Abstract
Objective
The current meta-analysis evaluated the association between vitamin B12 intake and blood vitamin B12 level and colorectal cancer (CRC) risk.
Design
The PubMed and EMBASE databases were searched. A dose–response analysis was performed with generalized least squares regression, with the relative risk (RR) and 95 % CI as effect values.
Setting
The meta-analysis included seventeen studies.
Subjects
A total of 10 601 patients.
Results
The non-linear dose–response relationship between total vitamin B12 intake and CRC risk was insignificant (P=0·690), but the relationship between dietary vitamin B12 intake and CRC risk was significant (P<0·001). Every 4·5 μg/d increment in total and dietary vitamin B12 intake was inversely associated with CRC risk (total intake: RR=0·963; 95 % CI 0·928, 0·999; dietary intake: RR=0·914; 95 % CI 0·856, 0·977). The inverse association between vitamin B12 intake and CRC risk was also significant when vitamin B12 intake was over a dosage threshold, enhancing the non-linear relationship. The non-linear dose–response relationship between blood vitamin B12 level and CRC risk was insignificant (P=0·219). There was an insignificant association between every 150 pmol/l increment in blood vitamin B12 level and CRC risk (RR=1·023; 95 % CI 0·881, 1·187).
Conclusions
Our meta-analysis indicates that evidence supports the use of vitamin B12 for cancer prevention, especially among populations with high-dose vitamin B12 intake, and that the association between CRC risk and total vitamin B12 intake is stronger than between CRC risk and dietary vitamin B12 intake only.
Keywords: Vitamin B12 , Colorectal cancer, Meta-analysis
Colorectal cancer (CRC) is one of the most commonly diagnosed cancers worldwide( 1 ). The development of CRC in the normal colorectal epithelium results from genetic alterations, epigenetic modifications and environmental factors( 2 – 4 ). Diet may be an important aetiological factor in CRC( 5 ) and several biological mechanisms may explain the relationship between diet and CRC risk( 5 ).
Vitamin B12 is an essential coenzyme for methionine synthase, which maintains adequate intracellular methionine levels in one-carbon metabolism( 6 , 7 ) for several intracellular biological processes, including methylation reactions, nucleotide biosynthesis and DNA repair( 8 ). Recent studies have shown that one-carbon metabolism may be related to colorectal carcinogenesis( 9 , 10 ). Therefore, vitamin B12 deficiency may increase CRC risk( 11 , 12 ). However, several studies have reported that vitamin B12 is not associated with the risk of CRC( 13 , 14 ), or that folic acid plus vitamin B12 is associated with increased cancer risk( 15 ).
The results on the association between vitamin B12 intake and CRC risk are still controversial and no relevant pooled analyses have been performed. Therefore the aim of our meta-analysis study was to quantitatively and comprehensively summarize whether vitamin B12 intake or blood vitamin B12 level is related to CRC risk.
Methods
Literature search
A literature search for relevant studies was performed using the PubMed and EMBASE databases up to April 2014. The search terms used were: (vitamin B12 OR cyanocobalamin OR cobalamins OR cobalamin OR hydroxocobalamin OR 5′-deoxyadenosyl cobalamin) AND (colorectal cancer OR colon cancer OR rectal cancer). Moreover, we manually screened the references of the relevant studies and reviews to check for other potentially relevant studies.
Eligibility criteria
The studies which met the following eligible criteria were included: (i) cohort study or case–control study; (ii) the exposure of interest was vitamin B12 intake, or blood vitamin B12 level, for three or more quantitative categorized levels; (iii) the outcome of interest was related to CRC; and (iv) the risk estimates (OR, risk ratio (RR) or hazard ratio (HR)) and corresponding 95 % CI were reported or calculated from published data with a category-specific number of cases and a category-specific number of either person-years or non-cases. When several studies were based on the same population, only the most informative study was included.
Data extraction and quality assessment
Data were independently extracted by two reviewers. For each study, the following data were extracted: first author, publication year, publication country, study design, study name, population characteristics (sex and age), follow-up period, sample size, type of vitamin B12 intake (total intake=dietary intake plus dietary supplements; dietary intake; supplemental intake), measures and ranges of exposure, adjusted variables, and risk estimates with corresponding 95 % CI for each category. Risk estimates that reflected the greatest degree of adjustment for potential confounders were extracted.
The quality of the included studies was assessed according to the Newcastle–Ottawa Scale criteria( 16 ). A funnel plot was used to assess publication bias. Any disagreement on the data extraction and quality assessment of the studies was resolved through comprehensive discussion.
Statistical analysis
A dose–response analysis was first utilized to assess the relationship between vitamin B12 intake and blood vitamin B12 level and CRC risk using the generalized least squares method because this method can resolve the problem that the included studies used different FFQ( 17 , 18 ). The analytical method required the distribution of case and person-years, median values of vitamin B12 intake or blood vitamin B12 levels, and corresponding risk estimates in each category for each study. If there were results on both dietary and total B12 intake in one study, we used the results on total vitamin B12 intake for the main analyses. For each study, the mean or median value of vitamin B12 intake and blood vitamin B12 level in each category was assigned to each corresponding risk estimate. The assigned value of the lowest category was designated as a reference level. If the study did not provide mean or median values of exposure, the midpoint of the upper and lower boundaries in each category was assigned the median value of exposure( 19 ). For the open-ended exposure categories, the length of the open-ended interval was assumed to be the same as that of the adjacent interval( 20 ).
We used random-effect restricted cubic splines with three knots at the 25 %, 50 % and 75 % percentiles of the distribution to examine a potential non-linear dose–response relationship between vitamin B12 intake and blood vitamin B12 level with CRC risk( 21 , 22 ). A P value for non-linearity was calculated by testing the null hypothesis that the regression coefficient of the second spline was equal to zero( 21 ). To test and verify the non-linear model, a meta-analysis comparing the appropriate open categories (or highest category) of exposure with the lowest category was performed. The non-linear dose–response relationship was also verified by several representative point values and the risk estimates of a subgroup analysis based on the range of exposure.
