1 From the Jean Mayer US Department of Agriculture Human Nutrition

Transcription

1 Circulating unmetabolized folic acid and 5-methyltetrahydrofolate in relation to anemia, macrocytosis, and cognitive test performance in American seniors 1 4 Martha Savaria Morris, Paul F Jacques, Irwin H Rosenberg, and Jacob Selhub ABSTRACT Background: Folate deficiency has serious consequences for the fetus. Folic acid fortification of food addresses this problem. However, clinical consequences of vitamin B-12 deficiency may be worsened by high folic acid intakes, perhaps as a direct result of unmetabolized folic acid, which does not occur naturally in body tissues. Objective: We attempted to attribute associations that we previously found between higher folate status and anemia and cognitive test performance to circulating unmetabolized folic acid or 5-methyltetrahydrofolate (5MeTHF). Design: The subjects (n = 1858) were senior participants in the US National Health and Nutrition Examination Survey ( ) who had normal renal function and reported no history of stroke, recent anemia therapy, or diseases of the liver, thyroid, or coronary arteries. Subjects had undergone a phlebotomy, a complete blood count, and cognitive and dietary assessments. Results: Circulating unmetabolized folic acid was detected in 33% of the subjects and was related to an increased odds of anemia in alcohol users. In seniors with a serum vitamin B-12 concentration,148 pmol/l or a plasma methylmalonic acid concentration 210 nmol/l, the presence compared with the absence of detectable circulating unmetabolized folic acid was related to lower cognitive test scores and lower mean cell volume. In the same subgroup, higher serum 5MeTHF was related to an increased odds of anemia and a marginally significantly decreased odds of macrocytosis. In seniors with a normal vitamin B-12 status, a higher serum 5MeTHF concentration was related to higher cognitive test scores. Conclusion: Results of this epidemiologic study were somewhat consistent with reports on the folic acid treatment of patients with pernicious anemia, but some findings were unexpected. Am J Clin Nutr 2010;91: INTRODUCTION Government-mandated folic acid fortification of enriched cereal grain products sold in the United States officially took effect in January 1998 (1). This intervention, which was aimed at reducing the incidence of neural tube defects, resulted in dramatic shifts in the distributions of folate intakes (2) and biochemical measures of folate status (3) in Americans. Controversy surrounds fortification, even now, because of the unknown consequences of possible long-term exposure to supraphysiologic folate concentrations and unmetabolized folic acid at any age (4 9). The elderly have been of particular concern because of their high risk of vitamin B-12 deficiency (10, 11) and the fear that folic acid might improve the blood picture (12 20) and either precipitate neurologic consequences (14, 16, 20 22) or allow neurologic function to deteriorate unchecked (14 19, 23). To avert such a scenario, the fortification level of 140 lg/100 g cereal grain product was selected to limit the regular daily consumption of total folic acid at no more than 1.0 mg from all sources combined (24). This upper safe intake level was derived by applying an uncertainty factor of 5 to the lowest observed adverse effect level of 5 mg/d (24). However, folic acid was predicted to appear unaltered in the circulation even by those adhering to the 1.0 mg/d rule if much of the 1.0 mg came from supplements (25). Indeed, studies conducted in the postfortification era in the United States, where supplement use is common, showed circulating unmetabolized folic acid in the vast majority of subjects (26, 27). Circulating unmetabolized folic acid implies that the body s capacity to convert folic acid to the metabolically active 5-methyltetrahydrofolate (5MeTHF) has been overwhelmed (28, 29) and that folic acid has passively diffused intact into the circulation (24). This may have adverse effects on persons who are vitamin B-12 deficient (21, 30). Using data collected in the National Health and Nutrition Examination Survey (NHANES), we recently investigated the interaction between vitamin B-12 status and the supraphysiologic folate status achievable in the age of food folate fortification and heavy supplement use in relation to anemia, macrocytosis, and cognitive impairment (31). Consistent with Reynolds s predictions (32, 33), we found a higher odds of 1 From the Jean Mayer US Department of Agriculture Human Nutrition Research Center on Aging at Tufts University, Boston, MA. 2 This material is based upon work supported by the US Department of Agriculture, Agricultural Research Service, under agreement no Any opinions, findings, conclusion, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the US Department of Agriculture. 3 Supported by USDA agreement no and USDA grant Address correspondence to MS Morris, Nutritional Epidemiology Program, Jean Mayer USDA Human Nutrition Research Center on Aging at Tufts University, 711 Washington Street, Room 901D, Boston, MA martha.morris@tufts.edu. Received September 14, Accepted for publication March 3, First published online March 31, 2010; doi: /ajcn Am J Clin Nutr 2010;91: Printed in USA. Ó 2010 American Society for Nutrition 1733

2 1734 MORRIS ET AL anemia and cognitive impairment in association with high (compared with lower) folate status in persons with low vitamin B-12 status. However, because only serum total folate was measured, the role that unmetabolized folic acid played in the results was unclear. In the present investigation, we assayed 5MeTHF and unmetabolized folic acid in blood samples obtained from NHANES participants and considered circulating concentrations of these 2 components of serum total folate and their interaction with vitamin B-12 status in relation to anemia, macrocytosis, and cognitive test performance. SUBJECTS AND METHODS Study population NHANES monitors the nation s health and nutritional status. The survey is currently implemented as a continuous annual survey and uses a complex multistage probability design to select a representative sample of the noninstitutionalized US civilian population. To increase the precision of estimates derived from the survey, adolescents, the elderly, Mexican Americans, and blacks are oversampled. The protocols for conducting the NHANES were approved by the Institutional Review Board of the National Center for Health Statistics, Centers for Disease Control and Prevention, and informed consent was obtained from all participants (34). Consistent with NHANES analytic guidelines (35), we combined data from the 2 most recent surveys in which cognitive data were collected into a single data set ( ). Trained interviewers used a computer-assisted personal interview system to interview participants in their homes. The participants were also asked to report to a mobile examination center (MEC) after a 6- or 9-h fast to provide further interview data and undergo a physical examination that included phlebotomy. A detailed description of blood collection and processing can be found in the NHANES Laboratory Procedures Manual (36). Although the survey included persons of all ages, we focused our attention on seniors (ie, those aged 60 y) the only group whose cognitive function was assessed. We excluded seniors with serum creatinine concentrations indicative of renal dysfunction (ie,.131 lmol/l in men and.115 lmol/l in women) and those who reported recent anemia therapy or a history of stroke or diseases of the liver, thyroid, or coronary arteries. Of 3706 senior survey participants, 1100 were excluded for these reasons, and eligibility status could not be determined for 545. Surplus serum samples for the measurement of serum 5MeTHF and serum folic acid were available for 1866 of the 2061 nonexcluded subjects, and vitamin B-12 status could be determined for 1858 of these subjects. The sample included 247 subjects whose cognitive function was not tested, and an additional 4 provided no educational data. Subjects with complete covariate data for analyses pertaining to cognition numbered Subjects with complete covariate data for analyses pertaining to hematologic abnormality numbered Assessment of cognitive function The cognitive function of seniors was assessed by using a version of the Digit-Symbol Substitution Test (DSST) of the Wechsler Adult Intelligence Scale III a screening test designed to detect cognitive impairment in adults and children (37). In the test, participants copy symbols that are paired with numbers. Using the key provided at the top of the exercise form, the participant draws the symbol under the corresponding number. The score, which declines with age (38), is the number of correct symbols drawn within 120 s. One point is given for each correctly drawn symbol completed within the time limit for a maximum score of 133. Use of the test in the NHANES was based on its reputation as a more sensitive measure of dementia than the Mini-Mental State Examination (39). According to NHANES documentation, aptitudes needed for a high score are response speed, sustained attention, visual spatial skills, associative learning, and memory (39). However, research suggests that information-processing speed is the prime determinant of performance on the test (38). For our main data analyses, we used the score on the test as a continuously scaled term. However, for the purpose of presenting results of preliminary data analyses, we classified subjects as having performed poorly or well using a score of 34 the 20th percentile of the distribution as the cutoff between the 2 categories. Assessment of anemia and macrocytosis Anemia and macrocytosis were assessed from results of the complete blood count. Anemia was based on hemoglobin concentrations that were measured at the MEC laboratory by using a MAXM hematology flow cytometer (Beckman Coulter Inc, Fullerton, CA). We defined anemia according to World Health Organization criteria (ie, hemoglobin,12 g/dl for women and,13 g/dl for men) (40). We defined macrocytosis as a mean cell volume 99 fl, and we also used mean cell volume as a continuously scaled outcome variable in some analyses. Biochemical measurements Detailed methods for all the biochemical assays can be found in the NHANES Laboratory Procedures Manuals (41). Most biochemical analyses were performed at the Inorganic Toxicology and Nutrition Branch of the Division of Laboratory Sciences, National Center for Environmental Health. These priority analyses included measurement of serum concentrations of vitamin B-12 and folate, which were carried out by using the Quantaphase II Radioassay Kit (Bio-Rad Laboratories, Anaheim, CA). Plasma methylmalonic acid (MMA) was measured by gas chromatography mass spectrometry with cyclohexanol derivatization. Plasma homocysteine was analyzed by using a commercially available fluorescence polarization immunoassay kit (Abbott Laboratories, Abbott Park, IL) on the Abbott IMx analyzer. Serum ferritin was measured by using the QuantaImmune Ferritin IRMA Kit (Bio-Rad Laboratories). Serum C-reactive protein and cystatin C were quantified by particle-enhanced nephelometry, and serum creatinine concentration was based on the Jaffe reaction. Cystatin C, folic acid, and 5MeTHF were measured as surplus serum projects. That is, these assays were conducted on serum samples that were initially used for NHANES priority analyses, but were subsequently stored at 270 C specifically to provide samples for future research projects approved by the Division of Health and Nutrition Examination Surveys of the National Center for Health Statistics (42). Before our use of the samples, they had been through 2 freeze-thaw cycles. Multiple freeze/

3 FOLATE DEFICIENCY AND COGNITIVE FUNCTION 1735 thaw cycles have been shown not to cause a noticeable loss of folates (43). We measured folic acid and 5MeTHF at the Jean Mayer Human Nutrition Research Center on Aging (HNRCA) at Tufts University using a modification of the affinity/hplc with electrochemical (coulometric) detection method previously developed at the HNRCA (27). Although serum total folate includes not only folic acid and 5MeTHF, but also 5-formyl-THF (43), which we did not measure, we used the terms radioassaydetermined serum total folate and HPLC-determined serum total folate in this report to distinguish the radioassay-determined values from the sum of the HPLC-determined folic acid and 5MeTHF concentrations. Classification of subjects according to vitamin B-12 status and folate fractions We defined low vitamin B-12 status as either a serum vitamin B-12 concentration below the conventionally applied cutoff for deficiency (ie, 148 pmol/l) or an elevated plasma MMA concentration. We defined elevated plasma MMA as a value 210 nmol/l, based on the published reference range for serum vitamin B-12 replete survey participants with normal serum creatinine concentrations (44). We divided subjects into 2 categories based on the presence or absence of detectable circulating unmetabolized folic acid. The limit of detection was nmol/l. For our main data analyses, we used serum 5MeTHF as a continuously