Vitamin B-12 Intake from Dairy but Not Meat Is Associated with Decreased Risk of Low Vitamin B-12 Status and Deficiency in Older Adults from Quebec, Canada

ABSTRACT

Background

Vitamin B-12 deficiency can result in irreversible neurologic damages. It is most prevalent among older adults (∼5%–15%), mainly due to impaired absorption. Vitamin B-12 bioavailability varies between food sources, so their importance in preventing deficiency may also vary.

Objectives

Using the NuAge Database and Biobank, we examined the associations between vitamin B-12 intake (total and by specific food groups) and low vitamin B-12 status and deficiency in older adults.

Methods

NuAge—the Quebec Longitudinal Study on Nutrition and Successful Aging—included 1753 adults aged 67–84 y who were followed 4 y. Analytic samples comprised 1230–1463 individuals. Dietary vitamin B-12 intake was assessed annually using three 24-h dietary recalls. Vitamin B-12 status was assessed annually as low serum vitamin B-12 (<221 pmol/L), elevated urinary methylmalonic acid (MMA)/creatinine ratio (>2 μmol/mmol), and a combination of both (deficiency). Vitamin B-12 supplement users were excluded. Multilevel logistic regressions, adjusted for relevant confounders, were used.

Results

Across all study years, 21.8%–32.5% of participants had low serum vitamin B-12, 12.5%–17.0% had elevated urine MMA/creatinine, and 10.1%–12.7% had deficiency. Median (IQR) total vitamin B-12 intake was 3.19 μg/d (2.31–4.37). Main sources were “dairy” and “meat, poultry, and organ meats.” The ORs (95% CIs) in the fifth quintile compared with the first of total vitamin B-12 intake were as follows: for low serum vitamin B-12, 0.52 (0.37, 0.75; P-trend < 0.0001); for elevated urine MMA/creatinine, 0.63 (0.37, 1.08; P-trend = 0.091); and for vitamin B-12 deficiency, 0.38 (0.18, 0.79; P-trend = 0.006). Similarly, ORs (95% CIs) in the fourth quartile compared with the first of dairy-derived vitamin B-12 intake were 0.46 (0.32, 0.66; P-trend < 0.0001), 0.51 (0.30, 0.87; P-trend = 0.006), and 0.35 (0.17, 0.73; P-trend = 0.003), respectively. No associations were observed with vitamin B-12 from “meat, poultry, and organ meats.”

Conclusions

Higher dietary vitamin B-12 intake, especially from dairy, was associated with decreased risk of low vitamin B-12 status and deficiency in older adults. Food groups might contribute differently at reducing risk of deficiency in older populations.

See corresponding editorial on page 2317.

Introduction

Vitamin B-12 deficiency is frequent in older adults with prevalence rates estimated at 5%–15% (1–3) and as high as 40% when mild cases are included (3). Causes of the deficiency are various, but in older adults, impaired absorption of food-bound vitamin B-12 is thought to be an important factor (4, 5). Vitamin B-12 deficiency can result in a range of symptoms, including neurologic manifestations such as cognitive decline, paresthesia, ataxia, and mood changes (6, 7). These neurologic manifestations may be observed even in mild cases (8, 9) and be partly or totally irreversible over time when the deficiency is left untreated (8). It is therefore important to understand how dietary recommendations may help prevent vitamin B-12 deficiency, especially in older populations.

In North America, the Institute of Medicine currently recommends an intake of 2.4 μg vitamin B-12/d for older adults, provided mainly by vitamin B-12–fortified foods or supplements to counteract potentially impaired absorption (10). This recommendation has been questioned though. Indeed, five large cross-sectional studies concluded that 5–10 μg/d is required for optimal vitamin B-12 status in older adults (11–15), including that in the United States, where vitamin B-12–fortified breakfast cereals contribute appreciably to total vitamin B-12 intake (12, 16). However, these studies might have overestimated the required intake since they were based on data derived from FFQs, which are known to produce higher estimates of vitamin B-12 intakes compared with weighted food records (17). Optimal dietary vitamin B-12 intake may also vary according to the contribution of vitamin B-12–fortified foods in the diet, which might be null or negligible in countries such as Canada, where fortification is restricted (18, 19).

The bioavailability of vitamin B-12 is known to vary between food categories where vitamin B-12 is naturally found (20–22), which may be of importance if the ability to absorb food-bound vitamin B-12 is impaired. To our knowledge, only four published studies, all cross-sectional, examined the associations between vitamin B-12 intake from specific food groups and vitamin B-12 status in older adults (11, 12, 15, 23). However, fortification aside, results were inconsistent across these studies. For instance, in 600 older adults in the Netherlands, high vitamin B-12 intakes from dairy, meat, fish, and shellfish were all significant protective factors against low vitamin B-12 status (23). Conversely, in one large study in Norway (N = 1310) and another in the United States (N = 2999), only vitamin B-12 intake from dairy was significantly protective (12, 15). Finally, in another US study (n = 449), neither dairy nor meat was associated with low vitamin B-12 status in non-Hispanic whites, whereas in Hispanics, only the middle tertile of dairy was (11).

In the present study, we aimed to examine whether dietary vitamin B-12 intake is associated with low vitamin B-12 status and deficiency in a well-characterized cohort of healthy older adults living in Quebec, Canada. Vitamin B-12 intake, total and from specific food groups, was assessed throughout 4 y using multiple 24-h dietary recalls (24HRs). Considering the lack of a gold standard test to assess vitamin B-12 status (24), different approaches were used on the basis of low serum vitamin B-12, elevated urine methylmalonic acid (MMA)/creatinine ratio, or a combination of both to diagnose deficiency. To our knowledge, this is the first study to investigate these associations using multiple 24HRs.