RR was used as a measure of the association between exposure and CRC risk. The OR provided by a case–control study was used as an estimate of the RR because the incidence of CRC was sufficiently rare( 23 ). When the category-specific risk estimates were reported separately for different polymorphisms in the 5,10-methylenetetrahydrofolate reductase gene, we combined the multiple risk estimates into a pooled estimate using a fixed-effects model for further meta-analysis( 24 ). When the RR of different population databases or population sex were presented in one study, we considered each population database as one study.
The heterogeneity of the studies was evaluated using the Cochran Q test and the I 2 statistic( 25 ). A P value <0·10 for the Q statistic and/or I 2>50 % were considered to indicate statistically significant heterogeneity. The random-effects model was then used if there was significant heterogeneity; otherwise, the fixed-effects model was used. A subgroup analysis and a Galbraith plot were used to explore the sources of heterogeneity. Publication bias was evaluated with Egger’s and Begg’s tests( 26 , 27 ), and a trim-and-fill analysis was conducted if publication bias was detected( 28 ). To explore the association between sample size and the RR for CRC, we constructed a sampling-based scatter plot graphically summarizing the association by modelling sample size as a continuous variable. A subgroup analysis was performed based on sample size. We also conducted subgroup analyses stratified by geographic region, type of vitamin B12 taken, range of exposure and tumour site.
A two-sided P value <0·05 was considered statistical significance. All statistical analyses were conducted using the statistical software package Stata version 12·0.
Results
Selection of studies
Figure 1 shows a detailed flowchart of the literature search and the study inclusion procedure. A total of 1362 studies were initially identified with this literature search, but 1231 studies were excluded after screening the titles and abstracts. Then 114 studies were excluded after a full text review. Finally, fourteen studies of vitamin B12 intake and three studies of blood vitamin B12 level were identified as eligible for our meta-analysis.
Fig. 1.
Flowchart showing the literature search and study selection
Study characteristics
The fourteen studies of vitamin B12 intake included a total of 9693 cases (Table 1)( 11 – 14 , 29 – 38 ). The studies were conducted in Asia, Europe and North America, and were published between 2008 and 2013. In terms of the study design, five studies were cohort studies( 13 , 29 , 31 , 34 , 36 ) and nine studies were case–control studies. Of these fourteen studies, eight studies provided results only for dietary vitamin B12 intake from food( 11 , 30 – 34 , 37 , 38 ), three studies provided results only for total vitamin B12 ( 13 , 35 , 36 ), and three studies provided both total and dietary vitamin B12 intakes( 12 , 14 , 29 ). Besides the main end point of CRC, six studies provided sub-cohort results for colon cancer and/or rectal cancer separately( 11 , 13 , 14 , 31 , 32 , 36 ). Three case–control studies on blood vitamin B12 levels were conducted in the USA and Sweden, and were published between 2008 and 2010, comprising a total of 908 cases and 2387 controls (Table 2)( 39 – 41 ).
Table 1.
The baseline characteristics of included studies on vitamin B12 intake and colorectal cancer
| Article | Country | Study name | Study type | Age (years) | Cancer site | No. of participants (M/F) | No. of cases | Vitamin B12 type* | Exposure dose (µg/d)† | Study quality‡ | Adjusted variables |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Zschabitz et al. (2013)( 29 ) | Germany | Women’s Health Initiative Observational Study | Cohort | 50–79 | CRC | F: 86 820 | 808 | TotalDietary | ≤5·13 (Q1), >12·87 (Q4)≤3·52 (Q1), >7·27 (Q4) | 7 | Age, BMI, race/ethnicity, past medical history of colonoscopy, smoking status, PA, postmenopausal hormone use |
| Morita et al. (2013)( 30 ) | Japan | Fukuoka Colorectal Cancer Study | CCS | 20–74 | CRC | M: 992 F: 639 | 816 | Dietary | 4·52 (Q1), 13·49 (Q5) | 6 | Sex, age, resident area, cigarette smoking, alcohol consumption, BMI, type of job, leisure-time PA, parental CRC, dietary intakes of Ca and n-3 PUFA |
| Bassett et al. (2013)( 31 ) | Australia | Melbourne Collaborative Cohort Study | Cohort | 27–80 | CRC | M: 17 045 F: 24 469 | 910 | Dietary | 5·85 (Q5), 1·75 (Q1) | 7 | Country of birth, sex, education, alcohol consumption, PA, smoking status, family history of cancer, intake of cereal fibre |
| CC | M: 17 045 F: 24 470 | 581 | |||||||||
| RC | M: 17 045 F: 24 471 | 326 | |||||||||
| Liu et al. (2013)( 32 ) | USA | Kaiser Permanente Medical Care Program | CCS | 30–79 | CC | M: 1944 F: 1639 | 982 | Dietary | ≤4·2 (T1), >6·9 (T3) | 6 | Age, sex, race/ethnicity, centre, BMI, lifetime vigorous activity, energy intake, dietary fibre, dietary Ca, cigarettes smoked, NSAID use, HRT. Participants with missing data for any of these variables were excluded |
| Key et al. (2012)( 33 ) | UK | UK Dietary Cohort Consortium | CCS | Median: 62 | CRC | M: 1246 F: 1270 | 565 | Dietary | M: <3·34 (Q1), >6·42 (Q4) F: <2·46 (Q1), >5·03 (Q4) | 6 | Age, date of diary and sex, and adjusted for exact age, height, weight, energy intake, alcohol intake, fibre intake, smoking, education, social class, PA |
