Alkylresorcinol Metabolite Concentrations in Spot Urine Samples ...

The Journal of Nutrition Nutritional Epidemiology

Alkylresorcinol Metabolite Concentrations in Spot Urine Samples Correlated with Whole Grain and Cereal Fiber Intake but Showed Low to Modest Reproducibility over One to Three Years in U.S. Women1–3 Rikard Landberg,4* Mary K. Townsend,5 Nithya Neelakantan,6 Qi Sun,5,7 Laura Sampson,7 Donna Spiegelman,8 and Rob M. van Dam7,6,9 4 Department of Food Science, Swedish University of Agricultural Sciences, Uppsala, Sweden; 5Channing Laboratory, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Harvard University, Boston; MA; 6Saw Swee Hock School of Public Health, National University of Singapore, Singapore; 7Department of Nutrition, and 8 Department of Epidemiology, Harvard School of Public Health, Harvard University, Boston, MA; and 9Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore

Abstract Two alkylresorcinol (AR) metabolites, 3, 5-dihydroxybenzoic acid (DHBA) and 3-(3,5-dihydroxyphenyl)-1-propanoic acid (DHPPA), in urine have been suggested as biomarkers of whole grain (WG) and cereal fiber intake but the long-term reproducibility and correlation with habitual intake has not been determined. Therefore, we evaluated the long-term reproducibility of AR metabolites in spot urine samples and investigated their correlation with habitual WG and cereal fiber intake in U.S. women. AR metabolites were analyzed in 104 women participating in the Nurses’ Health Study II and WG and fiber intakes were assessed using a FFQ. Long-term reproducibility was assessed by calculating the intra-class correlation coefficients (ICC) using samples taken 1–3 y (mean 1.8 y) apart. The observed Spearman correlation coefficients (rs) and rs adjusted for within-participant variation in the biomarker were calculated between WG and fiber intake and biomarkers. The long-term reproducibility was poor for DHBA [ICC = 0.17 (95% CI: 0.05, 0.43)] and modest for DHPPA [ICC = 0.31 (95% CI: 0.17, 0.51)]. The correlation between WG intake in 1995 and DHPPA measured 2 y later was 0.37 (P , 0.0001); the adjusted correlation was 0.60 (95% CI: 0.37, 0.76). Cereal fiber and WG intake were similarly correlated to the biomarkers. DHPPA in spot urine samples reflected WG intake despite relatively low intake of food sources of AR. The poor to modest reproducibility may limit the use of single measurements of these biomarkers in cohort studies in the US, where WG intake is relatively low and has changed over time. But DHPPA in repeated samples may be useful for validating WG intake and assessing compliance in WG intervention studies. J. Nutr. 142: 872–877, 2012.

Introduction A high whole grain (WG)10 intake has been consistently associated with a reduced risk for developing type 2 diabetes and coronary heart disease in epidemiological studies (1–6). In 1

Supported by NIH grant R01 DK082486, by Nordforsk Centre of Excellence in Nutrition program Helga (to R.L.), and by a career development award from the National Heart, Lung, and Blood Institute (K99HL098459 to Q.S.). 2 Author disclosures: R. Landberg, M. K. Townsend, N. Neelakantan, Q. Sun, L. Sampson, D. Spiegelman, and R. M. van Dam, no conflicts of interest. 3 Supplemental Table 1 is available from the “Online Supporting Material” link in the online posting of the article and from the same link in the online table of contents at jn.nutrition.org. 10 Abbreviations used: AR, alkylresorcinol; DHBA, 3, 5-dihydroxybenzoic acid; DHPPA, 3-(3,5-dihydroxyphenyl)-1-propanoic acid; ICC, intra-class correlation coefficient; WG, whole grain. * To whom correspondence should be addressed. E-mail: rikard.landberg@slu. se.

872

most studies, assessment of habitual WG intake was based on self-reports on FFQ. This method is subject to measurement error due to consumers’ difficulties in recognizing WG, the limited number of WG items usually included on the FFQ, and limited information on the WG content of representative foods in food composition databases (7–10). A biomarker of WG intake may have the potential to overcome some of the obstacles related to dietary assessment methods and provide an alternative or complementary ranking-tool of WG intake (4,8,11). Alkylresorcinols (AR) are phenolic lipids exclusively present in the outer parts of wheat and rye grains and have been suggested and evaluated as biomarkers of WG/bran of these cereals (4,8,11–13). However, the half-life of intact AR in plasma is rather short (4–5 h) and therefore they mainly reflect short- to medium-term intake in free-living populations (14,15).

ã 2012 American Society for Nutrition. Manuscript received December 13, 2011. Initial review completed January 13, 2012. Revision accepted February 9, 2012. First published online March 21, 2012; doi:10.3945/jn.111.156398.