scaled term. However, for the purpose of presenting results of preliminary data analyses, we classified subjects as having high or lower serum 5MeTHF concentrations using 50 nmol/l the 60th percentile of the distribution as the cutoff between the 2 categories. Folate status this high can be considered supraphysiologic in the sense that it would not occur from intakes of natural food folate alone. Serum total folate concentrations.50 nmol/l were observed in only 5% of participants in the prefortification NHANES III. Folate intake data Daily intakes of folic acid and food folate were available beginning with the NHANES. Therefore, dietary data are presented for those 2 y only. The dietary data were collected in a single 24-h dietary-recall interview administered by trained staff during the MEC examination (45). The US Department of Agriculture (USDA) was responsible for the survey s dietary data collection methods, maintenance of the databases used to code and process the data, and data review and processing. The USDA National Nutrient Database for Standard Reference Dietary Studies (version 1) was used to calculate daily nutrient intakes from food, including folic acid intake from fortified foods (46). Data on dietary supplement use were collected during the inhome interview (47). Subjects were asked whether they had used any vitamins, minerals, or other dietary supplements within 30 d of the interview. Subjects who had used such products were asked to show the interviewer the supplement containers and provide information on the amount, frequency, and duration of use. The amount of each ingredient in each product used was determined by matching the name and manufacturer of the supplement to those in a database developed by the National Center for Health Statistics in collaboration with the National Institutes of Health s Office of Dietary Supplements. The information in the database came from sources such as the manufacturer or retailer, the Internet, and company catalogs. We used the various dietary supplement files to identify all subjects who reported using any supplemental source of folic acid and to determine each such subject s average daily intake of folic acid from supplements. This amount was added to the amount from food to yield the daily folic acid intake for subjects who used supplemental folic acid. Statistical analyses Data analyses were performed by using SUDAAN release 9.0 (Research Triangle Institute, Research Triangle Park, NC) with appropriate sampling weights, pseudo primary sampling units, and stratification variables to account for the survey s complex sampling design (35). P, 0.05 was considered statistically significant for all tests. The main goal of preliminary data analyses was to identify correlates of the exposures and health problems of interest in this study. To accomplish this aim, we used a multiple linear regression analysis program (SUDAAN PROC REGRESS), with subject characteristics as continuously scaled outcome variables. Predictors in the model included age (continuous), sex, raceethnicity (ie, non-hispanic white, non-hispanic black, Mexican- American, Other Hispanic, and other), and the exposure or health problem of interest (dichotomous). For categorical characteristics, the continuously scaled outcome variable was created by assigning a 1 to each person with the characteristic and by assigning a 0 to each person who did not have the characteristic. The multivariate-adjusted least-squares means generated by the regression program were reported as means (for continuously scaled characteristics) and percentages (for categorical characteristics). The statistical significance of the association between each characteristic and the exposure or health problem of interest was assessed by using the P value from the t test output by the program. These P values were not adjusted for multiple comparisons. To graphically illustrate the relation between folic acid intake and the appearance of unmetabolized folic acid in the circulation, we plotted the multivariate-adjusted proportion 95% CI with detectable circulating unmetabolized folic acid for quintile categories of folic acid intake. The outcome variable in the multivariate linear regression model used for this purpose was the presence compared with absence of detectable circulating unmetabolized folic acid ( 0/1 coding), and the analysis was adjusted for sex, age, and race-ethnicity. The goal of the main data analyses was to evaluate the interaction between folate fractions and vitamin B-12 status in relation to anemia, macrocytosis, and cognitive test performance and to estimate the strength of associations between the 2 folate fractions and those health problems in subjects stratified by vitamin B-12 status. Results of the preliminary analyses described above were used to identify candidate terms for the multiple linear and logistic regression models used to accomplish these goals. All analyses were adjusted for sex, age, race-ethnicity, current smoking status, serum cystatin C, serum C-reactive protein, and an alcohol-intake variable (ie, current alcohol intake for anemia and macrocytosis and self-reported history of alcohol abuse for cognitive test performance). Analyses focused on

4 1736 MORRIS ET AL anemia were additionally adjusted for triceps skinfold thickness and self-reported diabetes status. Analyses focused on macrocytosis were additionally adjusted for body mass index, cancer history, and serum ferritin. Analyses focused on cognitive test performance were additionally adjusted for educational achievement and self-reported history of cancer and diabetes. To facilitate comparison between the results of our current investigation, in which folate fractions were measured by using HPLC, and the findings we aimed to clarify, which were based on radioassay-determined serum total folate, we also considered how vitamin B-12 status interacted with both radioassay-determined serum total folate and HPLC-determined serum total folate in relation to the odds of anemia, macrocytosis, and a low compared with a higher DSST score. We conducted these analyses by using multiple logistic regression as performed by SUDAAN PROC RLOGIST, and we defined high serum total folate as a value above the 75th percentile (ie, radioassay: 55 nmol/l; HPLC: 66 nmol/l). To graphically illustrate the trend in DSST scores with increasing 5MeTHF concentration in subjects with normal vitamin B-12 status, we used SUDAAN PROC REGRESS to estimate least-squares mean (95% CI) DSST scores for quintile categories of serum 5MeTHF using the multivariate model