Association of B12 deficiency and clinical neuropathy with metformin use in type 2 diabetes patientsAK Singh, A Kumar, D Karmakar, RK Jha
Department of Medicine, Rajendra Institute of Medical Sciences, Ranchi, Jharkhand, India
Context: Long-term metformin use has been hypothesized to cause B12 deficiency and neuropathy in Type 2 diabetes patients. However, there is a paucity of Indian data regarding the same. Aim: To compare the prevalence of B12 deficiency and peripheral neuropathy in patients with Type 2 diabetes mellitus treated with or without metformin. Materials and Methods: We recruited patients with Type 2 diabetes and divided them into metformin exposed and nonmetformin exposed groups. We measured baseline demographic variables like age, sex, vegetarian status, and HbA1c levels in both groups. We compared vitamin B12 levels and severity of peripheral neuropathy (using Toronto Clinical Scoring System (TCSS)) in both groups. Definite B12 deficiency was defined as B12 <150 pg/ml and possible B12 deficiency as <220 pg/ml. The difference in vitamin B12 levels and TCSS was calculated in both groups using independent samples t-test. Spearman's rank correlation between cumulative metformin use and B12 level was calculated. Odds ratio of vitamin B12 deficiency in metformin exposed group was also estimated. Results: Mean serum B12 levels was significantly lower in metformin exposed group (n=84) compared with nonmetformin exposed group (n=52) (410±230.7 versus 549.2±244.7, P=0.0011). Mean neuropathy score was significantly higher in metformin exposed group. (5.72±2.04 versus 4.62±2.12, P=0.0064). Odds ratio for possible B12 deficiency was 4.45 (95% CI 1.24-15.97). There was significant negative correlation between cumulative metformin dose and vitamin B12 level (r=−0.68, P<0.0001). Conclusion: Metformin use is associated with vitamin B12 deficiency and clinical neuropathy in Type 2 diabetes patients.
Keywords: Metformin, neuropathy, vitamin B12 deficiency
Vitamin B12 deficiency is common in vegans and can give rise to hematological and neurological manifestations. It can coexist with Type 2 diabetes mellitus and is an independent risk factor for cardiovascular disease. ,, Metformin use in Type 2 diabetes patients has also been shown to cause B12 deficiency in many studies. ,,,, Iatrogenic neuropathy caused by metformin-induced B12 deficiency can add to burden of peripheral neuropathy that already exists in diabetes patients.  In various surveys and cross sectional studies across the world, it has been associated with 10-50% reduction in B12 levels. ,,,,,,,,, This association was conclusively proved by two randomized placebo controlled trials, which showed a decrease in 22% over a period of 29 weeks  and 19% decrease over a period of 4 years, respectively.  Various authors have recommended different treatment approaches varying from annual treatment to routine prophylaxis for patients on long-term metformin. 
Some authors have also questioned the clinical relevance of these findings since B12 deficiency is asymptomatic at most times.  In the past decade many case reports have emerged showing that metformin-induced B12 deficiency can cause insidious neuropathy, which may be attributed to poor glycemic control in Type 2 diabetes patients and may be left untreated. ,
Recent studies have sought to determine prevalence of metformin-induced B12 deficiency and spectrum of its neuropathic burden by using validated neuropathy scales with conflicting results. , Due to paucity of Indian studies evaluating the prevalence of B12 deficiency and the neuropathic burden in Type 2 diabetes patients on metformin, we planned a cross sectional study comparing B12 levels and neuropathy scores in patients with or without metformin exposure.
Our study was a cross sectional study, which was carried out from March 2012 to November 2012, in a group of patients referred to Department of Medicine of a tertiary Hospital with diagnosis of type 2 diabetes mellitus, which was verified using World Health Organization (WHO) criteria.
Patients were excluded if they gave history of alcoholism, ongoing pregnancy, liver disease, renal disease, thyroid disorders or history suggestive of malabsorption disorders. Type 1 diabetes mellitus patients were also excluded from analysis. Patients were selected by adhering strictly to these inclusion and exclusion criteria.
This study was approved by the institutional ethics committee. Strict confidentiality of the study reports was maintained and all the queries and apprehensions of the patients and their families were addressed with utmost care. Prior to initiating the participants to the study, a thorough counseling of the patients and their families were also undertaken and written, informed consent taken.
