The Institute of Medicine (IOM) extensively reviewed the available data on vitamin B12 (B12) deficiency and requirements, and has identified biochemical B12 deficiency and effective methods to reduce risk of B12 deficiency as high priorities for research(1). The Recommended Daily Allowance (RDA) was set at 120 % of the daily requirement and is expected to protect 97–98 % of healthy persons from deficiency. Based on the lower bioavailability of food-bound B12 in older individuals, the IOM recommended adults over the age of 50 years meet the RDA for B12 of 2·4 μg by consuming synthetic vitamin B12 in B12-fortified foods or B12 supplements(1). Previously, three population-based studies examined biochemical indicators of B12 deficiency (serum B12 and/or homocysteine and methylmalonic acid (MMA) concentrations) in relation to vitamin supplement status and concluded that B12-containing supplements may protect against B12 deficiency(Reference Koehler, Romero, Stauber, Pareo-Tubbeh, Liang, Baumgartner, Garry, Allen and Stabler2–Reference Bates, Mansoor, van der Pols, Prentice, Cole and Finch4). However, all three studies used data from elderly, racially homogeneous cohorts with relatively small sample sizes, limiting the generalizability of their findings to the US adult population. Therefore, we used data from NHANES III (Third National Health and Nutrition Examination Survey) to further explore the relationship between B12 supplement use and biochemically defined B12 deficiency in a large, nationally representative population sample of the US adult population.
Methods
Subjects
NHANES III was conducted with a complex, multistage probability design(5). Persons older than 60 years as well as African-American and Mexican-Americans were oversampled to more precisely assess US health and nutritional measures. Importantly, NHANES III was the first NHANES survey without an upper age limit, thus improving the estimates of health and nutrition in older segments of the population. The survey design was approved by the National Center for Health Statistics institutional review board (IRB) and participants gave written informed consent prior to undergoing survey procedures(5); the present analysis was thus deemed exempt from review by the Emory University IRB. The survey was divided into two phases (Phase 1, 1988–91 and Phase 2, 1991–4), designed to allow individual or combined phase analysis. Because homocysteine concentrations were measured only in Phase 2, the current analysis focuses on data from non-pregnant adults (age 18 years and older) in Phase 2 (September 1991–October 1994) for whom supplement use information and laboratory analyses of serum B12 and homocysteine are available (n 8394). Laboratory measurements, including serum B12 and total serum homocysteine concentrations, were collected and measured in accordance with NHANES III procedures(5–Reference Pfeiffer, Caudill, Gunter, Osterloh and Sampson7).
Per NHANES protocol, persons who were institutionalized, had haemophilia or had undergone chemotherapy for cancer within the previous four weeks were excluded from phlebotomy and thus from the present analysis. Serum B12 concentrations were analysed with a commercially available radioprotein binding assay kit (Quantaphase II from Bio-Rad Laboratories, Hercules, CA, USA; during NHANES III, Phase 2). EDTA-treated plasma was not available from NHANES III and serum homocysteine concentration was measured by HPLC with fluorometric detection. Further survey procedure details, including how demographic, medical history and medication and supplement intakes were collected, may be found in the NHANES III plan and operation(5) and in the data documentation on the NHANES website (http://www.cdc.gov/nchs/nhanes.htm). In addition, pregnant females (n 167) and persons who reported taking antiretroviral prescription medication for HIV (n 7) were excluded because B12 nutriture is altered in these populations(Reference Hepburn, Dyal, Runser, Barfield, Hepburn and Fraser8, Reference Remacha and Cadafalch9).
Analysis
There is no gold standard test or definition for B12 deficiency. We used a conservative serum B12 cut-off point of <148 pmol/l (200 pg/ml)(Reference Carmel, Green, Rosenblatt and Watkins10) to define low B12 concentration. Subjects with serum B12 < 148 pmol/l and concomitant serum homocysteine >10·0 μmol/l (1·35 mg/l) were defined as biochemically defined B12-deficient.
