Science · Biomarkers · Dementia risk
What the Study Found and Did Not Show
Low hemoglobin does not cause dementia based on available scientific evidence. In a Korean national cohort of 316,542 adults aged 40 and older, researchers found an observational association between hemoglobin level and later dementia. They did not demonstrate a direct biological cause. The research was published in the Journal of Alzheimer's Disease in April 2026. Authors presented the findings in October 2026 at the 39th European College of Neuropsychopharmacology (ECNP) Congress in Munich.
Participants who completed a national health screening between 2004 and 2005 had their dementia diagnoses tracked through health insurance records through 2019. When researchers divided participants into five equal groups, those in the lowest 20 percent of hemoglobin had a 16 percent higher risk of all-cause dementia compared to reference groups. That same low group showed a 14 percent higher risk of Alzheimer's disease and a 19 percent higher risk of vascular dementia. Those fifths are positions inside this cohort, not a clinical anemia cut-off, so a reader cannot map them onto the flag on their own lab report.
Two structural limitations prevent these percentages from proving that low hemoglobin harms the brain. First, observational studies cannot demonstrate that one biomarker causes a clinical outcome. Lead researcher Dr Hyun Soo Kim, an assistant professor at Dong-A University College of Medicine in Busan, South Korea, stated directly that the findings "do not show that hemoglobin itself causes dementia".
Second, reverse causation offers an equally plausible explanation for the association. Developing dementia often causes subtle declines in appetite, dietary intake, and body weight years before a formal diagnosis occurs. These physical changes lead to nutritional deficits and anemia, which pull hemoglobin numbers downward. Lower hemoglobin in this cohort may simply reflect early neurodegeneration rather than cause it.
Why Risk Rose at Both Ends of the Range
Dementia risk was elevated in both the lowest and highest quintiles of hemoglobin, creating a non-linear curve across the study population. The lowest fifth of participants experienced a 16 percent higher risk of all-cause dementia, while the highest fifth saw a 7 percent higher risk. In vascular dementia, the risk elevation was almost identical at both extremes, rising 19 percent in the lowest fifth and 18 percent in the highest fifth.
The authors described the distribution across the five groups as complex and distinctly non-linear. Because risk sat higher at both ends, hemoglobin cannot be read on a scale where a bigger number means less risk.
The study also reported that the observed association was clearer in women than in men. Researchers have not established the biological mechanisms that explain why elevated hemoglobin correlates with vascular dementia diagnoses.
The Reverse-Causation Problem, Mechanically
Reverse causation occurs when an undiagnosed disease produces physical changes that lower blood test values years before a diagnosis. Neurodegenerative diseases develop slowly over many years before memory loss or cognitive symptoms trigger a clinical assessment. During this preclinical stage, developing dementia can reduce appetite and cause weight loss and frailty.
As frailty sets in, circulating hemoglobin drops as an indirect result of the neurological illness. The blood sample is drawn in that window. In the insurance records, the low value then looks like it came first.
Tracking participants for up to 15 years from the 2004 to 2005 screening period does not eliminate this possibility. Preclinical neurodegenerative pathology can advance silently across more than a decade before diagnosis codes appear in insurance records. Dr Kim and the research team acknowledged that early undiagnosed dementia could easily impair nutrition and produce the exact relationship observed in their data.
What the Coverage Does Not Publish
News coverage of the conference presentation does not provide the statistical parameters required to evaluate individual risk. The reporting published by Medical Xpress does not include hazard ratios or confidence intervals for any of the five hemoglobin groups. Without confidence intervals, readers cannot assess the statistical precision or range of uncertainty surrounding the reported percentages.
The coverage also omits absolute dementia rates, total case numbers, and person-years of follow-up. Relative risk figures like 16 percent or 19 percent describe proportional differences between groups, but they reveal nothing about baseline incidence. A 16 percent relative increase on an unpublished baseline cannot be turned into a personal risk.
Hemoglobin Is Already on Your Panel
Hemoglobin is the oxygen-carrying protein in red blood cells, and a routine complete blood count measures it. Related red cell indices evaluated on the same panel include hematocrit, red blood cell count, mean corpuscular volume (MCV), and red cell distribution width (RDW).
Readers will already have this number without having ordered it specially. Related markers covered on this site include ferritin. You can explore our broader library of blood test explainers in our biomarker directory.
None of these complete blood count indices measure dementia pathology. A routine blood count provides valuable information about general physiology, but it does not serve as a diagnostic screen for brain disorders.
Who This Does Not Apply To
The findings from this cohort cannot be generalized automatically to broader global populations. The entire study population consisted of 316,542 Korean adults aged 40 and older who participated in a specific national screening program. Dr Lindsey Sinclair, an independent researcher at the University of Southampton, pointed out that it remains uncertain whether this association exists in individuals from a wider range of backgrounds.
The study also does not apply to individuals currently undergoing medical treatment or diagnostic workups for known causes of anemia. Anyone with a known cause of anemia under investigation should follow their established clinical plan. This epidemiological correlation changes nothing about their existing clinical workup.
