GlycA as an Effect Modifier of Protein–Mortality Associations: A Prospective Cohort Study

Xinru Wang; Ziyan Qiao; Chengzhe Tao; Sijing Liao; Zhi Li; Qiaoqiao Xu; Yun Fan; Yufeng Wang; Yan Han; Xiuliang Dai; Chuncheng Lu

doi:10.1111/cts.70542

. 2026 Apr 21;19(4):e70542. doi: 10.1111/cts.70542

GlycA as an Effect Modifier of Protein–Mortality Associations: A Prospective Cohort Study

Xinru Wang ^1,^2,³, Ziyan Qiao ^1,^2,³, Chengzhe Tao ^2,^3,⁴, Sijing Liao ^1,^2,³, Zhi Li ^2,³, Qiaoqiao Xu ^1,^2,³, Yun Fan ^1,^2,³, Yufeng Wang ¹, Yan Han ^5,^✉, Xiuliang Dai ^1,^✉, Chuncheng Lu ^1,^2,^3,^✉

PMCID: PMC13097604 PMID: 42012021

ABSTRACT

The heterogeneity in health effects of circulating proteins remained unclear. Previous studies have identified that glycoprotein acetyls (GlycA), a stable biomarker of inflammation, were associated with risks of mortality and chronic diseases. However, it remained unclear whether the health risks of proteins may differ across people with varied GlycA levels. Based on the multi‐omics profiling of the UK Biobank prospective cohort, we evaluated whether GlycA statistically modifies the protein–mortality associations. In the discovery dataset (n = 24,134), we observed that GlycA significantly modified the associations of ANG, CRHBP, CXCL16, and PRRT3 with all‐cause mortality, and these findings were replicated in the validation dataset (n = 6081). Subgroup analyses further indicated that associations between these proteins and mortality can be modulated by GlycA levels. Through exploratory analyses focused on chronic diseases and cause‐specific mortality, we identified that cross‐products of GlycA with these proteins can be associated with cancer mortality and heart failure. Together, the findings will expand our understanding of heterogeneity in protein‐health associations and highlight several potential therapeutic targets for interventions across people with varied inflammation status.

Keywords: cohort, glycoprotein acetyls, mortality risk, proteomics, UK biobank

Study Highlights

What is the current knowledge on the topic?

GlycA is a nuclear magnetic resonance (NMR) signal serving as an emerging biomarker of systemic inflammation. Compared to acute‐phase proteins such as high‐sensitivity C‐reactive protein (hsCRP), GlycA exhibits greater stability and has been found to correlate with all‐cause mortality.

What question did this study address?

This study investigated the effect‐modifying of GlycA on the association between the proteome and mortality, and explored whether the association between protein levels and all‐cause mortality and chronic diseases differed among participants in varying inflammatory states.

What does this study add to our knowledge?

Our analysis observed GlycA might be a significant effect modifier in the associations of ANG, CRHBP, CXCL16, and PRRT3 with all‐cause mortality. Notably, these pairs of GlycA and proteins were also linked to heart failure and cancer mortality, with the direction of association consistent with that observed for all‐cause mortality.

How might this change clinical pharmacology or translational science?

This research may broaden our understanding of the relationship between proteins and health, potentially providing multiple therapeutic targets for future interventions in populations with different inflammatory status.

1. Introduction

Chronic inflammation is a key pathway to multiple chronic diseases and lower life expectancies, causing substantial health burden [1, 2]. Biochemical indicators such as C‐reactive protein (CRP) and systemic immune‐inflammation index (calculated by blood cell counts) are traditional biomarkers of inflammation [3, 4]. However, these indicators may not quantify the variance of inflammation accurately and lack stability. Glycoprotein Acetyls (GlycA) levels quantified through the nuclear magnetic resonance (NMR) signals of multiple acute‐phase proteins is a new inflammatory biomarker exhibiting a relatively lower response to acute environmental changes [5, 6]. Compared to acute‐phase proteins such as high‐sensitivity C‐reactive protein (hsCRP), GlycA had higher stability within the organism and may explain more variance of individual systems inflammation, which might be a better measure of chronic inflammation [7, 8].

Currently, several large‐scale cohort studies have found that elevated GlycA level can predict higher risk of mortality [9, 10]. However, there are still great knowledge gaps for the health effects of GlycA. Previous studies have primarily considered GlycA as an exposure, with its potential role as an effect modifier for risk factors of mortality not being thoroughly explored. Shaped by lifestyles, environmental factors and genetics, proteomics measured by high throughput approaches can reflect the overall physiological status more comprehensively than the traditional clinical variables [11, 12, 13]. Currently, the associations of protein biomarkers with mortality risks have been estimated [14, 15]. However, it remains unclear whether such associations may be modified by systemic inflammatory status. Taken together, analyzing how GlycA modifies the association of proteins with mortality and morbidity will expand our understanding of the role of GlycA in health. Moreover, investigating this effect modification could provide evidence for designing future clinical practice strategies and identifying potential therapeutic targets.

The UK biobank has now released data of GlycA measurements for around a quarter of a million people and proteomics data for around 53,000 people, offering opportunities for us to fill the current knowledge gap [16, 17]. We performed a prospective cohort study with the following aims: (1) to identify the effect‐modifying role of GlycA on the relation between proteome and mortality and (2) to explore if the association of protein levels with mortality and chronic diseases can vary among participants in different inflammatory states.

2. Methods

2.1. Study Participants

The study was conducted based on UK Biobank, a prospective cohort of 502,413 participants aged 40–69 years recruited from 2006 to 2010. Participants' demographic characteristics and a wide range of health‐related data were collected at baseline. Blood, urine, and saliva samples were collected at baseline [18, 19]. The detailed procedure for collecting the study data can be seen on this website (https://www.ukbiobank.ac.uk/). We restricted participants with complete GlycA data (n = 274,354) for analysis. We excluded a small number of participants who have withdrawn consent (n = 57) and loss of follow‐up (n = 1297). We randomly stratified these participants into a discovery cohort (80%) and a validation cohort (20%). We further merged the proteomics data with the discovery cohort and the validation cohort, separately. Of note, the baseline proteomics data was derived from the UKB‐PPP sub‐cohort, including participants randomly selected from the entire UKB cohort. We finally included 30,215 participants in the proteome‐wide analysis of effect modification, with 24,134 in the discovery cohort and 6081 in the validation cohort. The study was conducted in December 2024 under UK Biobank Project Approval 70522. All the participants have signed the informed consent. The UK Biobank study has received ethics approval from the Northwest Multi‐Centre Research Ethics Committee.

2.2. NMR Metabolomics Measurement

Plasma metabolomics (249 features) from blood samples of around half of the entire UK Biobank cohorts were measured at baseline (i.e., 2006–2010) through NMR (Nuclear Magnetic Resonance) technique [17, 20]. The metabolic profiles can be classified into 15 categories such as Apolipoproteins, Cholesterol, Cholesteryl esters, Fatty acids, Free cholesterol, Glycolysis related metabolites, Inflammation, Ketone Inflammation, Ketone and more [20]. Of these biomarkers, we standardized the GlycA levels by log(n + 1) transformation and scaled the data as z‐scores standardization for subsequent analyses [21].