described above for this outcome. RESULTS Subject characteristics By design, the subjects represented American seniors. Their mean age was 70 y, 15% smoked, 50% used alcohol, 66% used dietary supplements, 43% were educated beyond the 12th grade level, 14% reported a previous diagnosis of diabetes, 14% had a history of cancer, 4% were anemic, and 4% had macrocytosis. We detected unmetabolized folic acid in almost one-third of the surplus serum samples, and though values were as high as 111 nmol/l, 90% of values were,1.78 nmol/l. Of the 457 subjects with detectable circulating unmetabolized folic acid, folic acid accounted for, on average, 6% of the serum total folate (interquartile range: 1 8%). Ten percent of the seniors consumed.1 mg total folate/d. Furthermore, folic acid intake data available for the last 2 y of the study showed that 4% exceeded the tolerable upper intake level of 1 mg folic acid/d (24). About 25% of the seniors with folic acid and 5MeTHF measurements met our definition of low vitamin B-12 status (ie, serum vitamin B-12,148 pmol/l or plasma MMA 210 nmol/l), although the prevalence of a serum vitamin B-12 concentration,148 pmol/l was only 3%. Subject characteristics in relation to vitamin status and adverse health outcomes Many significant P values are shown in Tables 1 and 2. However, some of the apparent associations may be due to chance or confounding by factors other than sex, age, and raceethnicity. The only subject characteristic related to detectable circulating unmetabolized folic acid at P, 0.01, other than indicators of folate status and folate/folic acid intake, was Mexican American race-ethnicity, which was associated with protection against the exposure (Table 1). The prevalence of detectable circulating unmetabolized folic acid increased from 20% of subjects who consumed,84 lg folic acid/d to 50% of those with daily folic acid intakes 584 lg/d (Figure 1). The presence compared with the absence of detectable circulating unmetabolized folic acid remained significantly associated with Mexican American race-ethnicity at P, 0.01 even after folic acid intake was controlled for (data not shown). In contrast with the findings for detectable circulating unmetabolized folic acid, a high serum 5MeTHF concentration was strongly related to several correlates of good health, including female sex, non-hispanic white race-ethnicity, higher educational achievement, lower BMI, nonsmoking status, a low prevalence of vitamin B-12 deficiency, and lower circulating concentrations of cystatin C, homocysteine, and C-reactive protein. Despite the connection between aging and ill health, a high serum 5MeTHF concentration was linked to older age. Most of the characteristics that were significantly associated with the serum 5MeTHF concentration in these analyses were not related to the 5MeTHF concentration when a term for folate intake was in the multivariate model; however, age, sex, and vitamin B-12 status were exceptions to this rule (data not shown). Low vitamin B-12 status was associated with several indicators of poor health, including higher age, cigarette smoking, poor cognitive test performance, an elevated prevalence of anemia, and high circulating concentrations of cystatin C, creatinine, and homocysteine. Low vitamin B-12 status was also related to lower folate status and lower intakes of total folate, folic acid, and supplements containing folic acid. All 3 of the adverse health outcomes of interest in this investigation were associated with higher age (Table 2). Subjects with anemia and low DSST scores were also more likely than other subjects to be non-hispanic black. Poor cognitive test performance was additionally associated with Mexican-American race-ethnicity, low educational achievement, lower triceps skinfold thickness, cigarette smoking, a history of alcohol abuse, higher homocysteine concentration, higher cystatin C concentration, low vitamin B-12 status, lower folate status, and lower intakes of folate, folic acid, and supplements containing folic acid. The significant relation between female sex and protection against poor DSST performance was eliminated when triceps skinfold thickness was controlled for. Furthermore, homocysteine was not related to cognitive test performance (P = 0.936) when serum cystatin C and vitamin B-12 status were controlled for. In this older sample, when the demographics-adjusted model was used, anemia was not related at P, 0.01 to any subject characteristics other than higher age, non-hispanic black compared with non-hispanic white race-ethnicity, and low compared with normal vitamin B-12 status. It is worth mentioning, however, that, in a model including terms for these factors plus selfreported diabetes status and serum C-reactive protein, all factors remained significantly related to anemia, and P values for higher age, non-hispanic black compared with non-hispanic white race-ethnicity, low compared with normal vitamin B-12 status, and higher serum C-reactive protein were all, In comparison with the rest of the sample, subjects who met our definition of macrocytosis were not only older than other subjects, but also had a lower BMI and triceps skinfold thickness and higher serum ferritin concentrations. Moreover, they were much more likely to smoke and use alcohol. In multivariate

5 FOLATE DEFICIENCY AND COGNITIVE FUNCTION 1737 TABLE 1 Characteristics of senior participants in the National Health and Nutrition Examination Survey ( ) in relation to circulating folic acid concentrations, high 5-methyltetrahydrofolate (5MeTHF) concentrations, and low vitamin B-12 status 1 Circulating folic acid 2 5MeTHF (nmol/l) 2 Vitamin B-12 status 3 Characteristic Yes (n = 457) No (n = 1401) P.50 (n = 676) 50 (n = 1182) P Low (n = 456) Normal (n = 1402) P Female sex [n (%)] 202 (60) 692 (56) (66) 554 (51), (56) 731 (57) Age (y) 70.5 (70, 72) 70.2 (70, 71) (70, 72) 69.7 (69, 70) (72, 74) 69.5 (69, 70),0.001 Race-ethnicity [n (%)] Non-Hispanic White 310 (84) 699 (79) (86) 555 (76), (82) 730 (80) Black 83 (7.6) 239 (8.4) (3.8) 262 (11), (6.1) 266 (8.8) Mexican American 35 (1) 369 (4.1), (2.5) 288 (3.7) (2.8) 320 (3.4) Education [n (%)],High school 152 (26) 647 (31) (26) 567 (32) (34) 585 (28) High school 113 (26) 311 (30) (25) 275 (32) (31) 309 (29) High school 190 (48) 438 (39) (49) 338 (36), (35) 501 (44) BMI (kg/m 2 ) 28 (27.6, 28.5) 28.2 (27.8, 28.5) (26.9, 27.9) 28.7 (28.2, 29.1) (27.8, 29.1) 28 (27.6, 28.4) Triceps skinfold thickness (mm) 19.7 (19, 20.5) 19.1 (18.7, 19.6) (18.4, 19.8) 19.4 (18.9, 19.9) (17.8, 19.8) 19.5 (19, 19.9) 0.2 Cystatin C (mg/l) (0.97, 1) 0.98 (0.96, 