Approximately 600 different carotenoids are known to exist in natureReference Olson and Krinsky(1). Whereas plants, fungi, algae and bacteria can synthesise these compounds, animals cannot and therefore need to obtain them from their dietReference Mayne(2). The only clearly proven function of carotenoids in humans is their provitamin A activity. The major carotenoids found in human sera are α- and β-carotene, lycopene, lutein, zeaxanthin and cryptoxanthin. Of these, α- and β-carotene and β-cryptoxanthin have provitamin A activityReference Olson(3).
Recently, non provitamin A-related activities of carotenoids have received attention. These include possible contribution to reducing the risk for diseases such as cardiovascular disease (CVD)Reference Rissanen, Voutilainen, Nyyssonen, Salonen and Salonen(4–Reference Gaziano, Manson, Branch, Colditz, Willett and Buring6), cancerReference Ziegler, Mason, Stemhagen, Hoover, Schoenberg, Gridley, Virgo and Fraumeni(7–Reference Potischman, McCulloch and Byers9), cognitive declineReference Rinaldi, Polidori, Metastasio, Mariani, Mattioli, Cherubini, Catani, Cecchetti, Senin and Mecocci(10), age-related macular degeneration and cataractReference Seddon, Ajani and Sperduto(11–Reference Mares-Perlman, Fisher, Klein, Palta, Block, Millen and Wright13). However, many of these associations are yet to be proved conclusively.
CVD has been a dominant cause of death in the USA for the past 50 or more years. Heart disease and stroke are the first and third specific causes of death in this country. Within the USA, African Americans have the highest rates of morbidity and mortality from CVD, and the south-eastern part of the USA has the highest rates of hospitalisations due to stroke and heart failureReference Mensah, Mokdad, Ford, Greenlund and Croft(14). Much of the available data relating risk factors and CVD have come from predominantly white populationsReference Sempos, Bild and Manolio(15). The Jackson Heart Study (JHS) was therefore initiated to investigate the causes of CVD in an all African American cohort based in Jackson, MississippiReference Taylor(16).
Given the health disparities that exist among African Americans with respect to CVD and the protective role that carotenoid intake and status may play in the prevention of this group of diseases, the objectives of the current study were:
1. To assess the dietary intake and serum concentrations of α- and β-carotene, β-cryptoxanthin, lutein plus zeaxanthin, and lycopene among participants of the Diet and Physical Activity Sub-Study (DPASS) of the JHS using three dietary assessment tools – two regional food-frequency questionnaires (FFQs) and four 24 h recalls.
2. To evaluate the associations between carotenoid intake measures and serum concentrations.
3. To identify the relative contributions of foods to carotenoid intakes in this population.
Methods
Population
Participants were from the JHS, a single-site prospective epidemiologic investigation of CVD among African Americans from the Jackson, Mississippi metropolitan area. The JHS baseline data collection took place from late 2000 until early 2004. Data on conventional as well as new and emerging risk factors of CVD were collected on a randomly selected representative sample of African Americans aged 34–84 years residing in the Hinds, Madison and Rankin counties surrounding Jackson, Mississippi. A more detailed description of the original study has been published elsewhereReference Taylor, Wilson, Jones, Sarpong, Srinivasan, Garrison, Nelson and Wyatt(17).
Selection of study sample
A subset of participants (n = 499) from the JHS cohort (N = 5302) was selected for the JHS-DPASS. The aim of DPASS was to provide data for validation of the diet and physical activity instruments used for the entire cohort of the JHS. DPASS investigators identified potential participants for DPASS based on certain criteria. The goal was to include an equal number of men and women from younger (34–64 years) and older (65 years and older) age groups, lower and higher socio-economic status, as well as lower and higher physical activity groups. Participants were enrolled into DPASS on a rolling basis and data were collected throughout the year.
The data presented here include all DPASS participants with complete dietary and serum carotenoid data. The Institutional Review Board of the University of Mississippi approved the DPASS protocols, and all subjects gave written informed consent for their participation.