| Williams et al. (2010)( 12 ) | USA | North Carolina Colon Cancer Study–Phase II | CCS | 40–79 | CRC: WA | M: 904 F: 616 | 720 | Dietary | 2·5 (Q1), 9·5 (Q4) | 6 | Age, sex, education, BMI, family history of CRC, NSAID use, total energy |
| AA | M: 201 F: 183 | 225 | |||||||||
| WA | M: 904 F: 616 | 720 | Total | 3·3 (Q1), 13·5 (Q4) | |||||||
| AA | M: 201 F: 183 | 225 | |||||||||
| Shrubsole et al. (2009)( 34 ) | USA | Shanghai Women’s Health Study | Cohort | 40–70 | CRC | F: 74 942 | 431 | Dietary | 1·28 (Q1), 4·86 (Q5) | 5 | Age, educational attainment, baseline household income, smoking status, drinking status, PA, HRT, menopausal status, family history of CRC, BMI, NSAID use, use of a B-vitamin supplement, history of colorectal polyps, diabetes history, daily intakes of energy, vegetables, fruits, red meats and Ca |
| Sharp et al. (2008)( 35 ) | UK | Grampian Health Board | CCS | NR | CRC | M: 360 F: 312 | 264 | Total | ≤5·25 (Q1), ≥7·98 (Q4) | 7 | Sex, age, total energy, PA, family history of CRC, regular use of any NSAID, sex × NSAID interaction term; model for protein also adjusted for type of dietary supplement; model for alcohol also adjusted for type of dietary supplement and protein |
| Schernhammer et al. (2008)( 36 ) | USA | Health Professionals Follow-up StudyNurses’ Health Study | Cohort | Median: 54Median: 47 | CC | M: 47 371F: 88 691 | 277389 | Total | ≤6·0 (Q1), >16·1 (Q5)≤4·0 (Q1), >11·1 (Q5) | 6 | Age, energy intake, sex, screening sigmoidoscopy, family history of CRC, aspirin use, smoking, PA in MET, BMI, history of colon polyps, beef intake, Ca intake, multivitamin use, and baseline folate, vitamin B6, B12, methionine and alcohol if not primary exposure |
| Murtaugh et al. (2007)( 14 ) | USA | Kaiser Permanente Medical Care Program | CCS | 30–79 | RC | M: 1000 F: 730 | 941 | Total | ≤6·09 (T1), >11·17 (T3) | 7 | Age, sex BMI, activity, energy, fibre, Ca, ibuprofen use, smoking (pack-years) |
| Dietary | ≤3·92 (T1), >6·57 (T3) | ||||||||||
| Kune and Watson (2006)( 11 ) | Australia | Melbourne Colorectal Cancer Study | CCS | NR | CRCCCRC | 144211191050 | 715392323 | Dietary | <4·1 (Q1), >11·1 (Q5) | 5 | Age, sex, alcohol, BMI, energy intake, family history of CRC, oral contraceptive pill use, cigarette pack-years, aspirin use, non-aspirin NSAID use |
| Otani et al. (2005)( 38 ) | Japan | NR | CCS | 20–74 | CRC | M: 207 F: 124 | 107 | Dietary | ≤7·3 (T1), >11·2 (T3) | 7 | Smoking, alcohol consumption, BMI, total dietary fibre intake |
| Le Marchand et al. (2005)( 37 ) | USA | Multiethnic Cohort | CCS | 45–75 | CRC | M: 1605 F: 1238 | 822 | Dietary | ≤2·86 (T1), >4·99 (T3) | 5 | Age at blood draw, sex, race/ethnicity |
| Harnack et al. (2002)( 13 ) | USA | Iowa Women’s Health Study | Cohort | 55–69 | CCRC | F: 35 216 | 598123 | Total | <5·13 (Q1), >18·35 (Q5) | 6 | Age, pack-years of cigarettes, BMI, oestrogen use, intakes of Ca, vitamin E and energy |
| <7·18 (T1), >14·66 (T3) |
M, males; F, females; CCS, case–control study; NR, not reported; CRC, colorectal cancer; CC, colon cancer; RC, rectal cancer; WA, white Americans; AA, African Americans; Q, quartile/quintile; T, tertile; PA, physical activity; NSAID, non-steroidal anti-inflammatory drug; HRT, hormone replacement therapy; MET, metabolic equivalent of task.
Dietary vitamin B12 intake included vitamin B12 intake from foods only, and total vitamin B12 intake included vitamin B12 intake from foods and supplements.
Exposure dose means the cut-off points or distribution for the highest and lowest categories of vitamin B12 intake.
The quality of the included studies was assessed with the nine-star Newcastle–Ottawa Scale criteria.
Table 2.
The baseline characteristics of included studies on blood vitamin B12 level and colorectal cancer
| Study | Country | Study name | Study type | Age (years) | Cancer site | No. of cases/controls | Exposure dose (pmol/l)* | Study quality† | Adjusted variables |
|---|---|---|---|---|---|---|---|---|---|
| Van Guelpen et al. (2010)( 39 ) | Sweden | VIP Study, MONICA Study and MSP Study | CCS | NR | CRC: CIMP-low/high | 91/181 | M: ≤220 (Q1), >351 (Q4) F: ≤232 (Q1), >392 (Q4) | 6 | BMI, current smoking, recreational PA, alcohol intake |
| 62·5 (range 56·7–66·9) | CRC: CIMP-negative | 92/183 | M: ≤220 (Q1), >351 (Q4) F: ≤232 (Q1), >392 (Q4) | ||||||
| Le Marchand et al. (2009)( 40 ) | USA | Multiethnic Cohort | CCS | 63–76 | CRC | 224/411 | ≤361 (Q1), >722 (Q4) | 7 | Age at blood draw, hours of fasting prior to blood draw, hours of moderate or vigorous PA, processed meat, pack-years, BMI, ethanol, family history of CRC, history of CRC, plasma folate |
| Weinstein et al. (2008)( 41 ) | USA | ATBC Study | CCS | 54–62 | CRC | 275/275 | ≤243 (Q1), >529 (Q5) | 6 | Age, BMI, occupational and leisure PA, intakes of vitamin D and Fe |
VIP, Västerbotten Intervention Programme; MONICA, MONitoring of trends and determinants in CArdiovascular disease; MSP, Mammography Screening Project; ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention; CCS, case–control study; NR, not reported; CRC, colorectal cancer; CIMP, CpG island methylator phenotype; M, males; F, females; Q, quartile/quintile; PA, physical activity.
Exposure dose means the cut-off points or distribution for the highest and lowest categories of blood vitamin B12 level.
The quality of the included studies was assessed with the nine-star Newcastle–Ottawa Scale criteria.