2.3. Proteomic Profiling

The UK Biobank applied Olink 3072 platform plasma proteins from ~50,000 individuals at the Olink Analysis Service in Sweden. Standard quantification was performed using the proximity extension assay, and 2923 proteins were measured [16], with more details of samples collecting, processing and quality control found on the website (https://biobank.ctsu.ox.ac.uk/crystal/cats.cgi). Those proteins with a missing rate greater than 20% were excluded (n = 12). All the missing values of protein biomarkers were imputed by the medians. All protein levels were standardized using z scores (SD from the mean) in discovery and validation datasets to approximate a normal distribution [22].

2.4. Covariates

Covariates were obtained primarily through questionnaires, biochemical assessments and physical examinations. To exclude the effect of potential confounding factors, sex (female or male), age (years), ethnicity (white or non‐white), educational attainment (College or University degree, A levels/AS levels or equivalent, O levels/GCSEs or equivalent, CSEs or equivalent, NVQ/HND/HNC or equivalent, Other professional qualifications and None of the above), income level (below £18,000, £18,000–£30,999, £31,000–£51,999, £52,000–£100,000, above £100,000), Townsend deprivation index (derived using geographic matching corresponding to postcode, continuous variable), BMI (weight (kg) divided by height squared (m²), kg/m²), smoking (non‐smoker, past smoker or current smoker), alcohol consumption (never, lower than three times a month or over once a week), and dietary score (calculated by combining intake of fish, vegetables, fruit, red meat and processed meat, coded 0–5) were included in the analysis as covariates [23]. We used median imputation for missing data for continuous variables such as BMI, income, etc. Categorical variables such as smoking and alcohol consumption were defined as the category with the highest frequency of occurrence in that variable if missing data were present.

2.5. Outcomes

We used death record data obtained from the central registry for participant‐reported causes of death. The main result we were interested in was all‐cause mortality. We also focused on cause‐specific mortality including cardiovascular disease (CVD) deaths (ICD‐10 codes I00‐I99), respiratory disease deaths (ICD‐10 codes J00‐J99), and cancer deaths (ICD‐10 codes C00‐C99) [24]. To better understand how GlycA modifies protein‐disease associations, we further explored major CVD (ICD‐10 code I20–I25, I50, I60–I64), IHD (ICD‐10 code I20–I25), HF (ICD‐10 code I50), and diabetes (ICD‐10 code E10–E14) outcomes [25]. We used the time of first occurrence in defining disease, including disease incidence and mortality. With regard to censoring date, if the latest reported date of death exceeded the mid‐point of the current month, the mid‐point of the following month was considered as the review date. The censoring date was established as 15 January 2023 in this study. Follow‐up periods were calculated as the interval from the date of enrolment to the date of death or the date of first occurrence of disease (for the analysis of incident diseases).

2.6. Statistical Analysis

We estimated the associations between GlycA and all‐cause mortality in participants with complete data of metabolomics through Cox proportional hazards (CPH) regression. Models were adjusted for age and sex, as previously recommended. We further regressed GlycA by tertiles on the mortality to examine its potential non‐linear relationship in the entire cohort. In addition, we evaluated the non‐linear relationship between GlycA and mortality using restricted cubic spline (RCS) based on CPH models, with three knots established [26]. P for non‐linearity < 0.01 was considered as significant non‐linearity. To evaluate the robustness of the results, we further adjusted for ethnicity, BMI, smoking status, alcohol consumption, healthy diet score, Townsend deprivation index, education, and income level to estimate the association between GlycA and all‐cause mortality in both the discovery and validation datasets.

We initially applied CPH regression models to systematically assess modification by GlycA of the associations between 2911 plasma proteins and all‐cause mortality in the discovery dataset, with P values corrected as false discovery rate (FDR) using the Benjamini–Hochberg method [27]. We further replicated these modification patterns in the validation dataset. Age and sex were included as covariates for the analysis (hereafter: minimally adjusted models). Protein‐mortality association modifications by GlycA with FDR < 0.05 in the discovery cohort and p value < 0.05 in the validation cohort were regarded as robust. For those robust proteins, we further estimated the modification of protein‐mortality associations by GlycA in the entire cohort by CPH regression. These models were adjusted for age, sex, ethnicity, educational attainment, income level, Townsend deprivation index, BMI, smoking, alcohol consumption, and dietary score (hereafter: fully adjusted models), with an FDR < 0.05 defining statistical significance. We additionally modeled GlycA as a continuous variable using restricted cubic spline interaction models with three knots, implemented via the interactionRCS R package. In fully adjusted models, we assessed the effect measure modification of GlycA on the association between proteins and mortality using this approach, which enabled simultaneous evaluation of non‐linear and interactive associations. CI of the estimates were computed through delta methods.

Subgroup analyses were performed. First, we categorized GlycA into three levels based on its tertiles and divided protein levels into two groups based on the medians. We estimated the survival probability among individuals with varying levels of GlycA and protein using Kaplan–Meier (KM) survival analysis [28]. We stratified participants by tertiles of GlycA level and performed survival analyses to estimate the association between protein and all‐cause mortality based on fully adjusted models. We in addition assessed the relationship between GlycA and all‐cause mortality stratified by tertiles of protein levels. For all subgroup analyses, p values were further corrected for multiple testing using the Benjamini–Hochberg method to obtain FDR values.

To further validate our findings, we performed the following sensitivity analyses. First, we conducted the analyses with data of individuals without missing covariates. Second, we excluded participants with less than 3 years of follow‐up to avoid reverse causality and replicated the survival analyses [23]. Third, age was used as a time scale to better control for confounding of age [29]. Fourth, the quadratic term of age was additionally adjusted to more accurately model the influence of age [30]. Finally, the modification patterns were analyzed again in the dataset with proteins imputed by the method of multiple imputation [24]. All analyses were based on two models, including a minimally adjusted model and a fully adjusted model. In addition, we repeated the screening and validating steps using geographically defined discovery and validation datasets. Participants from the Birmingham, Nottingham, Reading, Croydon, and Oxford assessment centres were included in the validation dataset, and participants from the remaining assessment centres were included in the discovery dataset.

Based on fully adjusted models, we further explored whether GlycA modified the associations of proteins of interest with cause‐specific mortality (i.e., CVD, cancer and respiratory disease) and incident diseases (i.e., diabetes, major CVD, IHD and HF). We used cause‐specific hazard models because our aim was to evaluate etiological associations and to model the instantaneous risk of the event among individuals who remained event‐free [31]. Additionally, we applied subdistribution hazard models to account for competing risks. We further performed subgroup analyses to assess the association between proteins and outcomes in individuals stratified by GlycA tertiles. Similarly, all subgroup analyses in these explorative assessments also applied Benjamini–Hochberg correction to control the FDR.

All analyses in this study were performed in R v4.0 software. The study was reported following STROBE guidelines.