Methods

Data sources

The present study is based on the NuAge Database and Biobank, which includes data and biological samples of 1753 (98%) of the 1793 participants of the Quebec Longitudinal Study on Nutrition and Successful Aging (NuAge). A complete description of the NuAge Study can be found elsewhere (25). Briefly, participants were recruited in 2003–2005 and were globally healthy, community-dwelling men and women aged 67–84 y who lived in the Montreal, Laval, and Sherbrooke areas in the province of Quebec, Canada. Inclusion criteria were to be notably free of disabilities in activities of daily living and not cognitively impaired (Modified Mini-Mental State Examination score >79). Exclusion criteria included class II heart failure; chronic obstructive pulmonary disease requiring home oxygen therapy or oral steroids; inflammatory digestive diseases; and cancer treated by radiation therapy, chemotherapy, or surgery in the previous 5 y. Participants were then followed up annually for 3 subsequent years. At recruitment (labeled T1) and at each annual follow-up (labeled T2, T3, and T4), multiple nutritional, functional, medical, and social variables were collected by trained research dietitians and nurses, including all prescribed medication and the use of natural health products, vitamins, minerals, and other nutritional supplements. Dietary intake was assessed annually with 3 nonconsecutive 24HRs. Blood samples were collected each year of data collection, whereas urine samples were collected from T2 through T4. The NuAge Database and Biobank and the present study were approved by the Research Ethics Board of the Centre intégré universitaire de santé et de services sociaux de l'Estrie-Centre hospitalier universitaire de Sherbrooke (Quebec, Canada).

Analytic samples

We aimed to examine the associations between dietary vitamin B-12 intake (total and from specific food groups) and risk of 1) low serum vitamin B-12 concentration, 2) elevated urine MMA/creatinine ratio, and 3) a combination of both as an indicator of vitamin B-12 deficiency, over the 4 consecutive years of the NuAge Study. For objective 1, we first included in the analytic sample those who provided blood samples for at least one of the data collection years (n = 1719). Then, for each available blood sample of a given participant, we ascertained that three 24HRs were completed that year and that the participant reported no current or prior consumption of vitamin B-12 supplements, including multivitamin products containing vitamin B-12. Thus, the final sample for objective 1 included 1463 participants, 81% with ≥ 2 repeated measures of serum vitamin B-12 (Supplemental Figure 1a). Similarly for objective 2, we included those who provided urine samples for at least one of the data collection years (n = 1436); then, we applied the same inclusion criteria as objective 1. The final sample included 1231 participants, 87% with ≥ 2 repeated measures of urine MMA/creatinine ratio (Supplemental Figure 1b). Finally, for objective 3, we included those who provided both blood and urine samples for at least one of the data collection years (n = 1435); then, we applied the same inclusion criteria. The final sample included 1230 individuals, 87% with ≥ 2 repeated measures (Supplemental Figure 1c).

Assessment of vitamin B-12 status

There is no single gold standard biomarker to assess vitamin B-12 status (24). Therefore, in the present study, vitamin B-12 status was determined using two different biomarkers to ascertain the robustness of our conclusions—namely, serum total vitamin B-12 and urine MMA/creatinine ratio. Serum total vitamin B-12 (pmol/L) directly measures vitamin B-12 in blood and is commonly used in research and clinical practice. However, only ∼20% of serum total vitamin B-12 is available for cellular uptake (24), which can lead to equivocal results when used for diagnosis of vitamin B-12 deficiency (3, 26, 27). The urine MMA/creatinine ratio (μmol/mmol), for its part, was suggested as a useful and practical biomarker for vitamin B-12 deficiency in the 1980s (28). Specifically, MMA is a metabolite that accumulates in blood and excretes in urine in cases of vitamin B-12 deficiency. Therefore, blood MMA and urine MMA are markers of vitamin B-12 function in the body instead of vitamin B-12 “quantity.” In older adults, urine MMA/creatinine ratio was shown to be highly correlated with serum MMA and serum vitamin B-12 (29, 30).

The procedure of blood and urine sampling, processing, and storage in the NuAge study was described in detail elsewhere (25, 31). In short, 50 mL of blood and 6 mL of urine were obtained from each participant after an overnight fast. Blood and urine samples were processed, separated into aliquots in 1.5-mL opalescent-colored Eppendorf polypropylene tubes, stored in opaque cardboard at –20 °C, and then transferred to the NuAge biobank for long-term storage at –80 °C.

Serum total vitamin B-12 was analyzed at the Laboratoires CDL (Montreal, Canada) and Laboratoires Biron (Brossard, Canada) using an electrochemiluminescence immunoassay on a Cobas 8000/Module 602 with Elecsys Vitamin B-12 II reagents (Roche Diagnostic). Intra- and interassay CVs were all ≤ 10.4%. The urine MMA/creatinine ratio was assessed using reverse-phase ultraperformance LC coupled to tandem MS at the Waters-CHUS Expertise Centre in Clinical Mass Spectrometry (Sherbrooke, Canada). This procedure has been detailed elsewhere and validated against standard methods (30). Intra- and interday accuracy biases and precision coefficients were all ≤ 6.3% for urinary MMA and urinary creatinine. All blood and urine samples were shipped on dry ice to the laboratories where analyses were performed. None were thawed and refrozen before analyses.

Several cutoff values have been suggested to identity cases of vitamin B-12 deficiency when using serum vitamin B-12 as a biomarker (32). In the present study, we used < 148 pmol/L, which is the classical clinical cutoff to identify cases of deficiency (33). However, as previous studies showed that a significant proportion of older adults (≤56%) had elevated serum/plasma MMA or homocysteine when serum/plasma vitamin B-12 was within the range of 148–185 or 148–221 pmol/L, we used < 185 and < 221 pmol/L as indicators of low vitamin B-12 status as well (3, 34–36).

For the urine MMA/creatinine ratio, two cutoffs were suggested in published studies for diagnosis of metabolic vitamin B-12 deficiency: >1.5 and > 2.0 μmol/mmol (29, 37). In the NuAge cohort, the > 2.0 cutoff was shown to be sensitive (83%) and specific (93%) for serum MMA > 400 nmol/L, the latter being the cutoff suggested for serum MMA in older adults (29, 30). Therefore, values > 2.0 μmol/mmol were considered “elevated urine MMA/creatinine ratio” in the present study.

