Abstract

Objective:

To evaluate the associations of the hemoglobin-albumin-lymphocyte-platelet (HALP) score with baseline physical-cognitive comorbidity, incident physical-cognitive comorbidity, and all-cause mortality using two independent cohorts.

Methods:

Data were derived from the West China Health and Aging Trend (WCHAT) cohort and the UK Biobank (UKB). Participants aged ≥50 years with available laboratory, covariate, and functional data were included. Physical-cognitive comorbidity was defined as the coexistence of low grip strength and cognitive impairment. Multivariable logistic regression was used for cross-sectional analyses in both cohorts. Cox proportional hazards models were used to examine incident physical-cognitive comorbidity in WCHAT and all-cause mortality among participants with baseline comorbidity in UKB. Continuous-variable, subgroup, sensitivity, and restricted cubic spline analyses were additionally performed, and an exploratory incidence analysis was conducted in UKB participants without baseline comorbidity.

Results:

A total of 5,957 participants in WCHAT and 101,655 participants in UKB were included in the baseline analyses. In both cohorts, higher HALP scores were associated with a lower risk of baseline physical-cognitive comorbidity. In the fully adjusted models, the odds ratio comparing Quartile 4 (Q4) with Quartile 1 (Q1) was 0.79 (95% CI: 0.62–0.99) in WCHAT and 0.77 (95% CI: 0.66–0.89) in UKB. In WCHAT, among 2,782 participants free of comorbidity at baseline, 330 incident events occurred, and higher HALP was associated with lower incident risk (Q4 vs. Q1: HR = 0.70, 95% CI: 0.52–0.94; per 1-SD increase: HR = 0.88, 95% CI: 0.79–0.99). In UKB, among 1,393 participants with baseline comorbidity, 227 deaths occurred, and higher HALP was associated with lower all-cause mortality (Q4 vs. Q1: HR = 0.65, 95% CI: 0.45–0.93; per 1-SD increase: HR = 0.80, 95% CI: 0.70–0.92). Findings were generally consistent in subgroup and sensitivity analyses, whereas time-dependent ROC-AUC analysis showed limited discrimination of the HALP-only model.

Conclusion:

Higher HALP scores were associated with lower risks of baseline and incident physical-cognitive comorbidity and with better survival among participants with baseline physical-cognitive comorbidity. HALP may provide complementary information for immune-nutritional risk assessment, but it should not be interpreted as an independent predictive tool.

Introduction

With the continued acceleration of global population aging, the health burden arising from the coexistence of physical functional decline and cognitive impairment has become a major concern in geriatrics and public health (1). In recent years, concepts such as “cognitive frailty” have been used to describe this complex geriatric syndrome, emphasizing the coexistence and clinical relevance of physical vulnerability and cognitive impairment (2). Systematic reviews and meta-analyses have shown that middle-aged and older adults with concurrent physical dysfunction and cognitive impairment are more likely to experience disability, hospitalization, dementia, and all-cause mortality than those with only one functional abnormality (3, 4). Therefore, identifying simple indicators that can detect high-risk individuals before or during the development of physical-cognitive comorbidity is of important clinical and public health value.

Among indicators of physical function, grip strength is widely regarded as an important surrogate for overall muscle strength and physiological reserve because it is easy to measure, inexpensive, and highly reproducible (5). Previous prospective studies and meta-analyses have shown that lower grip strength is closely associated with increased risks of cognitive impairment, dementia, and Alzheimer’s disease (6). A pooled longitudinal analysis across multiple cohorts further showed that higher grip strength was associated with a slower rate of cognitive decline and a lower risk of cognitive impairment, suggesting that physical functional decline and cognitive deterioration may evolve in parallel (7). In addition, studies in community-dwelling middle-aged and older adults have shown that baseline physical frailty predicts subsequent cognitive decline (8). In Chinese middle-aged and older adults, the West China Health and Aging Trend (WCHAT) study also indicated that physical frailty is closely related to cognitive impairment and exhibits distinct co-occurrence patterns and influencing factors (9). Together, these findings suggest that physical functional decline and cognitive impairment should be studied not as two isolated problems, but as an integrated geriatric syndrome.

