The Association Between Neutrophil-Lymphocyte Ratio and Variability with New-Onset Dementia: A Population-Based Cohort Study

Study design and population

This study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (Reference No. UW 20–250) and Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (Reference No. 2018.309, 2018.643). This is a retrospective cohort study of patients attending family medicine clinics between January 1, 2000, to December 31, 2003, who have both neutrophil and lymphocyte measurements in the records. They were followed up until December 31, 2019. The patients were identified from the Clinical Data Analysis and Reporting System (CDARS), a territory-wide database that centralizes anonymized patient information from individual local hospitals. There was no adjudication of the outcomes as this relied on the International Classification of Diseases Ninth Edition (ICD-9) coding or a record on the death registry. The coding was performed by the clinicians or administrative staff, who were not involved in the model development. No blinding was performed for the predictor as the values were obtained from the electronic health records. The system has been used by both our team and other teams in Hong Kong [25–27]. Prior comorbidities at baseline were extracted. Charlson’s standard comorbidity index was also calculated. ICD-9 codes used to identify prior comorbidities and the outcomes are provided in Supplementary Table 1. Baseline NLR is calculated based on the absolute neutrophil and lymphocyte count on the initial admission date (from January 1, 2000, to December 31, 2003). The formula to calculate the NLR variability measures are presented in Supplementary Table 2.

Statistical analysis and primary outcomes

The primary outcomes were Alzheimer’s disease and related dementia and non-Alzheimer’s dementia [28, 29] (Supplementary Table 1). Mortality data were obtained from the Hong Kong Death Registry, a population-based official government registry with the registered death records of all Hong Kong citizens. Mortality was recorded using the International Classification of Diseases Tenth Edition (ICD-10) coding.

Descriptive statistics were used to summarise the baseline and clinical characteristics of patients. Continuous variables were presented as median (95% confidence interval [CI] or interquartile range [IQR]) and categorical variables were presented as count (%). The Mann-Whitney U test was used to compare continuous variables. The χ² test with Yates’ correction was used for 2×2 contingency data. We divided the cohorts according to stratified NLR into four quantiles: <1.87, 1.87–2.85, 2.85–5.44, and >5.44. Univariable and multivariable Cox regression was used to explore the possible relationships between NLR and adverse outcomes. Weibull model was conducted as a sensitivity analysis to demonstrate the relationship with the consideration of survival time and censoring. Restricted cubic spline was used to delineate the relationship between the NLR and the outcomes after adjustments, and patients with the top or bottom 5% NLR were trimmed. Hazard ratios (HRs) with corresponding 95% CIs and p values were reported. All significance tests were two-tailed and considered significant if p values were equal to or less than 0.05. No imputation was performed for missing data. Data analyses were performed using R-Studio software (Version: 1.1.456), STATA (Version: 16.1) and Python (Version: 3.6).

Basic characteristics

The cohort composed of an overall 9760 patients (41.08% males, median age: 70.2 years old [IQR: 56.68–77.78]; median follow-up duration: 4756.5 days [IQR: 1972.5–6636.5]) after excluding 655 patients with a prior diagnosis of Alzheimer’s disease and related dementia, non-Alzheimer’s dementia, steroid use, infection within a week at baseline, baseline haematological malignancies, and immunodeficiencies (Fig. 1) [30].

OPEN IN VIEWER

Fig. 1

Flow chart for the identification, inclusion and exclusion of study subjects. IR, incidence rate; NLR, neutrophil-lymphocyte ratio.

Among the included main cohort (N = 9760), 529 patients developed Alzheimer’s disease and related dementia (incidence rate [IR]: 5.72%), and 56 patients developed, non-Alzheimer’s dementia (IR: 0.61%) and 6,121 patients passed away (IR: 62.71%). The characteristics of patients were stratified into four NLR quantiles: <1.87 (N = 2,422), 1.87–2.85 (N = 2,428), 2.85–5.44 (N = 2,453), and >5.44 (N = 2,457) (Table 1). The number of patients with Alzheimer’s disease and related dementia increased as the NLR increased, however, the number of patients with non-Alzheimer’s dementia was not different. The characteristics of the patients who developed adverse outcomes are presented in Supplementary Table 3.

The association between NLR and dementia

Univariable Cox regression identified the risk factors for new onset Alzheimer’s disease and related dementia and non-Alzheimer’s dementia (Supplementary Table 4). After adjusting for the significant covariates in the univariate analysis (demographics, prior comorbidities, and medications), the patients with NLR >5.44 were associated with Alzheimer’s disease and related dementia (HR: 1.50, 95% CI: 1.17–1.93) but not non-Alzheimer’s dementia (HR: 1.33; 95% CI: 0.60–2.95) (Table 2). The result remained consistent in the Weibull model as a sensitivity analysis (Table 3).

The cumulative hazard of Alzheimer’s disease and related dementia was higher among patients with higher NLR (Fig. 2). This trend was also consistent if the NLR were stratified by the clinical NLR quantiles (Supplementary Figure 1). The restricted cubic splines models demonstrated that as the NLR increased, the HR of Alzheimer’s disease and related dementia also increased (Fig. 3). In the subgroup analysis, the association remained significant across gender and age (Supplementary Figures 2 and 3). Across different subgroups above 50 years old, NLR >5.44 were associated with Alzheimer’s disease and related dementia (all p < 0.05). Besides, among patients above 60 years old, NLR >5.44 were associated with non-Alzheimer’s dementia (all p < 0.05) (Supplementary Table 5).

