Low prognostic nutritional index tied to cognitive impairment in heart failure

A routine blood calculation that blends serum albumin with the circulating lymphocyte count has emerged as a statistical correlate of cognitive impairment in people living with heart failure. In a cross-sectional study of 153 patients treated at Fujian Medical University Union Hospital in Fuzhou, China, researchers found that the prognostic nutritional index, or PNI, was significantly lower among patients who scored below the cutoff for cognitive impairment on the Montreal Cognitive Assessment than among those who did not. After adjusting for age, sex, and educational attainment, every five-point increase in the PNI was associated with roughly 44 percent lower odds of prevalent cognitive impairment, with an odds ratio of 0.562 and a 95 percent confidence interval spanning 0.348 to 0.910.

The association survived a broader sensitivity analysis that additionally accounted for body mass index, coronary artery disease, left ventricular ejection fraction, and advanced New York Heart Association symptom class, yielding an odds ratio of 0.596 per five-point increment. The PNI is calculated as serum albumin in grams per liter plus five times the absolute lymphocyte count in units of ten to the ninth per liter. It has long been interpreted as a composite nutritional and inflammatory marker, and lower values have previously been linked to mortality and rehospitalization in both acute and chronic heart failure cohorts.

What makes the marker attractive in principle is its simplicity: both components come from laboratory tests that are ordered routinely in cardiac care, so no additional procedure, equipment, or specialist time is required. Yet the study's authors are careful to emphasize that the PNI is not a direct or nutrition-specific measure. Albumin levels are shaped by inflammation, hepatic synthesis, congestion, fluid status, and overall disease severity, while lymphocyte counts respond to immune activation, infection, medications, and physiologic stress.

The index therefore captures a mixture of nutritional, inflammatory, immune, congestion-related, and disease-severity signals rather than nutritional status alone. A recent global meta-analysis estimated that approximately 43 percent of older adults with the condition experience some degree of cognitive deficit, and affected patients tend to show poorer self-care, weaker medication adherence, reduced quality of life, and worse clinical outcomes. The Montreal Cognitive Assessment, or MoCA, is widely used to detect these deficits because it is sensitive to mild impairment, but routine cognitive testing is not universally available in cardiology settings.

That gap has fueled interest in inexpensive biomarkers that might flag patients who warrant formal neuropsychological evaluation. The new study set out to determine whether the PNI differed between heart failure patients with MoCA-defined cognitive impairment and those without it, and whether any difference persisted after adjustment for clinically relevant covariates. The investigation was designed as a cross-sectional observational study conducted between February and May 2025.

Researchers initially identified 189 patients with clinically diagnosed heart failure through an electronic medical record screening list on the inpatient cardiology ward. Twenty-one individuals declined participation, and after an eligibility assessment a further 15 patients were excluded, leaving 153 participants in the final analysis. Eligibility required that patients be at least 18 years old and carry a diagnosis of heart failure according to the 2024 Chinese Guidelines for the Diagnosis and Treatment of Heart Failure.

Exclusion criteria removed patients with gastrointestinal diseases including malignancies, Crohn's disease, inflammatory bowel disease, or ulcerative colitis; those with neurological disorders such as Alzheimer's disease or Parkinson's disease; patients with psychiatric conditions such as depression or schizophrenia; anyone with a history of cancer; and individuals unable to cooperate with cognitive testing because of severe visual or auditory impairments. The protocol was approved by the hospital's ethics committee, and all participants provided written informed consent. Cognitive status was classified using the MoCA total score, with one point added to the original score for any participant with fewer than 12 years of education, a prespecified correction applied before outcomes were derived and before any statistical analysis.

Patients were then divided into a cognitive impairment group, defined as an education-corrected MoCA score below 26, and a non-impairment group at or above that threshold. The result was striking: 93 of the 153 patients, or 60.8 percent, met criteria for cognitive impairment, a prevalence higher than the roughly 43 percent pooled estimate from the global meta-analysis, though the authors note it falls within the range reported across contemporary heart failure cohorts. They attribute the elevated figure to a potentially older and more clinically complex population, the sensitive MoCA threshold employed, and local referral patterns.

The baseline comparison revealed sharp demographic and clinical differences between the groups. Patients with cognitive impairment were older on average, at 68.0 versus 53.9 years, were more likely to be women, at 41.9 versus 21.7 percent, were less likely to have completed senior high school or above, and had lower body mass index. Coronary artery disease was more frequent in the impairment group, and median hemoglobin and albumin were both lower.

The median PNI stood at 44.10 in the impairment group against 47.08 in the non-impairment group, a difference that was highly statistically significant. Heart failure phenotype distribution, spanning reduced, mildly reduced, and preserved ejection fraction categories, did not differ significantly between the groups, and an exploratory interaction analysis found no evidence that the PNI-cognition association varied by phenotype, although the authors caution that the study was not powered for phenotype-specific inference. In unadjusted analyses, each five-point increase in the PNI was associated with lower odds of cognitive impairment, with an odds ratio of 0.410, but the researchers deliberately avoided using univariable results to select covariates.