3. Results

3.1. Baseline Characteristics of Study Participants

A total of 30,215 participants were included in this study, with 13,915 males and 16,300 females. The mean age was 56.87 years in the discovery cohort and 57.07 years in the validation cohort. During a median follow‐up of 13.89 years, 3348 deaths occurred, including 723 deaths from CVD, 1300 deaths from cancer and 289 deaths from respiratory diseases. Within the follow‐up periods, a total of 2695 cases of major CVD, 1906 cases of IHD, 649 cases of HF, and 1257 cases of diabetes were documented. The detailed design for inclusion and exclusion of the study population is shown in Figure 1, and the baseline characteristics of the study participants are shown in Table 1.

Flowchart of study population. We standardized both GlycA and protein data (less than 20% missing) for participants. The dataset was divided into discovery (80%) and validation (20%) sets, with more analytical details provided in Section 2.

TABLE 1.

Baseline characteristics of participants in the UK Biobank study.

Characteristics	Overall (N = 30,215)	Discovery cohort (N = 24,134)	Validation cohort (N = 6081)
Age, years	56.91 (8.19)	56.87 (8.19)	57.07 (8.17)
SBP, mmHg	137.81 (18.66)	137.78 (18.62)	137.93 (18.82)
DBP, mmHg	82.02 (10.18)	82.02 (10.17)	82.01 (10.21)
BMI, kg/m²	27.48 (4.78)	27.47 (4.77)	27.55 (4.80)
Ethnicity (%)
White	28,388 (94.4)	22,680 (94.4)	5708 (94.4)
Non‐white	1696 (5.6)	1355 (5.6)	341 (5.6)
Sex (%)
Female	16,300 (53.9)	13,018 (53.9)	3282 (54.0)
Male	13,915 (46.1)	11,116 (46.1)	2799 (46.0)
Healthy diet score (%)
0	564 (1.9)	449 (1.9)	115 (1.9)
1	2863 (9.5)	2312 (9.6)	551 (9.1)
2	6157 (20.4)	4937 (20.5)	1220 (20.1)
3	9436 (31.2)	7526 (31.2)	1910 (31.4)
4	7360 (24.4)	5869 (24.3)	1491 (24.5)
5	3835 (12.7)	3041 (12.6)	794 (13.1)
Smoker (%)
Non‐smoker	12,138 (40.2)	9764 (40.5)	2374 (39.0)
Past smoker	14,976 (49.6)	11,888 (49.3)	3088 (50.8)
Current smoker	3101 (10.3)	2482 (10.3)	619 (10.2)
Alcohol consumption (%)
Never	2582 (8.5)	2052 (8.5)	530 (8.7)
Lower than three times a month	6772 (22.4)	5398 (22.4)	1374 (22.6)
Over once a week	20,861 (69.0)	16,684 (69.1)	4177 (68.7)
Education level (%)
College or University degree	9416 (31.2)	7583 (31.4)	1833 (30.1)
A levels/AS levels or equivalent	3328 (11.0)	2671 (11.1)	657 (10.8)
O levels/GCSEs or equivalent	6709 (22.2)	5349 (22.2)	1360 (22.4)
CSEs or equivalent	1599 (5.3)	1280 (5.3)	319 (5.2)
NVQ, HND, HNC or equivalent	2036 (6.7)	1577 (6.5)	459 (7.5)
Other professional qualifications	1600 (5.3)	1263 (5.2)	337 (5.5)
None of the above	5527 (18.3)	4411 (18.3)	1116 (18.4)
Income level (%)
< 18,000	6393 (21.2)	5083 (21.1)	1310 (21.5)
18,000–30,999	11,485 (38.0)	9226 (38.2)	2259 (37.1)
31,000–51,999	6334 (21.0)	5036 (20.9)	1298 (21.3)
52,000–100,000	4772 (15.8)	3824 (15.8)	948 (15.6)
> 100,000	1231 (4.1)	965 (4.0)	266 (4.4)
Event of deaths (%)
All‐cause mortality	3348 (11.1)	2658 (11.0)	690 (11.3)
CVD mortality	723 (2.4)	572 (2.4)	151 (2.5)
Cancer mortality	1300 (4.3)	1036 (4.3)	264 (4.3)
Respiratory mortality	289 (1.0)	222 (1.0)	67 (1.1)
Event of disease (%)
Major CVD incidence	2695 (8.9)	2180 (9.0)	515 (8.5)
IHD incidence	1906 (6.3)	1539 (6.4)	367 (6.0)
HF incidence	649 (2.1)	513 (2.1)	136 (2.2)
Diabetes	1257 (4.2)	992 (4.1)	265 (4.4)

Open in a new tab

Abbreviations: CVD, cardiovascular disease; HF, heart failure; IHD, ischemic heart disease.

3.2. Effect Modification by GlycA on Protein‐Mortality Associations

GlycA was observed to be positively associated with mortality risk (HR per SD: 1.26, 95% CI: 1.25–1.28; Table S1). In the entire cohort, GlycA showed a linear relationship with all‐cause mortality after adjusting for age and sex (All p nonlinearity > 0.01; Figure 2a,b; Figures S1 and S2). In the discovery dataset, our proteome‐wide analysis of 2911 proteins revealed 30 with statistically significant effect modification by GlycA on their associations with all‐cause mortality, based on minimally adjusted models (Figure 2c; Table S2). Combining data from the discovery and validation datasets, we found that five of these modifications by GlycA were robust (Figure 2d; Table S3), including Angiogenin (ANG, HR: 1.10, p = 1.2E‐05), Corticotropin Releasing Hormone Binding Protein (CRHBP, HR: 1.11, p = 7.3E‐06), C‐X‐C Motif Chemokine Ligand 16 (CXCL16, HR: 1.12, p = 1.0E‐06), Proline Rich Transmembrane Protein 3 (PRRT3, HR: 1.11, p = 5.6E‐06), and Neurotrophic Receptor Tyrosine Kinase 3 (NTRK3, HR: 0.90, p = 8.1E‐06). In the fully adjusted model, the magnitudes of these GlycA modification terms were attenuated but remained significant. The Interaction RCS models further characterized the shape of effect modification by GlycA on the associations between proteins and mortality (Figure S3). ANG (HR: 1.33, 95% CI: 1.24–1.43 at GlycA mean + 2 SD), CXCL16 (HR: 1.35, 95% CI: 1.26–1.44 at GlycA mean + 2 SD), and CRHBP (HR: 1.28, 95% CI: 1.18–1.38 at GlycA mean + 2 SD) showed steadily increasing hazard ratios with higher GlycA concentrations, suggesting that elevated systemic inflammation strengthens their associations with mortality. NTRK3 (HR: 0.89, 95% CI: 0.82–0.96 at GlycA mean + 2 SD) exhibited a decreasing association at lower GlycA levels that subsequently plateaued, whereas PRRT3 (HR: 1.24, 95% CI: 1.18–1.31 at GlycA mean + 2 SD) displayed a distinct non‐linear pattern, with minimal change in risk until GlycA reached higher values.