Dose–response association between vitamin B12 intake and colorectal cancer risk
We first evaluated the non-linear dose–response relationship between vitamin B12 intake and CRC risk. Heterogeneity existed (P heterogeneity=0·017) in the overall analysis of vitamin B12 intake, with a significant non-linear dose–response relationship (P non-linearity=0·026; Fig. 2). In the non-linear dose–response relationship, there was a slight trend towards a reduction in CRC risk when vitamin B12 intake was less than 12·85 μg/d, although the reduction was not statistically significant, and a significantly reduced risk was observed when vitamin B12 intake was more than 12·85 μg/d. A slight reduction in CRC risk was also observed with every 4·5 μg/d increment in vitamin B12 intake (RR=0·961; 95 % CI 0·930, 0·994) when we evaluated the relationship in a separate analysis with a linear model, suggesting that similar trends were observed with linear and non-linear models.
Fig. 2.
Dose–response relationship between vitamin B12 intake and risk of colorectal cancer. Relative risks (RR; ———) and the corresponding 95 % CI (— — —) were summarized for the dose–response relationship between vitamin B12 intake (μg/d) and risk of colorectal cancer. Data were modelled with random-effects restricted cubic spline models, where ---- represents the linear trend
In addition, dose–response analyses for the association between vitamin B12 intake and CRC risk were stratified based on the type of vitamin B12 intake (total intake and dietary intake). There was no evidence of a non-linear association between total vitamin B12 intake and CRC risk (P non-linearity=0·690), and every 4·5 μg/d increment in total vitamin B12 intake was inversely associated with CRC risk (RR=0·963; 95 % CI 0·928, 0·999) without significant heterogeneity (P heterogeneity=0·138). For dietary vitamin B12 intake there was significant evidence of non-linearity, and every 4·5 μg/d increment in dietary vitamin B12 intake was inversely associated with CRC risk (RR=0·914; 95 % CI 0·856, 0·977) under the linear model.
The non-linear dose–response relationship (Fig. 2) graphically showed that vitamin B12 intake above a certain threshold (high dose, i.e. >12 μg/d) was inversely associated with CRC risk. We used several representative point values to test and verify the non-linear dose–response relationship by pooling the CRC risks for appropriate open categories of vitamin B12 intake when the lowest category in each study was regarded as a reference level. Our results indicated a significantly decreased risk of CRC when the vitamin B12 intake was above a certain threshold (8·5 μg/d: RR=0·898; 95 % CI 0·824, 0·979; 11 μg/d: RR=0·843; 95 % CI 0·731, 0·971; or 13 μg/d: RR=0·881; 95 % CI 0·802, 0·968), thereby enhancing the non-linear dose–response relationship.
Dose–response association between blood vitamin B12 level and colorectal cancer risk
There was no heterogeneity (P heterogeneity=0·327) in the overall analysis of blood vitamin B12 levels, with an insignificant non-linear dose–response relationship between blood vitamin B12 level and CRC risk (P non-linearity=0·219). The results of the linear model indicated that every 150 pmol/l increment in blood vitamin B12 level was not associated with CRC risk (RR=1·023; 95 % CI 0·881, 1·187) without significant heterogeneity (P heterogeneity=0·303).
High v. low vitamin B12 intake or blood vitamin B12 level
In the crude analyses, the highest vitamin B12 intake or blood vitamin B12 level v. the lowest intake or blood level was insignificantly associated with CRC risk (vitamin B12 intake: RR=0·942; 95 % CI 0·829, 1·070; blood vitamin B12 level: RR=0·927; 95 % CI 0·560, 1·534). We also performed in-depth subgroup analysis of the vitamin B12 intake. The results indicated that the association between vitamin B12 intake and CRC risk was stronger in studies with a wider range of vitamin B12 intake (>8 μg/d difference in assigned value compared with the reference level: RR=0·831; 95 % CI 0·711, 0·971; Fig. 3) compared with studies with a narrower range of vitamin B12 intake (≤8 μg/d: RR=1·016; 95 % CI 0·878, 1·175), thereby enhancing the non-linear dose–response relationship. The highest v. lowest total vitamin B12 intake was significantly associated with CRC risk (RR=0·870; 95 % CI 0·782, 0·967).
Fig. 3.
Adjusted relative risk (RR) of colorectal cancer for a wider range of vitamin B12 intake (range >8 μg/d); the adjusted RR was summarized for the association between a wider range of vitamin B12 intake and risk of colorectal cancer. The study-specific RR and 95 % CI are represented by the black dot and horizontal line, respectively; the area of the grey square is proportional to the specific-study weight to the overall meta-analysis. The centre of the open diamond presents the pooled RR risk and its width represents the pooled 95 % CI (NHS, Nurses’ Health Study; HPFS, Health Professionals Follow-up Study; CC, colon cancer; RC, rectal cancer)
The association between sample size and the RR of CRC is summarized in Fig. 4 with a sampling-based scatter plot. The subgroup analysis based on sample size showed that the sample sizes of the studies did not obviously influence our results. Detailed subgroup analyses based on CRC sites, sex of participants, regions of participants and the adjusted covariates were also conducted, and the results are summarized in Table 3.
Fig. 4.
The association between sample size and the relative risk (RR) of colorectal cancer; a sampling-based scatter plot summarized the association between sample size and the RR of colorectal cancer
Table 3.