Instead, the primary multivariable model was specified on clinical grounds to include the PNI alongside age, sex, and educational attainment. Because serum albumin and lymphocyte count are algebraic components of the PNI, they were never entered simultaneously with the index in the same model, a methodological decision that prevents the statistical artifacts of collinearity. In a separate alternative model adjusted for the same three covariates, each 5-gram-per-liter increase in albumin alone was independently associated with lower odds of cognitive impairment, with an odds ratio of 0.362.

The authors stress that these alternative-model estimates should not be read as mutually adjusted effects, since albumin is embedded within the composite index. The most sobering findings concern what the PNI adds beyond information already available from basic demographics. A base clinical model containing only age, sex, and educational attainment achieved an apparent area under the receiver operating characteristic curve of 0.868.

Adding the PNI raised this to 0.882, an increment of just 0.014 with a confidence interval crossing zero, and the paired DeLong test returned a p-value of 0.204, far from significance. Although the addition improved overall model fit by likelihood-ratio testing and nudged the Brier score down from 0.141 to 0.132, the authors explicitly state that these gains do not establish incremental clinical utility. In sensitivity comparisons, the PNI alone did not outperform albumin alone as a single marker, with areas under the curve of 0.713 versus 0.693, and the PNI-extended model did not outperform an albumin-extended model, at 0.882 versus 0.887.

After 1,000 bootstrap resamples corrected for optimism, the PNI-extended model's area under the curve settled at 0.869, and calibration checks, including a Hosmer-Lemeshow test with a p-value of 0.454, showed no evidence of poor fit. An exploratory decision-curve analysis suggested some net benefit over treat-all and treat-none strategies within a threshold probability range of 0.20 to 0.70, but the authors confine this to exploratory status. The authors argue that the direction of the association is biologically plausible.

Malnutrition and systemic inflammation frequently coexist in heart failure and may contribute to cognitive dysfunction through reduced physiologic reserve, endothelial dysfunction, immune activation, and impaired cerebral perfusion. The same processes that depress albumin and lymphocyte counts, including venous congestion, hemodilution, renal losses, and disease severity, may also be linked to cerebral hypoperfusion, vascular dysfunction, neuroinflammation, and diminished cognitive reserve. The observed prevalence of 60.8 percent, while above the meta-analytic average, is consistent with the spread across modern heart failure studies.

One seemingly counterintuitive detail, that patients with cognitive impairment had a higher median ejection fraction and a greater share of preserved-ejection-fraction phenotype, may partly reflect the older age and higher proportion of women in that group, characteristics frequently associated with preserved ejection fraction, alongside residual differences in comorbidities, disease duration, and treatment. The authors warn that this cross-sectional difference must not be interpreted as evidence that higher ejection fraction contributes to cognitive impairment. The study's limitations are substantial and clearly acknowledged.

The cross-sectional design precludes any conclusion about temporality, causality, or the prediction of future cognitive decline. The sample was modest and drawn from a single center, so selection bias from recruitment and exclusions cannot be excluded. Residual confounding by depression, frailty, prior stroke, sleep disorders, socioeconomic status, medication use, heart failure duration, and other severity measures remains possible.

The PNI was derived from single measurements, and no direct nutritional assessments such as dietary intake records, recent weight loss documentation, the Mini Nutritional Assessment, or the Global Leadership Initiative on Malnutrition criteria were available, so a nutrition-specific interpretation is not supported. Even with bootstrap correction, the model was developed and internally assessed in the same small cohort, meaning the performance estimates describe internal behavior rather than externally validated accuracy. The authors conclude that lower PNI is independently associated with prevalent cognitive impairment in heart failure, but that the association does not establish incremental clinical utility, and prospective multicenter external validation with direct nutritional assessment is required before the index could be considered for clinical classification or screening.

Subject of Research: Association between the prognostic nutritional index and cognitive impairment in patients with heart failure

Article Title: Association between the prognostic nutritional index and cognitive impairment in patients with heart failure: a cross-sectional study

Article References: Li, C., Gao, Y., Niu, Q., Chen, L., & Lin, Y. (2026). Association between the prognostic nutritional index and cognitive impairment in patients with heart failure: a cross-sectional study. Frontiers in Nutrition, 13, Article 1939333. https://doi.org/10.3389/fnut.2026.1939333

Image Credits: AI Generated

DOI: 10.3389/fnut.2026.1939333

Keywords: heart failure, cognitive impairment, prognostic nutritional index, serum albumin, lymphocyte count, Montreal Cognitive Assessment, cross-sectional study, nutritional inflammation marker, cardiology, biomarkers, logistic regression, Fujian Medical University