1) (0.94, 0.98) 1 (0.98, 1.02) (1.05, 1.1) 0.96 (0.94, 0.97),0.001 Creatinine (lmol/l) 77 (75, 79) 76 (74, 78) (73, 77) 77 (75, 79) (80, 84) 74 (73, 76),0.001 Smoker [n (%)] 69 (15) 228 (16) (13) 214 (18) (22) 213 (14) Past alcohol abuser [n (%)] 54 (11) 204 (13) (11) 194 (13) (13) 192 (12) Alcohol user [n (%)] 217 (51) 647 (49) (53) 542 (47) (45) 668 (51) Hypertension [n (%)] (40) 631 (46) (42) 524 (47) (45) 605 (45) Diabetes [n (%)] 78 (16) 234 (13) (13) 205 (14) (16) 235 (13) Cancer history [n (%)] 59 (14) 163 (14) (14) 134 (13) (13) 163 (14) Homosysteine (lmol/l) (8.7, 9.5) 9.3 (9.1, 9.6) (8.2, 8.8) 9.8 (9.6, 10.1), (10.8, 11.6) 8.7 (8.5, 8.9),0.001 Ferritin (lg/l) 6 97 (90, 105) 106 (97, 115) (89, 108) 107 (97, 118) (87, 113) 104 (99, 110) C-reactive protein (mg/dl) (0.22, 0.27) 0.27 (0.26, 0.29) (0.22, 0.26) 0.28 (0.26, 0.3) (0.21, 0.29) 0.27 (0.25, 0.28) Low vitamin B-12 status [n (%)] (24) 345 (25) (17) 328 (30), (100) 0 (0) 1.0 Anemia [n (%)] 7 39 (6.6) 85 (4.4) (5) 85 (5) (8.1) 77 (4),0.001 Macrocytosis [n (%)] 8 18 (3.1) 60 (4.3) (3.9) 49 (4.1) (6) 43 (3.3) Low DSST score [n (%)] (22) 408 (21) (18) 361 (23) (31) 358 (18), MeTHF (nmol/l) 2,6 48 (45, 50) 38 (36, 40), (67, 70) 28 (27, 29), (33, 39) 43 (41, 44),0.001 Folate intake,400 lg/d [n (%)] 161 (30) 743 (48), (20) 730 (59), (51) 662 (40) Folic acid user [n (%)] 237 (58) 460 (39), (72) 241 (25), (30) 567 (49),0.001 Food folic acid (lg/d) 6, (305, 376) 217 (183, 258), (370, 485) 174 (152, 198), (175, 245) 275 (242, 312) Values are n (%) or means (95% CIs), except where otherwise indicated. Subjects were men and women aged 60 y; reasons for exclusion were a serum creatinine concentration indicative of frank kidney dysfunction and self-report of any of the following: recent anemia therapy, coronary artery disease, liver disease, and thyroid disease. DSST, Digit-Symbol Substitution Test. All means, proportions, and P values were derived by using a multiple linear regression model. P values were derived from t tests. For characteristics other than sex, age, and race-ethnicity, the results were controlled for sex, age, and race-ethnicity. The results for age were controlled for sex and race-ethnicity. The results for sex were controlled for age and race-ethnicity. The results for race-ethnicity were controlled for sex and age. 2 Unmetabolized folic acid and 5MeTHF concentrations in serum were measured by using affinity/hplc with electrochemical (coulometric) detection. Yes indicates that unmetabolized folic acid was detected in serum (detection limit = nmol/l). 3 Low vitamin B-12 status was defined as a serum vitamin B-12 concentration,148 pmol/l or a serum methylmalonic acid concentration 210 nmol/l. All other subjects were considered to have normal status. 4 Values in this row were back transformed from least-squares means generated by using the inverse of the serum cystatin C concentration as the dependent variable in the multivariate model. 5 Hypertension was defined as systolic blood pressure.140 mm Hg. 6 Values in this row are geometric means; 95% CIs in parentheses. 7 Defined as a hemoglobin concentration,12 g/dl (women) or,13 g/dl (men). 8 Mean cell volume 99 fl. 9 A score,34 out of 133 points (20th percentile) was considered low. 10 The folic acid content of food was determined only for foods reported by participants in the survey. These values were additionally adjusted for food folate.

6 1738 MORRIS ET AL TABLE 2 Characteristics of senior participants in the National Health and Nutrition Examination Survey ( ) in relation to anemia, macrocytosis, and cognitive test performance 1 Anemia 2 Macrocytosis 3 DSST score 4 Characteristic Yes (n = 124) No (n = 1732) P Yes (n = 78) No (n = 1778) P,34 (n = 522) 34 (n = 1089) P Female sex [n (%)] 54 (40) 909 (58) (46) 932 (57) (51) 599 (59) Age (y) 73 (72, 75) 70 (70, 71), (71, 75) 70 (70, 71) (73, 75) 69 (68, 70),0.001 Race-ethnicity [n (%)] Non-Hispanic White 48 (60) 960 (81), (81) 959 (80) (61) 750 (88),0.001 Black 56 (29) 266 (7.1), (8.7) 309 (8.1) (17) 134 (4.5),0.001 Mexican American 13 (1.8) 390 (3.3) (2.3) 393 (3.2) (7.6) 147 (1.6) Education [n (%)],High school 69 (38) 728 (29) (36) 763 (29) (59) 248 (19),0.001 High school 23 (22) 401 (30) (25) 408 (29) (25) 314 (31) High school 31 (40) 597 (41) (39) 600 (41) (16) 526 (50),0.001 BMI (kg/m 2 ) 28.9 ( ) 28.1 (27.8, 28.4) (25.2, 27.6) 28.2 (27.9, 28.5) (27.2, 28.8) 28.1 (27.7, 28.5) Triceps skinfold thickness (mm) 19 (17.3, 20.7) 19.3 (18.9, 19.7) (15.9, 18.4) 19.4 (19, 19.8), (17.6, 19.1) 19.6 (19.2, 20.1) Cystatin C (mg/l) (1, 1.11) 0.98 (0.96, 0.99) (0.96, 1.07) 0.98 (0.97, 1) (0.99, 1.05) 0.97 (0.95, 0.98),0.001 Creatinine (lmol/l) 79 (74, 85) 76 (75, 78) (74, 82) 76 (75, 78) (74, 79) 76 (74, 78) Smoker [n (%)] 11 (8.1) 286 (16) (44) 268 (14), (24) 155 (13) Past alcohol abuse [n (%)] 26 (18) 232 (12) (17) 245 (12) (17) 119 (10) Alcohol user [n (%)] 50 (50) 814 (50) (69) 820 (49), (42) 559 (52) Hypertension [n (%)] 6 63 (47) 757 (44) (48) 778 (44) (48) 454 (43) Diabetes [n (%)] 41 (29) 271 (13) (7.8) 303 (14) (17) 150 (12) Cancer history [n (%)] 23 (20) 199 (13) (26) 204 (13) (15) 141 (13) 0.51 Homocysteine (lmol/l) (9.2, 10.4) 9.2 (9, 9.5) (9.2, 11.1) 9.2 (9, 9.4) (9.3, 10.2) 9 (8.8, 9.3) Ferritin (lg/l) 7 76 (55, 104) 105 (98, 112) (112, 201) 102 (95, 108) (95, 121) 100 (92, 107) C-reactive protein (mg/dl) (0.27, 0.5) 0.26 (0.25, 0.27) (0.24, 0.52) 0.26 (0.25, 0.27) (0.23, 0.33) 0.25 (0.24, 0.27) Low vitamin B-12 status [n (%)] 8 47 (40) 409 (24), (36) 421 (24) (37) 217 (21),0.001 Anemia [n (%)] (100) 0 (0) (7) 115 (4.9) (6.3) 48 (4) Macrocytosis [n (%)] 3 9 (5.6) 69 (3.9) (100) 0 (0) (3.1) 41 (3.6) Low DSST score [n (%)] 4 46 (29) 475 (21) (19) 499 (21) (100) 0 (0) 1.0 5MeTHF (nmol/l) 7,9 42 (36, 49) 41 (39, 42) (35, 45) 41 (39, 43) (33, 41) 42 (41, 44) Folate intake, 400 lg/d [n (%)] 66 (46) 837 (43) (55) 861 (42) (55) 438 (38),0.001 Folic acid user [n (%)] 40 (39) 656 (45) (44) 667 (44) (35) 491 (49),0.001 Folic acid intake (lg/d) 7, (140, 284) 261 (230, 296) (173, 267) 259 (230, 291) (155, 257) 278 (244, 315) Values are n (%) or means (95% CIs), except where otherwise indicated. Subjects were men and women aged 60 y; reasons for exclusion were a serum creatinine concentration indicative of frank kidney dysfunction and self-report of any of the following: recent anemia therapy, coronary artery disease, liver disease, and thyroid disease. 