Out of 160 patients screened, 136 cases of type 2 diabetes mellitus were selected. A large part of this group was on oral hypoglycemic drugs. A detailed history of the duration of diabetes and the type of oral hypoglycemic drug intake was taken. The type of diet was enquired into. Patients were divided into two groups: Those taking metformin, and those who were not taking metformin. Cumulative metformin dose was recorded in patients taking metformin using their dose and duration of treatment history.
In all the outpatients and admitted diabetes patients, a detailed history taking and clinical examination were performed. Blood pressure, body mass index (BMI), and HbA1c were recorded in every patient. All the included patients were subjected to vitamin B 12 assay done using the electrochemiluminescence immunoassay method (Roche Elecsys 2010 analyzer). Serum samples were stored at room temperature (15-30°C) for not longer than 8 hours. On the basis of results of B12 levels, patients were classified into normal level (>220 pg/ml), possible B12 deficiency (150-220 pg/ml), and definite deficiency (<150 pg/ml).
All patients were subjected to standardized neurological examination including power, tone, deep tendon reflexes, and sensory function. Romberg's test was performed in case of any gait abnormality. On the basis of data gathered from history of neurological symptoms and standardized examination, neuropathy in the patient was graded using Toronto clinical scoring system (TCSS) by an investigator blinded to the the laboratory results  TCSS is a clinical screening tool for presence and severity of neuropathy in diabetes patients that focuses on sensory deficits. Though it is a subjective score, it has advantages of including elements of both history and examination unlike other neuropathy screening scores. It has also been validated by sural nerve fiber density and is very easy to calculate from basic history and neurological examination even in a busy outpatient practice. On the basis of the score, patients were graded into classes of no neuropathy: 0-5; mild neuropathy: 6-9; moderate neuropathy: 10-12; and severe neuropathy >12.
Statistical analysis and sample size estimation
We did a sample size calculation at 80% power, with an alpha error of 0.05, while taking expected incidence of B12 deficiency at 25% in metformin exposed group and 5% in nonmetformin exposed group based on previous studies.  The minimum total sample size calculated was 96. Based on these estimates, we screened 160 patients and included 136 patients in our study so that it was adequately powered.
Baseline characteristics like age, duration of diabetes, mean blood pressure, A1C, BMI were compared by relevant independent samples t-test taking into account normality of data. Sex, vegan status, and proportion of other antidiabetic agent use were compared by Chi-square test.
The primary outcome variables were serum B12 levels and clinical neuropathic severity (TCSS). Bivariate correlations of B12 levels and TCSS were done with cumulative metformin dose using rank correlation test. Linear regression analysis was also performed using TCSS as dependent variable and HbA1C, duration of diabetes, Age and cumulative metformin dose as explanatory variables. F statistics and R-square was also calculated for this model. The significance level was taken as 5%.
Of the 136 patients, there were 84 Type 2 diabetes patients with exposure to metformin and 52 metformin naïve patients. We also assessed vitamin B12 levels in 20 apparently healthy controls. The mean age was 53±12.2 years in metformin exposed group and 49±13.6 in nonmetformin group. HbA1c levels were 8.2±1.02% in metformin exposed group and 8.4±0.81% in nonmetformin exposed group. Rest of the baseline characteristics are presented in [Table 1]. Most of the baseline characteristics were similar in both groups with no significant difference between the two. Insulin exposure in nonmetformin exposed group was significantly higher than metformin group (22/52 versus 14/84).
Serum B12 level (in pg/ml) was significantly higher in nonmetformin group as compared with metformin group. (549.2±244.7 versus 410±230.7, mean difference=139.6 95% CI 56.86-221.54, P=0.0011) [Table 2]. The proportion of patients with possible B12 deficiency (150-220 pg/ml) in metformin exposed group was significantly higher than proportion in nonmetformin exposed group [18/84 (21.4%) versus 3/52 (5.7%), mean difference=15.7% 95% CI 4.9-26.5%, P=0.026). While there were six patients with definite B12 deficiency (<150 pg/ml) in metformin exposed group there was nonmetformin group. Vitamin B12 level had significant negative correlation with cumulative metformin dose (r=−0.688, 95% CI −055 to −0.78, P<0.0001) [Figure 1] and duration of metformin treatment (r=−0.74, 95% CI −0.625 to −0.824, P<0.001). Odds ratio for possible B12 deficiency in metformin users was 4.45 (95% CI 1.24-15.97). On adjusting for confounding variables like age and diet by logistic regression, the adjusted odds ratio for possible B12 deficiency in metformin users was 4.27 (95% CI 1.15-15.84). There was no significant difference between vitamin B12 level in control group (510±86 pg/ml) and nonmetformin exposed Type 2 diabetes patients. Only 1 out of 20 patients in control group was possibly B12 deficient. There was significant difference in vitamin B12 levels in patient exposed to metformin for greater than 1 year (305.87±240) and patients exposed for less than 1 year (460.4±233) [mean difference 155.4 pg/ml, (95% CI 51.6-238 pg/ml, P=0.00138)].