From a mechanistic perspective, anemia, malnutrition, chronic low-grade inflammation, and immune imbalance are considered key shared pathways linking physical functional decline and cognitive impairment (10). Previous systematic reviews and meta-analyses have shown that anemia is significantly associated with cognitive decline and increased dementia risk, possibly through insufficient oxygen delivery, abnormal cerebral perfusion, and related pathological processes (11, 12). Low albumin levels not only indicate inadequate nutritional reserve but are also regarded as biological markers of systemic frailty and poor prognosis, and they have been associated with cognitive decline and adverse outcomes in middle-aged and older adults (13). Meanwhile, persistent low-grade inflammation and immune dysregulation are also thought to jointly promote muscle weakness, physical frailty, and neurocognitive decline (14). Therefore, compared with single indicators, a composite index integrating hematologic, nutritional, inflammatory, and immune information may better identify middle-aged and older adults at high risk for physical-cognitive comorbidity.

The HALP score comprises hemoglobin, albumin, lymphocytes, and platelets, and it was originally proposed as a prognostic index for patients with gastric cancer (15). Biologically, hemoglobin and albumin reflect oxygen-carrying capacity and nutritional reserve, respectively, whereas lymphocytes and platelets may, to some extent, reflect immune status, inflammatory activation, and thrombo-inflammatory responses (16, 17). In recent years, the application of HALP has gradually expanded to non-neoplastic diseases. Previous studies have shown that a lower HALP score is associated with an increased risk of post-stroke cognitive impairment, suggesting its potential value for risk stratification in neurological disorders as well (18).

Against this background, we used two independent large-scale cohorts, the West China Health and Aging Trend (WCHAT) study and the UK Biobank (UKB), to examine the associations of the HALP score with physical-cognitive comorbidity and its prognosis (19, 20). The WCHAT study is a prospective cohort of community-dwelling middle-aged and older adults in western China, with relatively complete baseline and follow-up assessments. The UK Biobank is a large population-based prospective cohort in the United Kingdom, with a large sample size and long-term follow-up. In the present study, WCHAT was mainly used to evaluate incident physical-cognitive comorbidity, whereas UK Biobank was mainly used to evaluate all-cause mortality among participants with baseline physical-cognitive comorbidity and to provide a supplementary exploratory incidence analysis. We hypothesized that a higher HALP score, as a composite marker of better immune-nutritional status, would be independently associated with lower risk of physical-cognitive comorbidity and better survival.

Materials and methodsStudy design and population

This study was based on two independent population-based cohorts: the West China Health and Aging Trend (WCHAT) study and the UK Biobank (UKB) (19, 20). The WCHAT study is a prospective cohort of community-dwelling middle-aged and older adults in western China. At baseline in 2018, the cohort enrolled 7,536 participants, including 7,439 individuals aged 50 years and older, and collected relatively complete information on demographic characteristics, lifestyle factors, laboratory indicators, physical function, and cognitive status, with repeated follow-up assessments. The UK Biobank is a large-scale population-based prospective cohort in the United Kingdom. Between 2006 and 2010, more than 500,000 participants aged 40–69 years were recruited from across England, Wales, and Scotland, with extensive baseline information and long-term follow-up for major health outcomes. The participant selection flowchart is shown in Figure 1.

Study design and flowchart of participant selection in the WCHAT and UK Biobank cohorts. This flowchart illustrates the multi-stage recruitment and exclusion process for the WCHAT cohort used for prospective incidence analysis, and for the UK Biobank cohort used for baseline cross-sectional analysis, mortality analysis among participants with baseline physical-cognitive comorbidity, and exploratory incidence analysis among participants without baseline comorbidity.

This study was conducted in accordance with the Declaration of Helsinki. The WCHAT protocol was approved by the Biomedical Ethics Review Committee of West China Hospital, Sichuan University, and the UK Biobank study was approved by the North West Multi-center Research Ethics Committee. Written informed consent was obtained from all participants prior to enrollment.

Among WCHAT participants, the present study focused on middle-aged and older adults; therefore, age ≥50 years was used as the primary inclusion threshold in order to capture an earlier stage at which physical functional decline and cognitive deterioration may begin to accumulate, while also maintaining consistency across the two cohorts. After excluding participants with missing or extreme HALP values, missing key covariates, or missing baseline functional assessments, 5,957 individuals were included in the baseline cross-sectional analysis. After further excluding those with baseline physical-cognitive comorbidity and those lost to follow-up or with indeterminate outcomes, 2,782 participants remained for the prospective incidence analysis.

In the UK Biobank cohort, the original sample consisted of 502,357 participants. After applying the age criterion (≥50 years), HALP quality control, and screening for covariates and baseline functional data, 101,655 participants were included in the baseline cross-sectional analysis. Among them, 1,393 participants with baseline physical-cognitive comorbidity were included in the all-cause mortality analysis, and 14,050 participants without baseline comorbidity but with available repeated grip strength, repeated cognitive assessments, and valid follow-up time were included in the exploratory incidence analysis.