Dietary assessment
The most widely used FFQs in the USA were designed to capture foods most commonly consumed in the general US populationReference Block, Hartman, Dresser, Carroll, Gannon and Gardner(18, Reference Willett, Sampson, Stampfer, Rosner, Bain, Witschi, Hennekens and Speizer19). This can lead to incorrect estimation of the dietary intakes of population subgroups and ethnicities who have dietary intakes and practices that are different from those of the ‘general’ population. These errors can ultimately lead to extensive misclassification.
The Lower Mississippi Delta Nutrition Intervention Research Initiative (LMD NIRI), funded by the US Department of Agriculture (USDA) Agricultural Research Service, conducted a telephone survey in the Delta region to collect representative dietary data using 24 h dietary recalls. These data were used to develop a new FFQ designed for use in the LMD region. Foods reported in the recalls were grouped into food groups based on nutrient content and culinary use. These food groups were then ranked based on their relative contributions to energy and nutrients. Foods contributing at least 0·5 % to any of the selected nutrients were included on the FFQ food list. Serving sizes were also modified to reflect those reported during the recalls. Data from the survey were also used to provide weighting and recipe information for the nutrient database used to analyse questionnaire responses. The Delta NIRI FFQ was also field-tested in the LMD and Jackson, Mississippi region using a modified cognitive interviewing methodology. The resulting Delta NIRI FFQ has 283 items (long FFQ). Further details regarding development of this regional FFQ are available elsewhereReference Tucker, Maras, Champagne, Connell, Goolsby, Weber, Zaghloul, Carithers and Bogle(20). A shortened version with 158 items (short FFQ) was specifically developed for use in the JHS. This version does not eliminate food groups, but rather includes less detail by collapsing similar items for efficiency in reporting. Foods within line items were weighted, based on the frequency of consumption in the Delta 24 h recalls, for calculation of nutrient content. To avoid overestimation of fruit and vegetable intake on the long FFQ, summary questions on overall fruit and vegetable intake were asked. Nutrient intakes from the responses on individual fruits and vegetables were then adjusted proportionally to these responses for overall frequency. Both FFQs were used as dietary assessment tools in the DPASS.
Four 24 h recalls were used to measure the ‘quantitative’ intake of DPASS participants. Trained and certified dietitians collected these recalls using the Nutrient Data System developed by the University of Minnesota. To stabilise variability of intake, participants provided two weekdays and two weekend days for the recalls.
The DPASS encounters included an initial administration of the short FFQ, followed by four 24 h recalls scheduled approximately one month apart, and finally administration of the long FFQ, scheduled approximately one week after administration of the last recall. The short FFQs were administered by trained clinic staff while the recalls and long FFQs were administered by trained dietitians. All FFQs were reviewed by DPASS staff for entry errors. Furthermore, 5 % of all FFQs and all recalls were audio-taped for quality control purposes, and were reviewed by the DPASS principal investigator. Retraining was conducted whenever problems with quality or completeness were identified by the review. Details regarding the methodology used for the DPASS have been published elsewhereReference Carithers, Dubbert, Crook, Davy, Wyatt, Bogle, Taylor and Tucker(21).
Laboratory analyses
Participants provided blood samples on the day of the baseline interview, which took place on the day of administration of the short FFQ and, on average, a year prior to administration of the long FFQ. Blood samples from fasting (12 h) participants were collected in Vacutainer tubes and centrifuged at 3000 g for 10 min at 4°C. Serum was separated and frozen at −70°C until analysed for carotenoids. Estimation of carotenoids was performed using high-performance liquid chromatography as described by Yeum et al.Reference Yeum, Ahn, Rupp de Paiva, Lee-Kim, Krinsky and Russell(22). Serum cholesterol concentrations were determined according to methods described previouslyReference Carpenter, Crow, Steffes, Rock, Heilbraun, Evans, Skelton, Jensen and Sarpong(23).
Assessment of covariates
Information on age and smoking status was collected by questionnaire at either the home induction visit or at the time of the participant’s clinic visit. Supplement use was obtained from responses on the FFQs. Both height and weight were measured by trained technicians at the time of the clinic visit. Anthropometric procedures have been detailed elsewhereReference Taylor, Wilson, Jones, Sarpong, Srinivasan, Garrison, Nelson and Wyatt(17). Body mass index (BMI) was calculated from these measurements as weight/height2 (kg/m2). Information on month of administration of the FFQ was obtained from the FFQ.