The primary exposure variable of interest, B12-containing supplement use, was ascertained by calculating the total (single dose) supplemental B12 intake among supplement users from the NHANES III PUVITMIN and SUPLCONC data sets, which define supplement use in this cohort. Supplement users were defined as having total oral B12 supplement intake >0 μg. Supplement dose categories were defined as: 0 μg (non-users), >0 to 6 μg (89 % of whom ingested 6 μg), >6 to 25 μg (88 % of whom ingested 9–25 μg) and >25 μg. The categories were chosen such that the lowest category of use included the dose found in many over-the-counter multivitamin supplements (6 μg). The second category of supplement users includes the 25 μg amount found in many senior daily multivitamin supplements, and the highest category included the higher doses (100, 500, 1000 μg) found in high-dose B-vitamin supplements.
Predictors of B12 and/or homocysteine concentrations which could potentially confound the analysis include age, race, impaired renal function, smoking(Reference Piyathilake, Macaluso, Hine, Richards and Krumdieck11), heavy alcohol intake(Reference Carmel and James12, Reference Ganji and Kafai13), hypothyroidism(Reference Hussein, Green, Jacobsen and Faiman14) and folate deficiency. Whites tend to have lower serum B12 concentrations than blacks or Mexican-Americans(Reference Wright, Bialostosky, Gunter, Carroll, Najjar, Bowman and Johnson15) and Mexican-American Hispanics tend to have lower homocysteine concentrations than non-Hispanic whites or non-Hispanic blacks(Reference Ganji and Kafai16). Except where noted, potential confounding exposure variables were dichotomized for the univariate analysis and multivariate logistic regression modelling.
After examining the distribution of cases within quartiles of age and ensuring adequate sample size, age was classified as a dichotomous variable to match the age recommendations discussed in the IOM report (18–50 years and >50 years)(1), and included as a continuous variable in the logistic regression models. Those participants who reported being current smokers of cigarettes, pipes or cigars were classified as current smokers. Study participants’ average daily alcohol consumption was estimated from the NHANES survey alcohol consumption questions (number of drinks on drinking days×number of days per year drinking/365). If the number of days per year drinking was missing, the daily average was calculated by multiplying the number of drinks on drinking days by the number days per year with 9+ drinks, and then dividing by 365 days per year. Based on the US Department of Agriculture’s dietary guidelines for alcohol (no more than one to two drinks of alcohol daily)(17, 18) and the definition of heavy drinking given on the Centers for Disease Control and Prevention (CDC) alcohol fact sheet, participants whose calculated average daily alcohol consumption was three or more drinks were classified as ‘heavy drinkers’(17, 18).
Due to possible alteration in B12 and/or homocysteine status in renal impairment, serum creatinine was included as a continuous variable in the multivariate analysis. Participants with thyroid-stimulating hormone concentration greater than 5 μIU/l were considered ‘hypothyroid’ for the present analysis(Reference Baskin, Cobin, Duick, Gharib, Guttler, Kaplan and Segal19). Race was classified into three groups, white, black and other, based on NHANES analysis guidelines and available crude sample size(20). Folate is a major determinant of homocysteine concentrations(Reference Pfeiffer, Caudill, Gunter, Osterloh and Sampson7, Reference Ganji and Kafai21, Reference Selhub, Jacques, Rosenberg, Rogers, Bowman, Gunter, Wright and Johnson22), and folate deficiency was therefore defined as red blood cell (RBC) folate <232 nmol/l (102·6 ng/ml; in home-examined participants, for whom RBC folate analysis was not done, serum folate <13·3 nmol/l (5·89 ng/ml))(Reference Gunter, Koncikowski and Lewis6, Reference Guralnik, Eisenstaedt, Ferrucci, Klein and Woodman23).
In all analyses, α = 0·05 was considered significant and 95 % confidence intervals are reported. To account for the complex survey design, SURVEY procedures in the SAS statistical software package version 9·1 (SAS Institute Inc., Cary, NC, USA) were used to estimate the population prevalence and prevalence ratio estimates. To calculate crude prevalence ratios, PROCSURVEY FREQ outcome prevalence estimates from the exposed (supplemented participants) are divided by the prevalence estimates in the unexposed (non-supplemented participants).
Unconditional logistic regression techniques were used to evaluate the univariate frequencies of each potential exposure variable and measures of association were computed, adjusting for supplement intake. Those exposures variables which were significantly related to the outcome variable were then included in a multivariate logistic regression model. Exposure variables were then removed from the model by backwards elimination and those exposures which altered the prevalence odds ratio by more than 10 % were kept in the model. The prevalence of biochemical B12 deficiency in the NHANES sample meets the rare disease assumption, and adjusted prevalence ratios are approximated from the multivariate logistic regression odds ratios using the method described by Zhang and Yu(Reference Zhang and Yu24).