In the epidemiologic setting, Carmel suggested the combination of ≥ 2 test result abnormalities as the most reliable approach for diagnosing vitamin B-12 deficiency, since reliance on one biomarker alone led to frequent misdiagnosis (36). We thus used the combination of serum vitamin B-12 < 221 pmol/L and urine MMA/creatinine ratio > 2.0 μmol/mmol as a way to identify cases of vitamin B-12 deficiency.

Assessment of dietary vitamin B-12 intake

Dietary information was collected annually using three nonconsecutive 24HRs. Each set included 2 weekdays and 1 weekend day. The first interview was conducted in person whereas subsequent 24HRs were conducted by telephone within the next 2 mo, at days randomly chosen without prior notice. Using the USDA five-step multiple-pass method (38), detailed description and portion sizes of all food items were recorded by registered dietitians who received formal training. All 24HRs were reviewed by an experienced research assistant. The 24HRs were processed using the CANDAT nutrient analysis program (Godin London) based on the Canadian Nutrient File (CNF) database version 2007b (Health Canada) and a database developed on-site of > 1200 additional foods.

In total, 18,054 24HRs were used in the present study. They resulted in 4997 individual food items that were expected to contain vitamin B-12, either because they were animal food items or because they contained at least one ingredient of animal origin. Of these, only 285 (5.7%) food items had missing data for vitamin B-12 content. Nonetheless, to limit underestimation, values were imputed to 283 of these food codes based on 1) values from similar foods in the CNF (2007b or 2015) or the USDA Food and Nutrient Database for Dietary Studies (2017–2018) or 2) calculations from recipes used in the CNF (39, 40). Total vitamin B-12 intake (μg/d) was then determined for each 24HR.

Sources of vitamin B-12 were categorized into predefined food groups—namely, “dairy products,” “meat, poultry, and organ meats,” “fish and shellfish,” and “other sources,” the latter notably including eggs and vitamin B-12–fortified foods. For food items including multiple ingredients (e.g., lasagna), the relative contribution of each ingredient to the total vitamin B-12 content was assessed using the recipes used in the CNF (39, 40); after which, each ingredient contributing to the vitamin B-12 content was categorized by food group. Vitamin B-12 intake (μg/d) from each food group was then determined for each 24HR.

Covariates

Sociodemographic data included age and self-reported biological sex. Calculated age at recruitment was based on the date of birth and categorized into three groups according to the stratified sampling design of the NuAge Study: 67–72, 73–77, and 78–84 y. BMI was calculated as the ratio of body weight to height squared and classified into standard categories: <25, ≥25 to < 30, and ≥ 30 kg/m². Alcohol consumption was assessed using a validated semiquantitative FFQ administered only at baseline (41) and categorized into three groups (0, >0 to ≤ 5, or > 5 drinks/wk) based on a standard drink of 13.6 g of alcohol. Smoking status (never vs current/former) was self-reported at baseline as well. Other covariates include renal function and the use of metformin and proton pump inhibitors (PPIs). Renal function is known to influence the rate of MMA and creatinine clearance and was associated with the urine MMA/creatinine ratio in previous studies (30, 42). Therefore, we used the CKD-EPI Creatinine Equation (2009) to calculate the estimated glomerular filtration rate (eGFR) annually throughout the NuAge Study. Serum creatinine was analyzed using a colorimetric assay on Cobas 8000/Modules 502 and 702 with creatinine Jaffé Gen.2 reagents (Roche Diagnostic). Intra- and interassay CVs were all ≤ 5.7%. Renal function was then categorized into 2 groups: eGFR < 60 mL · min^–1 · 1.73 m^–2 and eGFR ≥ 60 mL · min^–1 · 1.73 m^–2. Finally, as metformin and PPI were both associated with an increased risk of vitamin B-12 deficiency (43–46), likely because of impaired absorption, users of metformin and PPIs (omeprazole, esomeprazole, pantoprazole, lansoprazole, dexlansoprazole, rabeprazole) were identified (reported usage) at each data collection year of the NuAge Study.

Statistical analyses

Statistical analyses were performed using SPSS Statistics version 25 (SPSS Inc.). Participant characteristics are presented as proportions (%) or medians (IQRs), as appropriate. Due to the positively skewed distributions of total vitamin B-12 intake as well as from specific food groups, log-transformed daily vitamin B-12 intake values were used to calculate the geometric mean intake each study year (as a measure of the participants’ usual intake) based on the three 24HRs collected that year. Participants were then classified into quintiles of total dietary vitamin B-12 intakes; the cutoff points of quintiles were based on data distribution at baseline. Similarly, participants were categorized into quartiles of vitamin B-12 intake from “dairy products” and “meat, poultry, and organ meats” and into tertiles for vitamin B-12 intake from “fish and shellfish” to account for the lower and narrowly distributed values into these specific food groups. Participants classified into the lowest vitamin B-12 intake category (total or from specific food groups) were used as referent in analyses.

Multilevel logistic regression modeling was used to determine the associations between vitamin B-12 intake (total and from food groups) and vitamin B-12 status throughout the 4 y of data collection, allowing us to take into account time-varying explanatory variables such as dietary intakes or the use of metformin and PPI. Intraclass correlation coefficients were first determined using a variance component model, which varied 0.40–0.47 (P < 0.05) across models, indicating significant intraindividual variances. The association between total dietary vitamin B-12 (as an ordinal variable expressed in quintiles) was then determined for each definition of vitamin B-12 status described earlier. Models were all adjusted for year of biological sample collection (first year of collection as referent), age groups, renal function, BMI, use of PPI, and use of metformin as level 1 variables, in addition to sex, smoking status, and alcohol consumption as level 2 variables.