5-MeTHF, 5-methyltetrahydrofolate; DSST, Digit-Symbol Substitution Test. All means, proportions, and P values were obtained by using a multiple linear regression model. P values were obtained from t tests. For characteristics other than sex, age, and race-ethnicity, the results were controlled for sex, age, and race-ethnicity. The results for age were controlled for sex and race-ethnicity. The results for sex were controlled for age and race-ethnicity. The results for raceethnicity were controlled for sex and age. Anemia and macrocytosis data were missing for 2 subjects. 2 Defined as a hemoglobin concentration,12 g/dl (women) or,13 g/dl (men). 3 Mean cell volume 99 fl. 4 A score,34 out of 133 points (20th percentile) was considered low. 5 Values in this row were back transformed from least-squares means generated by using the inverse of the serum cystatin C concentration as the dependent variable in the multivariate model. 6 Hypertension was defined as systolic blood pressure.140 mm Hg. 7 Values in this row are geometric means (95% CI). 8 Low vitamin B-12 status was defined as a serum vitamin B-12 concentration,148 pmol/l or a serum methylmalonic acid concentration 210 nmol/l. All other subjects were considered to have normal status. 9 5MeTHF concentrations in serum were measured by using affinity/hplc with electrochemical (coulometric) detection. 10 The folic acid content of food was determined only for foods reported by participants in the survey. These values were additionally adjusted for food folate.

7 FOLATE DEFICIENCY AND COGNITIVE FUNCTION 1739 FIGURE 1. Prevalence (and 95% CIs) of detectable circulating unmetabolized folic acid by quintile category of estimated folic acid intake. Subjects were senior participants in the National Health and Nutrition Examination Survey ( ) who had no evidence of renal dysfunction or history of stroke, recent anemia therapy, or diseases of the liver, thyroid, or coronary arteries (n = 852). Points represent least-squares geometric means adjusted for sex, age, and race-ethnicity. Points are plotted at the category medians. Points with the same lowercase letter do not differ significantly. Folic acid in serum was measured by using affinity/hplc with electrochemical (coulometric) detection. modeling, cancer history and higher serum C-reactive protein concentration remained marginally significantly associated with this rare health problem, but self-reported diabetes diagnosis, which was somewhat related to macrocytosis in the preliminary modeling, did not. Interaction between vitamin B-12 status and serum total folate in relation to anemia, macrocytosis, and poor cognitive test performance Results obtained with the use of HPLC-determined serum total folate values generally resembled those obtained with use of the radioassay-determined serum total folate values (Table 3). However, 2 differences are worth noting. First, in contrast with the results obtained by using the radioassay-derived values, use of the HPLC-derived values showed not only a marginally significant interaction with vitamin B-12 status in relation to macrocytosis, but also a significant association between a high serum total folate concentration and protection against macrocytosis in subjects with low vitamin B-12 status. Second, although both assays supported the conclusion that high compared with lower folate status was associated with increased odds of anemia and poor cognitive test performance in subjects with low vitamin B-12 status, only the radioassay-determined values showed an association between high folate status and significantly decreased odds of poor cognitive test performance in subjects with normal vitamin B-12 status. However, use of both the DSST score and HPLC-determined serum total folate as continuous variables in a linear regression analysis showed that the score increased by 0.05 points per unit increase in serum total folate among all subjects combined (P = 0.004) and by 0.08 points in the subgroup with normal vitamin B-12 status (P, 0.001). Consequently, despite the discrepancy between the radioassay- and HPLC-determined results shown in Table 3, use of both assays showed strong evidence of an association between higher folate status and higher DSST scores among the subjects who were not vitamin B-12 deficient. Interaction between vitamin B-12 status and folate fractions in relation to anemia, macrocytosis, and the DSST score Investigation of the relations between the folate fractions and the 3 health outcomes (Table 4) helped to clarify the results obtained by using serum total folate. TABLE 3 Interaction between vitamin B-12 status and serum total folate concentration above the 75th percentile (compared with a lower value) in relation to the odds of anemia, macrocytosis, and a low Digit-Symbol Substitution Test (DSST) score in senior participants in the National Health and Nutrition Examination Survey ( ), by folate assay method 1 Vitamin B-12 status 2 by Quantaphase II Radioassay Kit 3 Vitamin B-12 status 2 by HPLC 4 Health problem Low Normal P for interaction Low Normal P for interaction Anemia (1.38, 7.2) 0.85 (0.34, 2.13) (1.43, 7.47) 0.72 (0.22, 2.29) P Macrocytosis (0.13, 1.17) 1.54 (0.46, 5.11) (0.11, 0.92) 1.6 (0.51, 5.04) 0.05 P Low DSST score (1.1, 3.71) 0.49 (0.28, 0.85) (1.03, 4.38) 0.77 (0.44, 1.36) P Except where otherwise indicated, values are odds ratios (95% CIs). Subjects were men and women aged 60 y; reasons for exclusion were a serum creatinine concentration indicative of frank kidney dysfunction and self-report of any of the following: recent anemia therapy, coronary artery disease, liver disease, and thyroid disease. The Quantaphase II Radioassay Kit was manufactured by Bio-Rad Laboratories, Anaheim, CA. 2 Low vitamin B-12 status was defined as a serum vitamin B-12 concentration,148 pmol/l or a serum methylmalonic acid concentration 210 nmol/l. All other combinations were considered normal. 3 The 75th percentile was 55 nmol/l. 4 The 75th percentile was 66 nmol/l. 5 Defined as a hemoglobin concentration,12 g/dl (women) or,13 g/dl (men). Results were controlled for sex, age, race-ethnicity, current smoking, current alcohol intake, triceps skinfold thickness, self-reported diabetes status, and serum concentrations of cystatin C and C-reactive protein. 6 Mean cell volume 99 fl. Results were controlled for sex, age, race-ethnicity, current smoking, current alcohol intake, BMI, self-reported cancer history, and serum concentrations of ferritin, cystatin C, and C-reactive protein. 7 A score,34 out of 133 points (20th percentile) was considered low. Results were controlled for sex, age, race-ethnicity, educational achievement, current smoking status, serum concentrations of cystatin C and C-reactive protein, and self-reported history of cancer, diabetes, and alcohol abuse.