Results
Prevalence of biochemical B12 deficiency
Applying the above criteria yielded a crude sample size of 8394 for the analysis of low B12 and 7404 for the analysis of biochemical B12 deficiency (Table 1). The overall estimated US adult population prevalence of biochemically defined B12 deficiency was 1·6 %. The prevalence was 1·2 % for those aged 18–50 years and 2·5 % for those aged >50 years. The prevalence of low serum B12 concentrations was 3·2 % for all adults, 2·6 % for adults aged 18–50 years and 4·4 % for those aged >50 years (Tables 1, 2 and 3).
Table 1 Prevalence and unadjusted and adjusted prevalence ratio of low serum vitamin B12 concentration and biochemically defined vitamin B12 deficiency by B12 supplement dose for all persons aged 18 years and older in NHANES III (Third National Health and Nutrition Examination Survey; Phase 2, 1991–4)
Unadj., unadjusted; Adj., adjusted; PR, prevalence ratio.
The PROCSURVEY FREQ procedure in SAS version 9·1 (SAS Institute Inc., Cary, NC, USA) and the NHANES III, Phase 2 data sets were used to obtain these US population prevalence proportion estimates in adults aged 18 years and older.
*Low serum B12: serum B12 < 148 pmol/l.
†Biochemically defined B12 deficiency: serum B12 < 148 pmol/l with serum homocysteine > 10 μmol/l.
‡Adj. PR = prevalence ratio adjusted for age (continuous), gender, race (black race or other), folate deficiency status and education (<8 years v. >8 years).
§Adj. PR = prevalence ratio adjusted for age (continuous), gender, race (black race or other), folate deficiency status, serum creatinine and education (<8 years v. >8 years).
||Nine hundred and seventy-two participants had missing homocysteine measurements, so estimates are based on 7404 survey participants for the biochemical B12 deficiency analysis and on 8394 survey participants for the low B12 analysis.
¶Rao–Scott P value = 0·001 compared with supplement non-users.
**Rao–Scott P value = 0·0002 compared with supplement non-users.
††PR = 0·02, rounded to 0·0.
Table 2 Prevalence and unadjusted and adjusted prevalence ratio of low serum vitamin B12 concentration and biochemically defined vitamin B12 deficiency by B12 supplement dose for adults aged 18–50 years in NHANES III (Third National Health and Nutrition Examination Survey; Phase 2, 1991–4)
Unadj., unadjusted; Adj., adjusted; PR, prevalence ratio; n/a, not applicable.
The PROCSURVEY FREQ procedure in SAS version 9·1 (SAS Institute Inc., Cary, NC, USA) and the NHANES III, Phase 2 data sets were used to obtain these US population prevalence proportion estimates in adults aged 18–50 years.
*Low serum B12: serum B12 < 148 pmol/l.
†Biochemically defined B12 deficiency: serum B12 < 148 pmol/l with serum homocysteine > 10 μmol/l.
‡Adj. PR = prevalence ratio adjusted for age (continuous), gender, race (black race or other), folate deficiency status and education (<8 years v. >8 years).
§Adj. PR = prevalence ratio adjusted for age (continuous), gender, race (black race or other), folate deficiency status, serum creatinine and education (<8 years v. >8 years).
||Estimates are based on 7404 survey participants for the biochemical B12 deficiency analysis and on 8376 survey participants for the low B12 analysis.
¶Rao–Scott P value = 0·2 compared with supplement non-users.
**Rao–Scott P value = 0·06 compared with supplement non-users.
Table 3 Prevalence and unadjusted and adjusted prevalence ratio of low serum vitamin B12 concentration and biochemically defined vitamin B12 deficiency by B12 supplement dose for adults aged >50 years in NHANES III (Third National Health and Nutrition Examination Survey; Phase 2, 1991–4)
Unadj., unadjusted; Adj., adjusted; PR, prevalence ratio.