The associations between 1) vitamin B-12 intake from “dairy products” (expressed as quartiles), “meat, poultry, and organ meats” (quartiles), and “fish and shellfish” (tertiles) and 2) vitamin B-12 status were subsequently assessed. Models were mutually adjusted for other food groups in addition to vitamin B-12 intake from “other sources,” which included miscellaneous food items, as well as for the same potential confounders described previously. Results are presented as ORs with the corresponding 95% CIs. Also, to examine potential dose-dependent associations (P for trend), analyses were repeated with the quintiles, quartiles, or tertiles entered as continuous variables. P < 0.05 was considered statistically significant.

Sensitivity analyses

When using 24HR to estimate usual nutrient intakes, accuracy of the estimates is closely related to the number of days of diet recording: the more days, the more accurate the estimates. For vitamin B-12, the number of days of diet recording needed to accurately assess usual intake is not known in older adults. Even so, three 24HRs might not be enough and may lead to measurement errors that would produce false-negative results in regression models. To correct these measurement errors (at least partially), statistical methods have been suggested (47, 48). However, among the NuAge participants, >700 provided three 24HRs annually over 4 y, for a total of 12 per individual. Therefore, using logistic regression modeling, we conducted sensitivity analyses aiming to determine the associations between vitamin B-12 intake (total and from food groups) estimated over 4 y using twelve 24HRs and vitamin B-12 status assessed at the last NuAge follow-up (T4). Models were all adjusted for sex, age groups, renal function, BMI, use of PPI and use of metformin, smoking status, and alcohol consumption.

In the present study, vitamin B-12 status was described as low or deficiency state (thus two-level variable) instead as a scale (continuous variable). Indeed, there are no reported health benefits at improving vitamin B-12 status beyond a certain stage. Nonetheless, to confirm our conclusions, we conducted additional analyses using multiple linear regression modeling to assess the associations between vitamin B-12 intake (total and from specific food groups) and serum vitamin B-12 levels at baseline (n = 1246). Values of vitamin B-12 intake and serum vitamin B-12 were log transformed for analysis. Models were adjusted for sex, age groups, renal function, BMI, use of PPI and use of metformin, smoking status, and alcohol consumption.

Results

Participant characteristics

Baseline characteristics of the three analytic samples used in the present study are shown in Table 1. No large discrepancies were noted between samples. Briefly, participants were aged 68–84 y, with an almost equal number of men and women. Over 70% of participants were overweight or obese according to the WHO classification of BMI values (49). Few were current smokers (6%–7%), and most had low to moderate alcohol consumption. Renal function was mostly normal, whereas the proportions of metformin and PPI users were lower than those normally found in older populations in Canada (50, 51), underlining the fact that the NuAge participants were globally healthy.

TABLE 1.

Baseline characteristics of community-dwelling older adults from the NuAge Study in each analytic sample with values of serum vitamin B-12, urine MMA/creatinine ratio, or both

	Analytic sample with values1
Characteristic	Serum vitamin B-12 (n = 1463)	Urine MMA/creatinine ratio (n = 1231)	Both biomarkers combined (n = 1230)
Age, y
67–72	38.1	39.8	39.8
73–77	32.9	33.6	33.7
78–84	29.0	26.6	26.5
Women	51.1	49.2	49.2
BMI, kg/m2
<25	26.0	26.1	25.4
≥25 to < 30	46.5	45.7	44.8
≥30	27.5	28.2	27.6
Alcohol consumption,2 drinks/wk
Never	28.8	27.9	27.9
>0 to ≤ 5	40.7	41.0	41.0
>5	30.5	31.1	31.1
Current or former smoker	47.4	48.3	48.3
Proton pump inhibitor usage	17.8	18.2	18.2
Metformin usage	6.5	6.2	6.2
eGFR,  mL · min−1 · 1.73 m−2
≥60	81.2	81.8	81.8
<60	18.8	18.2	18.2
Dietary vitamin B-12 intake, μg/d
Total	3.19 (2.31–4.37)	3.22 (2.32–4.42)	3.23 (2.31–4.42)
Dairy products	1.00 (0.50–1.62)	0.99 (0.51–1.62)	0.99 (0.51–1.63)
Meat, poultry, and organ meats	0.68 (0.10–1.40)	0.72 (0.11–1.40)	0.72 (0.11–1.40)
Fish and shellfish	0.002 (0.002–0.023)	0.002 (0.002–0.023)	0.002 (0.002–0.023)
Other sources	0.14 (0.043–0.31)	0.14 (0.046–0.32)	0.14 (0.046–0.32)
Serum vitamin B-12, pmol/L	300 (229–382)	303 (233–381)	303 (233–381)
Low serum vitamin B-12, pmol/L
<148	3.5	NA	3.2
<185	10.4	NA	9.2
<221	21.8	NA	21.0

¹Values are presented as proportions (%) for categorical variables and medians (IQRs) for continuous variables. eGFR, estimated glomerular filtration rate; MMA, methylmalonic acid; NA, nonapplicable; NuAge, Quebec Longitudinal Study on Nutrition and Successful Aging.

²Based on a standard drink of 13.6 g of alcohol.

Median (IQR) total dietary vitamin B-12 intake at baseline was comparable to the published estimates from a 2004 national survey conducted in the province of Quebec and using 24HR as well: 2.9 μg/d (2.2–3.5) and 4.0 μg/d (2.4–5.5) for women and men aged ≥71 y, respectively (52). For instance, 17.6% of participants in analytic samples had total vitamin B-12 intake below the estimated average requirement, which had been set at 2.0 μg/d by the Institute of Medicine (10). Main dietary sources of vitamin B-12 were “dairy products” and “meat, poultry, and organ meats,” as expected in a typical Western diet. Also, in accordance with Canadian regulations in force at the time of data collection, vitamin B-12–fortified foods contributed negligibly to total intake in this cohort: median, 0.000 μg/d (IQR: 0.000–0.0001). As users of vitamin B-12 supplements were excluded from our analytic samples, median serum vitamin B-12 at baseline was slightly lower than values reported (303–361 pmol/L) in Canada's older population (53).