People who fall outside the cohort's age threshold cannot apply these findings to their personal lab results. The investigation enrolled adults aged 40 and older, so it provides no data on hemoglobin variations in younger individuals. Medical decisions regarding blood values must always rely on individual clinical history rather than demographic cohort analyses.
What Would Change This Read
Replication across diverse international populations with published baseline incidence rates would clarify this topic. Publication of hazard ratios and confidence intervals alongside the original paper by Ji Hee Song, Dr Kim, and colleagues would allow independent epidemiologists to verify the strength of the association. Clear documentation of absolute event rates would also show whether the relative differences translate into meaningful clinical changes.
Evaluating true causality would require prospective interventional trials. In our guide on homocysteine and dementia risk, we examined an identical dilemma involving a routine marker, an association, and no trial.
Specialized diagnostic technologies take an entirely different regulatory and clinical path. As covered in our report on the Food and Drug Administration (FDA) clearance of the PrecivityAD2 blood test, that cleared diagnostic is restricted to individuals with existing cognitive impairment. By contrast, large population screenings like this Korean cohort or the UK Biobank observational study on hormone replacement therapy (HRT) and dementia identify population trends that should never guide self-treatment.
The study authors state clearly that people should not take iron supplements or attempt to manipulate hemoglobin levels to protect cognitive function. If a low hemoglobin dementia headline has you looking at your own complete blood count, take that result to the clinician who holds your record.
The short version
Sources
- Journal of Alzheimer's Disease. Source of the peer-reviewed paper 'The association between hemoglobin level and risk of developing dementia', authors Song Ji Hee, Kim Hyeon Ji, Son Minkook, Kim Hye Ryeon, Kim Seong Hwan, Kim Hyun Soo, DOI 10.1177/13872877261443702, published April 23, 2026, confirmed on Crossref October 13, 2026.
- Medical Xpress. Source of the reporting on the 316,542 Korean adult cohort, the 15-year follow-up, the quintile breakdown, the risk percentages (16% all-cause, 14% Alzheimer's, 19% and 18% vascular), Dr Hyun Soo Kim's quote and affiliations, Dr Lindsey Sinclair's commentary, and the lack of published hazard ratios or absolute counts (accessed October 13, 2026).
- European College of Neuropsychopharmacology. Source of the 39th ECNP Congress in Munich, where the study findings were presented in October 2026 (accessed October 13, 2026).
Frequently Asked Questions
Does low hemoglobin cause dementia?
No. The study shows an observational association. It does not establish a causal relationship. Dr Hyun Soo Kim, the lead researcher from Dong-A University College of Medicine, stated directly that the findings do not show that hemoglobin causes dementia. Observational data cannot establish that a blood value alters brain disease risk directly. In addition, early stages of undiagnosed dementia can cause frailty, reduced appetite, and weight loss, which in turn lower hemoglobin levels. This reverse causation means lower hemoglobin may simply reflect underlying health changes rather than initiating cognitive decline.
How much higher was the dementia risk in the lowest hemoglobin group?
Participants in the lowest 20 percent of hemoglobin had a 16 percent higher risk of all-cause dementia compared to reference groups. For specific dementia subtypes, the lowest group showed a 14 percent higher risk of Alzheimer's disease and a 19 percent higher risk of vascular dementia. These percentages reflect relative risk across 316,542 Korean adults tracked up to 15 years. The study coverage does not publish baseline incidence rates or absolute counts. A relative increase without a denominator cannot be translated into an individual's personal likelihood of developing dementia.
Should I take iron if my hemoglobin is low?
No. The study authors explicitly state that their findings should not be used as advice to take iron supplements or to manipulate hemoglobin levels to prevent dementia. Any evaluation of abnormal blood counts belongs in a clinical setting with a physician who can identify the specific underlying cause and recommend appropriate medical care.
Why was dementia risk also higher in the group with the highest hemoglobin?
The Korean cohort demonstrated a non-linear relationship where risk increased at both extremes of the hemoglobin distribution. While the lowest fifth had a 16 percent higher rate of all-cause dementia, the highest fifth also showed a 7 percent higher risk. For vascular dementia, the highest fifth showed an 18 percent higher risk, nearly matching the 19 percent higher risk in the lowest group. The study authors noted that the pattern across all five groups was not simple. Higher hemoglobin cannot be viewed as protective, and the underlying biological mechanisms behind this distribution remain unconfirmed.
Is this study new?
The underlying research paper was published in the Journal of Alzheimer's Disease on April 23, 2026. The findings received broad news coverage in October 2026 because researchers presented them at the 39th European College of Neuropsychopharmacology (ECNP) Congress in Munich. While the conference presentation generated headlines, the dataset and analysis reflect the previously published peer-reviewed paper based on health screenings conducted between 2004 and 2005. The coverage reports on an existing observational study rather than newly discovered trial data.