The results of subgroup analyses for the relationship between vitamin B12 intake and colorectal cancer risk
| RR‡ | |||||||
|---|---|---|---|---|---|---|---|
| Subgroups | No. of studies | No. of independent cohorts* | Effect model† | Estimate | 95 % CI | P § | I 2 (%)|| |
| All studies | 14 | 17 | Random | 0·942 | 0·829, 1·070 | 0·357 | 55·60 |
| Type of vitamin B12 intake | |||||||
| Total | 6 | 9 | Fixed | 0·870 | 0·782, 0·967 | 0·010 | 22·90 |
| Food | 11 | 12 | Random | 0·960 | 0·810, 1·137 | 0·635 | 67·00 |
| Range of exposure | |||||||
| ≤8 µg/d | 10 | 11 | Random | 1·016 | 0·878, 1·175 | 0·835 | 49·50 |
| >8 µg/d | 7 | 10 | Random | 0·831 | 0·711, 0·971 | 0·020 | 53·20 |
| Study type | |||||||
| Cohort | 5 | 7 | Fixed | 0·962 | 0·862, 1·074 | 0·488 | 32·30 |
| Case–control | 9 | 10 | Random | 0·925 | 0·748, 1·144 | 0·472 | 66·00 |
| Cancer type | |||||||
| Colon | 5 | 6 | Random | 0·867 | 0·655, 1·149 | 0·321 | 73·30 |
| Rectal | 4 | 4 | Random | 0·958 | 0·669, 1·372 | 0·815 | 70·20 |
| Sample size | |||||||
| ≤3000 | 9 | 10 | Random | 0·925 | 0·748, 1·144 | 0·472 | 66·00 |
| >3000 | 5 | 7 | Fixed | 0·962 | 0·862, 1·074 | 0·488 | 32·30 |
| Sex | |||||||
| Female | 5 | 6 | Fixed | 0·995 | 0·872, 1·135 | 0·940 | 35·70 |
| Regions of participants | |||||||
| USA (all¶) | 7 | 10 | Random | 0·939 | 0·802, 1·100 | 0·436 | 53·60 |
| USA (total**) | 5 | 8 | Fixed | 0·866 | 0·777, 0·966 | 0·010 | 31·80 |
| USA (food††) | 5 | 6 | Random | 0·979 | 0·761, 1·260 | 0·868 | 70·60 |
| Europe (all) | 4 | 4 | Random | 0·849 | 0·601, 1·198 | 0·352 | 77·00 |
| Europe (total) | 1 | 1 | NA | 0·950 | 0·560, 1·630 | NA | NA |
| Europe (food) | 3 | 3 | Random | 0·823 | 0·538, 1·261 | 0·371 | 84·70 |
| Asia (all) | 3 | 3 | Fixed | 1·123 | 0·872, 1·445 | 0·369 | 0·00 |
| Asia (total) | NA | NA | NA | NA | – | NA | NA |
| Asia (food) | 3 | 3 | Fixed | 1·123 | 0·872, 1·445 | 0·369 | 0·00 |
| Adjustment for confounders | |||||||
| Smoking | |||||||
| Yes | 11 | 13 | Random | 0·954 | 0·830, 1·097 | 0·511 | 56·50 |
| No | 3 | 4 | Random | 0·903 | 0·634, 1·286 | 0·572 | 61·60 |
| Alcohol | |||||||
| Yes | 7 | 7 | Random | 0·943 | 0·747, 1·188 | 0·617 | 63·20 |
| No | 7 | 10 | Random | 0·939 | 0·802, 1·100 | 0·436 | 53·60 |
| Physical activity | |||||||
| Yes | 8 | 9 | Random | 0·967 | 0·875, 1·069 | 0·517 | 39·60 |
| No | 6 | 8 | Random | 0·894 | 0·700, 1·141 | 0·368 | 67·00 |
| Family history | |||||||
| Yes | 5 | 7 | Random | 0·806 | 0·614, 1·058 | 0·120 | 69·20 |
| No | 9 | 10 | Random | 1·002 | 0·911, 1·103 | 0·966 | 0·00 |
| Vitamin B12 intake (μg/d) | |||||||
| ≥8 | 10 | 27 | Random | 0·919 | 0·842, 1·004 | 0·061 | 42·20 |
| ≥8·5 | 9 | 25 | Random | 0·898 | 0·824, 0·979 | 0·014 | 35·90 |
| ≥9 | 8 | 23 | Random | 0·896 | 0·816, 0·985 | 0·023 | 41·20 |
| ≥9·5 | 8 | 20 | Random | 0·901 | 0·812, 1·001 | 0·051 | 46·30 |
| ≥10 | 8 | 17 | Random | 0·890 | 0·793, 1·000 | 0·050 | 51·00 |
| ≥10·5 | 8 | 16 | Random | 0·889 | 0·786, 1·005 | 0·061 | 54·10 |
| ≥11 | 8 | 13 | Random | 0·843 | 0·731, 0·971 | 0·018 | 53·70 |
| ≥11·5 | 8 | 13 | Random | 0·843 | 0·731, 0·971 | 0·018 | 53·70 |
| ≥12 | 8 | 13 | Random | 0·843 | 0·731, 0·971 | 0·018 | 53·70 |
| ≥12·5 | 8 | 13 | Random | 0·843 | 0·731, 0·971 | 0·018 | 53·70 |
| ≥13 | 7 | 12 | Fixed | 0·881 | 0·802, 0·968 | 0·008 | 34·30 |
| ≥13·5 | 5 | 9 | Random | 0·858 | 0·736, 1·000 | 0·050 | 49·50 |
RR, relative risk; NA, not applicable.
The actual number of ‘independent study cohorts’ that can be included in the corresponding analysis because we considered each population database as one study cohort for statistical analyses if the RR of different populations or population sex were available (one study); thus the number of independent study cohorts can be larger than the number of studies.
Random-effects model was used if there was significant heterogeneity; otherwise, fixed-effects model was used.
RR for colorectal cancer risk of the highest v. lowest categories of vitamin B12 intake.
P for the RR.
I 2 shows the degree of heterogeneity among studies.
‘All’ means that the analysis included either total vitamin B12 intake or dietary vitamin B12 intake.
‘Total’ means that the analysis included only total vitamin B12 intake.
‘Food’ means that the analysis included only dietary vitamin B12 intake.
Publication bias
Begg’s test and Egger’s test showed no evidence of publication bias for the overall analysis of vitamin B12 intake (P Begg=0·192, P Egger=0·266) and blood vitamin B12 level (P Begg=1·000, P Egger=0·447), and the subgroup analysis of total vitamin B12 intake (P Begg=0·348, P Egger=0·616).
Discussion
The one-carbon metabolism pathway requires adequate vitamin B12 and this raises the possibility that vitamin B12 may have an important role in CRC risk. However, the association between vitamin B12 and CRC risk is still under debate due to a lack of sufficient evidence. To the best of our knowledge, our present dose–response meta-analysis is the first study to systematically evaluate the association between vitamin B12 and CRC risk.
In crude overall analyses, vitamin B12 intake or blood vitamin B12 level was insignificantly associated with CRC risk. Interestingly, current results found a non-linear dose–response relationship between vitamin B12 intake and CRC risk. The association was insignificant if vitamin B12 intake was under a certain threshold (low dosage, i.e. <7 μg/d), whereas vitamin B12 intake above a certain threshold (high dosage, i.e. >12 μg/d) was inversely associated with CRC risk. Moreover, this non-linear dose–response relationship was enhanced by several representative point values and the risk estimates of subgroup analysis based on range of exposure (Fig. 3).