As expected, the proportion of participants with low serum vitamin B-12 levels at baseline (T1) increased sharply as the cutoff values increased 148–221 pmol/L. Throughout the follow-up years of NuAge (T2 to T4), these proportions varied ≤4.5%, ≤15.1%, and ≤32.5% of each cutoff (i.e., <148, <185, and < 221 pmol/L), respectively. As indicated earlier, no baseline data were available for urine MMA/creatinine ratio, as collection of urine samples started at the first annual follow-up of the NuAge Study (T2). Yet, throughout T2 to T4, the proportion of participants with elevated urine MMA/creatinine ratios varied 12.5%–17.0%. Similarly, the proportion of individuals with vitamin B-12 deficiency (diagnosed by the combination of the two biomarkers) varied 10.1%–12.7% throughout T2 to T4.

Association between total vitamin B-12 intake and vitamin B-12 status

Multilevel logistic regression models were used to examine whether total vitamin B-12 intake was associated with vitamin B-12 status defined using serum vitamin B-12 values (Table 2), urine MMA/creatinine ratios (Table 3), or the combination of both biomarkers (Table 4). Higher total dietary vitamin B-12 intake was associated significantly, in a dose-dependent manner, with lower risk of having serum vitamin B-12 < 221 or < 185 pmol/L. In both models, the ORs were significant in quintiles 4 and 5, so for total vitamin B-12 intake ≥ 3.60 μg/d. Though nonsignificant, a marginal dose-dependent association between total dietary vitamin B-12 intake and elevated urine MMA/creatinine ratio was also observed (Table 3).

TABLE 2.

Associations between dietary vitamin B-12 intake (total and from selected food groups) and low serum vitamin B-12 in healthy older adults from the NuAge Study

	No. of subjects					Serum vitamin B-12,1 OR (95% CI)
	T1	T2	T3	T4	Vitamin B-12 intake,2 μg/d	< 221 pmol/L	< 185 pmol/L
Total vitamin B-12 intake
Quintile 1	248	202	208	153	1.60 ± 0.38 (0.23 to < 2.09)	1	1
Quintile 2	247	224	171	142	2.47 ± 0.21 (≥2.09 to < 2.83)	0.885 (0.642, 1.221)	0.881 (0.606, 1.281)
Quintile 3	250	190	178	140	3.21 ± 0.22 (≥2.83 to < 3.60)	0.919 (0.659, 1.283)	0.808 (0.545, 1.197)
Quintile 4	245	196	151	134	4.12 ± 0.33 (≥3.60 to < 4.76)	0.629 (0.442, 0.895)**	0.553 (0.360, 0.849)**
Quintile 5	256	174	186	165	7.36 ± 5.31 (≥4.76 to 24.39)	0.524 (0.366, 0.750)***	0.474 (0.305, 0.735)***
P-trend						<0.0001	<0.0001
Dairy products
Quartile 1	303	248	237	178	0.26 ± 0.15 (0.00 to < 0.50)	1	1
Quartile 2	314	262	252	192	0.74 ± 0.14 (≥0.50 to < 1.00)	0.836 (0.621, 1.126)	0.742 (0.525, 1.050)
Quartile 3	312	226	206	179	1.29 ± 0.18 (≥1.00 to < 1.63)	0.762 (0.554, 1.050)	0.663 (0.456, 0.964)*
Quartile 4	317	250	199	185	2.49 ± 0.89 (≥1.63 to 9.08)	0.460 (0.322, 0.658)***	0.393 (0.255, 0.605)***
P-trend						<0.0001	<0.0001
Meat, poultry, and organ meats
Quartile 1	307	265	220	149	0.039 ± 0.029 (0.00 to < 0.11)	1	1
Quartile 2	325	244	230	211	0.42 ± 0.17 (≥0.11 to < 0.71)	0.995 (0.731, 1.355)	0.895 (0.618, 1.297)
Quartile 3	307	251	242	183	1.01 ± 0.19 (≥0.71 to < 1.38)	0.941 (0.685, 1.293)	0.781 (0.533, 1.145)
Quartile 4	307	226	202	191	2.71 ± 3.96 (≥1.38 to 18.24)	0.852 (0.609, 1.193)	0.723 (0.483, 1.083)
P-trend						0.299	0.080
Fish and shellfish
Tertile 1	676	556	475	398	0.0018 ± 0.00043 (0.00 to < 0.0020)	1	1
Tertile 2	135	97	256	86	0.014 ± 0.0035 (≥0.0020 to < 0.019)	1.144 (0.840, 1.558)	0.944 (0.647, 1.376)
Tertile 3	429	333	163	249	0.37 ± 1.18 (≥0.019 to 8.33)	0.811 (0.623, 1.056)	0.634 (0.455, 0.883)**
P-trend						0.173	0.011

¹ORs were obtained by multilevel logistic regression analyses, adjusted for data collection year, age groups, sex, BMI, alcohol consumption, smoking status, estimated glomerular filtration rate, and usage of proton pump inhibitor and metformin. Results from each food group are mutually adjusted for vitamin B-12 intake from other sources. *P < 0.05. **P < 0.01. ***P < 0.001. NuAge, Quebec Longitudinal Study on Nutrition and Successful Aging.

²Mean ± SD (range).

TABLE 3.