Recently, 2 main metabolites of AR 3,5-dihydroxybenzoic acid (DHBA) and 3-(3,5-dihydroxyphenyl)-1-propionic acid (DHPPA) showed longer apparent half-lives compared with intact AR homologs (16,17). The longest apparent half-life was observed for DHPPA (12 h), suggesting that it may better reflect more long-term WG intake (17). Aubertin-Leudre et al. (18,19) showed that DHBA and DHPPA concentrations in plasma and their excretion in three 2-h urine collections correlated with cereal fiber intake estimated by weighed food records in a Finnish group of free-living individuals (n = 56). The correlation coefficients ranged from 0.41 to 0.46 for plasma and from 0.38 to 0.40 for urine (18,19). In a cross-sectional study in U.S. men and women, every serving per day increase of WG wheat and rye food products was associated with 94% higher creatinine-adjusted DHPPA concentrations in morning spot urine samples (20). However, none of the previous studies evaluated the long-term reproducibility (1–3 y) of DHBA and DHPPA in spot urine samples or investigated the correlation with long-term (1 y) WG intake. These are important features of a biomarker that need to be investigated before applying the biomarker in epidemiological studies. In the present study, we evaluated the long-term reproducibility of the 2 AR metabolites in spot urine samples in U.S. women over a period of 1–3 y. We also evaluated the relative validity of the 2 metabolites as biomarkers of WG, bran, or dietary fiber intake estimated by FFQ.

Participants and Methods Study population. The Nurses’ Health Study II was established in 1989 and included 116,430 female nurses aged 25–42 y residing in the United States. Since then, participants have been mailed biennial follow-up questionnaires regarding their medical history, lifestyle, and other risk factors. FFQ were mailed to participants every 4 y starting in 1991. Between 1996 and 1999, 29,611 participants donated blood and urine samples. Of these women, 18,521 were premenopausal and had blood and urine samples that were timed within the menstrual cycle. Samples were shipped via overnight courier with an ice pack to the laboratory where urine was aliquoted and stored in liquid nitrogen freezers (#1308C). Of these participants, 412 women were invited to participate in the “hormone stability study” and asked to provide a second and third set of blood and urine samples. A second set of samples was provided by 74% of the women (n = 304) and 57% (n = 236) provided 3 sets in total. Of the women providing a third urine sample, 110 women provided samples 3–11 d before the start of the next menstrual cycle in all 3 collections. Samples from the first and second collections from 104 women, among whom sufficient samples remained, were included in the present study. The second sample was taken 1–3 y after the first sample (mean 1.8 y). The Harvard School of Public Health and Brigham and Women’s Hospital Human Subjects Committee Review Board approved the study protocol. Laboratory methods. Urinary AR metabolites DHBA and DHPPA were analyzed by a validated GC-MS method previously described (21,22). The lower limit of quantification (signal:noise ratio of 10:1) was 0.1 and 0.01 mmol/L for DHPPA and DHBA, respectively. Two laboratory quality control samples, containing low (5.0 mmol/L) or high (258 mmol/L) total AR-metabolite concentrations were included in quadruplicate in every batch analyzed. The AR metabolite within- and between-batch CV were similar for DHBA and DHPPA and total AR metabolite CV were 6.5 and 4.8% for low and 8.6 and 7.9% for high concentration samples, respectively. Samples were analyzed in 3 batches. Moreover, blinded quality control samples (n = 24) were included in triplicate to ensure that appropriate precision was achieved within the range of samples analyzed. The within- and between-batch CV for total AR metabolites for these 2 blinded samples were ,8%. To evaluate the stability of AR metabolites in urine during sampling and mailing, we