Exposure, covariates, and outcome definitions

The primary exposure was the HALP score, which was used to characterize immune-nutritional status. Based on fasting venous blood measurements, HALP was calculated as follows:

In the primary analyses, HALP was categorized into quartiles (Q1–Q4), with the lowest quartile (Q1) serving as the reference group. In supplementary analyses, HALP was entered into the models as a continuous variable and standardized as a Z score, with effect estimates reported per 1-standard deviation (SD) increase. To reduce the influence of outliers, extreme HALP values were defined according to cohort-specific distributions as observations below the 1st percentile or above the 99th percentile.

Physical-cognitive comorbidity was defined as the coexistence of low grip strength and cognitive impairment. In the WCHAT cohort, low grip strength was defined according to the Asian Working Group for Sarcopenia (AWGS 2019) criteria as <28.0 kg in men and <18.0 kg in women (21), and cognitive impairment was assessed using the Short Portable Mental Status Questionnaire (SPMSQ), with a score >2 indicating cognitive impairment (22). In the UK Biobank cohort, low grip strength was defined according to the European Working Group on Sarcopenia in Older People 2 (EWGSOP2) criteria as <27.0 kg in men and <16.0 kg in women (23). Cognitive impairment was based on an operational definition used in previous UK Biobank studies and was defined as a fluid intelligence score more than 1 SD below the mean of the baseline analytical sample (24). In both WCHAT and UK Biobank, maximum grip strength from both hands was used as the analytic grip strength variable.

This study included three levels of analysis. First, both WCHAT and UK Biobank were used for baseline cross-sectional analyses to evaluate the association between HALP and baseline physical-cognitive comorbidity. Second, WCHAT was used to evaluate the association between HALP and incident physical-cognitive comorbidity. Third, UK Biobank participants with baseline comorbidity were used to evaluate the association between HALP and all-cause mortality. In addition, an exploratory incidence analysis was conducted among UK Biobank participants without baseline comorbidity.

Covariate adjustment strategy

The same sequential covariate adjustment strategy was applied to both the logistic regression analyses for baseline physical-cognitive comorbidity and the Cox proportional hazards regression analyses for longitudinal outcomes. Model 1 was unadjusted. Model 2 was adjusted for age and sex. Model 3 was the fully adjusted model, including age, body mass index, sex, education, race/ethnicity, smoking status, alcohol drinking status, hypertension, diabetes, and coronary heart disease. For the WCHAT incidence analysis, baseline single impairment status, defined as the presence of either low grip strength alone or cognitive impairment alone at baseline, was additionally included in Model 3 to account for pre-existing functional or cognitive vulnerability before the onset of physical-cognitive comorbidity.

Statistical analysis

All statistical analyses were performed using Python (version 3.12). Continuous variables are presented as means ± standard deviations, and categorical variables as counts (percentages). For baseline comparisons, continuous variables were analyzed using Student’s t-test with Welch’s correction, and categorical variables using the χ2 test.

Multivariable logistic regression was used to assess the association between HALP and baseline physical-cognitive comorbidity, with results reported as odds ratios (ORs) and 95% confidence intervals (CIs). Cox proportional hazards models were used for incident physical-cognitive comorbidity in WCHAT and all-cause mortality in UK Biobank, with results reported as hazard ratios (HRs) and 95% CIs. In the UK Biobank mortality analysis, Kaplan–Meier survival curves were plotted and compared across HALP quartiles using the log-rank test; pairwise post hoc comparisons between quartiles were further adjusted using the Bonferroni method.

In addition to quartile-based analyses, HALP was also modeled as a continuous variable, and the associations with the outcomes were estimated per 1-SD increase. To explore potential nonlinear dose–response relationships, restricted cubic spline (RCS) models with four knots were fitted on the basis of the fully adjusted Cox models (25). Considering the right-skewed distribution of HALP, HALP was first natural-log transformed to reduce skewness and the influence of extreme high values (26), and the log-transformed HALP was then standardized before RCS modeling. The RCS plots were presented on the same transformed scale used for model fitting, namely the Z-score of ln (HALP). For each RCS analysis, both the p-value for the overall association and the p value for non-linearity were evaluated.