Associations between dietary vitamin B-12 intake (total and from selected food groups) and elevated urine MMA/creatinine ratio in healthy older adults of the NuAge cohort

					OR (95% CI)1
	No. of subjects				MMA/creatinine Urine
	T2	T3	T4	Vitamin B-12 intake,2 μg/d	ratio > 2 μmol/mmol
Total vitamin B-12 intake
Quintile 1	145	196	145	1.60 ± 0.37 (0.23 to < 2.09)	1
Quintile 2	164	149	127	2.46 ± 0.22 (≥2.09 to < 2.83)	0.822 (0.506, 1.335)
Quintile 3	125	173	135	3.21 ± 0.23 (≥2.83 to < 3.60)	0.949 (0.581, 1.549)
Quintile 4	157	142	121	4.14 ± 0.33 (≥3.60 to < 4.76)	0.771 (0.462, 1.286)
Quintile 5	131	170	153	7.62 ± 6.36 (≥4.76 to 24.62)	0.634 (0.374, 1.075)
P-trend					0.091
Dairy products
Quartile 1	186	223	168	0.26 ± 0.15 (0.00 to < 0.50)	1
Quartile 2	182	232	176	0.74 ± 0.14 (≥0.50 to < 1.00)	1.155 (0.750, 1.780)
Quartile 3	163	187	164	1.30 ± 0.18 (≥1.00 to < 1.63)	0.885 (0.555, 1.411)
Quartile 4	191	188	173	2.50 ± 0.90 (≥1.63 to 9.08)	0.513 (0.304, 0.865)*
P-trend					0.006
Meat, poultry, and organ meats
Quartile 1	180	206	139	0.039 ± 0.030 (0.00 to < 0.11)	1
Quartile 2	186	213	194	0.42 ± 0.18 (≥0.11 to < 0.71)	1.210 (0.766, 1.912)
Quartile 3	179	220	170	1.01 ± 0.20 (≥0.71 to < 1.38)	1.193 (0.742, 1.918)
Quartile 4	177	191	178	2.81 ± 4.85 (≥1.38 to 18.24)	0.728 (0.432, 1.228)
P-trend					0.296
Fish and shellfish
Tertile 1	402	438	369	0.0016 ± 0.00048 (0.00 to < 0.0020)	1
Tertile 2	71	244	80	0.013 ± 0.0035 (≥0.0020 to < 0.019)	0.633 (0.392, 1.025)
Tertile 3	249	148	231	0.42 ± 1.46 (≥0.019 to 8.33)	0.924 (0.626, 1.365)
P-trend					0.449

¹ORs were obtained by multilevel logistic regression analyses, adjusted for data collection year, age groups, sex, BMI, alcohol consumption, smoking status, estimated glomerular filtration rate, and usage of proton pump inhibitor and metformin. Results from each food group are mutually adjusted for vitamin B-12 intake from other sources. *P < 0.05. MMA, methylmalonic acid; NuAge, Quebec Longitudinal Study on Nutrition and Successful Aging.

²Mean ± SD (range).

TABLE 4.

Associations between dietary vitamin B-12 intake (total and from selected food groups) and vitamin B-12 deficiency defined by serum vitamin B-12 < 221 pmol/L and elevated urine MMA/creatinine ratio in healthy older adults of the NuAge Study

					OR (95% CI)1
	No. of subjects				Serum vitamin B-12 < 221 pmol/L and urine MMA/creatinine
	T2	T3	T4	Vitamin B-12 intake, μg/d	ratio > 2 μmol/mmol
Total vitamin B-12 intake
Quintile 1	93	128	83	1.60 ± 0.37 (0.23 to < 2.09)	1
Quintile 2	97	108	93	2.46 ± 0.22 (≥2.09 to < 2.83)	0.845 (0.442, 1.618)
Quintile 3	86	118	90	3.21 ± 0.23 (≥2.83 to < 3.60)	0.763 (0.388, 1.498)
Quintile 4	117	111	88	4.14 ± 0.33 (≥3.60 to < 4.76)	0.630 (0.320, 1.242)
Quintile 5	97	136	120	7.62 ± 6.36 (≥4.76 to 24.62)	0.380 (0.183, 0.792)**
P-trend					0.006
Dairy products
Quartile 1	117	147	99	0.26 ± 0.15 (0.00 to < 0.50)	1
Quartile 2	117	165	119	0.74 ± 0.14 (≥0.50 to < 1.00)	0.898 (0.492, 1.636)
Quartile 3	113	137	122	1.30 ± 0.18 (≥1.00 to < 1.63)	0.781 (0.418, 1.460)
Quartile 4	143	152	134	2.50 ± 0.90 (≥1.63 to 9.08)	0.352 (0.170, 0.726)**
P-trend					0.003
Meat, poultry, and organ meats
Quartile 1	118	155	93	0.039 ± 0.030 (0.00 to < 0.11)	1
Quartile 2	124	151	131	0.42 ± 0.18 (≥0.11 to < 0.71)	1.086 (0.567, 2.082)
Quartile 3	125	159	122	1.01 ± 0.20 (≥0.71 to < 1.38)	1.325 (0.688, 2.552)
Quartile 4	123	136	128	2.81 ± 4.85 (≥1.38 to 18.24)	0.764 (0.372, 1.569)
P-trend					0.575
Fish and shellfish
Tertile 1	266	313	246	0.0016 ± 0.00048 (0.00 to < 0.0020)	1
Tertile 2	48	179	51	0.013 ± 0.0035 (≥0.0020 to < 0.019)	0.729 (0.385, 1.380)
Tertile 3	176	109	176	0.42 ± 1.46 (≥0.019 to 8.33)	0.785 (0.457, 1.349)
P-trend					0.312

¹ORs were obtained by multilevel logistic regression analyses, adjusted for data collection year, age groups, sex, BMI, alcohol consumption, smoking status, estimated glomerular filtration rate, and usage of proton pump inhibitor and metformin. Results from each food group are mutually adjusted for vitamin B-12 intake from other sources. **P < 0.01. MMA, methylmalonic acid; NuAge, Quebec Longitudinal Study on Nutrition and Successful Aging.

The association with vitamin B-12 deficiency defined as serum vitamin B-12 < 148 pmol/L was not possible to determine, as the multilevel logistic regression model did not converge due to the limited number of cases. However, when using the combination of markers (serum vitamin B-12 < 221 pmol/L and elevated urine MMA/creatinine ratio), higher total dietary vitamin B-12 intake was significantly associated, in a dose-dependent manner, with lower risk of vitamin B-12 deficiency (Table 4). The OR was significant in quintile 5, so for total vitamin B-12 intake ≥ 4.76 μg/d.