simulated the conditions of overnight and next-day mail service used for sample collection within the Nurses’ Health Study. Urine samples from 10 individuals were placed with ice packs for 0, 24, and 48 h before they were prepared into aliquots and stored in liquid nitrogen freezers (#1308C). We found no significant difference in urinary DHBA, DHPPA, or total metabolite concentrations over 0–48 h (P = 0.28–0.38), i.e., AR metabolites are stable under the tested conditions. Creatinine was determined with an Architect ci8200 (Abbott). Within- and betweenbatch CV were ,5%. Assessment of WG and fiber intakes and BMI. We used 145 and 151-item semiquantitative FFQ to assess diet in 1995 and 1999, respectively. Briefly, the FFQ were designed to assess average food intake during the previous year. For each food, a commonly used portion size was specified together with 9 possible response categories for frequency of intake, ranging from never to $6/d. WG intake was derived from questions on the FFQ (14 in 1995 and 16 in 1999). Open-ended questions regarding the usual serving size and frequency of consumption of foods not listed on the FFQ were included and participants were asked to add brand name on breakfast cereal products consumed. Portions were converted to gram weights per serving and intake of nutrients and WG were computed by multiplying the frequency of consumption of each unit of food by the nutrient or WG content in grams. A database for WG content of cereal food products used in the US has been developed and food items in the FFQ were assigned total WG, total and added bran, and total and added germ content values (23). The following cereals or ingredients were used to calculate the content values: barley, brown rice, brown rice flour, buckwheat, bulgur, corn bran, whole corn flour, whole cornmeal, millet, oats, oat bran, oat fiber, whole oat flour, psyllium, rye, whole rye flour, wheat bran, wheat germ, and whole wheat flour. In the present study, we excluded added bran and germ in the total WG content value, because these are not considered as WG according to the American Association of Cereal Chemist’s definition of WG commonly used in the US and Europe (24). The amount of WG in each grain and ingredient was assigned by setting it equal to its dry weight. Bran and germ values were assigned using typical percentages of these constituents for different cereal grains reported in the literature. Product labels and nutrient and ingredient information declared by the manufacturers were used to derive amounts of WG in 96 brand-name breakfast cereal products. For the question “dark bread” in the FFQ, WG content in one bread was used in 1995 and a composite of 7 different commercially available breads was used in 1999 by calculating their weighed mean based on observed shelf space in supermarkets from a pilot study. Intakes of total WG, total bran, and total germ were calculated in grams per day. In addition, intakes of AOAC total and cereal fiber were calculated. Body weight was self-reported every second year. Statistical analysis. We tested the stability of AR metabolites over 48 h of storage by random effect models using PROC MIXED in SAS ver. 9.1 (SAS Institute). Response variables (DHBA, DHPPA, and their sum representing total AR metabolite concentration) were log-transformed to ensure validity of inference. Participant was entered as a random factor and time (0, 24, 48 h) was entered as class variable. For the assessment of long-term reproducibility and relative validity, the AR metabolite concentrations in urine were adjusted for creatinine by dividing the AR metabolite concentrations by the concentration of creatinine to correct for differences in urine volumes between individuals. Concentrations were log-transformed before analysis to ensure validity of inference. Differences in intake variables were evaluated by Wilcoxon’s matched pairs Signed Rank Sum test (because these variables were not normally distributed after log-transformation) and differences in urinary biomarker values between occasion 1 and 2 were evaluated using paired t tests. Between-person and within-participant variances were estimated from 2 sampling occasions by random effect models using PROC MIXED in SAS ver. 9.1 (SAS Institute). Reproducibility, which is a measure of the within-participant stability in biomarker concentration over time, was assessed over a 1- to 3-y period by the intra-class correlation coefficient (ICC). ICC is defined as the betweenparticipant variance divided by the total variance (i.e., the sum of withinand between participant variance). The 95% CI of estimated ICC were Alkylresorcinol metabolites in U.S. women

873

calculated as described by Hankinson et al. (25). Between- and withinparticipant CV on the original scale were estimated by taking the square root of the between- and within-participant variance, respectively. The residual method was used to calculate energy-adjusted nutrient intake variables (26). Spearman’s rank correlation coefficients were calculated to examine the correlation between dietary intake variables and creatinine-adjusted biomarker concentrations. Adjusted Spearman’s rank correlation coefficients were used to estimate the correlation between dietary variables and urinary biomarker concentrations, correcting for random measurement errors in the biomarker. Methods proposed by Rosner et al. (27) were used to calculate adjusted correlation coefficients and their 95%CI. All statistical analyses were carried out by using SAS version 9.1 (SAS Institute). P values , 0.05 were considered significant. Values in the text are mean 6 SD unless otherwise stated.

Results The mean age of participants in the present study (n = 104) was 41.7 6 3.5 y. BMI was 24.6 6 5.2 and 25.0 6 5.3 kg/m2 for 1995 and 1999, respectively. All samples contained DHBA and DHPPA above the lower limit of quantification. The urinary DHBA concentrations ranged from 0.2 to 15 mmol/L with IQR of 3.6 and 4.9 mmol/L for the first and second collections, respectively. The DHPPA concentration ranged from 1.6 to 47.6 mmol/L (IQR1st = 11.3 mmol/L, IQR2nd = 10.9). DHBA was significantly higher at the second occasion, whereas no significant difference was observed for DHPPA (Table 1). The ICC were higher for DHPPA compared with DHBA, because of both lower within-participant variation and higher between-participant variation (Table 1). For comparison, we also evaluated within-participant variation in reported WG and fiber intakes using data from FFQ administered in 1995 and 1999. The reported mean energy or macronutrient intakes did not differ between these 2 occasions, except for cereal fiber, WG, bran, and germ intakes, which increased significantly between 1995 and 1999 (Table 2). The long-term reproducibility was high for total dietary fiber and modest for cereal fiber, WG, bran, and germ intake, which in part may be explained by the increase in WG intakes between 1995 and 1999 (Table 2). The lower ICC observed for cereal fiber, bran germ, and WG compared with total fiber was due to larger within-participant variation rather than smaller betweenparticipant variation (Table 2). The between-participant variation in cereal fiber, WG, bran, and germ intakes was considerably higher compared with total dietary fiber. For the entire cohort in 1995 and 1999 (n = 85,101–85,122), main food