Statistical analyses
DPASS participants for whom there were no serum samples provided for carotenoid analysis (n = 39), or who reported energy intake outside the plausible range (<2510 or >16 736 kJ/d) on any of the FFQs or mean of the four 24 h recalls (n = 53), or who had more than 10 % of the questions blank on any FFQ (n = 5), were excluded from analyses. This resulted in a sample of 402 individuals.
Both intake and biochemical measures of carotenoids were skewed, and were log-transformed prior to analyses. Descriptive analyses were performed to assess sex differences using general linear models for continuous variables and χ 2 analyses for categorical data for demographics of the population and serum carotenoid and lipid measurements. Pearson’s correlations were used to examine associations between intake and serum measures for each of the carotenoids. Of these carotenoids, only β-carotene was commonly present in dietary supplements at the time of this study. Therefore, we estimated the total (diet plus supplement) and dietary associations between this carotenoid and its corresponding serum measure. Partial correlations were used to adjust for age, sex, BMI, energy intake from the corresponding dietary instrument used, serum total cholesterol concentrations and smoking status. Day-to-day within-person variation across 24 h dietary recalls can attenuate correlations between nutrient intakes derived from the mean of the dietary recalls and the serum carotenoid estimates. We therefore calculated the intra- to inter-person variance for the nutrient intakes from the four 24 h recalls. The following formula was used to calculate the deattenuated correlation coefficients: 
, where r o is the observed correlation coefficient between the nutrient intake as determined from the mean of the four 24 h recalls and the corresponding serum measure, intrax is the intra-subject variation, interx is the inter-subject component of variance for each nutrient and n is the number of days of recalls, which in the present study was four days. Deattenuated correlations for adjusted variables were calculated after adjustment by the residual method.
Using the General Linear Models (GLM) procedure, with adjustment for age, sex, energy intake from the appropriate assessment method, month of administration (for the FFQ only), BMI, serum cholesterol concentrations and smoking status, we estimated the mean carotenoid concentrations for respective quartiles of total carotenoid intake. We also ranked the main sources of carotenoids for the long FFQ by calculating the percentage contribution of each food item (or supplement) to the total intake of each carotenoid. All α values were set at the 0·05 level. The SAS statistical software package, release version 9.1 (SAS Institute, Cary, NC, USA) was used for all analyses.
Results
The mean age of subjects did not differ significantly between men and women (approximately 60 vs. 62 years) (Table 1). Mean age-adjusted BMI was higher for women than men (approximately 32 vs. 29 kg/m2, P < 0·05). A lower percentage of women than men were current smokers. As reported on either FFQ, significantly more women than men reported taking supplements. Also, use of supplements containing β-carotene was higher in women compared with men.
Table 1 Characteristics of the JHS-DPASS participants‡
JHS, Jackson Heart Study; DPASS, Diet and Physical Activity SubStudy; sem, standard error of the mean; BMI, body mass index; FFQ, food-frequency questionnaire.
*Significantly different from men: P < 0·05.
‡Sex groups were compared by Generalised Linear Models after adjusting for age.
Women had higher serum concentrations of high-density lipoprotein cholesterol and total cholesterol compared with men (P < 0·05) (Table 2). No differences were observed for serum low-density lipoprotein cholesterol or triacylglycerol concentrations. Women had significantly lower serum lycopene concentrations (P < 0·05) than to men. There were no other statistically significant differences for serum carotenoids.
Table 2 Serum concentrations of carotenoids and lipoproteins of the JHS-DPASS participants‡,§,∥
JHS, Jackson Heart Study; DPASS, Diet and Physical Activity Sub-Study; sem, standard error of the mean; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol.
*Mean values were significantly different from those for men: P < 0·05.
‡Sex groups were compared by Generalised Linear Models after adjusting for age.
§Pairwise comparisons, using log-transformed variables, were done by sex.
∥β-Carotene supplement use was based on reported use on the short food-frequency questionnaire.