Association between vitamin B-12 intake from specific food groups and vitamin B-12 status

Multilevel logistic regression models were used to examine whether vitamin B-12 intake from specific food groups—namely, “dairy products,” “meat, poultry, and organ meats,” and “fish and shellfish”—were associated with vitamin B-12 status defined as low serum vitamin B-12 values (Table 2), elevated urine MMA/creatinine ratios (Table 3), or the combination of biomarkers to diagnose vitamin B-12 deficiency (Table 4). Higher vitamin B-12 intake from “dairy products” was significantly and consistently associated, in a dose-dependent manner, with lower risk of having low serum vitamin B-12, elevated urine MMA/creatinine ratio, or vitamin B-12 deficiency. Specifically, in all models, at least half as many NuAge participants in quartile 4 (i.e., with dairy-derived vitamin B-12 intake ≥ 1.63 μg/d) had vitamin B-12 deficiency compared with the first quartile (i.e., dairy-derived vitamin B-12 intake < 0.50 μg/d).

In contrast, vitamin B-12 intake from “meat, poultry, and organ meats” was not associated with vitamin B-12 status in any models. Higher vitamin B-12 intake from “fish and shellfish” was significantly associated with serum vitamin B-12 < 185 pmol/L only. Again, the multilevel logistic regression model did not converge when vitamin B-12 deficiency was defined as serum vitamin B-12 < 148 pmol/L.

Sensitivity analyses

Results from sensitivity analyses led to similar conclusions compared with our main analyses. Higher total dietary vitamin B-12 intake assessed using twelve 24HRs was significantly associated, in a dose-dependent manner, with lower risk of low serum vitamin B-12, elevated urine MMA/creatinine ratio, and vitamin B-12 deficiency assessed at the last NuAge follow-up (Supplemental Tables 1–3). In all models, the ORs were significant in quintiles 4 and 5, so for total vitamin B-12 intake ≥ 3.51 μg/d. Likewise, multiple linear regression revealed a significant positive association at baseline between total vitamin B-12 intake (log transformed) and serum vitamin B-12 (log transformed) after adjustment for covariates (Supplemental Table 4).

As observed in our main analyses, vitamin B-12 derived from dairy products was significantly associated with lower risk of low serum vitamin B-12, elevated urine MMA/creatinine ratio, and vitamin B-12 deficiency assessed at the last NuAge follow-up, as well as serum vitamin B-12 expressed as a continuous variable (Supplemental Tables 1–4). Also, vitamin B-12 from “meat, poultry, and organ meats” was not associated with vitamin B-12 status, except in one model when the outcome was serum vitamin B-12 < 185 pmol/L. Finally, higher vitamin B-12 intake from fish and shellfish was associated with lower risk of low serum vitamin B-12 but not with elevated urine MMA/creatinine ratio nor with vitamin B-12 deficiency.

Discussion

To our knowledge, this is the first study examining the associations between dietary vitamin B-12 intake and vitamin B-12 status in older adults living in Canada, where food fortification with vitamin B-12 is restricted. Higher total dietary vitamin B-12 intake, in particular from dairy products, was associated with a decreased risk of low vitamin B12 status and deficiency. Also, despite the fact that meat products were major dietary vitamin B-12 sources, higher intake was not associated with vitamin B-12 status. Vitamin B-12 intake from fish and shellfish was associated with low serum vitamin B-12 but not with urine MMA and deficiency per se.

Our results imply that vitamin B-12 naturally found in foods can contribute significantly to prevent low vitamin B-12 status and deficiency in older adults who are not supplement users, even when vitamin B-12 intake from fortification is negligible. Specifically, participants with total dietary vitamin B-12 intakes in the fourth quintile (≥3.60 to < 4.76 μg/d) and fifth (≥4.76 μg/d) showed a marked decrease in risk of having low vitamin B-12 status. Not surprising, published studies consistently reported that higher total intakes of vitamin B-12, including supplements and fortification where relevant, were associated with better vitamin B-12 status in adults and older adults (11–15, 23). One of these studies was similar to ours, though, as it was conducted in a sample of 600 adults aged ≥ 65 y for whom vitamin B-12 was virtually all provided by nonfortified foods (23). Interestingly, the authors also observed a significantly decreased risk of impaired vitamin B-12 status with intake of 3.20–4.40 μg/d (second tertile) and ≥ 4.41 μg/d (third tertile). These data raise questions about the current recommendation for vitamin B-12 intakes of older adults published by the Institute of Medicine in 1998 (10). The need for vitamin B-12 supplements and fortification might not be as necessary as thought at the time, especially when there are no risk factors for vitamin B-12 malabsorption, such as the use of metformin (54). Also, evidence is accumulating for the need to review the current recommended dietary allowance of 2.4 μg/d. Indeed, regardless of the study design and vitamin B-12 sources, authors consistently reported that higher intake is required for optimal vitamin B-12 status in adults and older adults (11–15, 23), including the present study based on data from 24HR.

In the present study, dairy products appeared to be the most potent dietary source of vitamin B-12 for prevention of low vitamin B-12 status and deficiency in older adults. To our knowledge, four studies examined whether the likelihood of low vitamin B-12 status was related to vitamin B-12 intake from dairy in adults. All reported a decreasing probability of deficiency as vitamin B-12 intake from dairy increased (11, 12, 15, 23). For example, in a sample of 2866 subjects aged 71–74 y, men and women in the top tertile of vitamin B-12 intake from dairy were 59% less likely to have low vitamin B-12 status than those in the bottom tertile, after adjustment for other food sources of vitamin B-12 (15). The US study from Tucker et al. (12) found that the proportion of individuals with low vitamin B-12 status decreased 14.1%–6.8% across tertiles of dairy intake. In Finland, older adults reporting avoidance of milk products were 2.3 times more likely to have vitamin B-12 deficiency (55).