Proteome‐wide statistical effect modification by GlycA on associations between proteins and all‐cause mortality. (a) Association of the tertiles of GlycA with all‐cause mortality in the entire cohort after adjusting for age and sex. (b) Nonlinear association between GlycA and all‐cause mortality adjusted for age and sex in the entire cohort. (c) The association of GlycA's effect modification of proteins with all‐cause mortality in the discovery dataset after adjustment for age and sex was considered significant at FDR < 0.05. (d) The −log₁₀(p value) of robust GlycA‐modified protein associations in the discovery and validation datasets, adjusted for age and sex. The dashed line corresponds to −log₁₀(0.05).

3.3. Subgroup Analysis

Based on the identified proteins, subgroup analysis was performed. For ANG, CXCL16 and CRHBP, KM analyses exhibited that all‐cause mortality risks were highest among participants with both high protein and GlycA levels. Both low NTRK3 levels in high inflammation states and high NTRK3 levels in low inflammation states were associated with an increased risk of all‐cause mortality. Of note, PRRT3 in those with high GlycA levels showed a stronger correlation with increased all‐cause mortality risk (Figure 3a). We analyzed the association between five plasma proteins and all‐cause mortality within participants of different GlycA levels (Figure 3b). We estimated that in subgroup with higher inflammation, the associations of ANG (HR per SD: 1.26, 95% CI: 1.20–1.33), CRHBP (HR: 1.24, 95% CI: 1.16–1.31) and CXCL16 (HR per SD: 1.34, 95% CI: 1.27–1.41) with all‐cause mortality were stronger. In addition, PRRT3 (HR per SD: 1.18, 95% CI: 1.12–1.24) showed a significant association with all‐cause mortality, but no such association was found in the medium and low GlycA groups. We also observed that NTRK3 (HR per SD: 1.12, 95% CI: 1.03–1.22) was associated with a higher risk of all‐cause mortality in the low GlycA group (Figure 3b). We also examined the association of GlycA with all‐cause mortality risk in participants with different protein levels. In individuals with different protein levels (i.e., ANG, CRHBP, CXCL16, and PRRT3), we noted the relationship between GlycA levels and all‐cause mortality risks were stronger with those with higher protein levels. Specifically, GlycA was associated with higher mortality risks in individuals with low NTRK3 levels (Figure 3c).

The survival probability and mortality risks for participants with varied GlycA and protein levels. (a) Kaplan–Meier survival plots were used to evaluate the survival rates across different subgroups stratified by GlycA and protein levels. The hazard ratios and FDR values for (b) the association of protein with all‐cause mortality stratified by GlycA tertiles and (c) the association of GlycA with all‐cause mortality stratified by protein tertiles, after adjusting for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

3.4. Sensitivity Analysis

After excluding individuals with follow‐up durations of less than 3 years (Table S4) and missing values of covariables (Table S5), the association related to GlycA's modification of protein‐mortality associations remained consistent with the main analyses. In survival analyses using age as the time scale (Table S6) or adjusting for the quadratic term of age (Table S7), the results remained stable. Results from the sensitivity analyses of proteins with multiple imputations were consistent with the primary findings (Table S8). Results from the sensitivity analyses were largely consistent with those from the main analysis. In the geographically defined internal validation, associations for four proteins (ANG, CRHBP, CXCL16, and PRRT3) were consistent with the primary findings, whereas the association for NTRK3 did not reach statistical significance (Table S9).

3.5. GlycA‐Modified Protein Associations in the Risk of Cause‐Specific Mortality and Disease Incidence

We further examined the associations between the identified GlycA‐modified proteins and cause‐specific mortality and chronic disease incidence (Figure 4a). The analysis revealed that these associations remained significant in cancer mortality, with proteins including ANG (HR: 1.08, 95% CI: 1.03–1.14), CRHBP (HR: 1.09, 95% CI: 1.03–1.14), CXCL16 (HR: 1.07, 95% CI: 1.01–1.12), NTRK3 (HR: 0.94, 95% CI: 0.89–0.99), and PRRT3 (HR: 1.12, 95% CI: 1.07–1.17). We also observed a significant effect modification by GlycA in relation to CVD mortality for CXCL16. However, no significant associations were found between GlycA‐modified proteins and respiratory disease‐specific mortality. Notably, results from the subdistribution hazard models were consistent with those from the cause‐specific hazard models (Table S10). Upon examining the cancer cause‐specific mortality risk across different inflammation levels in significant protein groups, we observed results similar to those for all‐cause mortality. ANG, CRHBP, and CXCL16 were linked to a stronger risk of cancer deaths in participants with higher GlycA levels. Among individuals with high levels of GlycA, PRRT3 was significantly associated with cancer mortality risks. Apart from this, we found that NTRK3 shows protective associations with cancer mortality among people with medium or high GlycA levels. The direction of such associations becomes reverse for people with low GlycA levels (Figure 4b). As shown in Figure S4, after stratifying GlycA by tertiles, we found that ANG, CRHBP, and CXCL16 had stronger associations with CVD and respiratory disease death in the highest GlycA level group. Similar to the results of all‐cause mortality, PRRT3 was only associated with an increased risk of these two cause‐specific deaths in the high GlycA level group.

Effect modification by GlycA on proteins in relation to cause‐specific mortality and incident chronic diseases. (a) Association between effect modification of proteins by GlycA and risk of death from cancer, cardiovascular disease, and respiratory disease. (b) Hazard ratios and FDR values for the association of protein with cancer‐specific mortality in individuals stratified by GlycA tertiles. (c–f) The modification terms of proteins by the GlycA inflammatory marker on the risk of major cardiovascular diseases (CVD), ischemic heart disease (IHD), heart failure (HF), and diabetes (including both incidence and mortality). The above were adjusted for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

For incident diseases, we discovered that the association between ANG and the risk of major CVD (HR: 1.06, 95% CI: 1.02–1.10), IHD (HR: 1.05, 95% CI: 1.00–1.10), and HF (HR: 1.16, 95% CI: 1.08–1.25) was significantly modified by GlycA. The association between CXCL16 and incident HF was significantly modified by GlycA (HR: 1.10, 95% CI: 1.03–1.18), while GlycA showed an opposite modification on the NTRK3‐HF association (HR: 0.91, 95% CI: 0.84–0.98; Figure 4c–e). Additionally, we observed significant effect modification by GlycA on the association of PRRT3 with diabetes risk (HR per SD: 1.08, 95% CI: 1.03–1.13; Figure 4f). In subgroup analyses, we observed that the highest GlycA level group, ANG, CRHBP, and CXCL16 were associated with increased risk of major CVD, IHD, and HF. Interestingly, NTRK3 showed a protective role in major CVD and IHD in participants with high and low GlycA levels but was inversely associated with the risk of HF and diabetes only in individuals with high GlycA levels. We found no significant association between PRRT3 and the onset of these chronic diseases in the cohort of GlycA stratification (Figure S5).