TABLE 1

items likely to contain AR (cold cereals and dark bread) accounted for ;43–45% and 44–48% of the total WG and bran intakes, respectively (Table 3). Wheat was the predominant grain in cold cereal products rich in WG ($50% WG) (23). To evaluate the relative validity of creatinine-adjusted DHBA and DHPPA in spot urine samples as biomarkers of long-term WG and fiber intakes, we compared biomarker concentrations with intakes estimated from the FFQ. The median time between the 1995 FFQ and urine samples from occasion 1 and 2 was 2.0 (IQR: 1.7, 2.3) y and 3.7 (IQR: 3.6, 3.8) y, respectively. The corresponding figures between the 1999 FFQ and urine samples was 22 (IQR: 22.3, 21.8) y and 20.3 (IQR: 20.4, 20.2) y, respectively. In general, biomarker measurements (DHBA, DHPPA, and total AR metabolites) from both occasions and their mean were significantly correlated with WG and cereal and total fiber intake derived from both FFQ. Overall, DHPPA had higher correlations than DHBA with dietary variables (Table 4; Supplemental Table 1). As expected, the corrected correlation coefficients between intakes and urinary DHPPA concentrations adjusted for withinparticipant variation in the biomarker over the period 1997– 1999 were stronger than crude correlation coefficients (Table 4).

Discussion In this study, we examined the feasibility of using AR metabolites in spot urine samples as long-term biomarkers of WG, bran, and cereal fiber intakes in a U.S. population where WG intake is typically low and wheat, oats, rice, and corn are the primary sources of WG. This was accomplished through the evaluation of the long-term reproducibility (1–3 y) and by investigating the correlations of the intakes of WG and cereal fiber and the creatinine-adjusted AR metabolite concentration in spot urine samples. The evaluation described is necessary before applying AR metabolites in spot urine as biomarkers of WG and/ or fiber intake in epidemiological studies. Long-term WG, bran, germ, and fiber intakes. AR and their metabolites have mainly been evaluated as biomarkers in Nordic populations (8,18,19,28). Compared with Nordic populations where WG intake is high and mainly from AR-rich sources such as wheat and rye (29,30), participants in the present study had low WG intakes and wheat was the predominant source of AR, because rye intake is low in the US. Moreover, food products

AR metabolite concentrations, within- and between person variation and the long-term reproducibility in spot urine samples Urine collection1 First

DHBA, mmol/L DHPPA, mmol/L Total metabolites,4 mmol/L

Second

Mean (95% CI) 3.3* (2.9, 3.9) 4.2* (3.5, 4.9) 8.4 (7.2, 9.8) 9.0 (7.8, 10.4) 12.0 (10.3, 13.9) 13.4 (11.6, 15.5)

CV2within-person

CV2between-person

ICC3 (95% CI)

27.1 34.2 28.4

0.17 (0.05, 0.43) 0.31 (0.17, 0.51) 0.21 (0.08, 0.45)

% 58.8 50.1 54.1

Values are geometric mean (95% CI), n = 104. *Log-transformed creatinine-adjusted biomarker concentrations from 2 collections differed, P , 0.05 (paired t test). AR, alkylresorcinol; CVwithin, within-participant CV; CVbetween, between-participant CV; DHBA, 3, 5-dihydroxybenzoic acid; DHPPA, 3-(3,5-dihydroxyphenyl)-1-propanoic acid; ICC, intra-class correlation coefficient [defined as between-participant variance/ (total variance)]. 2 Variance components were estimated on log-transformed values by a random effect model. 3 Variance components were estimated on log-transformed values using a random effects model. Creatinine-adjusted metabolite concentrations were used. 4 The sum of DHBA and DHPPA. 1

874

Landberg et al.

