Frontiers in Aging Neuroscience · Oct 30, 2026
Routine vitamin B12, folate and thyroid biomarkers, cognitive trajectory and mortality in older adults with neurocognitive disorders: a retrospective cohort study
By Pei-Hao Chen, Fang-Yu Cheng, Kai Lo
Background: Vitamin B12, folate, and thyroid-function testing are routinely included in dementia work-ups, but their prognostic meaning after a neurocognitive-disorder diagnosis remains uncertain. Methods: We analysed a de-identified Taiwanese tertiary-hospital cohort of patients aged ≥65 years with neurocognitive disorders (dementia or mild cognitive impairment; n=7,969). The prespecified primary analyses were mixed-effects models of annual MMSE and CDR-SB trajectories in which baseline B12, folate, TSH and free thyroxine were entered jointly, restricted to patients with all four biomarkers and at least two assessments (MMSE n=818; CDR-SB n=1,009). CDR progression and all-cause mortality were secondary outcomes, and prescription-record analyses were exploratory. Sensitivity analyses addressed missing data, selection into the analysis sets, competing risks, confounding, laboratory timing and prior treatment. Results: Biomarker testing and repeated cognitive assessment were clinically driven, so all estimates apply to a selected tested-and-followed subgroup that was younger, less impaired and less likely to die than untested or unfollowed patients. Higher baseline B12 was associated with slower MMSE decline (+0.146 points/year per SD, 95% CI +0.038 to +0.255; FDR q=0.017), an association retained after multiple imputation and selection weighting but attenuated after excluding patients treated before baseline. The full-range association of higher TSH with slower decline on both scales was not robust: it depended on a few patients with TSH >10 mIU/L and was absent within the euthyroid range. Folate deficiency was associated with faster CDR progression (HR 1.57, 95% CI 1.13–2.19, p=0.008), including in competing-risk models; the estimate for B12 deficiency was imprecise (HR 1.22, 95% CI 0.88–1.69, p=0.23) and does not exclude a clinically relevant increase. Higher B12 was associated with mortality (1.035 per 100 pg/mL, 95% CI 1.016–1.055), but this was attenuated after flexible adjustment for renal function. Documented B12 supplementation showed no detectable association with CDR-SB trajectory in a small exploratory matched analysis. Conclusion: In patients who underwent biomarker testing and repeated cognitive assessment, routine reversible-workup biomarkers provided modest, context-dependent prognostic information. We interpret these associations as possible indicators of clinical state rather than evidence that the biomarkers, or their correction, influence the progression of neurocognitive disorders.