8 1740 MORRIS ET AL TABLE 4 Interaction between vitamin B-12 status and serum folate fraction in relation to Digit-Symbol Substitution Test score and the odds of anemia and macrocytosis in senior participants in the National Health and Nutrition Examination Survey ( ) 1 Vitamin B-12 status 2 Health problem Exposure Statistic Low Normal P for interaction Anemia 3 Folic acid 4,5 OR (95% CI) 1.91 (1.01, 3.61) P Anemia 3 5MeTHF 5,7 OR (95% CI) 1.01 (1.0, 1.02) 0.99 (0.98, 1.01) P Macrocytosis 8 5MeTHF 5,7 OR (95% CI) 0.98 (0.97, 1.0) 1.01 (1.0, 1.02) P Mean cell volume Folic acid 4,5 b-coefficient (95% CI) (22.84, 20.03) 0.23 (20.27, 0.72) P Cognitive score 9 Folic acid 4,5 b-coefficient (95% CI) (29.09, 20.63) 2.05 (20.6, 4.7) P Cognitive score 9 5MeTHF 5,7 b-coefficient (95% CI) (20.1, 0.04) 0.07 (0.04, 0.11) P 0.408, Subjects were men and women aged 60 y; reasons for exclusion were a serum creatinine concentration indicative of frank kidney dysfunction and self-report of any of the following: recent anemia therapy, coronary artery disease, liver disease, and thyroid disease. 5MeTHF, 5-methyltetrahydrofolate; OR, odds ratio. 2 Low vitamin B-12 status was defined as a serum vitamin B-12 concentration,148 pmol/l or a serum methylmalonic acid concentration 210 nmol/l. All other combinations were considered normal. 3 Defined as a hemoglobin concentration,12 g/dl (women) or,13 g/dl (men). Results were controlled for sex, age, race-ethnicity, current smoking, current alcohol intake, triceps skinfold thickness, self-reported diabetes status, and serum concentrations of cystatin C and C-reactive protein. 4 Presence compared with the absence of detectable circulating unmetabolized folic acid (detection limit = nmol/l). 5 Unmetabolized folic acid and 5MeTHF in serum were measured by using affinity/hplc with electrochemical (coulometric) detection. 6 These values apply to subjects with low and normal vitamin B-12 status combined, because the interaction between the presence compared with the absence of detectable circulating folic acid did not interact with vitamin B-12 status in relation to anemia. 7 5MeTHF as a continuous term. 8 Mean cell volume 99 fl. Results were controlled for sex, age, race-ethnicity, current smoking, current alcohol intake, BMI, self-reported cancer history, and serum concentrations of ferritin, cystatin C, and C-reactive protein. 9 Score on the Digit-Symbol Substitution Test. Results were controlled for sex, age, race-ethnicity, educational achievement, current smoking status, serum concentrations of cystatin C and C-reactive protein, and self-reported history of cancer, diabetes, and alcohol abuse. Anemia The positive association between the serum total folate concentration and anemia in subjects with low vitamin B-12 status occurred through associations with both folate fractions. Among subjects with normal vitamin B-12 status, the results for serum total folate hid a positive association between the presence compared with the absence of detectable circulating unmetabolized folic acid and anemia. The similarity between the subgroups with low and normal vitamin B-12 status in terms of the association between the presence compared with the absence of detectable circulating unmetabolized folic acid and anemia suggested to us that this association might reflect effect modification by some factor other than vitamin B-12 status. To explore this possibility, we tested interactions between the presence compared with the absence of detectable circulating unmetabolized folic acid and the other covariates in the multivariate model in relation to the odds of anemia. We found that current alcohol intake was an effect modifier (P for interaction = 0.015), because the presence compared with the absence of detectable circulating unmetabolized folic acid was associated with increased odds of anemia among alcohol users [odds ratio (OR): 3.37; 95% CI: 1.42, 8.01], but not among nonusers of alcohol (OR: 0.77; 95% CI: 0.41, 1.45). The serum 5MeTHF concentration interacted with vitamin B-12 status, but not with alcohol intake, in relation to the odds of anemia. Here again, the exposure was associated with increased odds of anemia in one group (ie, those with low vitamin B-12 status) and was not associated with the odds of anemia in the other group (ie, those with normal vitamin B-12 status). Clearly, the results did not support the hypothesis that high folate status or circulating unmetabolized folic acid was associated with the curing of anemia in persons with low vitamin B-12 status. They suggested, instead, that exposure to circulating unmetabolized folic acid was associated with an increased odds of anemia in alcohol users and that a higher 5MeTHF concentration was associated with an increased odds of anemia in persons with low vitamin B-12 status. Macrocytosis For both folate fractions, the results suggested that higher exposure in subjects with low vitamin B-12 status was associated with less evidence of macrocytosis. In other words, these results were consistent with the idea that macrocytosis was masked in association with circulating unmetabolized folic acid and high 5MeTHF concentrations in persons with low vitamin B-12 status. It is important to note that we found a significant trend of decreasing odds of macrocytosis with increasing 5MeTHF concentration, even in subjects with no detectable serum folic acid (P = 0.047). The restriction of these inverse associations to subjects with low vitamin B-12 status suggests that macrocytosis from causes other than low vitamin B-12 status is not masked by the high folic acid intakes that result in high 5MeTHF concentrations and the appearance of unmetabolized folic acid in the circulation.