Surprisingly, our analyses did not reveal any significant association between vitamin B-12 intake from meat products and vitamin B-12 status. These results are nonetheless in line with two US studies (11, 12) and another in Norway (15). Only one study, from Brouwer-Broslma et al. (23), found a significant dose-dependent association with meat product consumption, a discrepancy that they explained by the lower meat consumption in the Netherlands such that the maximal load of vitamin B-12 that can be absorbed in one meal is not reached. Conversely, it can be expected that the lower meat consumption often reported in older adults could have impeded our capacity to reveal any significant association with vitamin B-12 status, due notably to measurement errors (56, 57). However, sensitivity analyses conducted in the present study using twelve 24HRs to estimate dietary intake led to mostly similar conclusions.

Results across models were not consistent for vitamin B-12 from fish and shellfish, as we observed significant associations with low serum vitamin B-12 but not with less sensitive outcomes, such as urine MMA or deficiency. In our cohort, fish and shellfish consumption was very low in comparison with others (15, 23), which could explain these results. For example, one study in Norway reported median vitamin B-12 intake from fish and shellfish to be 2.1 μg in older women and 3.4 μg in older men, and the authors observed significant associations with low serum vitamin B-12 as well as with impaired vitamin B-12 function (15). Given the low consumption, measurement errors of dietary intakes might have also contributed to our results for fish and shellfish. However, results of our sensitivity analyses based on twelve 24HRs did not support this explanation since they led to similar conclusions.

An increased bioavailability of vitamin B-12 in dairy might explain our findings. Indeed, in one study, radioactive vitamin B-12 mixed in milk was 18% more bioavailable than when mixed with water (58). Also, the bioavailability of vitamin B-12 in cow milk was 8%–10% in pig models and significantly greater than that of a similar amount provided by supplements (59). Moreover, vitamin B-12 in milk appears to resist pasteurization and remain very stable during storage in retail stores as well as under domestic conditions (60). In contrast, vitamin B-12 in meat is partly degraded by heating, and ≤60% losses were reported during cooking (61, 62), which could explain why no significant associations with this food source were observed in the present study and others (11, 12, 15). In a narrative review, Matte et al. (63) hypothesized that a specific component resisting pasteurization could facilitate vitamin B-12 absorption in dairy. This component could simply be calcium ions, since vitamin B-12 absorption is a calcium-dependent process (64). In fact, calcium ions have been shown to restore vitamin B-12 absorption in subjects with steatorrhea, hyperparathyroidism, or metformin treatment (65–67). Calcium supplements have also been identified as a protective factor against vitamin B-12 deficiency in frail older adults using potent gastric acid inhibitors (68).

Regardless of the mechanism, the fact that dairy products could be quantitatively and qualitatively important sources of vitamin B-12 is of particular interest for preventing vitamin B-12 deficiency in older populations. This is even truer considering that, in the present study, a significant decrease in risk was observed in participants with mean ± SD dairy-derived vitamin B-12 intakes of 2.50 ± 0.90 μg/d (≥1.63 to 9.08), which can be provided by only two servings of fluid milk or 75 g of Swiss cheese. Bearing in mind that dairy products are rich in other key nutrients that are beneficial during aging (protein, calcium, vitamin D), increasing dairy consumption in older adults with an omnivorous diet could lead to multiple benefits.

Strengths of the present study include its unique design based on a large sample size and repeated measures over 4 y. Moreover, vitamin B-12 status was assessed using two distinct biomarkers—one quantitative marker in serum and one functional marker in urine—adding robustness to our conclusions. Also, multiple 24HRs were used for more accurate estimation of dietary vitamin B-12 intake (total and from specific food groups) in addition to having been collected using the multiple-pass method, which improves consistency across interviewers and limits memory bias (38). The NuAge cohort included adults aged > 75 y, but participants were mostly Caucasian in addition to being overall healthy at recruitment, which limits generalization of our results and likely explains why few individuals had serum vitamin B-12 < 148 pmol/L. Also, as vitamin B-12 intake from fortified foods was negligible in the NuAge cohort, no conclusion can be made in this regard. Finally, although analyses included a number of potential confounders, we cannot rule out possible residual confounding.

In conclusion, increasing vitamin B-12 intake, most particularly from dairy, was associated with lowering risk of low vitamin B-12 status and deficiency in healthy older adults. Notably, a significant decrease in risk was observed at levels of vitamin B-12 intakes easily achievable in a typical Western diet. Our results suggest that prevention of low vitamin B-12 status and deficiency in older populations can be possible through diet, without relying on vitamin B-12 supplementation or fortification. This certainly underlines the need to review the current recommended dietary allowance in Canada and the United States.

Notes

The present study was made possible through funding provided by the Dairy Farmers of Canada awarded to NP and PG; the Canadian Institutes of Health Research (CIHR; MOP-153011) awarded to AAC, PG, and NP; a starting grant for new investigators awarded to NP by the Université de Sherbrooke; and private funding from CA-B. The NuAge Study was supported by a research grant from the CIHR (MOP-62842). The NuAge Database and Biobank is supported by the Fonds de recherche du Québec (2020-VICO-279753); by the Quebec Network for Research on Aging, funded by the Fonds de Recherche du Québec–Santé (FRQS); and by the Merck-Frosst Chair, funded by La Fondation de l'Université de Sherbrooke. NP is a Junior 1 Research Scholar of the FRQS. AAC is a Senior Research Scholar of the FRQS. PG is a fellow of the Canadian Academy of Health Sciences. None of the funding organizations had a role in the design and conduct of the study or in the collection, management, analysis, and interpretation of the data.

Author disclosures: AAC is president and CEO at Oken Health. All other authors report no conflicts of interest.

Supplemental Figure 1 and Supplemental Tables 1–4 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: CNF, Canadian Nutrient File; MMA, methylmalonic acid; NuAge, Quebec Longitudinal Study on Nutrition and Successful Aging; 24HR, 24-h dietary recall.