TABLE 2

Estimated intake, within- and between-person variation, and long-term reproducibility of dietary fiber, WG, bran, and germ estimated by FFQ (distributed in 1995 and 1999) in participants in the Nurses’ Health Study II 1995 FFQ

Total dietary fiber, g/d Cereal fiber, g/d Total WG, g/d Bran,4 g/d Germ,5 g/d

1999 FFQ

Mean3 (95% CI) 19.0 (17.9, 20.2) 19.7 (18.6, 21.0) 5.8* (5.3, 6.3) 6.4* (5.8, 7.0) 15.9* (13.7, 18.4) 23.0* (20.0, 26.5) 3.8* (3.2, 4.6) 4.9* (4.2, 5.7) 0.8* (0.6, 1.0) 1.3* (1.1, 1.5)

CV1within-person

CV1between-person

ICC2

27.1 33.0 50.8 52.9 72.9

(95% CI) 0.76 (0.67, 0.84) 0.50 (0.35–0.65) 0.46 (0.31, 0.62) 0.40 (0.24, 0.58) 0.40 (0.25, 0.57)

% 15.0 32.6 54.1 64.6 88.3

1 Variance components were estimated on log-transformed values by a random effect model. *Reported intakes from 1995 and 1999 were significantly different (P , 0.05) when tested with Wilcoxon’s matched pairs signed rank sum test. CVwithin, within-participant CV; CVbetween, between-participant CV; ICC, intra-class correlation coefficient; WG, whole grain. 2 ICC defined as between-participant variance/(total variance). Variance components were estimated on log-transformed values by a random effect model. 3 Values are energy-adjusted geometric mean (95% CI), n = 84–104 (depending on variable). The residual method was used to calculate energy-adjusted nutrient intake variables (26). 4 The sum of natural bran + added bran. 5 The sum of natural germ + added germ.

anticipated to contain AR (cold breakfast cereals and dark bread) contributed to only ;45% of the WG intake over the years 1995–1999, whereas non-AR–containing sources such as popcorn, brown rice, tortillas, and oatmeal were the main

TABLE 3

Main foods contributing to daily WG, bran, germ, and total and cereal fiber intakes estimated by FFQ in 1995 and 1999 in the Nurses’ Health Study II1 1999 FFQ2

1995 FFQ % WG2 Cold breakfast Brown rice Dark bread Popcorn Bran2,3 Cold breakfast Oatmeal bran Dark bread Germ2,4 Cold breakfast Popcorn Tortillas Total fiber5 Cold breakfast Apple Potato Beans Cereal fiber5 Cold breakfast Pasta Dark bread English muffin Oatmeal bran Popcorn Bran muffin

cereals

31 18 14 10

24 12 19 18

cereals

39 32 9

31 33 13

cereals

25 20 18

17 32 12

cereals

8 6 3 4

7 5 6 5

cereals

23 12 9 9 6 3 —

20 10 12 7 7 8 7

n = 85,101–85,122, depending on variable. WG, whole grain. Foods that contributed to .10% in the 1999 FFQ were reported for both occasions. The sum of natural bran + added bran. 4 The sum of natural germ + added germ. 5 Foods that contributed to .5% in the 1999 FFQ were reported for both occasions. 1 2 3

contributors. The WG intake in Norwegian, Swedish, and Danish women was 44, 35, and 31 g/d, respectively, as estimated by validated FFQ, and the main sources of WG intake, wheat and rye, accounted for 73–90% of the total WG intake (29). AR metabolite concentrations and long-term reproducibility. To our knowledge, no previous study has investigated the creatinine-adjusted AR metabolite concentration in morning spot urine samples. However, DHPPA concentrations in overnight urine samples (12 h) were determined in a U.S. population of 100 men and women and were similar to those in our study (20). The poor to modest long-term reproducibility for creatinineadjusted DHBA and DHPPA is likely due to a relatively short apparent elimination half-life of these biomarkers and to the increase in WG intake over the period from 1995 to 1999 in our study population. To evaluate the impact of the anticipated increase in true WG intake between 1995 and 1999, we calculated the ICC for biomarkers after excluding 10% of the women who reported the largest increase in their WG intake over the period, with no difference in the results (data not shown). DHPPA has a slightly longer apparent elimination half-life (12 h) compared with DHBA (10 h), which partly may explain the difference in reproducibility compared to DHBA, although it is likely that other, so-far-unidentified factors may contribute to the difference as well (17). Moreover, DHPPA concentrations were typically higher and had a wider range than DHBA and this probably also contributed to a higher ICC. Interestingly, 2 recent studies reported higher reproducibility of (ICC = 0.45–0.55) intact AR in plasma conducted 2–4 mo apart in Swedish and German women despite the fact that the apparent elimination half-lives of AR in plasma are shorter compared with those of the metabolites (4,11). The lower reproducibility of AR metabolites in urine found in the present study may be due to the longer time period between measurements, incomplete correction of differences in urine dilution by creatinine, or enterohepatic circulation of AR metabolites and a shift in elimination of AR metabolites from urine to bile at higher intake levels. This would introduce variation in the biomarker concentrations that is not related to intake (15). Moreover, WG consumption was higher in the Swedish and German populations and may be more stable compared with the present population. Alkylresorcinol metabolites in U.S. women