To evaluate the consistency of the findings across different populations, subgroup analyses were conducted and multiplicative interactions between HALP and stratification variables were tested. Stratification variables included age, sex, BMI, smoking, alcohol drinking, and chronic disease status. Two types of sensitivity analyses were also performed: first, analyses excluding events that occurred within the first year of follow-up to reduce potential reverse causation; and second, analyses restricted to participants aged ≥60 years to evaluate the robustness of the age threshold used in the main analyses.

In addition, we compared baseline characteristics between participants included in and excluded from the prospective analyses and conducted an exploratory incidence analysis among UK Biobank participants without baseline comorbidity. To evaluate the predictive performance boundary of HALP as a single indicator, HALP-only Cox proportional hazards models were fitted, and model discrimination was evaluated using Harrell’s C-index and time-dependent receiver operating characteristic curve analysis. Because the log-transformed and standardized RCS analyses did not show statistically significant nonlinearity or visually obvious turning points, exploratory threshold effect analyses were not further performed. A two-sided p < 0.05 was considered statistically significant.

ResultsParticipant selection and baseline characteristics

The participant selection flowchart is shown in Figure 1. In WCHAT, 7,536 baseline participants were initially considered. After excluding those aged <50 years and those with missing or extreme HALP values, missing covariates, or missing functional assessments, 5,957 participants were included in the baseline cross-sectional analysis. After further excluding participants with baseline physical-cognitive comorbidity and those lost to follow-up or with missing outcome data, 2,782 participants were ultimately included in the prospective incidence analysis. In UK Biobank, 502,357 baseline participants were initially screened. After the same quality-control procedures, 101,655 participants were included in the baseline cross-sectional analysis; among them, 1,393 participants with baseline physical-cognitive comorbidity entered the mortality analysis, and 14,050 participants without baseline comorbidity entered the supplementary exploratory incidence analysis. Variable-level missingness, the definition of extreme HALP values, and the stepwise exclusion process are shown in Supplementary Table S1; comparisons between included and excluded participants in the prospective analyses are shown in Supplementary Table S2.

Baseline characteristics of the two cohorts are presented in Table 1. In WCHAT, compared with the non-comorbidity group, the comorbidity group was older, had a lower BMI, lower HALP scores, worse cognitive scores, lower grip strength, and higher proportions of women, participants with lower education, and participants from minority ethnic groups. In UK Biobank, the comorbidity group was likewise older, had poorer cognitive performance and lower grip strength, and had higher proportions of hypertension, diabetes, and coronary heart disease. In both cohorts, HALP scores were lower in the comorbidity group than in the non-comorbidity group, although this difference did not reach statistical significance in UK Biobank.

VariablesWCHAT (N = 5,957)UK Biobank (N = 101,655)Non-comorbidity (n = 5,137)Comorbidity (n = 820)P-valueNon-comorbidity (n = 100,262)Comorbidity (n = 1,393)P-valueAge (years)61.62 ± 7.9466.79 ± 8.91<0.00160.22 ± 5.4061.73 ± 5.44<0.001BMI (kg/m2)25.57 ± 4.0024.49 ± 4.58<0.00127.47 ± 4.7128.47 ± 5.50<0.001HALP score73.50 ± 30.9266.25 ± 31.06<0.00154.67 ± 19.5454.39 ± 21.270.624Cognitive scorea0.89 ± 1.304.35 ± 1.43<0.0016.05 ± 2.112.45 ± 0.73<0.001Grip strength (kg)23.53 ± 8.7214.15 ± 4.59<0.00131.14 ± 10.7016.08 ± 6.09<0.001Sex, n (%)<0.0010.005Female3,123 (60.8%)622 (75.9%)53,968 (53.8%)803 (57.6%)Male2014 (39.2%)198 (24.1%)46,294 (46.2%)590 (42.4%)Education, n (%)<0.001<0.001Below junior high school2,891 (56.3%)763 (93.0%)30,575 (30.5%)889 (63.8%)Junior high school and above2,246 (43.7%)57 (7.0%)69,687 (69.5%)504 (36.2%)Race, n (%)<0.001<0.001Recognized minority group3,049 (59.4%)582 (71.0%)11,643 (11.6%)467 (33.5%)Majority/dominant group2088 (40.6%)238 (29.0%)88,619 (88.4%)926 (66.5%)Smoking, n (%)<0.001<0.001No4,126 (80.3%)712 (86.8%)91,675 (91.4%)1,217 (87.4%)Yes1,011 (19.7%)108 (13.2%)8,587 (8.6%)176 (12.6%)Alcohol use, n (%)<0.001<0.001No4,112 (80.0%)718 (87.6%)7,694 (7.7%)323 (23.2%)Yes1,025 (20.0%)102 (12.4%)92,568 (92.3%)1,070 (76.8%)Hypertension, n (%)0.577<0.001No2,317 (45.1%)379 (46.2%)69,530 (69.3%)814 (58.4%)Yes2,820 (54.9%)441 (53.8%)30,732 (30.7%)579 (41.6%)Diabetes, n (%)0.016<0.001No4,410 (85.8%)730 (89.0%)94,288 (94.0%)1,172 (84.1%)Yes727 (14.2%)90 (11.0%)5,974 (6.0%)221 (15.9%)Coronary heart disease, n (%)0.123<0.001No4,739 (92.3%)743 (90.6%)95,193 (94.9%)1,223 (87.8%)Yes398 (7.7%)77 (9.4%)5,069 (5.1%)170 (12.2%)