The PROCSURVEY FREQ procedure in SAS version 9·1 (SAS Institute Inc., Cary, NC, USA) and the NHANES III, Phase 2 data sets were used to obtain these US population prevalence proportion estimates in adults aged >50 years.
*Low serum B12: serum B12 < 148 pmol/l.
†Biochemically defined B12 deficiency: serum B12 < 148 pmol/l with serum homocysteine > 10 μmol/l.
‡Adj. PR = prevalence ratio adjusted for age (continuous) gender, race (black race or other) and folate deficiency status.
§Rao–Scott P value = 0·001 compared with supplement non-users.
||Rao–Scott P value = 0·0002 compared with supplement non-users.
¶Point estimate for PR was 0·03, rounded to 0·0.
**There were no cases of biochemical B12 deficiency in the 149 survey participants over age 50 who were taking >25 μg of supplemental B12.
B12-containing supplement consumption and prevalence ratios
Supplement consumption was associated with lower prevalence of both low serum B12 concentration and biochemically determined B12 deficiency in the total population and in both age categories (18–50 and >50 years, Tables 1–3). Among the over-50 s taking >0–6 μg or >6–25 μg, supplements reduced the occurrence of both outcomes by 60–70 %. The prevalence was even lower among those consuming >25 μg.
Among all B12-supplement users, the distribution of total supplemental intakes was skewed, with a mean of 31·4 μg/d and a median of 6·0 μg/d (range 0·19–2075 μg/d). Of participants who consumed B12-containing supplements, 99·8 % consumed at least 2·4 μg B12/d (the RDA suggested by the IOM) and 80·7 % consumed 6–25 μg B12/d (the amounts most commonly found in multivitamin supplements). Of supplement users in the second supplement dose category, >0 to 6 μg, most (89 %) participants took 6 μg, the dose found in ‘regular’ (non-geriatric) multivitamin supplements. Only 2·1 % of supplement users took large doses (250 μg or more) of supplemental B12 and no cases of biochemical B12 deficiency occurred in supplement users taking >75 μg of B12 daily.
Univariate analysis and multivariate logistic regression modelling
Logistic regression analysis of each potentially confounding exposure variable (with adjustment for NHANES design and supplement status) revealed that age >50 years and 8 years or less of education were both significantly associated with biochemical B12 deficiency. Odds ratio point estimates of biochemical B12 deficiency were above unity for male gender, non-white or African-American race, as well as hypothyroid state, but none of these associations was significant. For each demographic exposure category (old and young age categories, gender, white and black races, and all education levels) B12 supplement consumption was associated with reduced odds of biochemical B12 deficiency.
After adjusting for survey design, 17·9 % of persons had elevated serum creatinine. However, the prevalence of low serum B12 concentration and biochemically defined B12 deficiency in persons with and without elevated creatinine (>1·1 mg/dl for females, >1·4 mg/dl for males) was not significantly different (P = 0·5 for low B12 and P = 0·09 for biochemically determined B12 deficiency). Similarly, the prevalence of low B12 in participants aged >50 years with normal creatinine (4·1 %) was not significantly different from that in older participants with elevated creatinine (5·2 %, P = 0·26).
Multivariate logistic regression model analysis revealed that hypothyroidism, heavy alcohol use and smoking status were not significant predictors of low serum B12 or biochemical B12 deficiency. Although lower educational level showed a statistically significant association with biochemical B12 deficiency in the univariate analysis, removing it from the model did not change the prevalence ratio estimates. Similarly, removing serum creatinine concentration from the multivariate model did not affect the point estimates or confidence intervals.
Discussion
Consistent with our a priori hypothesis, we found that intake of supplements providing 6 or 25 μg of vitamin B12 was associated with reductions in the prevalence of biochemical B12 deficiency and low serum B12 concentrations in US adults (aged both 18 years and older and above 50 years). The majority of supplement users met or exceeded the IOM RDA of B12 intake (2·4 μg) because the supplemental B12 quantities commonly found in over-the-counter multivitamin supplements range from 6 to 25 μg. Although we cannot conclude that there is a significant benefit to consuming 6 μg v. 2·4 μg supplement, or 25 μg v. 6 μg supplement, our data suggest that the prevalence of B12 deficiency (defined as low serum B12 or biochemical deficiency) among persons aged >50 years who consume B12 supplements is half to two-thirds the prevalence in persons who do not consume B12 supplements; the difference between prevalence for supplement users v. non-users in persons aged 18–50 years is about a third. The prevalence reduction was even higher in persons taking >25 μg supplemental B12 daily.