The biological mechanisms responsible for the protective effect of high-dosage vitamin B12 are unclear. One possible explanation is that vitamin B12 is an essential coenzyme in one-carbon metabolism for methylation reactions, nucleotide biosynthesis and DNA repair through transforming homocysteine into methionine. Indeed, Choi et al. reported that the colonic DNA of vitamin B12-deficient rats has a 35 % decrease in genomic methylation and a 105 % increase in uracil incorporation, which might increase susceptibility to carcinogenesis( 42 ). Furthermore, B-group vitamin supplementation, including vitamin B12, may have antioxidant and anti-inflammatory effects( 43 , 44 ). In addition, animal experiments have shown that vitamin B12 can inhibit the proliferation of cancer cells( 45 ).
We did not find statistically significant evidence for a non-linear relationship between blood vitamin B12 level and CRC risk (P non-linearity=0·219). Every 150 pmol/l increment in blood vitamin B12 level was not associated with CRC risk, enhanced by the analysis of the highest blood vitamin B12 levels v. lowest blood levels. One explanation may be that all included studies roughly assessed blood vitamin B12 level in relation to CRC risk while ignoring the fact that only methylcobalamin and 5′-deoxyadenosylcobalamin are the principal active coenzyme forms of vitamin B12 compared with other forms of vitamin B12 in blood (e.g. cyanocobalamin and hydroxocobalamin)( 46 – 48 ). Thus, these inactive forms of blood vitamin B12 may underestimate the effect of active forms of vitamin B12 on CRC risk. Similar to studies regarding pyridoxal 5′-phosphate (PLP, the active form of vitamin B6), the association between the active forms of vitamin B12 and CRC risk should be deeply investigated by further multicentre clinical studies.
Our meta-analysis indicated no association between blood vitamin B12 level and CRC risk, different from the association found between vitamin B12 intake and CRC risk. One potential reason may be that plasma vitamin B12 level has been found to be only slightly correlated with dietary vitamin B12 intake (multivariable Pearson correlation coefficient, r=0·08) and total vitamin B12 intake (r=0·25)( 49 ).
The strength of the current meta-analysis study was that our results and conclusions were enhanced by an in-depth subgroup analysis. A subgroup analysis based on the type of vitamin B12 intake indicated that the association between total vitamin B12 and CRC risk was stronger than the association between dietary vitamin B12 intake alone and CRC risk (Table 3). As expected, heterogeneity was reduced to low levels in the assessment of total vitamin B12 intake. Supplemental vitamin B12 intake may play an important role in CRC risk, and Zschabitz et al. showed that supplemental vitamin B12 intake was marginally associated with CRC risk( 29 ). However, the interaction between supplemental vitamin B12 intake and dietary vitamin B12 intake is unclear. Thus, the type of vitamin B12 intake may be important for the assessment of vitamin B12 intake and CRC risk, and future clinical studies should not ignore the role of supplemental vitamin B12 intake. The association between vitamin B12 intake and CRC risk may be various among different races of population (Table 3), in agreement with the study by Williams et al. in which the associations between vitamin B12 and distal CRC differed for whites compared with African Americans( 12 ). This may be due to polymorphisms in genes related to one-carbon metabolism, supported by a previous study in which gene–nutrient interactions in folate-mediated one-carbon metabolism played an important role in modifying the risk of CRC( 32 ). No evidence of publication bias was detected, indicating that the entire pooled results may be unbiased.
Several limitations of our meta-analysis must be acknowledged. First, the controls were not always of uniform ascertainment in case–control studies, although most studies matched controls to cases based on age and sex. Some controls might therefore have had benign diseases or other risk factors that contribute to CRC risk and misclassifications of controls or vitamin B12 intake possibly exist among several included studies. Second, some important confounding factors were not measured in several studies and some inherently confounding factors in included studies could not be solved perfectly in our study. Thus, an inadequate control of confounding factors may lead to an underestimation of risk estimates of vitamin B12 on CRC. Third, the limited number of included studies may impact the statistical power and performance of some in-depth subgroup analysis. The assigned value of vitamin B12 in each category (midpoint, mean and median) was not available in all included studies for dose–response analysis. Therefore, various assigned values might affect the accuracy of the dose–response relationship. In addition, a considerable degree of heterogeneity was observed among studies and could not be explained completely. Our subgroup analysis indicated that study design may have been one source of heterogeneity. After the case–control studies were excluded, the heterogeneity among the studies decreased (I 2=32·3 %) and the fixed-effects model was used. Similarly, our results also showed that the types of vitamin B12 intake and population sex contributed to the heterogeneity (total vitamin B12 intake group: I 2=22·9 %; female group: I 2=35·7 %). For individual studies, the result of the Galbraith plot (Fig. 5) indicated that the studies by Liu et al.( 32 ) and Kune and Watson( 11 ) contributed substantial heterogeneity. Indeed, the heterogeneity was clearly reduced after the removal of one or both of these studies (without Liu et al.: I 2=49·5 %; without Kune and Watson: I 2=32·7 %; without Liu et al. and Kune and Watson: I 2=9·4 %). The remaining unexplained heterogeneity may have been caused by differences in the population characteristics and methodological differences (e.g. differences in the estimates of vitamin B12 intake in FFQ and the control of confounding factors). The heterogeneity did not influence or dominate the quality or stability of the results. Despite these limitations, the present study is the first dose–response meta-analysis to quantitatively assess the association between vitamin B12 and CRC risk.
Fig. 5.
Galbraith plot for exploring the sources of heterogeneity in the fourteen studies examining the relationship of vitamin B12 intake and risk of colorectal cancer. ——— represent fitted lines; those at ±2 from the fitted (regression-through-the-origin) line represent the approximate 95 % confidence region; the studies are denoted by the first author’s surname
Conclusion
In conclusion, results from the present meta-analysis indicate that vitamin B12 intake is inversely associated with CRC risk when the dosage of vitamin B12 intake is above a certain threshold, and that the association between total vitamin B12 and CRC risk is stronger than the association between dietary vitamin B12 intake alone and CRC risk. In addition, there was an insignificant association between blood vitamin B12 level and CRC risk. Further studies are required to investigate the associations between the active forms of blood vitamin B12 and CRC risk and to evaluate the influence of vitamin B12 supplementation.