Baseline characteristics of participants in the WCHAT and UK Biobank cohorts according to baseline physical-cognitive comorbidity status.

Continuous variables are presented as mean ± standard deviation (SD), and categorical variables are presented as number (percentage). aCognitive score indicates the SPMSQ score in the WCHAT cohort (higher scores indicate worse cognitive function) and the Fluid Intelligence score in the UK Biobank cohort (lower scores indicate worse cognitive function). Grip strength indicates maximum grip strength derived from both hands in both the WCHAT cohort and the UK Biobank cohort. BMI, body mass index; HALP, hemoglobin-albumin-lymphocyte-platelet score.

To quantify heterogeneity in the operational definitions of physical-cognitive comorbidity across cohorts, we further summarized the baseline prevalence and follow-up event rates under the cohort-specific definitions used in WCHAT and UK Biobank (Supplementary Table S3). The results showed clear differences between WCHAT and UK Biobank in the definitions of low grip strength and cognitive impairment, indicating that the identified comorbidity phenotypes were not entirely equivalent in severity. These findings support focusing on the direction of association, robustness, and biological consistency rather than mechanically comparing absolute effect sizes across the two cohorts.

Cross-sectional associations between HALP and baseline physical-cognitive comorbidity

Cross-sectional results are shown in Table 2. In WCHAT, using the lowest HALP quartile (Q1) as the reference, both the unadjusted and minimally adjusted models showed that higher HALP quartiles were associated with lower risk of baseline physical-cognitive comorbidity. In the fully adjusted model, the association was attenuated but remained significant: compared with Q1, the ORs for Q3 and Q4 were 0.75 (95% CI: 0.60–0.94) and 0.79 (95% CI: 0.62–0.99), respectively. In UK Biobank, after full adjustment, the ORs for Q2, Q3, and Q4 versus Q1 were 0.83 (95% CI: 0.71–0.96), 0.72 (95% CI: 0.62–0.84), and 0.77 (95% CI: 0.66–0.89), respectively, indicating that higher HALP scores were independently associated with lower risk of baseline physical-cognitive comorbidity.

ModelHALP quartilesWCHAT cohort OR (95% CI)P-valueUK Biobank cohort OR (95% CI)P-valueModel 1Q1 (Ref)1–1–Q20.72 (0.59, 0.87)<0.0010.83 (0.72, 0.96)0.012Q30.56 (0.46, 0.69)<0.0010.76 (0.65, 0.88)<0.001Q40.52 (0.42, 0.64)<0.0010.92 (0.80, 1.07)0.274Model 2Q1 (Ref)1–1–Q20.77 (0.63, 0.94)0.0110.84 (0.72, 0.97)0.018Q30.62 (0.50, 0.76)<0.0010.78 (0.67, 0.90)0.001Q40.59 (0.47, 0.73)<0.0010.96 (0.83, 1.10)0.535Model 3Q1 (Ref)1–1–Q20.88 (0.71, 1.09)0.2380.83 (0.71, 0.96)0.014Q30.75 (0.60, 0.94)0.0110.72 (0.62, 0.84)<0.001Q40.79 (0.62, 0.99)0.040.77 (0.66, 0.89)<0.001

Cross-sectional associations between HALP score and baseline physical-cognitive comorbidity in the WCHAT and UK Biobank cohorts.

Results are presented as odds ratios (ORs) and 95% confidence intervals (CIs). Model 1 was unadjusted. Model 2 was adjusted for age and sex. Model 3 was fully adjusted for age, BMI, sex, education, race/ethnicity, smoking, alcohol use, hypertension, diabetes, and coronary heart disease (CHD). HALP, hemoglobin-albumin-lymphocyte-platelet score.