4. Discussion

Overall, our prospective cohort study systematically estimated GlycA's effect modification on proteins at proteome‐wide scale. We observed that the ANG, CRHBP, and CXCL16, and PRRT3 may be modified by GlycA, and such results indicated that there might be heterogeneity in associations between protein and mortality risks among participants with varied inflammation levels. Furthermore, we observed that these pairs of GlycA and proteins may have relations with heart failure and cancer mortality, with the same direction of effect sizes compared to those of all‐cause mortality risks. Given the observational nature of our study, these findings should be interpreted as hypothesis‐generating, but the study might expand our understanding of the health effects of GlycA on mortality and morbidity. Despite modest effect sizes, we observed substantial heterogeneity in associations between proteins and mortality risk across different GlycA strata. These results can provide clinical insights into the health effects relevant to inflammation and will likely inform precision risk classification and management among individuals with diverse inflammation levels.

Studies from several cohorts have shown that high levels of GlycA were positively correlated with the risk of all‐cause and cause‐specific mortality [9, 10, 32]. We also found that GlycA levels were associated with an elevated risk of all‐cause mortality in individuals with available protein measurements. Previous studies based on the UK Biobank proteomic cohort have revealed that plasma proteins are associated with disease and death [15, 33]. Our study further found that the proteins ANG, CRHBP, and CXCL16 contribute to excess mortality risk in individuals in the inflammatory state. These three proteins have been linked to elevated risks of kidney disease, cardiovascular disease, and cancer in previous studies [33], and we noted that the association between these proteins and mortality and morbidities might differ in participants' varied inflammatory status. A population‐based study showed that NTRK3 was negatively associated with all‐cause mortality [34], yet we identified a positive relationship between NTRK3 and mortality risk in individuals with low inflammation. PRRT3 levels in vivo have also been correlated with a variety of chronic diseases [33], but evidence of association with death is lacking. Our findings additionally showed that PRRT3 was specifically associated with the risk of death in those with relatively higher inflammation.

ANG is a family of growth factors that regulate vascular remodeling and angiogenesis. ANG binds to endothelial and smooth muscle cells and stimulates the generation of new blood vessels, which leads to coronary plaque instability [35]. Apart from this, several studies have shown that ANG is also involved in host defense, innate immune response, and tumorigenesis [36, 37]. Overall, the ANG‐linked pathway may be typically modulated by inflammatory responses, and both the increment of ANG and GlycA may cause excess risks of inflammation associated with an increased risk of death, cancer, and cardiovascular diseases. CXCL16 is produced by a variety of inflammatory cells, including macrophages, and is important in inducing tumor cell proliferation and migration, intercellular communication in the tumor microenvironment, angiogenesis, and atherosclerotic injury [38, 39, 40]. The presence of inflammatory response also increases CXCL16 expression, which in turn affects the survival time of individuals [38]. Mouse experiments also demonstrated that CXCL16 promotes the proliferation of cardiac fibroblasts while inhibiting the synthesis of collagen [41]. This evidence may explain part of the underlying mechanism through which the joint effect of CXCL16 and GlycA contributes to an increased risk of heart failure and related mortality.

CRHBP, a modulator of the Hypothalamic–Pituitary–Adrenal (HPA) axis, binds to and inactivates Corticotropin‐Releasing Hormone (CRH) and may prevent inappropriate pituitary–adrenal stimulation [42]. The dysregulation of the HPA axis and inflammation in the body may jointly result in disruption of the immune system and thus may cause excessive risks of mortality. Immune cell infiltration plays an integral role in cancer progression [43]. CRHBP expression positively correlates with the numbers of CD8+ T cells, macrophages, and cancer‐associated fibroblasts (CAFs) in various cancers [44]. CRHBP involved in immune cell infiltration may promote GlycA secretion, thereby accelerating tumor progression. Overall, this evidence supports our observed higher risks of CRHBP associated with all‐cause and cancer mortality in hyperinflammatory individuals.

NTRK3 expression has been reported to be associated with better prognosis in numerous cancers (melanoma, neuroblastoma, and colorectal cancer) [45, 46]. Previous studies have shown that the expression and activation of NTRK3 can trigger apoptosis in neuroblastoma cells [47]. Our analyses observed that NTRK3 levels were associated with a reduced risk of all‐cause mortality in those with medium or higher inflammatory status but associated with higher mortality risks in those with lower inflammation. However, this association was not consistently replicated in the geographically defined internal validation cohort. The variability observed across analyses suggests that the robustness of this finding warrants cautious interpretation. From a biological perspective, prior studies suggest that NTRK3 activity in inflammatory environments may play a role in inhibiting tumor progression, which might partially explain the observed associations with mortality [48, 49]. Nevertheless, these mechanistic considerations remain speculative and require further validation. In our study, NTRK3 was associated with a lower risk of heart failure in the inflammatory condition, and the exact mechanism needs to be further explored. It is important to note that NTRK3 is a membrane‐bound receptor tyrosine kinase whose biological function depends primarily on ligand binding and subsequent activation of downstream signaling pathways [50], rather than on its circulating protein abundance. Therefore, associations derived from plasma NTRK3 measurements should be interpreted with caution, as blood concentrations may not accurately reflect tissue‐specific expression, receptor activation status, or functional signaling activity. Future studies integrating tissue‐level quantification, ligand availability, and receptor activation models will be necessary to elucidate the specific mechanisms underlying the observed associations between NTRK3 and mortality risk in the context of systemic inflammation.

The detailed function of PRRT3 is unknown, and it has been used as a marker for diagnosing tumors in population studies [33]. We have found that PRRT3 was associated with an increased risk of cancer death under inflammatory conditions. PRRT3 may serve as a possible new biomarker for predicting future death in participants with high levels of inflammation. We also identified that PRRT3 was positively correlated with diabetes under inflammatory conditions, suggesting that PRRT3 may be a therapeutic target, and future studies are required to explore the underlying mechanisms.

The results of our study may provide implications for future clinical practice. In the study, we detected four proteins that were associated with stronger mortality risks among those with an increment in systemic inflammation based on a large‐scale cohort. Such observed results may deepen our understanding of the heterogeneity of proteomic‐associated mortality risk among people with varied inflammation, and the evidence we presented may help identify people within high‐risk strata more precisely and provide proof‐of‐principle evidence for personalized therapeutic targets for individuals with varied inflammatory levels.

Our study had several limitations. First, we estimated GlycA's effect modification of protein‐outcome associations based on data from the UK Biobank only for. Although this may not affect the overall validity of the study, it remains unclear whether the findings of our study can be generalizable to general population or minorities since the participants enrolled by the UK Biobank were mostly white and healthy participants. Moreover, external validation was not feasible because no other available large‐scale cohorts currently provide sufficiently large samples with both proteomic and metabolomic data alongside adequate follow‐up for mortality outcomes. Second, the proteomics profiling was measured by Olink, which was a targeted platform. Future studies may require non‐targeted techniques to better understand the modifications of GlycA on protein‐outcome associations. Third, considering the observational nature of this study, causal relationships cannot be established. Despite extensive covariate adjustment, residual confounding may exist, and several Bradford Hill criteria (e.g., experimental evidence, mechanistic confirmation) remain unevaluated. Future studies incorporating randomized controlled trial or mechanistic investigations are needed to assess potential causality. Last, our study identifies associations with modest effect sizes that are statistically significant. Therefore, the robustness of our findings should be considered preliminary and verified in external studies.