Although there was overlap in the confidence intervals, the trend we observed of decreasing prevalence of low serum B12 and biochemical B12 deficiency with increasing B12 supplement dose is consistent with results seen in vitamin supplement intervention trials in the elderly(Reference Eussen, de Groot, Clarke, Schneede, Ueland, Hoefnagels and van Staveren25–Reference Tucker, Olson, Bakun, Dallal, Selhub and Rosenberg28). In three of these studies, 10–500 μg of supplemental B12 increased serum B12 concentrations by 18·5–166 pmol/l (25–225 pg/ml)(Reference Eussen, de Groot, Clarke, Schneede, Ueland, Hoefnagels and van Staveren25–Reference Seal, Metz, Flicker and Melny27). Along with our findings, these studies indicate that daily doses of supplemental B12 as low as 6 μg can shift the distribution of B12 concentrations upwards in the general population and reduce biochemical B12 deficiency prevalence.
About 2 % of people aged >50 years will still have low B12 concentrations despite consuming 6–25 μg of synthetic B12. Daily consumption of more than 75 μg B12 may nearly eliminate low B12 and biochemical B12 deficiency, but more research is needed to determine how much supplement is needed by those individuals for whom 6 or 25 μg/d does not prevent low B12 or biochemical B12 deficiency.
Given the success of a proof of concept study by Winkels et al.(Reference Winkels, Brouwer, Clarke, Katan and Verhoef29), our data suggest that a B12 fortification programme that delivered 6 to 25 μg of synthetic vitamin B12 to most of the population may be highly effective in decreasing the prevalence of both low B12 concentrations and biochemically defined B12 deficiency. Were such a programme implemented, there would be fewer patients with low serum B12 and biochemically defined B12 deficiency for clinicians to investigate. Several studies have reported that adults with suboptimal B12 status have poorer cognitive functioning or could improve with supplementation(Reference Lökk30–Reference van Asselt, Pasman, van Lier, Vingerhoets, Poels, Kuin, Blom and Hoefnagels32). However, results from other B-vitamin supplement trials have failed to yield clinically measurable benefit(Reference Toole, Malinow, Chambless, Spence, Pettigrew, Howard, Sides, Wang and Stampfer33, Reference Spence, Bang, Chambless and Stampfer34). Further investigations are required to determine whether improving B12 status with fortification would result in improved functional status or clinical benefit.
The prevalence of low B12 in NHANES III (3·2 %) was similar to that previously observed in NHANES III (3 %)(Reference Pfeiffer, Caudill, Gunter, Osterloh and Sampson7), but was lower than that observed in other cohort studies in the USA (5·3–14·5 %)(Reference Koehler, Romero, Stauber, Pareo-Tubbeh, Liang, Baumgartner, Garry, Allen and Stabler2, Reference Lindenbaum, Rosenberg, Wilson, Stabler and Allen3, Reference Pennypacker, Allen, Kelly, Matthews, Grigsby, Kaye, Lindenbaum and Stabler35). However, those studies focused on racially homogeneous, elderly cohorts rather than the entire adult population, and they had relatively small sample sizes. Increased age and white race have both been associated with lower B12 concentrations(Reference Wright, Bialostosky, Gunter, Carroll, Najjar, Bowman and Johnson15) and estimates from predominantly older, white cohorts may therefore overestimate the national prevalence of B12 deficiency. The higher cut-off points for defining low B12 (<221 pmol/l or 300 pg/ml) in three non-institutionalized adult populations(Reference Koehler, Romero, Stauber, Pareo-Tubbeh, Liang, Baumgartner, Garry, Allen and Stabler2, Reference Lindenbaum, Rosenberg, Wilson, Stabler and Allen3, Reference Pennypacker, Allen, Kelly, Matthews, Grigsby, Kaye, Lindenbaum and Stabler35) and two outpatient source populations(Reference Pennypacker, Allen, Kelly, Matthews, Grigsby, Kaye, Lindenbaum and Stabler35, Reference Yao, Yao, Yao, Yao and Lou36) may also explain the lower prevalence of B12 deficiency in our study population. Overall, our study supports the results of previous studies, but widens their generalizability for the US population.