Acknowledgements
Financial support: This work was supported by the Natural Science Foundation of Liaoning Province (grant number 2014029201), Program of Education Department of Liaoning Province (grant number L2014307). The sponsors had no role in study design, data collection, data analysis, data interpretation or writing of the report. Conflict of interest: None. Authorship: N.-H.S., X.-Z.H., S.-B.W. and Z.-N.W. were responsible for the conception and design of the study. N.-H.S. and X.-Z.H. conducted the statistical analyses and wrote the article. S.-B.W., Y.L., L.-Y.W. and H.-C.W. contributed to the literature search, acquisition of data, tables and figures. Z.-N.W. provided clinical expertise and interpretation of the data. C.-W.Z., C.Z. and H.-P.L. provided the draft of the article and statistical expertise. All the investigators have read and approved the final manuscript. Ethics of human subject participation: Ethical approval was not required.
References
- 1. Jemal A, Bray F, Center MM et al. (2011) Global cancer statistics. CA Cancer J Clin 61, 69–90. [DOI] [PubMed] [Google Scholar]
- 2. Potter JD (1999) Colorectal cancer: molecules and populations. J Natl Cancer Inst 91, 916–932. [DOI] [PubMed] [Google Scholar]
- 3. Cunningham C & Dunlop MG (1996) Molecular genetic basis of colorectal cancer susceptibility. Br J Surg 83, 321–329. [DOI] [PubMed] [Google Scholar]
- 4. Hill MJ (1998) Gene–environment interactions in the pathogenesis of colorectal cancer. Eur J Cancer Prev 7, 351–352. [DOI] [PubMed] [Google Scholar]
- 5. Steinmetz KA & Potter JD (1996) Vegetables, fruit, and cancer prevention: a review. J Am Diet Assoc 96, 1027–1039. [DOI] [PubMed] [Google Scholar]
- 6. Vidal AC, Grant DJ, Williams CD et al. (2012) Associations between intake of folate, methionine, and vitamins B-12, B-6 and prostate cancer risk in American veterans. J Cancer Epidemiol 2012, 957467. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Lim U, Schenk M, Kelemen LE et al. (2005) Dietary determinants of one-carbon metabolism and the risk of non-Hodgkin’s lymphoma: NCI-SEER case–control study, 1998–2000. Am J Epidemiol 162, 953–964. [DOI] [PubMed] [Google Scholar]
- 8. Kim YI (2004) Folate and DNA methylation: a mechanistic link between folate deficiency and colorectal cancer? Cancer Epidemiol Biomarkers Prev 13, 511–519. [PubMed] [Google Scholar]
- 9. Williams EA (2012) Folate, colorectal cancer and the involvement of DNA methylation. Proc Nutr Soc 71, 592–597. [DOI] [PubMed] [Google Scholar]
- 10. Strohle A, Wolters M & Hahn A (2005) Folic acid and colorectal cancer prevention: molecular mechanisms and epidemiological evidence. Int J Oncol 26, 1449–1464. [PubMed] [Google Scholar]
- 11. Kune G & Watson L (2006) Colorectal cancer protective effects and the dietary micronutrients folate, methionine, vitamins B6, B12, C, E, selenium, and lycopene. Nutr Cancer 56, 11–21. [DOI] [PubMed] [Google Scholar]
- 12. Williams CD, Satia JA, Adair LS et al. (2010) Antioxidant and DNA methylation-related nutrients and risk of distal colorectal cancer. Cancer Causes Control 21, 1171–1181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Harnack L, Jacobs DR Jr, Nicodemus K et al. (2002) Relationship of folate, vitamin B-6, vitamin B-12, and methionine intake to incidence of colorectal cancers. Nutr Cancer 43, 152–158. [DOI] [PubMed] [Google Scholar]
- 14. Murtaugh MA, Curtin K, Sweeney C et al. (2007) Dietary intake of folate and co-factors in folate metabolism, MTHFR polymorphisms, and reduced rectal cancer. Cancer Causes Control 18, 153–163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Ebbing M, Bonaa KH, Nygard O et al. (2009) Cancer incidence and mortality after treatment with folic acid and vitamin B12 . JAMA 302, 2119–2126. [DOI] [PubMed] [Google Scholar]
- 16. Stang A (2010) Critical evaluation of the Newcastle–Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 25, 603–605. [DOI] [PubMed] [Google Scholar]
- 17. Greenland S & Longnecker MP (1992) Methods for trend estimation from summarized dose–response data, with applications to meta-analysis. Am J Epidemiol 135, 1301–1309. [DOI] [PubMed] [Google Scholar]
- 18. Orsini N, Bellocco R & Greenland S (2006) Generalized least squares for trend estimation of summarized dose–response data. Stata J 6, 40–57. [Google Scholar]
- 19. Berlin JA, Longnecker MP & Greenland S (1993) Meta-analysis of epidemiologic dose–response data. Epidemiology 4, 218–228. [DOI] [PubMed] [Google Scholar]
- 20. Keum N, Aune D, Greenwood DC et al. (2014) Calcium intake and colorectal cancer risk: dose–response meta-analysis of prospective observational studies. Int J Cancer 135, 1940–1948. [DOI] [PubMed] [Google Scholar]
- 21. Orsini N, Li R, Wolk A et al. (2012) Meta-analysis for linear and nonlinear dose–response relations: examples, an evaluation of approximations, and software. Am J Epidemiol 175, 66–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Bagnardi V, Zambon A, Quatto P et al. (2004) Flexible meta-regression functions for modeling aggregate dose–response data, with an application to alcohol and mortality. Am J Epidemiol 159, 1077–1086. [DOI] [PubMed] [Google Scholar]
- 23. Greenland S (1987) Quantitative methods in the review of epidemiologic literature. Epidemiol Rev 9, 1–30. [DOI] [PubMed] [Google Scholar]
- 24. Orsini N (2010) From floated to conventional confidence intervals for the relative risks based on published dose–response data. Comput Methods Programs Biomed 98, 90–93. [DOI] [PubMed] [Google Scholar]
