Vitamin B6 and Cancer Risk: A Field Synopsis and Meta-analysis

Simone Mocellin; Marta Briarava; Pierluigi Pilati

Disclosures

J Natl Cancer Inst. 2016;109(3) 

In This Article

Results

Search Findings

The literature search yielded 949 articles: Their screening led to the identification of 132 eligible articles reporting on 155 data sets (146 from observational studies and nine from randomized controlled trials) (Figure 1). The full list of included studies along with their main features and their references is available online in the Supplementary Materials (available online): as Supplementary Table 1 (vitamin intake), Supplementary Table 2 (blood PLP), and Supplementary Table 3 (RCTs).

Figure 1.

Flow chart of the literature search.

Overall, data were available from 1 959 417 participants, including 98 975 cases affected with cancers originating from the following 19 sites: breast (n = 33 934, 34.3%), colorectal (26 220, 26.5%), ovary (n = 5816, 5.9%), prostate (n = 5135, 5.2%), immune system (n = 3837, 3.9%), unspecified site (n = 3502, 3.5%), endometrium (n = 3030, 3.1%), lung (n = 3024, 3.1%), stomach (n = 2937, 3.0%), esophagus (n = 2684, 2.7%), pancreas (n = 2290, 2.3%), kidney (n = 2256, 2.3%), bladder (n = 1559, 1.6%), oral cavity (n = 754,0.8%), nasopharynx (n = 600,0.6%), larynx (n = 527,0.5%), cervix (n = 357,0.4%), liver (n = 297,0.3%), and brain (n = 216,0.2%). Data came from 121 observational studies (participants, n = 1 924 506, cases, n = 96 436) and nine randomized controlled trials (RCTs; participants, n = 34 911; cases, n = 2539).

Observational Studies: Vitamin B6 Intake

Vitamin B6 intake and cancer risk were assessed in 121 studies (participants, n = 1 902 712; cases, n = 85 734), the design being retrospective and prospective in 73 (60.3%) and 48 studies, respectively (Supplementary Table 1, available online). Based on the Newcastle-Ottawa Scale tool, study quality was maximal (nine stars) in most cases (n = 70, 57.8%); lower-quality studies were graded with eight (n = 42, 34.7%), seven (n = 6, 4.9%), and six stars (n = 3, 2.5%), respectively, the reasons for downgrading being suboptimal representativeness (n = 33, 64.7%) or comparability (n = 15, 29.4%), and mixed (n = 3, 5.9%), respectively.

High vs Low Category Meta-analysis

Considering dietary intake (studies, n = 95; cases, n = 34 861), a full list of results is available online (see Supplementary Table 4 and Supplementary Figure 1, available online), whereas a summary of main findings is reported in Figure 2. Higher vitamin B6 doses were associated with a statistically significantly reduced risk of any type of cancer (RR = 0.78, 95% CI = 0.73 to 0.84). There was no evidence of small study effect (Egger's test P = .23), suggesting no publication bias. In the light of the remarkable between-study heterogeneity (I-squared = 77%), we investigated its potential sources.

Figure 2.

Main results from data meta-analysis of observational studies investigating the association between cancer risk and vitamin B6 dietary intake (high vs low category comparison). The category "Any" refers to all cancer sites combined together. RR = relative risk.

Subgroup analysis by cancer site showed that higher intake of dietary vitamin B6 was linked to a statistically significantly lower risk of esophageal (RR = 0.57, 95% CI = 0.47 to 0.69), pancreatic (RR = 0.64, 95% CI = 0.44 to 0.93), gastric (RR = 0.66, 95% CI = 0.57 to 0.76), colorectal (RR = 0.72, 95% CI = 0.61 to 0.84), and breast (RR = 0.88, 95% CI = 0.78 to 0.98) cancer, heterogeneity being low only for esophageal and gastric tumors (Figure 2). In contrast, meta-analysis suggested no evidence of association for the following tumor sites: lung, prostate, kidney, endometrium, ovary, and immune system. Pooling studies by tumor site, we observed a statistically significant association with the following groups: upper airway (oral cavity and nasopharynx and larynx: RR = 0.46, 95% CI = 0.33 to 0.63), urinary tract (kidney and prostate and bladder: RR = 0.88, 95% CI = 0.78 to 0.99), and gastrointestinal (esophagus and stomach and colorectal and pancreatic: RR = 0.68, 95% CI = 0.61 to 0.75), heterogeneity being low, moderate, and high, respectively.

In many circumstances, the evidence in favor of an association with cancer risk was weaker among prospective as compared with retrospective studies; for example, data on gastrointestinal tumors showed a lower risk among retrospective (RR = 0.62, 95% CI = 0.54 to 0.71) as compared with prospective (RR = 0.81, 95% CI = 0.70 to 0.94) series (meta-regression P interaction =.02) (Supplementary Table 4, available online). Nevertheless, between-study heterogeneity did not decrease remarkably across studies with the same design (Supplementary Table 4, available online).