Median intakes of the carotenoids estimated by the two FFQs and the average of four 24 h recalls are shown in Table 3. For illustration, median values from either of the FFQs that were 15 % higher or lower than those reported by the recalls are marked with ‘††’. For men, as compared with the recalls, both the short and the long FFQ overestimated α-carotene, β-cryptoxanthin and lycopene and underestimated lutein plus zeaxanthin. The long FFQ underestimated both dietary and total β-carotene intakes. For women, both the long and the short FFQ overestimated α-carotene and lycopene while underestimating lutein plus zeaxanthin. The long FFQ underestimated dietary β-carotene intakes and overestimated β-cryptoxanthin intakes compared with the recalls.
Table 3 Median (25th, 75th percentile) intakes of carotenoid nutrients of the JHS-DPASS participants
JHS, Jackson Heart Study; DPASS, Diet and Physical Activity Sub-Study; FFQ, food-frequency questionnaire.
†Median value is >15 % different from that for men.
††Median value is >15 % different from that for the recalls.
Median values for nutrient differences between men and women higher or lower than 15 % as estimated by each assessment tool are indicated by ‘†’. Men had higher intakes of α-carotene on the recalls and the long FFQ and of lycopene for all the assessment tools. Energy intakes were also higher for men as estimated by the recalls and the short FFQ (Table 3).
We examined simple and adjusted Pearson’s correlations between carotenoid intakes and serum concentrations (Table 4). Correlations were adjusted for age, BMI, serum cholesterol concentration, smoking status, month of administration (only for the FFQs) and energy intake from the respective assessment tool. In general, the adjusted coefficients were higher for the questionnaires than their crude counterparts; particularly for the short FFQ. For the average of the recalls, deattenuated adjusted correlations either remained the same or decreased slightly compared with their deattenuated crude counterparts. For the short FFQ, adjusted correlations ranged from 0·35 for α-carotene to 0·15 for lutein plus zeaxanthin. For the long FFQ, adjusted correlations ranged from 0·28 for total β-carotene to 0·14 for lycopene. For the average of the recalls, deattenuated adjusted correlations ranged from 0·42 for β-cryptoxanthin to 0·25 for dietary β-carotene.
Table 4 Crude and adjusted Pearson’s correlations between serum carotenoid nutrient biomarkers and carotenoid intakes in the JHS-DPASS‡,§,∥,¶
JHS, Jackson Heart Study; DPASS, Diet and Physical Activity Sub-Study; FFQ, food-frequency questionnaire.
*P < 0·05.
‡Both intake and serum carotenoid variables were log-transformed.
§For recalls, both crude and adjusted correlations calculated are deattenuated. Adjustment is for age, sex, energy intake from the appropriate assessment tool, body mass index (BMI), serum cholesterol and smoking status.
∥For FFQs, adjustment is for age, sex, energy intake from the appropriate assessment tool, BMI, serum cholesterol, smoking status and month of FFQ administration.
¶For crude correlations, N = 402 for all assessment methods. For adjusted correlations, because of missing data, N = 401 for the short FFQ and recalls and N = 400 for the long FFQ.
We also examined the mean serum concentrations of carotenoids by quartiles of total intake as measured by each of the dietary assessment instruments. For α-carotene and β-carotene, serum measures were each higher with increasing intake and levelled at the highest intake quartiles. For β-cryptoxanthin, there was a linear trend between intake and serum concentration. For lycopene, weaker trends were seen. The associations for lutein and zeaxanthin were weaker than those seen for the other carotenoids (Table 5).
Table 5 Mean serum concentrations of carotenoids (μmol/l) for the JHS-DPASS participants by quartile of carotenoid intake as estimated by the different dietary assessment instruments‡,§
JHS, Jackson Heart Study; DPASS, Diet and Physical Activity Sub-Study; FFQ, food-frequency questionnaire.
*Test for trend across quartiles: P < 0·05.
‡Mean serum carotenoid concentrations adjusted for age, sex, energy intake from the appropriate assessment tool, body mass index, serum cholesterol, current smoker (Y/N), month of administration (only for the FFQs).
§(1) For α-carotene; median intakes in the quartile categories were 173, 289, 446, 733 μg/d on the short FFQ; 109, 227, 354, 689 μg/d on the long FFQ; and 21, 67, 290, 782 μg/d on the average of the recalls.
(2) For β-carotene; median intakes in the quartile categories were 1950, 2781, 3372, 4302 μg/d on the short FFQ; 1413, 2312, 3347, 5165 μg/d on the long FFQ; and 863, 2444, 4256, 7874 μg/d on the average of the recalls.