875

TABLE 4

Observed and adjusted Spearman rank correlation coefficients (95% CI) between WG and fiber variables derived from FFQ (distributed in 1995 and 1999) and creatinine-adjusted urinary DHPPA concentration from the first and second spot urine samples and their mean from women in the Nurses’ Health Study II1 Urine collection

Food

FFQ

First

WG

1995 1999 1995 1999 1995 1999 1995 1999 1995 1999

0.37 (0.19,0.53) 0.31 (0.12, 0.47) 0.22 (0.02, 0.39) 0.13 (20.07,0.31) 0.31 (0.12, 0.48) 0.22 (0.03, 0.40) 0.40 (0.22, 0.55) 0.35 (0.17, 0.51) 0.33 (0.15, 0.50) 0.25 (0.06, 0.43)

Bran Germ Total fiber Cereal fiber

Second

0.25 0.17 0.29 0.10 0.17 0.06 0.34 0.40 0.30 0.15

(0.06, 0.43) (20.03, 0.35) (0.10, 0.46) (20.09, 0.29) (20.03, 0.35) (20.13, 0.25) (0.16, 0.50) (0.22, 0.55) (0.12, 0.47) (20.05, 0.33)

Mean rs (95% CI) 0.39 0.30 0.32 0.15 0.30 0.18 0.46 0.47 0.40 0.25

(0.22, 0.55) (0.11, 0.47) (0.13, 0.48) (20.05, 0.33) (0.11, 0.47) (20.02, 0.36) (0.30, 0.60) (0.30, 0.61) (0.22, 0.55) (0.06, 0.42)

Adjusted

0.60 (0.37, 0.76) 0.45 (0.18, 0.66) 0.48 (0.22, 0.68) 0.22 (20.07, 0.48) 0.45 (0.18, 0.66) 0.27 (20.02, 0.52) 0.71 (0.54, 0.83) 0.72 (0.56, 0.83) 0.61 (0.38, 0.77) 0.38 (0.10, 0.60)

1 Observed Spearman rank correlation coefficients were adjusted for within-participant variation in the biomarker using methods described by Rosner et al. (27). n = 95 (only participants who reported intakes in both 1995 and 1999 were included). DHPPA, 3-(3,5-dihydroxyphenyl)1-propanoic acid; WG, whole grain.

WG, bran, germ, and fiber intake in relation to AR metabolites in urine. We investigated the relative validity of AR metabolites as biomarkers of habitual WG intake as reflected by the FFQ. We realize that both the biomarker and the FFQ have limitations as measures of true WG intake, but because the measurement errors are likely uncorrelated, this comparison is still informative. We observed correlations between AR metabolites in spot urine samples and cereal fiber intakes estimated from FFQ that were of the same magnitude as those reported for AR metabolites in 24-h urine collections and cereal fiber intake estimated from 5-d weighed food records in Finnish men and women [r = 0.37, P , 0.01 (DHBA) and r = 0.41, P , 0.01 (DHPPA)] (18). The correlations were somewhat higher between the biomarkers and WG and cereal fiber intake when using intakes from 1995 compared with those from 1999 despite the shorter time between biomarker measurements and the 1999 FFQ. This may be due to lower intake of AR-containing WG foods and/or less reflection of such foods by the 1999 FFQ. For example, the intake of popcorn, which contains no AR, was higher in 1999 compared with 1995 and the intake of cold breakfast cereals (rich in WG wheat) was lower. As expected, deattenuated correlation coefficients between WG or fiber intake and DHPPA were considerably stronger when adjusting for within-participant in the biomarker, showing that a substantial part of unexplained variance in the association between WG intake and DHPPA is due to imprecision in the biomarker. The CI were broad because of the modest biomarker ICC, increased intake over the period, and the limited number of participants. Previous studies have shown that AR, the precursors of DHBA and DHPPA, are exclusively found in WG and the bran of wheat, rye, and barley (31). Low contents or traces have been measured in refined wheat (32) due to contamination of bran during the milling process, but this provides only a minor contribution to the plasma AR concentrations. Therefore, intact AR have been suggested and evaluated as a biomarker of WG and bran intake of wheat and rye (11). In a recent Finnish study, where AR intake was high, AR metabolite excretions in 24-h urine were significantly associated with cereal fiber intake but not with fruit or vegetable fiber (18). As a result of this finding, 876

Landberg et al.