5. Conclusion

Our study found that GlycA exerted a significant effect modification on the associations between four proteins and mortality risk. The findings will expand our understanding of heterogeneity in associations of proteins and health, highlighting several potential therapeutic targets for interventions across people with varied inflammation status. Future studies are required to explore the underlying mechanisms and validate the results.

Author Contributions

X.W., Z.Q., and C.T. wrote the manuscript; X.W., C.T., Y.W., Y.H., X.D., and C.L. designed the research; X.W., Z.Q., Z.L., S.L., Q.X., and Y.F. performed the research; X.W., Z.Q., and C.T. analyzed the data.

Funding

This work was supported by the National Key Research and Development Program of China (No. 2023YFC2705700), the National Natural Science Foundation of China (No. 82271691) and “333 High‐level Talent Training Project” of the Jiangsu Province ((2022) 3‐16‐425).

Ethics Statement

The UK biobank study was approved by the Northwest Multicenter Research Ethical Committee. All individuals signed informed consent.

Consent

All the authors agreed to publish this article.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Association between GlycA by tertile and all‐cause mortality in individuals with and without proteomics data in the UK Biobank. Model was adjusted for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

Figure S2: Nonlinear association between GlycA and all‐cause mortality based on two models in both the discovery and validation datasets. Figures (a) and (b) were adjusted for age and sex. Figures (c) and (d) were adjusted for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

Figure S3: Interaction‐based restricted cubic spline analyses showing how GlycA levels modify the associations of proteins with all‐cause mortality. Models were adjusted for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

Figure S4: Forest plots of protein associations with CVD/respiratory disease mortality in populations with different levels of GlycA. Subgroup analyses were adjusted for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

Figure S5: Forest plots of protein associations with Major CVD/IHD/HF/diabetes events in populations with different levels of GlycA. Subgroup analyses were adjusted for age, sex, ethnicity, BMI, alcohol consumption, smoking status, educational level, healthy diet score, income level, and Townsend deprivation.

Table S1: Association between GlycA by tertile and all‐cause mortality based on two models.

Table S2: Significant proteins in statistical effect modification by GlycA on proteins associated with all‐cause mortality in discovery dataset.

Table S3: Performance of proteins of interest and their GlycA modification terms associated with all‐cause mortality.

Table S4: Association of identified GlycA's statistical effect modification of proteins with all‐cause mortality risk in the dataset excluding participants with follow‐up periods below 3 years.

Table S5: Association of identified statistical effect modification of proteins by GlycA with all‐cause mortality risk in the dataset including only participants with complete covariate data.

Table S6: Association of identified GlycA's statistical effect modification of proteins with all‐cause mortality risk using age as the time scale.

Table S7: Association of identified statistical effect modification of proteins by GlycA with all‐cause mortality risk in datasets with further adjustment for age‐squared.

Table S8: Association of identified statistical effect modification of proteins by GlycA with all‐cause mortality risk in datasets with multiple imputations.

Table S9: Association of identified statistical effect modification of proteins by GlycA with all‐cause mortality risk in geographically defined discovery and validation datasets.

Table S10: Association of identified statistical effect modification of proteins by GlycA with cause‐specific mortality risk based on subdistribution hazard models.

CTS-19-e70542-s001.docx^{(947.9KB, docx)}

Acknowledgments

The authors thank all the participants and staff in the UK biobank cohort for their substantial contributions for preparing the underlying data. This research has been conducted using the UK Biobank Resource under Application 70522, with Professor Chuncheng Lu (Nanjing Medical University) serving as the designated Principal Investigator.

Contributor Information

Yan Han, Email: hany@ncstdlc.org.

Xiuliang Dai, Email: daixiuliang@njmu.edu.cn.

Chuncheng Lu, Email: chunchenglu@njmu.edu.cn.

Data Availability Statement

The datasets used in the current study are available on the website listed in the methods section. The UK biobank data will be made available directly from UK biobank website (https://www.ukbiobank.ac.uk/) upon researchers' requests. The code used in this study is available at the following website: https://doi.org/10.5281/zenodo.18253922.