(3) For β-cryptoxanthin; median intakes in the quartile categories were 40, 86, 147, 230 μg/d on the short FFQ; 37, 83, 151, 268 μg/d on the long FFQ; and 16, 62, 139, 254 μg/d on the average of the recalls.
(4) For lutein plus zeaxanthin; median intakes in the quartile categories were 1436, 1896, 2465, 3575 μg/d on the short FFQ; 1143, 1591, 2119, 3334 μg/d on the long FFQ; and 912, 1893, 3991, 7697 μg/d on the average of the recalls.
(5) For lycopene; median intakes in the quartile categories were 1220, 2472, 3646, 6247 μg/d on the short FFQ; 1219, 2265, 3318, 5858 μg/d on the long FFQ; and 36, 918, 2342, 6999 μg/d on the average of the recalls.
We identified and ranked the foods that contributed to the majority of the dietary intakes of these carotenoids in this population as assessed by the long FFQ (Table 6). Top dietary contributors of α-carotene were orange vegetables (both raw and cooked), whereas dietary sources of β-carotene were mustard, turnip and collard greens followed by sweet potato. When supplements were considered, they provided 22 % of β-carotene. Orange juice provided most of the β-cryptoxanthin. Leafy greens like mustard, turnip and collard were the top contributors of lutein and zeaxanthin, providing almost 40 % of the carotenoid. Tomato juice and pasta preparations were the main sources of lycopene.
Table 6 Main contributors to carotenoid intakes among the JHS-DPASS participants‡
JHS, Jackson Heart Study; DPASS, Diet and Physical Activity Sub-Study.
‡Those contributing at least 5 % of intake on the long food-frequency questionnaire are listed.
Discussion
CVD is a major health concern among African Americans. Dietary carotenoids have been implicated in decreased risk for CVD. The purpose of this study was to assess the intake and serum concentrations of α- and β-carotene, β-cryptoxanthin, lutein plus zeaxanthin, and lycopene among participants of the JHS-DPASS using three dietary assessment tools (two regional FFQs and the average of four 24 h recalls) and to assess associations between intake measures and serum concentrations. In addition, we identified the relative contributions of foods to carotenoid intakes in this population.
Despite reports of low fruit and vegetable intake in this regionReference Champagne, Bogle, McGee, Yadrick, Allen, Kramer, Simpson, Gossett and Weber(24), the carotenoid intakes of the JHS-DPASS participants were within the 50th–90th percentile values from the Third National Health and Nutrition Examination Survey (NHANES III) for similar age and sex groups for all carotenoids except lycopene, where the intakes were within the 50th–75th percentile range(25). In general, the biochemical status of the carotenoids was also within the 50th–75th percentile range for African American participants of the NHANES III examinationReference Ford(26) and similar to those reported for other populationsReference Ford(26–Reference Bermudez, Ribaya-Mercado, Talegawkar and Tucker31). Serum lycopene concentrations appeared higher than other published values. This did not appear to be due to a higher intake of dietary lycopene, but rather could be attributed to the inclusion of both cis and trans isomers of lycopene in our estimation of serum lycopene concentrations. Generally, the cis isomers are not included in lycopene estimationsReference Vogt, Mayne and Graubard(32). When we included only trans-lycopene in the assessment, the age-adjusted serum trans-lycopene concentrations were 0·50 (standard error of the mean (sem) 0·02) μmol/l for men and 0·45 (sem 0·01) μmol/l for women, which were closer to values reported by others.
The FFQs used in this study were specifically developed for use with a southern US population. Correlations comparing intake measured with other assessment tools including the Block or Harvard FFQ and biochemical measures of carotenoids have demonstrated correlation coefficients ranging from 0·09 to 0·45 for various carotenoidsReference Tucker, Chen, Vogel, Wilson, Schaefer and Lammi-Keefe(28, Reference Tangney, Bienias, Evans and Morris30, Reference Jacques, Sulsky, Sadowski, Phillips, Rush and Willett33–Reference Kabagambe, Baylin, Allan, Siles, Spiegelman and Campos37), similar to our range of 0·12 to 0·35. Some of the lowest correlations seen were for lutein plus zeaxanthin. Lutein and zeaxanthin, unlike nutrients such as β-cryptoxanthin, are present in a wider variety of foods. This could make its estimation using FFQs with a limited number of food items difficult(38).