OPEN IN VIEWER

Fig. 2

Cumulative incidence curves stratified by NLR quantiles to predict Alzheimer’s disease and related dementia and non-Alzheimer’s dementia. NLR, neutrophil-lymphocyte ratio.

OPEN IN VIEWER

Fig. 3

Restricted cubic spline curves for predicting Alzheimer’s disease and related dementia and non-Alzheimer’s dementia.

The relationship between NLR variability and dementia was also explored. The NLR variability measures, including NLR variance, NLR standard deviation [SD], SD/initial-NLR and NLR coefficient of variation (CV) were not associated with Alzheimer’s disease and related dementia following adjustments. However, NLR CV, but not other NLR variability measures, was associated with increased risks of non-Alzheimer’s dementia (HR: 4.93; 95% CI: 1.03–23.61) (Table 4).

Comparison with previous studies

Generally, NLR seems to have a role in predicting dementia in the future. Positive associations between NLR and dementia risk were demonstrated in several studies [23, 31]. Similarly, studies in the past have shown that patients with Alzheimer’s disease had raised NLR as compared to the healthy controls [32–34]. Similarly, an association between NLR and cognitive impairment in community-dwelling older patients was also demonstrated in one study [19]. Additionally, there is some evidence for the prognostic value of NLR for dementia. Halazun et al. suggested that patients undergoing carotid endarterectomy with NLR ≥5 had a higher risk of dementia post-operation [35].

Our results support the use of NLR as a predictor for Alzheimer’s and related dementia, which are consistent with several published studies that demonstrated higher NLRs in patients with Alzheimer’s disease compared with healthy controls [32, 37]. Moreover, Zhang et al. also found patients with higher NLR were at greater risk of developing Alzheimer’s dementia [23]. On the other hand, we did not find an association between NLR and non-Alzheimer’s dementia. Although this contradicts the findings of Zhang et al. that higher NLR is related to an increased risk of vascular dementia, our results seem to be consistent with existing literature [23]. Furthermore, Schröder et al. did not find significant differences between the NLRs of patients with different types of dementia (Alzheimer’s disease, vascular dementia, mixed dementia, and frontotemporal dementia) compared to those without dementia [38]. From the evidence gathered, NLR could be a good predictor of Alzheimer’s and related dementia. However, the role of NLR remains uncertain for its predictive value for non-Alzheimer’s dementia.

Inflammation and dementia

Several mechanisms is responsible for the association of NLR with dementia. Recent evidence suggests that all forms of dementia, including Alzheimer’s disease, and non-Alzheimer’s dementia, are characterized by neuroinflammation [39, 40]. Patients with dementia had higher levels of proinflammatory cytokines and lower levels of antioxidants [41].

Neutrophils play an important role in the development of dementia. Glial cells activate in response to neutrophil granule proteins, resulting in deterioration of Alzheimer’s disease [42–44]. Meanwhile, higher levels of lymphocytes, particularly regulatory T cells, may help suppress the inflammatory response, hence, dampening the progression of Alzheimer’s disease [45]. As such, the NLR may reflect the imbalance of immune response involved in dementia.

Atherosclerosis may also explain the association between NLR and dementia. Some studies have shown that NLR is an independent predictor for atherosclerosis [46, 47]. Additionally, an association between dementia and atherosclerosis of intracranial arteries has been established [48, 49]. An efficient blood supply to the brain is necessary for its normal function, thus, reduced blood flow is a contributor to the development of dementia.

Clinical relevance

This study provided substantial evidence to support the inclusion of NLR into the risk assessment of dementia within the family medicine patients cohort. Our study design included all adult patients that attend family medicine clinics in Hong Kong. Thus, the results of this study may be reflective of daily practice and applicable to clinical practice. NLR and its variability markers are emerging marker that contains valuable information on both neutrophils and lymphocytes, and are not altered by fluid imbalance in comparison to single components of the white blood cells [50]. These measures may also serve as inexpensive and quick markers that could be easily incorporated into risk stratification assessments. NLR may predict the development of Alzheimer’s disease and related dementia, particularly at early stages. Early diagnosis of dementia is crucial, as it helps delay the progression and reduce the severity of dementia. Our findings provided further evidence regarding the role of inflammation in the development of dementia, supporting the need for future research to evaluate the role of anti-inflammatory therapy for the prevention and treatment of dementia.

Limitations

Several limitations are present in this study which should be acknowledged. Firstly, the derived populations may differ from others due to comorbidities and demographic differences. The model should be externally validated using patient data from other regions. Secondly, given the retrospective nature of this study, residual or unmeasured confounders were not addressed. However, the established risk factors of dementia were included in the multivariable regression models. Besides, the observational study was also subjected to bias secondary to under-coding and documentation errors. As such, some covariates associated with dementia, such as smoking, alcohol uses, obesity, and substance use, were not included in this study. Moreover, the patients that were lost to follow-up apart from mortality were not documented in the study. Furthermore, as NLR was not routinely recorded in clinical practice, thus, variations between the frequency and timings of NLR measurements were not standardised in all patients. Last but not least, this retrospective study could only demonstrate the association but not the causation relationship between NLR and dementia.

Conclusions

This cohort study demonstrated that the baseline NLR was associated with the risks of Alzheimer’s disease and related dementia among the family medicine patients. Baseline NLR may allow earlier diagnosis of dementia, and patients may benefit from prompt intervention.