We did not examine the limited physical examination data in NHANES III addressing whether persons with biochemically determined B12 deficiency exhibited clinical signs of B12 deficiency. Although 6 to 25 μg of supplemental B12 appears to prevent most preclinical deficiency indicators in those aged 18 years and older, higher doses may be warranted in persons over 50 years old or with clinical signs such as anaemia and/or neuropathy(Reference Eussen, de Groot, Clarke, Schneede, Ueland, Hoefnagels and van Staveren25). We remind clinicians that patients with clinical symptoms consistent with clinical B12 deficiency disease should be worked up and treated with an appropriate dose of vitamin B12.
The primary strength of our analysis is that it examines a nationally representative population sample of ambulatory persons in the USA. With the largest sample size reported to date (over 8000 participants), we were able to adjust prevalence ratio estimates for the most common potential confounding exposures, including age, race, educational level, renal insufficiency and folate status.
A potential weakness of our study is the absence of MMA data to use as a criterion for biochemical determination of B12 deficiency. An elevated MMA concentration is typically considered a more specific confirmatory indicator of B12 deficiency than elevated homocysteine. Because NHANES III MMA data are not publicly available, we chose a homocysteine cut-off that is within the range expected in B12- and folate-replete individuals(Reference Guralnik, Eisenstaedt, Ferrucci, Klein and Woodman23, Reference Hackam, Peterson and Spence37). Another potential weakness of our study is the incomplete exclusion of HIV-positive individuals(Reference McQuillan, Khare, Karon, Schable and Vlahov38). Given that low B12 concentrations are reportedly less likely in HIV after antiretroviral therapy is started(Reference Hepburn, Dyal, Runser, Barfield, Hepburn and Fraser8), those excluded were least likely to be cases. HIV patients use supplements at a higher rate (over 70 %) than the general population and the participants not excluded would likely increase the number of exposed controls, potentially resulting in bias towards the null. However, at the estimated prevalence of HIV of 0·34 %(Reference McQuillan, Khare, Karon, Schable and Vlahov38), unidentified HIV-positive individuals would have virtually no impact on the prevalence of B12 deficiency and prevalence ratio estimates in our data.
In conclusion, consumption of supplements with 6 or 25 μg of vitamin B12 was associated with about a 50 % reduction in the prevalence of low B12 concentrations and biochemically defined B12 deficiency. These findings suggest that approximately three million cases of low serum B12 and biochemical B12 deficiency in the US population may be prevented if all adults consumed 6–25 μg B12/d from supplements or fortified foods. We also found low vitamin B12 concentrations and biochemically defined vitamin B12 deficiency in 1–2 % of persons aged >50 years who were consuming 6 to 25 μg of B12 from supplements. This relatively high prevalence in older supplement users suggests that the current RDA of 2·4 μg/d may be too low for this age group. As Berry et al. noted, people who have low serum concentrations of B12 may have problems in absorbing free B12(Reference Berry, Carter and Yang39) and have the preclinical pernicious anaemia that Carmel et al. have identified among those over 50 years of age(Reference Carmel, Green, Rosenblatt and Watkins10). Clinicians should be vigilant for subtle signs and symptoms of pernicious anaemia and clinical vitamin B12 deficiency, particularly in older adults who have low concentration of serum B12 and/or biochemically defined vitamin B12 deficiency in spite of consuming vitamin B12 supplements, and manage these patients appropriately.
Acknowledgements
The study was supported by NIH/NCRR grant K12 RR017643 Emory Mentored Clinical Research Scholars Program (MLE) and the Emory General Clinical Research Center NIH/NCRR M01 RR00039. There was no funding directly for this manuscript; the authors did it on their own time. M.L.E., P.D.T. and T.R.Z. have no conflicts of interest. G.P.O. is co-inventor (while at CDC) of a patent that covers adding folic acid to contraceptive pills and is a paid consultant to Ortho McNeil on this issue. Compensation, if any, will be under the regulations of CDC. All authors participated in designing the analyses and writing the article. G.P.O. arranged the use of the data; G.P.O., M.L.E. and P.D.T. conducted statistical analyses.