The mean lycopene intakes obtained from both FFQs in general were higher than those from the recalls. This has also been previously reported by other researchersReference VandenLangenberg, Brady, Nebeling, Block, Forman, Bowen, Stacewicz-Sapuntzakis and Mares-Perlman(39). Relatively weak correlations were also seen for lycopene with the long and the short FFQ compared with the average of the 24 h recalls. It is possible that the recalls capture processing techniques that may reflect availability of lycopene from tomato-containing products better than the FFQs, which make assumptions about the tomato content in foods such as pasta with meat dishesReference Porrini and Riso(40). Besides this, researchers have suggested that the lack of correlation between intake and serum measures may be affected by several other factors including recent lycopene intake, age and genetics, as well as individual absorption capacityReference Porrini and Riso(40).
The design of the DPASS included administration of the short FFQ at the time of the clinic visit. If the participant met the criteria for the diet sub-study, he or she was asked to join, and if the participant agreed, the first DPASS visit with administration of the first 24 h recall took place on average a month or two after the clinic visit. The next three 24 h recalls were scheduled to be administered approximately one month apart, and a week after the last recall, the participant was administered the long FFQ. As recruitment to the JHS was on a rolling basis, participants enrolled into the DPASS on a rolling basis as well. Because the short FFQ was administered on the day of the blood draw (date of clinic visit), it could be expected that the correlations between carotenoid intake and biochemical status may be artefactually higher than for the FFQ administered later; however this was not the case. Correlations for β-carotene and lutein plus zeaxanthin were higher for the long FFQ than for the short, despite the greater distance in time between measures.
In this population of African Americans, orange vegetables were the top contributors to α-carotene intake and greens and sweet potato were the top contributors to β-carotene intake. In a study conducted in the LMD regionReference Tucker, Maras, Champagne, Connell, Goolsby, Weber, Zaghloul, Carithers and Bogle(20), we previously reported that sweet potatoes were the top contributor to vitamin A intakes in the African American subgroup. Citrus fruit juices were the main contributors of β-cryptoxanthin intakes in this population. This has been reported in several diverse study populationsReference Bermudez, Ribaya-Mercado, Talegawkar and Tucker(31, Reference Garcia-Closas, Berenguer and Jose Tormo41, Reference Manzi, Flood, Webb and Mitchell42). Dark green leafy vegetables were the main source of lutein plus zeaxanthin. Nebeling et al.Reference Nebeling, Forman, Graubard and Snyder(43) reported similar findings in the African American subgroup of a national survey. In our study, tomato and tomato products were the top contributors of lycopene intake. This is in agreement with other studies conducted in the USA, where most of the lycopene is accounted for by tomato and tomato-containing productsReference Clinton(44).
Contrary to our expectation, carotenoid intakes and status did not appear to be lower in this population of southern African American adults than in the general US population. The deattenuated adjusted correlations obtained using the 24 h recalls were higher than those obtained from either FFQ. This could indicate that multiple recalls may be a better way to capture carotenoid intake. However, the range of correlations seen with these FFQs is similar to those published from validation studies of carotenoids with other FFQs. This, taken with the advantages of single administration and lower cost, suggests that FFQs provide relatively valid and useful measures of carotenoid intake in this population.
Acknowledgements
The authors have no conflicts of interest to report. This research was supported by NIH contracts N01-HC-95170, N01-HC-95171 and N01-HC-95172 from the National Heart, Lung, and Blood Institute and the National Center for Minority Health and Health Disparities; and by USDA Agricultural Research Service contracts #6251-53000-003-00D and #58-1950-7-707. S.A.T. performed the analysis and drafted the paper as part of her PhD thesis at Tufts University, E.J.J. directed the laboratory analysis and contributed to the interpretation of laboratory measures, T.C.C. directed the dietary data collection, H.A.T. is the principal investigator of the Jackson Heart Study, M.L.B. contributed to the dietary data collection and interpretation, K.L.T. oversaw the design, analysis and writing for this manuscript. All authors reviewed earlier drafts for content and form and approved the final version.