urinary AR metabolites were also suggested to reflect cereal fiber (18). In the present study, where intake of AR was much lower, AR metabolites in spot urine samples were somewhat more strongly correlated with total fiber than with cereal fiber. We speculate that this may be due to minor dietary sources of DHBA and DHPPA in unknown plant-based foods, which subsequently lead to higher correlations between total fiber intake than to cereal fiber intake and urinary AR metabolites in populations with low intake of AR-rich foods. This is supported by the fact that small amounts of the AR metabolites DHBA and DHPPA are present in urine from participants reporting no intake of ARrich foods (15). In addition, consumption of WG food products with high WG and AR contents may be more likely in persons with higher fruit and vegetable fiber intakes. These persons may also better recognize such products and more accurately report their intake. We conclude that AR metabolites, particularly DHPPA in spot urine samples, had good relative validity as a biomarker of WG intake in this population. However, the long-term reproducibility was modest, indicating that multiple measurements of spot urine DHPPA will be more useful than single spot urine in epidemiological studies. The long-term reproducibility may be higher in populations with a higher WG wheat and rye intake and where intake is frequent and stable over time, such as in Scandinavian populations. Methods such as those proposed by Spiegelman et al. (33) can be used in dietary validation studies of long-term WG intake and may include this biomarker to enable the estimation of the attenuation factor taking into account correlated participant-specific biases and correlated random within-participant errors when 2 surrogate methods are compared. In addition, AR metabolites can be used as a measure of compliance in future dietary intervention studies of carbohydrate quality. Acknowledgments R.L. and R.M.v.D. conceived and designed the study; R.L. analyzed the AR metabolites in urine samples; R.L., M.K.T., and N.N. performed statistical analyses; R.L., M.K.T., N.N., Q.S., L.S., D.S., and R.M.v.D. contributed to the interpretation of the data and critically revised the manuscript for important

intellectual content; and R.L. and R.M.v.D. drafted the paper. All authors read and approved the final manuscript.

Literature Cited 1.

2.

3.

4.

5.

6.

7. 8.

9.

10.

11.

12.

13.

14.

15.

16.

Mellen PB, Walsh TF, Herrington DM. Whole grain intake and cardiovascular disease: a meta-analysis. Nutr Metab Cardiovasc Dis. 2008;18:283–90. Priebe MG, van Binsbergen JJ, de Vos R, Vonk RJ. Whole grain foods for the prevention of type 2 diabetes mellitus. Cochrane Database Syst Rev. 2008;CD006061. Jensen MK, Koh-Banerjee P, Franz M, Sampson L, Grønbaek M, Rimm EB. Whole grains, bran, and germ in relation to homocysteine and markers of glycemic control, lipids, and inflammation. Am J Clin Nutr. 2006;83:275–83. Montonen J, Landberg R, Kamal-Eldin A, A˚man P, Knueppel S, Boeing H, Pischon T. Reliability of fasting plasma alkylresorcinol concentrations measured 4 months apart. Eur J Clin Nutr. 2010;64:698–703. He M, van Dam RM, Rimm E, Hu FB, Qi L. Whole-grain, cereal fiber, bran, and germ intake and the risks of all-cause and cardiovascular disease-specific mortality among women with type 2 diabetes mellitus. Circulation. 2010;121:2162–8. Kelly SAM, Summerbell CD, Brynes A, Whittaker V, Frost G. Wholegrain cereals for coronary heart disease. Cochrane Database Syst Rev. 2007;CD005051. Bingham SA. Biomarkers in nutritional epidemiology. Public Health Nutr. 2002;5:821–7. Landberg R, Kamal-Eldin A, Andersson A, Vessby B, A˚man P. Alkylresorcinols as biomarkers of whole grain wheat and rye intake: plasma concentration and intake estimated from dietary records. Am J Clin Nutr. 2008;87:832–8. Schatzkin A, Kipnis V. Could exposure assessment problems give us wrong answers to nutrition and cancer questions? J Natl Cancer Inst. 2004;96:1564–5. Thane CW, Jones AR, Stephen AM, Seal JC, Jebb SA. Comparative whole-grain intake of British adults in 1986–7 and 2000–1. Br J Nutr. 2007;97:987–92. Andersson A, Marklund M, Diana M, Landberg R. Plasma alkylresorcinol concentrations correlate with whole grain wheat and rye intake and show moderate reproducibility over a 2–3 month period in freeliving Swedish adults. J Nutr. 2011;141:1712–8. Ross AB, Oineau N, Kochhar S, Bourgeois A, Beaumont M, Decarli B. Validation of a FFQ for estimating whole-grain cereal food intake. Br J Nutr. 2009;102:1547–51. Linko AM, Juntunen K, Mykkanen H, Adlercreutz H. Whole-grain rye bread consumption by women correlates with plasma alkylresorcinols and increases their concentration compared with low-fiber. J Nutr. 2005;135:580–3. Landberg R, Linko AM, Kamal-Eldin A, Vessby B, Adlercreutz H, A˚man P. Human plasma kinetics and relative bioavailability of alkylresorcinols after intake of rye bran. J Nutr. 2006;136:2760–5. Landberg R, A˚man P, Adlercreutz H, Vessby B, Kamal-Eldin A. Dose response of whole-grain biomarkers: alkylresorcinols in human plasma and their metabolites in urine in relation to intake. Am J Clin Nutr. 2009;89:290–6. So¨derholm PP, Koskela AH, Lundin JE, Tikkanen MJ, Adlercreutz HC. Plasma pharmacokinetics of alkylresorcinol metabolites: new candidate biomarkers for whole-grain rye and wheat intake. Am J Clin Nutr. 2009;90:1167–71.