Vitamin B12 levels were significantly higher in vegan population (n=18) than nonvegans (529.7±162.3 versus 450±170, P=0.048, 95% CI 4.41-164.75). However, there was no significant difference in proportion of vegans in metformin and nonmetformin users.
There was a significant difference in TCSS between metformin exposed and nonmetformin exposed group (5.72±2.04 versus 4.62±2.12, 95% CI 0.36-2.096, P=0.0064) [Figure 2]. Based on classification used in this score the proportion of the patients with no neuropathy was significantly higher in nonmetformin exposed group [39/52 (75%) versus 39/84 (46.4%), mean difference=28.6%, 95% CI 12.72-44.48%, P=0.0019) while proportion of patients with mild neuropathy was higher in metformin exposed group 41/84 (48.4%) versus 11/52 (21.2%), mean difference=27.6%, 95% CI 12.2-43.02%, P=0.0024). The proportion of patients with moderate neuropathy was higher in metformin group but not significant due to small number of patients in both groups (4/84 (4.8%) versus 2/52 (3.8%), P=0.87). There were no patients with severe neuropathy in both the groups. There was a positive correlation between cumulative metformin dose and neuropathy score (r=0.529, 95% CI 035-0.65, P<0.001). There was also a significant negative correlation between B12 levels and TCSS score in metformin exposed group r=−0.341, 95% CI −0.14-0.51,P=0.002).
In regression model using stepwise enter method, a linear model was observed (F ratio=23.39, R square=0.45, P<0.001) and cumulative metformin dose and HbA1C were independently associated with neuropathy score while age did not assume significance in this model.
There was no significant difference between hemoglobin and mean corpuscular volume (MCV) in both the groups.
We found a significant association between cumulative dose of metformin and lower serum B12 levels in diabetes patients at our hospital-based set up. We also found that metformin exposed group had clinically more severe peripheral neuropathy (as calculated by TCSS score). TCSS also showed significant correlation with cumulative metformin dose, HbA1c, and duration of therapy. This has important implications because India has a high burden of uncontrolled Type 2 diabetes mellitus patients and vegetarian population who are at risk of neuropathy. Long-term metformin use can possibly add to this burden by decreasing serum B12 levels and worsening neuropathy if proper monitoring is not done. Thus our study estimates the prevalence of B12 deficiency in Indian Type 2 diabetes mellitus patients exposed to metformin and demonstrates a significant association between metformin use, decreasing B12 levels, and clinical neuropathy.
While there have been many studies showing lowering of B12 with metformin use ,,,, and some case reports showing worsening neuropathy with metformin, , only two recent studies have systematically explored association of metformin with B12 levels and neuropathy , with conflicting results. Our study shows a significant decrease in serum B12 levels of 132.64 pg/ml in metformin group, and this effect is dose-dependent as indicated by significant negative correlation (r=−0.68). In metformin exposed group, almost one-third of the patients: 24/84 (28.5%) had B12 deficiency (<220 pg/ml) compared with 3/52 (4.7%) in nonmetformin exposed group and 1/40 (2.5%) in control group of healthy volunteers. These findings are in concordance with findings seen in a recent placebo-controlled trial showing decrease of B12 with metformin use leading to low B12 levels in 18% of patients.  A retrospective analysis of NHANES survey published recently  also shows that 13.1% of diabetes patients on metformin had low B12 levels (150-220 pg/ml) and 5.6% were clearly B12 deficient (<150 pg/ml), and compared with 6.2% and 2.2% in control group, respectively. Thus metformin exposure was associated with almost three-fold risk of severe B12 deficiency and two-fold risk of possibly low B12 levels. Other cross sectional studies in Canada  and Hong Kong  also showed low B12 levels of in 31% and 37%, respectively, in metformin exposed group. Thus even according to conservative estimates, almost one-fifth to one-third Type 2 diabetes patients on metformin have low B12 levels.