- 25. Higgins JP, Thompson SG, Deeks JJ et al. (2003) Measuring inconsistency in meta-analyses. BMJ 327, 557–560. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Egger M, Davey Smith G, Schneider M et al. (1997) Bias in meta-analysis detected by a simple, graphical test. BMJ 315, 629–634. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Begg CB & Mazumdar M (1994) Operating characteristics of a rank correlation test for publication bias. Biometrics 50, 1088–1101. [PubMed] [Google Scholar]
- 28. Duval S & Tweedie R (2000) Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 56, 455–463. [DOI] [PubMed] [Google Scholar]
- 29. Zschabitz S, Cheng TY, Neuhouser ML et al. (2013) B vitamin intakes and incidence of colorectal cancer: results from the Women’s Health Initiative Observational Study cohort. Am J Clin Nutr 97, 332–343. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Morita M, Yin G, Yoshimitsu S et al. (2013) Folate-related nutrients, genetic polymorphisms, and colorectal cancer risk: the Fukuoka colorectal cancer study. Asian Pac J Cancer Prev 14, 6249–6256. [DOI] [PubMed] [Google Scholar]
- 31. Bassett JK, Severi G, Hodge AM et al. (2013) Dietary intake of B vitamins and methionine and colorectal cancer risk. Nutr Cancer 65, 659–667. [DOI] [PubMed] [Google Scholar]
- 32. Liu AY, Scherer D, Poole E et al. (2013) Gene–diet-interactions in folate-mediated one-carbon metabolism modify colon cancer risk. Mol Nutr Food Res 57, 721–734. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Key TJ, Appleby PN, Masset G et al. (2012) Vitamins, minerals, essential fatty acids and colorectal cancer risk in the United Kingdom Dietary Cohort Consortium. Int J Cancer 131, E320–E325. [DOI] [PubMed] [Google Scholar]
- 34. Shrubsole MJ, Yang G, Gao YT et al. (2009) Dietary B vitamin and methionine intakes and plasma folate are not associated with colorectal cancer risk in Chinese women. Cancer Epidemiol Biomarkers Prev 18, 1003–1006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Sharp L, Little J, Brockton NT et al. (2008) Polymorphisms in the methylenetetrahydrofolate reductase (MTHFR) gene, intakes of folate and related B vitamins and colorectal cancer: a case–control study in a population with relatively low folate intake. Br J Nutr 99, 379–389. [DOI] [PubMed] [Google Scholar]
- 36. Schernhammer ES, Giovannuccci E, Fuchs CS et al. (2008) A prospective study of dietary folate and vitamin B and colon cancer according to microsatellite instability and KRAS mutational status. A prospective study of dietary folate and vitamin B and colon cancer according to microsatellite instability and KRAS mutational status. Cancer Epidemiol Biomarkers Prev 17, 2895–2898. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Le Marchand L, Wilkens LR, Kolonel LN et al. (2005) The MTHFR C677T polymorphism and colorectal cancer: the multiethnic cohort study. Cancer Epidemiol Biomarkers Prev 14, 1198–1203. [DOI] [PubMed] [Google Scholar]
- 38. Otani T, Iwasaki M, Hanaoka T et al. (2005) Folate, vitamin B6, vitamin B12, and vitamin B2 intake, genetic polymorphisms of related enzymes, and risk of colorectal cancer in a hospital-based case–control study in Japan. Nutr Cancer 53, 42–50. [DOI] [PubMed] [Google Scholar]
- 39. Van Guelpen B, Dahlin AM, Hultdin J et al. (2010) One-carbon metabolism and CpG island methylator phenotype status in incident colorectal cancer: a nested case–referent study. Cancer Causes Control 21, 557–566. [DOI] [PubMed] [Google Scholar]
- 40. Le Marchand L, White KK, Nomura AMY et al. (2009) Plasma levels of B vitamins and colorectal cancer risk: the multiethnic cohort study. Cancer Epidemiol Biomarkers Prev 18, 2195–2201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Weinstein SJ, Albanes D, Selhub J et al. (2008) One-carbon metabolism biomarkers and risk of colon and rectal cancers. Cancer Epidemiol Biomarkers Prev 17, 3233–3240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Choi SW, Friso S, Ghandour H et al. (2004) Vitamin B-12 deficiency induces anomalies of base substitution and methylation in the DNA of rat colonic epithelium. J Nutr 134, 750–755. [DOI] [PubMed] [Google Scholar]
- 43. Ullegaddi R, Powers HJ & Gariballa SE (2004) B-group vitamin supplementation mitigates oxidative damage after acute ischaemic stroke. Clin Sci (Lond) 107, 477–484. [DOI] [PubMed] [Google Scholar]
- 44. Ullegaddi R, Powers HJ & Gariballa SE (2006) Antioxidant supplementation with or without B-group vitamins after acute ischemic stroke: a randomized controlled trial. JPEN J Parenter Enteral Nutr 30, 108–114. [DOI] [PubMed] [Google Scholar]
- 45. Nishizawa Y, Yamamoto T, Terada N et al. (1997) Effects of methylcobalamin on the proliferation of androgen-sensitive or estrogen-sensitive malignant cells in culture and in vivo . Int J Vitam Nutr Res 67, 164–170. [PubMed] [Google Scholar]
- 46. Donaldson MS (2004) Nutrition and cancer: a review of the evidence for an anti-cancer diet. Nutr J 3, 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47. Suarez-Suarez A, Tovar-Sanchez A & Rossello-Mora R (2011) Determination of cobalamins (hydroxo-, cyano-, adenosyl- and methyl-cobalamins) in seawater using reversed-phase liquid chromatography with diode-array detection. Anal Chim Acta 701, 81–85. [DOI] [PubMed] [Google Scholar]
- 48. Jones AR, Russell HJ, Greetham GM et al. (2012) Ultrafast infrared spectral fingerprints of vitamin B12 and related cobalamins. J Phys Chem A 116, 5586–5594. [DOI] [PubMed] [Google Scholar]
- 49. Zhang SM, Willett WC, Selhub J et al. (2003) Plasma folate, vitamin B6, vitamin B12, homocysteine, and risk of breast cancer. J Natl Cancer Inst 95, 373–380. [DOI] [PubMed] [Google Scholar]