9 FOLATE DEFICIENCY AND COGNITIVE FUNCTION 1741 DSST performance In the subgroup with low vitamin B-12 status, the mean DSST score for the subjects with detectable circulating unmetabolized folic acid was almost 5 points lower than that of the unexposed subjects; although the association with the 5MeTHF concentration was in the same direction, it was far from statistically significant. The observation that a high compared with a lower serum total folate concentration was associated with a doubling of the odds of a low DSST score (Table 3) reflects the doubling of the prevalence of exposure to unmetabolized folic acid associated with being in the top 25% of the distribution for serum total folate. Consistent with this conclusion, we found that, when subjects with detectable circulating unmetabolized folic acid were excluded from analysis, the OR relating a high compared with a lower serum total folate concentration to poor DSST performance was 1.33 (95% CI: 0.45, 3.93) (P = 0.592) in subjects with low vitamin B-12 status. In the subgroup with normal vitamin B-12 status, the mean DSST score was 2 points higher in association with the presence compared with the absence of detectable circulating unmetabolized folic, but that difference was not statistically significant. Results for the 5MeTHF concentration, on the other hand, showed a significant positive association between that exposure and the DSST score (Figure 2). Regardless of vitamin B-12 status, the creation of a model that included terms for both folate fractions resulted in b coefficients closer to zero for both terms. However, for subjects with low vitamin B-12 status, the term for folic acid remained statistically significant, and the term for the 5MeTHF concentration did not; whereas, for subjects with normal vitamin B-12 status, the term for the 5MeTHF concentration remained statistically significant, and the term for folic acid did not. DISCUSSION In a recent editorial, Carmel (48) speculated that the biochemical vitamin B-12 deficiency (18) identified in population FIGURE 2. Mean (and 95% CI) Digit-Symbol Substitution Test (DSST) score (with 95% CIs) by quintile category of serum 5-methyltetrahydrofolate concentration. Subjects were senior participants in the National Health and Nutrition Examination Survey ( ) who had no evidence of renal dysfunction or history of stroke, recent anemia therapy, or diseases of the liver, thyroid, or coronary arteries (n = 1536). Points represent least-squares geometric means adjusted for sex, age, race-ethnicity, educational achievement, current smoking status, self-reported cancer history and diabetes status, and serum concentrations of cystatin C and C-reactive protein. Points are plotted at the category medians. Points with the same lowercase letter do not differ significantly. 5-Methyltetrahydrofolate was determined by using affinity/hplc with electrochemical (coulometric) detection. surveys might be benign and not susceptible to harm from folic acid intake. However, our results for American seniors suggest that biochemical deficiency, which has different causes from pernicious anemia (PA) (48, 49), is clinically relevant and may have consequences similar to those observed when intrinsic factor is lacking. Specifically, our study linked low vitamin B-12 status to increased odds of hematologic and cognitive problems. Those findings corroborate a high prevalence of anemia and impaired mental functioning noted in seniors with established food-cobalamin malabsorption (50), which affects up to 30% of seniors and can cause vitamin B-12 deficiency (10, 51, 52). Our findings were also somewhat consistent with the reputed dissociation between the hematologic and neurologic effects of vitamin B-12 deficiency (53 55). Specifically, macrocytosis was related to a decreased odds of a low DSST score in subjects with low vitamin B-12 status (OR: 0.15; 95% CI: 0.03, 0.87). Our previous investigation was undertaken to shed light on Carmel s (48) question concerning the susceptibility of biochemical vitamin B-12 deficiency to adverse effects of folic acid intake. We found positive associations between a high serum total folate concentration and both anemia and poor cognitive test performance that were restricted to subjects with low vitamin B-12 status (31). Two other groups that subsequently attempted to fully (56) or partially (57) replicate our study failed to corroborate these results. One study was conducted in the United Kingdom, where food folic acid fortification is not mandatory, and where the prevalence of serum total folate.30 nmol/l was only 9% (56) as compared with 65% in the general US population. The other investigation used data collected in the Sacramento Area Latino Study on Aging (SALSA) (57). SALSA differed from our study in many important ways (eg, target population, cognitive assessment methods, inclusion and exclusion criteria, and definition of low vitamin B-12 status). Our findings for folate fractions were consistent with reports of mitigating effects of folic acid treatment on macrocytosis in PA cases (20), but our results did not corroborate the observation that a high folic acid intake cures vitamin B-12 deficiency anemia (58). Note, however, that the curing of anemia was observed mainly in PA cases with macrocytosis (60) and that anemia frequently occurs without macrocytosis in vitamin B-12 deficiency (61 64). Our findings for anemia are consistent with Reynolds s warning that high folic acid intakes precipitate anemia in vitamin B-12 deficiency (32, 33), and they also generate the hypothesis that alcohol and circulating unmetabolized folic acid interact synergistically to precipitate anemia even in the absence of vitamin B-12 deficiency. Similar effects of high folate status on alcohol users and vitamin B-12 deficient persons would be understandable if alcohol inhibited methionine synthase activity similarly to vitamin B-12 deficiency, as Mason and Choi (65) have suggested. However, the specificity of the association for different folate fractions in alcohol users and subjects with low vitamin B-12 status tends to argue against a common mechanism. Our results for macrocytosis also conflicted with mechanistic hypotheses invoking a special ability of folic acid, as opposed to 5MeTHF, to drive DNA synthesis despite the block in methionine synthase activity imposed by a lack of vitamin B-12 (30, 66). Specifically, we found an inverse association between the 5MeTHF concentration and the odds of macrocytosis that was statistically independent of circulating unmetabolized folic acid. Whether the key to the observed effects of high folic acid intake is