasma measurements (r=0.09 – r=0.29, p varies).

4. Discussion

The current study examined the cross-sectional associations between SCS using RS and individual carotenoid species (measured in plasma and via self-report) among adults from racially and ethnically diverse backgrounds. The study found that SCS was positively correlated with plasma concentrations and intakes of all major dietary carotenoids except lycopene, in agreement with our initial hypothesis. Our findings are in contrast with findings of Casperson et al,[19] and Faraji et al,[20] who found that Veggie Meter ^®-assessed skin carotenoids were positively and statistically significantly correlated with plasma lycopene. The low correlations between SCS and plasma lycopene concentration and intake observed in the current study could be due to (1) differential carotenoid deposition,(2) preferential lycopene oxidation/degradation in the skin, and/or (3) measurement bias of the Veggie Meter ^®. These three possibilities are discussed in detail below.

With regard to differential carotenoid deposition, studies have actually reported that lycopene is preferentially deposited in the skin, relative to other carotenoids in relation to plasma carotenoid concentrations.[11,27] In both cases, carotenoids were measured in excised skin and plasma using HPLC and lycopene was the most strongly correlated carotenoid. Thus, it is unlikely that decreased lycopene deposition is responsible for the poor correlation between SCS and plasma lycopene concentration.

Another possible explanation for poor correlation between SCS and lycopene intake and plasma lycopene concentration is preferential oxidation or degradation of lycopene in the skin. Research indicates that lycopene degrades more rapidly than other carotenoids when skin is exposed to UV radiation.[28] In the present study, participants were from three geographically different locations throughout the USA. In Study 1, there was a significant site effect (p < 0.001) so that participants from Minnesota had higher SCS (363 (99)) compared to participants from North Carolina (286 (114)). In Study 2, there was not a significant site effect (p = 0.337) with SCS means in Texas, North Carolina and Minnesota of 283 (97) vs. 305 (99) vs. 308 (94), three locations with different UV exposure from sunlight. These observations are mixed, and thus, UV exposure resulting in preferential degradation of lycopene cannot be ruled out as one of the potential explanations for lower correlation of SCS with plasma lycopene concentrations.

A third explanation for poor correlation between SCS and plasma lycopene concentrations is a bias in the detection of different carotenoids by the current device. Lycopene is a red carotenoid with a blue-shifted absorbance spectrum (Lambda max= 472 nm) relative to the other yellow-orange carotenoids found in the diet which have a red-shifted absorbance spectrum (Lambda max = 444-452).[29] The current study found that lycopene intake had a low correlation with plasma lycopene when compared to the correlations between intake of all other carotenoids and plasma measures of the same. The partial correlations (r-1) are noteworthy as they show the correlation between total skin carotenoids and individual plasma carotenoid species, adjusting for covariates. The partial correlations for all individual plasma and intake carotenoid species were positive and statistically significantly different from zero, except for lycopene. Nonsignificant correlations close to zero were observed between SCS and plasma and intake lycopene. This suggests that the tested device may not detect lycopene as well as other major dietary carotenoids in the skin. This result was echoed in partial correlations (r-2), adjusting for covariates and β-carotene, where SCS and plasma lycopene showed a negative correlation. In fact, percent lycopene in plasma showed a significantly negative correlation with SCS, while other percent carotenoid species in plasma showed significantly positive correlations with SCS.

Scarmo, et al.,[11] found a moderate correlation between both lycopene and β-carotene in the skin and plasma. They, however, were using RRS to measure skin carotenoids; RRS is a method which had been validated with excised skin samples measured using HPLC.[30] In contrast to the current study, Ermakov, et al. found the highest correlation between plasma lycopene and RRS-assessed SCS.[31] The current study results indicate that there was not a correlation between plasma lycopene and RS-assessed SCS. Because we did not find a positive relationship between lycopene and RS-assessed skin carotenoids, and Scarmo, et al.[11] and Ermakov, et al.[31] did find a positive relationship between plasma lycopene and skin carotenoids using RRS, the current RS device likely performs differently from the RRS device. The disagreement between the current study findings and those of Casperson, et al.[19] and Faraji, et al.,[20] warrant further validation using tissue biopsies.