References

1. GBD , “Diseases and Injuries Collaborators. Global Burden of 369 Diseases and Injuries in 204 Countries and Territories, 1990–2019: A Systematic Analysis for the Global Burden of Disease Study 2019,” Lancet 396, no. 10258 (2019): 1204–1222, 10.1016/S0140-6736(20)30925-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
2. Guo H., Callaway J. B., and Ting J. P.‐Y., “Inflammasomes: Mechanism of Action, Role in Disease, and Therapeutics,” Nature Medicine 21, no. 7 (2015): 677–687, 10.1038/nm.3893. [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Vashist S. K., Venkatesh A. G., Marion Schneider E., Beaudoin C., Luppa P. B., and Luong J. H. T., “Bioanalytical Advances in Assays for C‐Reactive Protein,” Biotechnology Advances 34, no. 3 (2016): 272–290, 10.1016/j.biotechadv.2015.12.010. [DOI] [PubMed] [Google Scholar]
4. Hu B., Yang X.‐R., Xu Y., et al., “Systemic Immune‐Inflammation Index Predicts Prognosis of Patients After Curative Resection for Hepatocellular Carcinoma,” Clinical Cancer Research: An Official Journal of the American Association for Cancer Research 20, no. 23 (2014): 6212–6222, 10.1158/1078-0432.CCR-14-0442. [DOI] [PubMed] [Google Scholar]
5. Crick D. C. P., Khandaker G. M., Halligan S. L., Burgner D., Mansell T., and Fraser A., “Comparison of the Stability of Glycoprotein Acetyls and High Sensitivity C‐Reactive Protein as Markers of Chronic Inflammation,” Immunology 171, no. 4 (2024): 497–512, 10.1111/imm.13739. [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Würtz P., Kangas A. J., Soininen P., Lawlor D. A., Davey Smith G., and Ala‐Korpela M., “Quantitative Serum Nuclear Magnetic Resonance Metabolomics in Large‐Scale Epidemiology: A Primer on ‐Omic Technologies,” American Journal of Epidemiology 186, no. 9 (2017): 1084–1096, 10.1093/aje/kwx016. [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Connelly M. A., Otvos J. D., Shalaurova I., Playford M. P., and Mehta N. N., “GlycA, a Novel Biomarker of Systemic Inflammation and Cardiovascular Disease Risk,” Journal of Translational Medicine 15, no. 1 (2017): 219, 10.1186/s12967-017-1321-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
8. Ritchie S. C., Würtz P., Nath A. P., et al., “The Biomarker GlycA Is Associated With Chronic Inflammation and Predicts Long‐Term Risk of Severe Infection,” Cell Systems 1, no. 4 (2015): 293–301, 10.1016/j.cels.2015.09.007. [DOI] [PubMed] [Google Scholar]
9. Deelen J., Kettunen J., Fischer K., et al., “A Metabolic Profile of All‐Cause Mortality Risk Identified in an Observational Study of 44,168 Individuals,” Nature Communications 10, no. 1 (2019): 3346, 10.1038/s41467-019-11311-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Otvos J. D., Shalaurova I., May H. T., et al., “Multimarkers of Metabolic Malnutrition and Inflammation and Their Association With Mortality Risk in Cardiac Catheterisation Patients: A Prospective, Longitudinal, Observational, Cohort Study,” Lancet. Healthy Longevity 4, no. 2 (2023): e72–e82, 10.1016/S2666-7568(23)00001-6. [DOI] [PubMed] [Google Scholar]
11. Emilsson V., Ilkov M., Lamb J. R., et al., “Co‐Regulatory Networks of Human Serum Proteins Link Genetics to Disease,” Science (New York, N.Y.) 361, no. 6404 (2018): 769–773, 10.1126/science.aaq1327. [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Walker K. A., Chen J., Zhang J., et al., “Large‐Scale Plasma Proteomic Analysis Identifies Proteins and Pathways Associated With Dementia Risk,” Nature Aging 1, no. 5 (2021): 473–489, 10.1038/s43587-021-00064-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Deng S., Mao R., and He Y., “Unveiling New Protein Biomarkers and Therapeutic Targets for Acne Through Integrated Analysis of Human Plasma Proteomics and Genomics,” Frontiers in Immunology 15 (2024): 1452801, 10.3389/fimmu.2024.1452801. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Zhu K., Li R., Yao P., et al., “Proteomic Signatures of Healthy Dietary Patterns Are Associated With Lower Risks of Major Chronic Diseases and Mortality,” Nature Food 6, no. 1 (2025): 47–57, 10.1038/s43016-024-01059-x. [DOI] [PubMed] [Google Scholar]
15. Argentieri M. A., Xiao S., Bennett D., et al., “Proteomic Aging Clock Predicts Mortality and Risk of Common Age‐Related Diseases in Diverse Populations,” Nature Medicine 30, no. 9 (2024): 2450–2460, 10.1038/s41591-024-03164-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
16. Sun B. B., Chiou J., Traylor M., et al., “Plasma Proteomic Associations With Genetics and Health in the UK Biobank,” Nature 622, no. 7982 (2023): 329–338, 10.1038/s41586-023-06592-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
17. Ritchie S. C., Surendran P., Karthikeyan S., et al., “Quality Control and Removal of Technical Variation of NMR Metabolic Biomarker Data in ~120,000 UK Biobank Participants,” Scientific Data 10, no. 1 (2023): 64, 10.1038/s41597-023-01949-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Bycroft C., Freeman C., Petkova D., et al., “The UK Biobank Resource With Deep Phenotyping and Genomic Data,” Nature 562, no. 7726 (2018): 203–209, 10.1038/s41586-018-0579-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. Sudlow C., Gallacher J., Allen N., et al., “UK Biobank: An Open Access Resource for Identifying the Causes of a Wide Range of Complex Diseases of Middle and Old Age,” PLoS Medicine 12, no. 3 (2015): e1001779, 10.1371/journal.pmed.1001779. [DOI] [PMC free article] [PubMed] [Google Scholar]
20. Julkunen H., Cichońska A., Tiainen M., et al., “Atlas of Plasma NMR Biomarkers for Health and Disease in 118,461 Individuals From the UK Biobank,” Nature Communications 14, no. 1 (2023): 604, 10.1038/s41467-023-36231-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
21. Zhuo B., Ran S., Qian A. M., et al., “Air Pollution Metabolomic Signatures and Chronic Respiratory Diseases Risk: A Longitudinal Study,” Chest 166, no. 5 (2024): 975–986, 10.1016/j.chest.2024.06.3809. [DOI] [PubMed] [Google Scholar]
22. Carrasco‐Zanini J., Pietzner M., Davitte J., et al., “Proteomic Signatures Improve Risk Prediction for Common and Rare Diseases,” Nature Medicine 30, no. 9 (2024): 2489–2498, 10.1038/s41591-024-03142-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Wang M., Zhou T., Song Q., et al., “Ambient Air Pollution, Healthy Diet and Vegetable Intakes, and Mortality: A Prospective UK Biobank Study,” International Journal of Epidemiology 51, no. 4 (2022): 1243–1253, 10.1093/ije/dyac022. [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Inoue‐Choi M., Ramirez Y., Cornelis M. C., Berrington de González A., Freedman N. D., and Loftfield E., “Tea Consumption and All‐Cause and Cause‐Specific Mortality in the UK Biobank: A Prospective Cohort Study,” Annals of Internal Medicine 175, no. 9 (2022): 1201–1211, 10.7326/M22-0041. [DOI] [PMC free article] [PubMed] [Google Scholar]
25. Cao Z., Xu C., Zhang P., and Wang Y., “Associations of Sedentary Time and Physical Activity With Adverse Health Conditions: Outcome‐Wide Analyses Using Isotemporal Substitution Model,” EClinicalMedicine 48 (2022): 101424, 10.1016/j.eclinm.2022.101424. [DOI] [PMC free article] [PubMed] [Google Scholar]