When HALP was analyzed as a continuous variable, the results were generally consistent with the quartile analyses (Supplementary Table S4). In the fully adjusted model, each 1-SD increase in HALP was associated with a 9% lower risk of baseline comorbidity in WCHAT (OR = 0.91, 95% CI: 0.84–0.99) and an 8% lower risk in UK Biobank (OR = 0.92, 95% CI: 0.87–0.97). Cross-sectional subgroup analyses are presented in Supplementary Table S5. No significant interactions were observed in WCHAT. In UK Biobank, statistically significant interactions were observed in strata defined by education, BMI, and hypertension, whereas the direction of association was generally consistent across the remaining strata.

Prospective associations between HALP and incident physical-cognitive comorbidity in WCHAT

In WCHAT, 2,782 participants free of physical-cognitive comorbidity at baseline were included in the prospective analysis, during which 330 incident events occurred over a median follow-up of 5 years (IQR: 3–5 years). The Cox proportional hazards results are shown in Table 3. After full adjustment, compared with Q1, the risks of incident physical-cognitive comorbidity were reduced by 40, 31, and 30% in Q2, Q3, and Q4, respectively, corresponding to HRs of 0.60 (95% CI: 0.44–0.81), 0.69 (95% CI: 0.51–0.92), and 0.70 (95% CI: 0.52–0.94). These findings indicate a stable association between higher HALP levels and lower risk of incident comorbidity.

ModelHALP quartilesHR (95% CI)P-valueModel 1Q1 (Ref)1-Q20.62 (0.46, 0.84)0.002Q30.73 (0.54, 0.97)0.031Q40.72 (0.54, 0.96)0.026Model 2Q1 (Ref)1-Q20.60 (0.44, 0.81)<0.001Q30.70 (0.52, 0.94)0.017Q40.68 (0.51, 0.91)0.010Model 3Q1 (Ref)1-Q20.60 (0.44, 0.81)0.001Q30.69 (0.51, 0.92)0.013Q40.70 (0.52, 0.94)0.018

Prospective associations between baseline HALP score and incident physical-cognitive comorbidity in the WCHAT cohort.

Among 2,782 participants free of physical-cognitive comorbidity at baseline, 330 incident events occurred during a median follow-up of 5 years (IQR: 3–5 years). Results are presented as hazard ratios (HRs) and 95% confidence intervals (CIs) from Cox proportional hazards models. Model 1 was unadjusted. Model 2 was adjusted for age and sex. Model 3 was fully adjusted for age, BMI, sex, education, race/ethnicity, smoking, alcohol use, hypertension, diabetes, coronary heart disease, and baseline single impairment status. HALP, hemoglobin-albumin-lymphocyte-platelet score.

When HALP was modeled as a continuous variable, the results remained robust (Supplementary Table S6). In the fully adjusted model, each 1-SD increase in HALP was associated with a 12% lower risk of incident physical-cognitive comorbidity (HR = 0.88, 95% CI: 0.79–0.99). RCS analysis using log-transformed and standardized HALP further supported these findings, showing a significant overall association with incident comorbidity risk (P for overall = 0.025), without evidence of significant nonlinearity (P for non-linearity = 0.238) (Figure 2). The RCS curve was displayed on the log-transformed and standardized HALP scale used for model fitting. In addition, no significant interactions were observed in the prospective subgroup analyses (Supplementary Table S5).

Restricted cubic spline association between log-transformed and standardized HALP and incident physical-cognitive comorbidity in the WCHAT cohort. Restricted cubic spline (RCS) analysis was performed using a Cox proportional hazards model to evaluate the association between baseline HALP and incident physical-cognitive comorbidity during follow-up in the WCHAT cohort. HALP was natural-log transformed and then standardized before RCS modeling. The x-axis is displayed as the Z-score of ln (HALP), consistent with the scale used for model fitting. The solid line represents the adjusted hazard ratio (HR), and the shaded area indicates the 95% confidence interval (CI). The reference value was set at 0 on the log-HALP Z-score scale, corresponding to HR = 1.00. Models were fully adjusted for age, body mass index, sex, education, race/ethnicity, smoking status, alcohol use, hypertension, diabetes, coronary heart disease, and baseline single impairment status. The overall association and nonlinearity were assessed by Wald tests.

In sensitivity analyses excluding events that occurred within the first year of follow-up, the protective association for Q4 versus Q1 remained in WCHAT (HR = 0.60, 95% CI: 0.39–0.91)