Overall, subgroup analysis by ethnicity was not informative because of the small number of Asian studies (Supplementary Table 4, available online). Analogously, meta-regression, including number of cases examined, range of exposure (highest minus lowest category values), year of publication, and study quality, was not informative.

We then considered total (dietary and supplements) vitamin B6 intake (studies, n = 49; cases, n = 20 796). A full list of results is available online (Supplementary Table 5, available online). As compared with dietary intake data, the association with cancer risk was weaker and limited to a smaller number of tumor sites: any (RR = 0.95, 95% CI = 0.91 to 0.99), colorectal (RR = 0.92, 95% CI = 0.86 to 0.99), gastrointestinal (RR = 0.91, 95% CI = 0.85 to 0.98), and urinary tract (RR = 0.85, 95% CI = 0.75 to 0.97). Statistically significant associations were characterized by a between-study heterogeneity that was low or absent. Neither subgroup analysis nor meta-regression was informative.

Dose-response Meta-analysis

Considering dietary intake (studies, n = 63; cases, n = 52 970), a full list of results is available online (Supplementary Table 6, available online). Trend meta-analysis showed a statistically significant inverse dose-response relationship between vitamin B6 and any type of cancer (6% risk reduction per milligram of vitamin daily intake, RR = 0.94, 95% CI = 0.92 to 0.96), without statistically significant deviation from linearity (Figure 3, upper panel). In the light of the statistically significant heterogeneity, we investigated its potential sources.

Figure 3.

Dose-response relationship between vitamin B6 dietary intake and cancer risk. Upper panel: all tumor sites (meta-analysis of 63 observational studies; cases, n = 52 970). Lower panel: gastrointestinal tumors (meta-analysis of 25 observational studies; cases, n = 17 998).

Subgroup analysis showed a dose-response relationship for the following tumor sites: stomach (RR = 0.90, 95% CI = 0.81 to 0.99), pancreas (RR = 0.88, 95% CI = 0.82 to 0.94), colorectal (RR = 0.93, 95% CI = 0.90 to 0.95), breast (RR = 0.95, 95% CI = 0.91 to 0.99), and ovary (RR = 0.93, 95% CI = 0.87 to 0.99), with lack of evidence for heterogeneity only among gastric cancer studies (Supplementary Table 6, available online). In contrast, meta-analysis suggested no evidence of association for esophageal, lung, prostate, kidney, endometrial, and hematological tumors (Supplementary Table 6, available online).

Pooling tumors by site, only gastrointestinal cancers were inversely associated with vitamin dietary intake (RR = 0.94, 95% CI = 0.92 to 0.96), with evidence of statistically significant heterogeneity (Figure 3, lower panel). No such a relationship could be demonstrated for gynecological or urologic cancers. In some cases, the evidence in favor of a dose-response relationship was weaker among prospective as compared with retrospective studies, but the risk difference was statistically significant only when all sites where considered, which showed homogeneity across prospective studies. Finally, subgroup analysis by ethnicity was not informative.

As regards total vitamin intake (studies, n = 43; participants, n = 33 927), all results are available online (Supplementary Table 7, available online). Trend meta-analysis showed a statistically significant association only with esophageal (RR = 0.87, 95% CI = 0.78 to 0.98) and gastrointestinal (RR = 0.98, 95% CI = 0.97 to 0.99) cancers, without evidence of heterogeneity or deviation from linearity. Subgroup analysis did not provide further insights.

Overall, in the light of the large protective effect coupled with the dose-response relationship but in the presence of between-study heterogeneity, the level of evidence supporting the association between vitamin B6 dietary intake and the risk of gastrointestinal cancers was graded as moderate.

Observational Studies: PLP Blood Levels

The relationship between circulating levels of vitamin B6 (measured as PLP concentration in the peripheral blood) and cancer risk was investigated in 25 studies (participants, n = 20 858; cases, n = 9671), the design being prospective in all studies but one and the ethnicity being Caucasian in all studies but one. Based on the Newcastle-Ottawa Scale tool, study quality was maximal (nine stars) in most cases (n = 14, 56.0%); the remaining 11 studies were graded with eight stars, the reason for downgrading being suboptimal representativeness of the studied population.

High vs Low Category Meta-analysis

A full list of results is available online (Supplementary Table 8, available online), whereas a summary of main findings is reported in Figure 4. Data meta-analysis from all 25 studies (cases, n = 2545) showed that higher vitamin B6 plasma levels were statistically significantly and inversely associated with the risk of any cancer site (RR = 0.66, 95% CI = 0.58 to 0.76). As the heterogeneity was remarkable (I-squared = 51%), we searched for its potential sources.

Figure 4.

Main results from data meta-analysis of observational studies investigating the association between cancer risk and vitamin B6 (pyridoxal-5'-phosphate, PLP) blood levels (high vs low category comparison). The category "Any" refers to all cancer sites combined together. RR = relative risk.

Subgroup analysis by tumor site led to the identification of a statistically significant association with colorectal (RR = 0.56, 95% CI = 0.46 to 0.67), pancreatic (RR = 0.56, 95% CI = 0.41 to 0.80), and lung (RR = 0.50, 95% CI = 0.38 to 0.67) cancers, with no between-study heterogeneity (I-squared =0%) (Supplementary Table 8, available online). No statistically significant association was found for breast, prostate, or kidney tumor sites. When pooling cancer sites, the association between PLP levels and the risk of gastrointestinal tumors was large and highly statistically significant (RR = 0.56, 95% CI = 0.48 to 0.65), with no heterogeneity (I-squared =0%) (Supplementary Table 8, available online).

Dose-response Meta-analysis

Dose-response data for PLP levels were available in 21 studies (cases, n = 9310). Full results are reported online (Supplementary Table 9, available online). Trend meta-analysis demonstrated a statistically significant inverse association between vitamin blood levels and all tumor sites (30% risk reduction per 100 nmol/L of blood PLP, RR = 0.70, 95% CI = 0.65 to 0.76), without evidence of heterogeneity or deviation from linearity (Figure 5, upper panel). Similarly, a statistically significant risk reduction was observed for the following tumor sites: colorectal (RR = 0.52, 95% CI = 0.43 to 0.64), pancreas (RR = 0.37, 95% CI = 0.19 to 0.71), breast (RR = 0.79, 95% CI = 0.78 to 0.93), and kidney (RR = 0.71, 95% CI = 0.60 to 0.84), without evidence of heterogeneity. No statistically significant association was found for lung and prostate cancers.

Figure 5.

Dose-response relationship between vitamin B6 (pyridoxal-5'-phosphate, PLP) blood levels and cancer risk. Upper panel: all tumor sites (meta-analysis of 63 observational studies; cases, n = 52 970). Lower panel: gastrointestinal tumors (meta-analysis of 25 observational studies; cases, n = 17 998).

Pooling tumor sites, PLP levels were inversely and statistically significantly associated with the risk of both gastrointestinal (RR = 0.49, 95% CI = 0.41 to 0.59) (Figure 5, lower panel) and urinary tract (RR = 0.75, 95% CI = 0.65 to 0.87) cancers, without evidence of heterogeneity.

Including predefined covariates in a meta-regression model did not yield noteworthy results.

Overall, in light of the large protective effect coupled with the dose-response relationship and the lack of heterogeneity, the level of evidence supporting the association between PLP levels and the risk of gastrointestinal cancers was graded as high.

Intervention Studies

Nine randomized controlled trials (RCTs) enrolling 34 911 participants (cases, n = 2539) reported cancer incidence data among participants randomly assigned to regularly take supplemental vitamin B6 (see Supplementary Table 3, available online, for details). In all trials, vitamin B6 (daily dose = 3 to 100 mg; mean treatment duration = 2 to 7.3 years) was administered in combination with vitamin B12 and folate, vitamin combination being compared with placebo in all cases except one RCT, where high vs low doses of micronutrients (including vitamin B6) were compared.[18]

In no trial was the primary endpoint cancer risk; all studies focused instead on the incidence of cardiovascular events; in addition, participants were not taken from the general population but were rather patients affected with specific diseases (cardiovascular disease [n = 8] or chronic renal failure [n = 1]). This is the reason why no adjustment for cancer-related risk factors was applied in data analysis. As regards quality assessment, no trial was deemed at high risk of bias.

No single study found a statistically significant association between vitamin B6 intake and incidence of cancer (any site). Data meta-analysis showed no protective effect on the risk of any tumor (RR = 1.03, 95% CI = 0.94 to 1.13, P = .47, I-squared = 0.0%). Data on single tumor sites were available only in two RCTs, the findings of which were reported in a single pooled analysis,[19] and in a third study[20] reporting on different tumor sites; therefore, no site-specific meta-analysis could be performed.

In the RCT with the longest treatment duration and follow-up (7.3 years)—where women only were recruited—the investigators reported a 17% risk reduction for breast cancer (cases, n = 154), although the result did not reach statistical significance (RR = 0.83, 95% CI = 0.60 to 1.14) (Supplementary Table 3, available online). In the only two studies[19] where one of the four arms consisted of vitamin B6 alone, the intake of this vitamin was not associated with cancer (any site) risk (RR = 1.11, 95% CI = 0.87 to 1.41); however, during the second half of the follow-up time (ie, 3.2 years), participants were not treated with vitamin B6.

Overall, the level of evidence supporting the lack of protective effect of vitamin B6 against any cancer was graded as low because of the high risk of bias in terms of representativeness of the population, type of treatment, and study outcomes.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....