Though B12 levels in metformin users have been studied across the world, only few studies have explored the additional burden of neuropathy it might impose on them. , Our study revealed a significantly higher neuropathy scores in metformin group and a positive correlation (r=0.53) between neuropathy and cumulative metformin dose. The percentage of the patients with no neuropathy was significantly higher in nonmetformin exposed group (75 versus 46.4), while percentage of patients with mild neuropathy was higher in metformin exposed group [48.4 versus 21.2). Our results are similar to the study by Wile et al.,  although there is smaller but significant difference in neuropathy scores in our study. Since, our study included patients with both asymptomatic and symptomatic neuropathies to make it more representative, this might have led to a smaller difference in our study. Our study differs from a recently published study  in which no difference in neuropathy scores were found using NTSS score. However, we hypothesize this might have to do with the nature of NTSS score,  which is based on history of sensory symptoms alone, rather than reliable and well validated TCSS used in our study, which is easy to calculate, highly reproducible, and includes elements of both history and examination.  No difference in hemoglobin levels and other hematological parameters were seen in our study, but it is expected since neurological manifestations of B12 deficiency appear much earlier than hematological changes. 
Vitamin B12 forms a complex with cubulin (endocytic) receptor at ileum for absorption This B-12 endocytic receptor complex is normally taken by ileal cell surface by calcium dependent process. Metformin with its protonated biguanide group binds to the B12-cubulin complex and imparts positive charge to it, alters membrane potential and competitively repels the divalent calcium ions thus preventing calcium dependent uptake, leading to malabsorption of B12  Owing to this mechanism, some of the patients with metformin-induced B12 deficiency have been treated by calcium supplementation as well. , It has also been proposed to act by increasing bacterial overgrowth, altering bowel motility, and by direct inhibition of B12 absorption. 
Our study has many strengths: We included patients of all age groups from 18 to 60 years, we did prior power calculations, we included patients with both short exposure (<16 week) and long exposure to metformin, and did neuropathy screening and neurological examination on both symptomatic and asymptomatic patients. We also had a control group of patients with no comorbidities in which B12 levels were recorded. Thus, we were able to address limitations of previous studies. The limitations of this study are that it is a hospital-based cross sectional study and we did not calculate more sensitive and expensive biomarkers of B12 deficiency like MMA and homocysteine due to resource constraints that makes our study underpowered to detect deficiency. However, this emphasizes the fact that B12 deficiency in metformin users is likely to be more prevalent than 28.5% observed in our study. We also did not do electro diagnostic tests routinely in every patient and instead screened for neuropathy by valid scale like TNSS, which is a simple reproducible scale using just history and neurological examination. We could also not assess all other confounding causes of neuropathy, however, HbA1c levels, duration of diabetes and vegan status were similar in both groups and unlikely to have contributed to worse neuropathy in diabetic group. The cumulative dose of metformin was an independent predictor on regression analysis of neuropathy even on controlling for HbA1c and duration of diabetes.
The findings in this study suggest that a cumulative dose of metformin is associated with a fall in B12 levels and worsening neuropathy. However, metformin is corner stone of therapy in type 2 diabetes mellitus and has also been seen to have multiple beneficial effects on diabetes patients by altering Advanced Glycosylation end products (age)  and neurodegenerative process.  Hence, we suggest that serum B12 levels should be regularly monitored on Type 2 diabetes patients already at higher risk of neuropathy like vegans and those on long-term metformin therapy. The issue of routine prophylaxis of vitamin B12 in patients on metformin is a complicated one, since in a recently published study using data from the NHANES survey, even people on metformin taking even B12 supplements were three-fold more likely to be B12 deficient.  We suggest that this issue should be addressed in a prospective controlled clinical trial. In view of our findings and on the basis of current evidence, we suggest that B12 levels should be monitored in patients on long-term metformin, and in cases with neuropathic symptoms and B12 deficiency should be treated by intramuscular injections as is routinely done. The frequency of screening should also be explored in a prospective study.
In conclusion, metformin use in Type 2 diabetes patients is associated with B12 deficiency and worsening clinical neuropathy, which is a dose-dependent effect. Hence we suggest for screening of B12 levels in long-term metformin users and treating them if they are deficient.
[Figure 1], [Figure 2]
[Table 1], [Table 2]