**Background:** The prevalence of older age and type-2 diabetes mellitus (T2DM) is rising in Singapore, predisposing the population to frailty and complications that can be aggravated by vitamin B12 deficiency — a prevalent but potentially modifiable condition. B12 deficiency adversely impacts neuro-cognitive, haematological, cardiovascular, and immune systems. Despite a consistently reported prevalence >20% in primary-care T2DM populations, screening remains lacking due to unclear guidelines. This study aimed to identify determinants of B12 deficiency in community-dwelling T2DM patients to profile those most in need of screening.
**Methods:** A retrospective cohort of 592 primary-care patients with T2DM was recruited from a Singapore polyclinic between 2008 and 2011. Patients were registered under the Family Physician Clinic (FPC) service, which had piloted periodic B12-folate testing since 2008 without a formal screening protocol. Fourteen patients were excluded for having no prior B12-folate test, and 2 were excluded (one type-1 diabetes, one missing LDL). The lowest recorded B12 level was used as the reference. Covariates were extracted from electronic records for the 12 months preceding the reference B12 test date, including metformin exposure (daily dose and 1-year cumulative dose), age, gender, ethnicity, BMI, vegetarianism, prescribed B12 and calcium supplementation, PPI/H2A use, serum folate, creatinine, LDL cholesterol, HbA1c, and diabetes duration. B12 deficiency was defined as serum B12 <150 pmol/L; folate deficiency as ≤13.4 nmol/L. Univariate and multivariable logistic regression were performed, with variables having P<0.1 in univariate analysis and four clinically important variables (age, PPI/H2A use, gender, ethnicity) selected for the multivariable model. The study was adequately powered (>80%) with 592 patients.
**Key Results:** B12 deficiency was present in 164 of 592 patients (27.7%). Folate deficiency was present in 132 patients (22.3%), and was more prevalent in B12-deficient patients (31.7%, 52/164) than in B12-replete patients (18.7%, 80/428; P=0.001). In multivariable logistic regression, the following were independently associated with B12 deficiency: metformin daily dose (adjusted OR 2.79 per g/day; 95% CI 2.22–3.48; P<0.001); age ≥80 years (adjusted OR 2.86; 95% CI 1.31–6.25; P=0.008); vegetarianism (adjusted OR 21.61; 95% CI 4.47–104.44; P<0.001); and folate deficiency (adjusted OR 2.04; 95% CI 1.27–3.28; P=0.003). Prescribed B12 supplementation was associated with lower odds of deficiency (adjusted OR 0.37; 95% CI 0.22–0.61; P<0.001). The area under the ROC curve for the multivariable model was 0.803 (95% CI 0.765–0.842). The univariate ROC cut-point for metformin daily dose was 1.425 g/day (sensitivity 83.86%, specificity 64.95%). The estimated probability of B12 deficiency rose from 0.074 at 0 g/day metformin to 0.631 at 3.0 g/day. Metformin daily dose was highly correlated with metformin 1-year cumulative dose (r=0.960; P<0.01) and linearly associated with duration of diabetes (B=0.113; P<0.0001). Among the 164 B12-deficient patients, 48 (29.3%) required >1 screening test before detection, independent of diabetes duration (P=0.344). Age ≥80 years was a negative confounder in univariate analysis because older patients took lower metformin doses (mean 0.69 g/day vs. 1.69 g/day for those <60 years; P<0.001). PPI/H2A use and calcium supplementation were not significantly associated with B12 deficiency in multivariable analysis.
**Clinical Implications:** This study provides a framework for case-finding B12 deficiency in primary-care T2DM patients. The authors recommend that all T2DM patients in primary care be screened for B12 deficiency at least once, and that screening be incorporated into annual routine laboratory reviews — especially for those on ≥1.5 g/day metformin, aged ≥80 years, vegetarian, or not on prescribed B12 supplementation. Concurrent evaluation for folate deficiency is essential given the association between B12 and folate deficiencies, which may amplify clinical complications. Metformin daily dose serves as an accurate proxy for both cumulative metformin exposure and duration of T2DM, simplifying screening decisions in busy clinical settings. Limitations include the observational cross-sectional design (cannot establish causality), lack of confirmatory biomarkers for metabolic B12 deficiency, reliance on patient-reported vegetarianism, and potential confounding by indication for prescribed B12 supplementation.