26. Yu B., Li M., Yu Z., et al., “The Non‐High‐Density Lipoprotein Cholesterol to High‐Density Lipoprotein Cholesterol Ratio (NHHR) as a Predictor of All‐Cause and Cardiovascular Mortality in US Adults With Diabetes or Prediabetes: NHANES 1999‐2018,” BMC Medicine 22, no. 1 (2024): 317, 10.1186/s12916-024-03536-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
27. Benjamini Y. and Hochberg Y., “Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing,” Journal of the Royal Statistical Society, Series B: Statistical Methodology 57, no. 1 (1995): 289–300, 10.1111/j.2517-6161.1995.tb02031.x. [DOI] [Google Scholar]
28. Dou J., Guo C., Wang Y., et al., “Association Between Triglyceride Glucose‐Body Mass and One‐Year All‐Cause Mortality of Patients With Heart Failure: A Retrospective Study Utilizing the MIMIC‐IV Database,” Cardiovascular Diabetology 22, no. 1 (2023): 309, 10.1186/s12933-023-02047-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
29. Strak M., Weinmayr G., Rodopoulou S., et al., “Long Term Exposure to Low Level Air Pollution and Mortality in Eight European Cohorts Within the ELAPSE Project: Pooled Analysis,” BMJ 374 (2021): n1904, 10.1136/bmj.n1904. [DOI] [PMC free article] [PubMed] [Google Scholar]
30. van Kleef L. A., de Knegt R. J., and Brouwer W. P., “Metabolic Dysfunction‐Associated Fatty Liver Disease and Excessive Alcohol Consumption Are Both Independent Risk Factors for Mortality,” Hepatology (Baltimore, Md.) 77, no. 3 (2023): 942–948, 10.1002/hep.32642. [DOI] [PMC free article] [PubMed] [Google Scholar]
31. Lau B., Cole S. R., and Gange S. J., “Competing Risk Regression Models for Epidemiologic Data,” American Journal of Epidemiology 170, no. 2 (2009): 244–256, 10.1093/aje/kwp107. [DOI] [PMC free article] [PubMed] [Google Scholar]
32. Lian J. and Vardhanabhuti V., “Metabolic Biomarkers Using Nuclear Magnetic Resonance Metabolomics Assay for the Prediction of Aging‐Related Disease Risk and Mortality: A Prospective, Longitudinal, Observational, Cohort Study Based on the UK Biobank,” GeroScience 46, no. 2 (2024): 1515–1526, 10.1007/s11357-023-00918-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
33. Deng Y.‐T., You J., He Y., et al., “Atlas of the Plasma Proteome in Health and Disease in 53,026 Adults,” Cell 188, no. 1 (2025): 253–271.e7, 10.1016/j.cell.2024.10.045. [DOI] [PubMed] [Google Scholar]
34. Gadd D. A., Hillary R. F., Kuncheva Z., et al., “Blood Protein Assessment of Leading Incident Diseases and Mortality in the UK Biobank,” Nature Aging 4, no. 7 (2024): 939–948, 10.1038/s43587-024-00655-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
35. Sheng J. and Xu Z., “Three Decades of Research on Angiogenin: A Review and Perspective,” Acta Biochimica et Biophysica Sinica 48, no. 5 (2016): 399–410, 10.1093/abbs/gmv131. [DOI] [PMC free article] [PubMed] [Google Scholar]
36. Weng C., Dong H., Bai R., et al., “Angiogenin Promotes Angiogenesis via the Endonucleolytic Decay of miR‐141 in Colorectal Cancer,” Molecular Therapy—Nucleic Acids 27 (2022): 1010–1022, 10.1016/j.omtn.2022.01.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
37. Turgunova L., Baidildina B., Laryushina Y., Koichubekov B., Turmukhambetova A., and Akhmaltdinova L., “Association of Biomarker Level With Cardiovascular Events: Results of a 4‐Year Follow‐Up Study,” Cardiology Research and Practice 2020 (2020): 8020674, 10.1155/2020/8020674. [DOI] [PMC free article] [PubMed] [Google Scholar]
38. Korbecki J., Bajdak‐Rusinek K., Kupnicka P., et al., “The Role of CXCL16 in the Pathogenesis of Cancer and Other Diseases,” International Journal of Molecular Sciences 22, no. 7 (2021): 3490, 10.3390/ijms22073490. [DOI] [PMC free article] [PubMed] [Google Scholar]
39. Smith C., Halvorsen B., Otterdal K., et al., “High Levels and Inflammatory Effects of Soluble CXC Ligand 16 (CXCL16) in Coronary Artery Disease: Down‐Regulatory Effects of Statins,” Cardiovascular Research 79, no. 1 (2008): 195–203, 10.1093/cvr/cvn071. [DOI] [PubMed] [Google Scholar]
40. Wuttge D. M., Zhou X., Sheikine Y., et al., “CXCL16/SR‐PSOX Is an Interferon‐Gamma‐Regulated Chemokine and Scavenger Receptor Expressed in Atherosclerotic Lesions,” Arteriosclerosis, Thrombosis, and Vascular Biology 24, no. 4 (2004): 750–755, 10.1161/01.ATV.0000124102.11472.36. [DOI] [PubMed] [Google Scholar]
41. Dahl C. P., Husberg C., Gullestad L., et al., “Increased Production of CXCL16 in Experimental and Clinical Heart Failure: A Possible Role in Extracellular Matrix Remodeling,” Circulation. Heart Failure 2, no. 6 (2009): 624–632, 10.1161/CIRCHEARTFAILURE.108.821074. [DOI] [PubMed] [Google Scholar]
42. Ketchesin K. D., Stinnett G. S., and Seasholtz A. F., “Corticotropin‐Releasing Hormone‐Binding Protein and Stress: From Invertebrates to Humans,” Stress (Amsterdam, Netherlands) 20, no. 5 (2017): 449–464, 10.1080/10253890.2017.1322575. [DOI] [PMC free article] [PubMed] [Google Scholar]
43. Liu Y., Liu Z., Yang Y., Cui J., Sun J., and Liu Y., “The Prognostic and Biology of Tumour‐Infiltrating Lymphocytes in the Immunotherapy of Cancer,” British Journal of Cancer 129, no. 7 (2023): 1041–1049, 10.1038/s41416-023-02321-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
44. Yoo W., Choi H., Lee J., Lee Y., Park K. C., and Noh K., “CRHBP, a Novel Multiple Cancer Biomarker Connected With Better Prognosis and Anti‐Tumorigenicity,” Cancer Cell International 24, no. 1 (2024): 391, 10.1186/s12935-024-03562-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
45. Xu X., Tahan S. R., Pasha T. L., and Zhang P. J., “Expression of Neurotrophin Receptor Trk‐C in Nevi and Melanomas,” Journal of Cutaneous Pathology 30, no. 5 (2003): 318–322, 10.1034/j.1600-0560.2003.00068.x. [DOI] [PubMed] [Google Scholar]
46. Bouzas‐Rodriguez J., Cabrera J. R., Delloye‐Bourgeois C., et al., “Neurotrophin‐3 Production Promotes Human Neuroblastoma Cell Survival by Inhibiting TrkC‐Induced Apoptosis,” Journal of Clinical Investigation 120, no. 3 (2010): 850–858, 10.1172/JCI41013. [DOI] [PMC free article] [PubMed] [Google Scholar]
47. Kim J. Y., Sutton M. E., Lu D. J., et al., “Activation of Neurotrophin‐3 Receptor TrkC Induces Apoptosis in Medulloblastomas,” Cancer Research 59, no. 3 (1999): 711–719. [PubMed] [Google Scholar]
48. Noga O., Englmann C., Hanf G., Grützkau A., Guhl S., and Kunkel G., “Activation of the Specific Neurotrophin Receptors TrkA, TrkB and TrkC Influences the Function of Eosinophils,” Clinical and Experimental Allergy: Journal of the British Society for Allergy and Clinical Immunology 32, no. 9 (2002): 1348–1354, 10.1046/j.1365-2745.2002.01442.x. [DOI] [PubMed] [Google Scholar]
49. Fox A. J., Patel H. J., Barnes P. J., and Belvisi M. G., “Release of Nerve Growth Factor by Human Pulmonary Epithelial Cells: Role in Airway Inflammatory Diseases,” European Journal of Pharmacology 424, no. 2 (2001): 159–162, 10.1016/s0014-2999(01)01138-4. [DOI] [PubMed] [Google Scholar]
50. Ogura K., Somwar R., Hmeljak J., et al., “Therapeutic Potential of NTRK3 Inhibition in Desmoplastic Small Round Cell Tumor,” Clinical Cancer Research: An Official Journal of the American Association for Cancer Research 27, no. 4 (2021): 1184–1194, 10.1158/1078-0432.CCR-20-2585. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials