Serum vitamin B12, distal symmetrical polyneuropathy and anaemia in type 2 diabetes: the Fremantle Diabetes Study Phase 2

There are limited data relating to the effects of metformin‐associated vitamin B12 deficiency on the risk of distal symmetrical polyneuropathy (DSPN) and megaloblastic anaemia in well‐characterised community‐based cohorts.


Introduction
There is evidence that metformin use, especially when prolonged and at higher doses, is associated with the risk of vitamin B12 deficiency in type 2 diabetes. 1Multiple underlying mechanisms have been proposed, but most are related to intestinal malabsorption. 2 The initial presentation of vitamin B12 deficiency is typically with non-specific symptoms, such as fatigue, weakness, paraesthesia and vertigo, 3 and there are also a variety of associated dermatological and oral conditions. 4The most clinically concerning consequences are, however, adverse neurological effects and megaloblastic anaemia.
Distal symmetrical polyneuropathy (DSPN) is a readily assessable and common chronic neurological complication of diabetes.A range of studies have investigated whether metformin-associated vitamin B12 deficiency may be a risk factor.][11][12][13][14][15][16] Studies assessing the relationship between metformin-associated vitamin B12 deficiency and anaemia have all proved negative. 6,11,16he available data from studies examining the relationship between vitamin B12 and DSPN and/or anaemia have limitations.All relevant studies have recruited from outpatient clinics rather than utilising communitybased samples, and some important potentially confounding variables have not been available.These include the use of vitamin B12 supplementation and proton pump inhibitor (PPI) therapy.The use of PPIs is common in type 2 diabetes, 17 and they have the potential to interact with metformin to increase the risk of B12 deficiency by further reducing absorption. 18In addition, almost all studies published to date have assessed deficiency from total vitamin B12 concentrations rather than measurement of the transcobalamin-vitamin B12 complex (holotranscobalamin), which represents the biologically active form taken up through specific cell surface receptors. 19In light of these considerations, we have assessed the inter-relationships between metformin therapy, vitamin B12 deficiency assessed from active B12 concentrations and both DSPN and megaloblastic anaemia in well-characterised community-based people with type 2 diabetes.

Study site, participants and approvals
The Fremantle Diabetes Study Phase 2 (FDS2) is a prospective observational study of diabetes in a postcode-defined urban population of approximately 153 000 people living in and around the port city of Fremantle in Western Australia (WA). 20Of 4639 people identified as living with diabetes between 2008 and 2011, 1668 (36.0%) were recruited together with 64 participants of the Fremantle Diabetes Study Phase 1 who had moved out of the study area.The present study included 1492 out of 1499 (99.5%) participants with clinically diagnosed type 2 diabetes who had a baseline serum total vitamin B12 measured.The FDS2 was approved by the South Metropolitan Area Health Service Human Research Ethics Committee and all participants gave written informed consent.

Clinical and laboratory methods
All FDS2 participants had a comprehensive assessment at baseline. 20Demographic and clinical information were documented, physical examinations and associated investigations were carried out, and fasting blood and urine samples were obtained.Medication use, including dose regimens, was documented.For the purposes of the present study, and given that metformin tablet strengths are 500, 850 and 1000 mg, daily doses were categorised into four groups as ≤1000, >1000 and ≤1700, >1700 and ≤2000, and >2000 mg.Microvascular and macrovascular complications of diabetes at study entry were identified using standard criteria, including DSPN (a score of >2/8 on the clinical portion of the Michigan Neuropathy Screening Instrument 21 ).All haematological and biochemical tests were performed in a single nationally accredited laboratory.A full blood count was obtained using an automated haematology analyser with the generation of standard parameters, including mean corpuscular volume (MCV).Participants were considered to have anaemia if they had a haemoglobin concentration ≤130 g/L in males and ≤120g/L in females based on World Health Organization criteria. 22itamin B12 was measured using an E170 immunoassay analyser and reagents supplied by Roche Diagnostics (Castle Hill, NSW, Australia) with an imprecision (expressed as the coefficient of variation) of 3.1% at 245 pmol/L and 2.7% at 660 nmol/L.Holotranscobalamin was measured using 'Active B12' reagents on an Alinity analyser (Abbott Diagnostics, North Ryde, NSW, Australia).Imprecision was 3.5% at 34.7 pmol/L and 3.1% at 51.5 pmol/L.For holotranscobalamin, the normal range is >35 pmol/L.In the present study, active B12 was measured in those with serum total vitamin B12 <250 pmol/L.There is no universally accepted biochemical definition of vitamin B12 deficiency. 23,24In the present study, to capture participants with definite and potentially clinically severe vitamin B12 deficiency, 10,23 deficiency was defined as serum total vitamin B12 <80 pmol/L and active B12 <23 pmol/ L. Because many studies employ total serum vitamin B12 cut point ≤200 pmol/L with or without an active B12 ≤35 pmol/L, 23 adequate vitamin B12 status was defined as serum concentrations of these analytes above these thresholds.Those with values intermediate between the definite deficiency and adequate vitamin B12 status were categorised as having borderline deficiency.

Statistical methods
Data are presented as percentages (%), mean ± SD, geometric mean (SD range) or median (interquartile range).Two independent samples were compared using Fisher's exact test for proportions, Student's t test for normally distributed variables and Mann-Whitney U test for variables that were not normally or log e -normally distributed.More than two independent samples were compared with the Fisher-Freeman-Halton exact test for proportions, ANOVA for normally or log e -normally distributed variables, or the Kruskal-Wallis test otherwise.Multiple logistic or linear regression analyses using backward stepwise conditional modelling (P < 0.05 for entry, P ≥ 0.05 for removal), with all clinically plausible variables at P < 0.20 in bivariable analyses considered for entry, were used to identify independent associates of variables of interest.

Participant characteristics
The characteristics of FDS2 participants with type 2 diabetes categorised by vitamin B12 status are summarised in Table 1.One in 50 (2.0%)had vitamin B12 deficiency, and a further 3.6% had borderline deficiency.In bivariable analyses, those who were deficient were older and more likely to be metformin-treated than participants in the other two groups, but there were no other statistically significant differences over a range of demographic, clinical and laboratory variables.These included the use of a PPI with or without metformin (P ≥ 0.395).The prevalence of vitamin B12 deficiency in metformintreated participants was 4.2%, and a further 3.1% of metformin-treated participants had borderline deficiency.

Associates of vitamin B12 deficiency
In multiple logistic regression analysis, age (odds ratio (OR) (95% confidence interval (CI)): 1.05 (1.01-1.09)for each year increase; P = 0.009) and metformin use (19.2 (2.6, 142.0); P = 0.004) were the only independent associates of vitamin B12 deficiency.When metformin dose categories were included instead of metformin therapy as a binary variable, there was a graded increase in the likelihood of vitamin B12 deficiency with increasing dose (trend P = 0.002), with those treated with >2000 mg daily at nearly 40 times the odds compared with those not taking metformin (Table 2).Diabetes duration did not add to this model (OR 0.96 (0.91, 1.01); P = 0.151).

Relationship between vitamin B12 deficiency, metformin therapy and DSPN
There was no statistically significant association between vitamin B12 status and prevalence of DSPN (P = 0.603; Table 1).In metformin-treated participants considered by increasing dose category, 61.1%, 61.5%, 59.0% and 52.2% respectively, had DSPN (P = 0.279), and there was no significant association after adjustment for age in a logistic regression model (trend P = 0.505).

Relationship between vitamin B12 deficiency, metformin therapy and anaemia
There was no statistically significant association between vitamin B12 status and prevalence of anaemia, haemoglobin concentration or MCV (P ≥ 0.147; Table 1).In metformin-treated participants considered by increasing metformin dose category, 7.8%, 7.3%, 16.0% and 17.8% respectively, had anaemia (P < 0.001).In a logistic regression model incorporating the whole cohort and including the variables we found previously to be significantly associated with prevalent anaemia in our participants 25 (diabetes duration, thiazolidinedione therapy, diastolic blood pressure, MCV, renal function and peripheral arterial disease), the two higher metformin dose categories (>1700 mg daily) had increased odds of anaemia (OR 2.45 (1.40-4.22)and 2.46 (1.31-4.59)respectively; P ≤ 0.005) compared to no metformin therapy while there was no significant difference for the two lower categories (P ≥ 0.341).Vitamin B12 deficiency was not a significant independent variable in the model (OR 1.46 (0.52-4.11) and 1.19 (0.408-2.98) for deficiency and borderline vitamin B12 respectively; P ≥ 0.470).
Using haemoglobin concentration as a continuous variable and after adjusting for all other significant associates in a linear regression model, there was an increasing inverse relationship between metformin dose and haemoglobin concentration (hazard ratios (95% CI) À2.31 (À3.84 to À0.79), À2.63 (À4.98 to À0.29), À5.57(À7.28 to À3.86) and À 6.56 (À8.64 to À4.48) respectively; P ≤ 0.028) for dose categories from ≤1000 mg daily to >2000 mg daily compared with no metformin therapy.Neither deficiency nor borderline vitamin B12 was a significant independent variable in the model (P ≥ 0.639).In a linear regression model with MCV as the dependent variable and incorporating all other significant independent associates, neither the metformin dose category nor vitamin B12 status was significant (P ≥ 0.239).

Discussion
The present data show that community-based people with type 2 diabetes from an urban Australian setting are at low risk of metformin-associated vitamin B12 deficiency, including those treated concomitantly with a PPI.In those with vitamin B12 deficiency, whether definite or borderline, there was no statistically significant relationship with DSPN or anaemia in our cohort.These findings argue against regular or opportunistic measurement of serum vitamin B12 in metformin-treated patients, which should be reserved for situations with a clear clinical indication.These include the investigation of suggestive symptoms, in ruling out causes of a DSPN other than diabetes, 26 and in the investigation of a megaloblastic anaemia.
The prevalence of vitamin B12 deficiency in our metformin-treated participants was 4.2%.If those with borderline deficiency were included, this figure increased to 7.3%.Although some of our participants with a total vitamin B12 <200 pmol/L had an active B12 >35 pmol/L and were thus not categorised as deficient, this is below the 7.4-28.8%range in published studies employing a similar <200 pmol/L total serum vitamin B12 concentration cut point. 5,6,14,16In a developed country such as Australia, greater nutritional sources likely counteract metformin-associated vitamin B12 malabsorption (and the additional potential influence of PPI use) compared to the situation in the developing world where the highest rates of metformin-associated deficiency have been reported. 14,16As a corollary to this, vitamin B12 deficiency severe enough to cause DSPN and/or megaloblastic anaemia is (as in the present study) likely to be relatively rare in Australia compared with other countries.The use of mostly over-the-counter vitamin B12 supplementation by FDS2 participants was low (<4%) and not significantly different by serum vitamin B12 status, suggesting that this was not a confounding factor.Although consistent with most previous studies, [9][10][11][12][13][14][15][16] we found no evidence that metformin therapy was associated with DSPN, whether through vitamin B12 deficiency or other mechanisms, there was a graded relationship between metformin dose category and both vitamin B12 deficiency and anaemia that was independent of confounding factors in our participants.Such a dose-response relationship has been found previously for serum vitamin B12 deficiency, 1 especially daily doses ≥2000 mg, but the association between metformin therapy and anaemia is less clear.Large-scale trials and realworld data suggest that metformin is associated with anaemia within the first few years of treatment but that this effect wanes thereafter. 27Although this latter study did not include serum vitamin B12 concentration data, the authors considered that the time course would be inconsistent with vitamin B12 deficiency as a cause, reflecting the lack of an association between vitamin B12 status and both anaemia and MCV in the present study.Why metformin might cause a generally mild anaemia that does not persist or worsen with continued therapy is unclear.Since iron 28 in addition to B12 deficiency is unlikely, haemodilution, bone marrow suppression and/or haemolysis may be contributory, but an examination of aetiology was beyond the scope of the present study.
In contrast to most similar epidemiological studies, we utilised active and total vitamin B12 in our deficiency assessment.Nevertheless, other biomarkers may have greater validity.These include plasma methylmalonic acid (MMA) and homocysteine, both elevated in vitamin B12 deficiency. 23Although specific to vitamin B12 status, measurement of MMA is expensive and not available as a routine test in many laboratories, while homocysteine can be affected by the presence of folate deficiency and may have questionable clinical application. 29Both these biomarkers can be influenced by renal impairment. 23Perhaps because of these considerations, they have only been used in a few studies of metforminassociated vitamin B12 deficiency. 7,16he present study had limitations.It had a crosssectional design which reflected, in part, the fact that a full blood count was done only at the baseline FDS2 assessment so that analysis of the serial changes observed in other studies 27 was not possible.In the case of DSPN, longitudinal analysis would be unlikely to be informative given the negative associations at study entry.However, the average diabetes duration of our participants at entry was around 10 years and the majority would have been taking metformin from soon after diagnosis.Any effects of metformin on vitamin B12 status and its consequences would likely have been realised by the time of recruitment. 1 The dosedependent effects of metformin on the likelihood of anaemia were still evident in our participants who had relatively long-duration diabetes (mean > 10 years).This is consistent with a metformin dose-related cumulative risk of anaemia that has been previously reported in a large administrative database analysis 27 and would argue against chronic haematological adaptation mechanisms.The strengths of the present study include its community-based cohort, which is larger than all but one study to have examined the association between vitamin B12 deficiency and DSPN and/or anaemia, 6 and the detailed baseline assessment of each participant.

Conclusion
Metformin-associated vitamin B12 deficiency is uncommon in community-based Australians with type 2 diabetes, and deficiency was not associated with prevalent DSPN or anaemia.Although clinical features of vitamin B12 deficiency should prompt measurement of total serum vitamin B12 and, if this is low (<200 pmol/L as in the present study and close to the middle of the wide range of levels used to define deficiency in the literature 23 ), active B12 with a view to replacement therapy, more intensive screening appears unnecessary.This includes the routine annual screening that has been recommended by some authors 30,31 but which, based on the present data, is not cost-effective.

Table 2
Logistic regression analysis of variables associated with vitamin B12 deficiency