17. So¨derholm PP, Lundin JE, Koskela AH, Tikkanen MJ, Adlercreutz HC. Pharmacokinetics of alkylresorcinol metabolites in human urine. Br J Nutr. 2011;106:1040–4. 18. Aubertin-Leheudre M, Koskela A, Marjamaa A, Adlercreutz H. Plasma alkylresorcinols and urinary alkylresorcinol metabolites as biomarkers of cereal fiber intake in Finnish women. Cancer Epidemiol Biomarkers Prev. 2008;17:2244–8. 19. Aubertin-Leheudre M, Koskela A, Samaletdin A, Adlercreutz H. Plasma alkylresorcinol metabolites as potential biomarkers of whole-grain wheat and rye cereal fibre intakes in women. Br J Nutr. 2010;103:339– 43. 20. Guyman LA, Adlercreutz H, Koskela A, Li L, Beresford SAA, Lampe JW. Urinary 3-(3,5-Dihydroxyphenyl)-1-propanoic acid, an alkylresorcinol metabolite, is a potential biomarker of whole-grain intake in a U. S. population. J Nutr. 2008;138:1957–62. 21. Marklund M, Landberg R, A˚man P, Kamal-Eldin A. Determination of alkylresorcinol metabolites in human urine by gas chromatographymass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci. 2010;878:888–94. 22. Marklund M, Landberg R, A˚man P, Kamal-Eldin A. Comparison of gas chromatography-mass spectrometry and high-performance liquid chromatography with coulometric electrode array detection for determination of alkylresorcinol metabolites in human urine. J Chromatogr B Analyt Technol Biomed Life Sci. 2011;879:647–51. 23. Franz M, Sampson L. Challenges in developing a whole grain database: definitions, methods and quantification. Journal of Food Composition and Analysis. 2006;19:S38–44. 24. AACC International. Definition of whole grain; 1999 [cited 2011 Sep 9]. Available from: www.aaccnet.org/definition/wholegrain.asp. 25. Hankinson SE, Manson JE, Spiegelman D, Willett WC, Longcope C, Speizer FE. Reproducibility of plasma hormone levels in postmenopausal women over a 2–3-year period. Cancer Epidemiol Biomarkers Prev. 1995;4:649–54. 26. Willett W, Stampfer MJ. Total energy intake: implications for epidemiologic analyses. Am J Epidemiol. 1986;124:17–27. 27. Rosner B, Glynn RJ. Interval estimation for rank correlation coefficients based on the probit transformation with extension to measurement error correction of correlated ranked data. Stat Med. 2007;26:633–46. 28. Landberg R, Kamal-Eldin A, A˚man P, Christensen J, Overvad K, Tjønneland A, Olsen A. Determinants of plasma alkylresorcinol concentration in Danish post-menopausal women. Eur J Clin Nutr. 2011;65:94–101. 29. Kyrø C, Skeie G, Dragsted L, Christensen J, Overvad K, Hallmans G, Johansson I, Lund E, Slimani N, Johnsen NF, et al. Intake of wholegrain in Scandinavia- intake, sources and recommendations. Scand J Public Health. 2012;40:76–84. 30. Mejborn H, Biltoft-Jensen A, Trolle E, Tetens I. Fuldkorn. Definition og vidensgrundlag for anbefaling af fuldkornsindtag i Danmark. [Whole grain. Definition and scientific evidence for recommendation of whole grain intake in Denmark]. DTU Fødevareinstituttet, Afdeling for Ernæring; 2008 [cited 2011 June 22]. Available from: http://www. food.dtu.dk/Default.aspx?ID=8319. 31. Landberg R, Kamal-Eldin A, Salmenkallio-Martilla M, Rouau X, A˚man P. Localization of alkylresorcinols in barley, wheat and rye kernels. J Cereal Chem. 2008;48:401–6. 32. Ross AB, Kochhar S. Rapid and sensitive analysis of alkylresorcinols from cereal grains and products using HPLC2coularray-based electrochemical detection. J Agric Food Chem. 2009;57:5187–93. 33. Spiegelman D, Zhao B, Kim J. Correlated errors in biased surrogates: study designs and methods for measurement error correction. Stat Med. 2005;24:1657–82.

Alkylresorcinol metabolites in U.S. women

877