Non-destructive spectroscopic approaches to measure carotenoids in skin have been reported for decades, as reviewed by Darvin et al.[32] They list five techniques along with the advantages and disadvantages of each. Most of the definitive and developmental work on the measurement of carotenoids in skin has been performed using RRS.[33,34] It permitted the selective monitoring of carotenoids with structures similar to β-carotene (9 conjugated double bonds and closed rings) to be distinguished from lycopene (11 conjugated double bonds and open rings).[33,34] (Figure 1) The validation of RRS for skin carotenoid measurements was performed by comparison to HPLC measurement of carotenoids in excised skin samples from the same subjects.[30,33] The correlations between total carotenoids by RRS and HPLC were strongly positive. Due to the invasive nature of collecting skin samples for validation studies, RS has only been compared to skin measurements by RRS and plasma carotenoids, however it has been noted that future studies should examine validity of RS against excised skin tissue samples[22]

Figure 1.

Absorbance Spectra and Chemical Structures of Lycopene and β-carotene, with reflection spectroscopy (RS) and resonance Raman spectroscopy (RRS) absorbance lines delineated.

Despite the lack of direct validation against excised skin, RS methods have become available as faster, easier, less expensive and more portable proxies for RRS.[13,14] RS methods rely on the absorbance characteristics of carotenoid molecules using an absorbance line at 480 nm, unlike RRS with selective excitation wavelengths which include 488 nm for β-carotene and similar carotenoids, and/or 514.5 nm for lycopene, the only major dietary carotenoid with excitation near 514.5 nm. Figure 1 illustrates the absorbance spectra of β-carotene and lycopene. The vertical lines mark the wavelengths used by RRS (black) and RS (blue) RRS also has selective emission spectra with lines at 1005, 1156, and 1523 cm-1. In the skin matrix, carotenoid spectra are shifted upward approximately 10-40 nm which would shift β-carotene-like carotenoids into the RS detection window (blue line in Figure 1), while shifting lycopene beyond the window of detection.[22,37] Thus, while RS has many advantages over RRS, including cost, ease of use, and portability, it lacks full validation when compared to excised skin carotenoids. The potential for less sensitive detection of lycopene is a limitation of skin carotenoid measurements using the current device.

The study has both strengths and limitations. One significant strength includes the extensive diversity in the large sample that encompasses individuals from geographically widespread areas across the United States with varied dietary intake. This broad representation offers a unique holistic and comprehensive approach to the association between and across different populations. Other strengths worth noting include the use of validated methods, and the use of non-invasive and objective measurements. Thirdly, lycopene intake and plasma concentrations were independent of other carotenoid intakes and plasma concentrations, bolstering the confidence in the lycopene findings.

Limitations include the fact that although blood plasma carotenoids are the standard gold biomarker for objectively measuring FV intake, skin biopsies were not available to directly compare actual skin carotenoid concentrations with SCS. Also, because we did not use an intervention design, we could not compare the dose response effect of independent carotenoid species on SCS. Furthermore, we used two different food frequency questionnaires in the two studies, which could create bias or systematic errors in our results. However, we got similar results for both studies, which suggests there was limited negative impact of using two different questionnaires. Finally, despite the inclusion of a large and diverse sample, there are challenges in generalizing the findings to broader populations, given considerable variations in the dietary and lifestyle behaviors within different regional, racial, and ethnic groups.

In conclusion, this study underscores the associations between SCS and individual plasma carotenoid species in a racially and ethnically diverse sample of adults. This study adds to the literature by looking at correlations between SCS and individual plasma carotenoids by racial/ethnic, age, and BMI status. The study sheds light on potential limitations and challenges with lycopene detection when utilizing the Veggie Meter^®. That is, lycopene-rich food intake may not be as well captured as foods rich in other dietary carotenoids. To address and enhance lycopene detection, future research studies should test multiple wavelength monitoring or an expanded window of detection. This could help provide a better assessment of skin lycopene. Furthermore, more studies should be conducted to evaluate commonly-used SCS devices and how the skin carotenoid scores from these devices correlate with individual plasma carotenoids.

Highlights:

• Skin carotenoid scores correlate with all plasma carotenoids, except lycopene.

• There was no confounding of the relationship by race, age, or body mass index.

• Lycopene-rich food intake may not be detected as well as other carotenoids.

List of abbreviations:

BMI

Body mass index

DHQ

Diet History Questionnaire

FFQ

Food frequency questionnaire

FV

Fruits and vegetables

HPLC

High performance liquid chromatography

MN

Minnesota

NCI

National Cancer Institute

NC

North Carolina

RS

Reflection spectroscopy

SCS

Skin carotenoid score

UV

Ultraviolet

US

United States