Inflammation and Perceived Social Support in Elite Athletes (A Quantile Regression Approach)

Abstract

Introduction

Inflammation and social relationships are bidirectionally linked, yet evidence in young, nonclinical populations is scarce. Given that elite athletes face continuous immune challenges and unique social conditions, this cohort provides a model to explore this association.

Methods

Several quantile regressions were computed across 422 elite athletes. To account for both the magnitude of inflammation and the characteristics of social relationships, quantiles were determined based on the concentrations of tumor necrosis factor alpha, interleukin-6, and the systemic inflammatory response index (SIRI) as well as on measures of the total, general perceived social support. The corresponding predictors were social support and the inflammatory markers.

Results

A higher total, general perceived social support predicted a significantly lower SIRI in the lowest quartile as well as lower concentrations of the cytokines in the highest three quartiles. Effects were small but robust. Moreover, higher inflammation predicted a lower total, general perceived social support in elite athletes, when the perceived social support was relatively moderate. However, this effect was not robust when covariates, such as urea or the living situation, were added to the models.

Conclusion

Results suggest that elite athletes’ social relationships are a small yet important factor influencing the inflammatory response. The social support elite athletes perceive, however, appears to be influenced by inflammation only under specific biopsychosocial conditions.

Introduction

Investing the link between elite athletes’ inflammation and their social relationships is well grounded within the framework of psychoneuroimmunology (PNI). PNI posits a bidirectional communication between the immune system and the brain mediated by cytokines, immune cells, hormones, neurotransmitters, and neuropeptides [1]. Moreover, PNI provides an insight on how psychosocial states and conditions affect a persons’ inflammatory status, and – the other way round – how the immune system influences thinking, feeling, and behavior [1].

Among the investigated psychosocial states and conditions in the context of inflammation is the concept of social relationships. Social relationships are commonly found to be linked to health and according to current PNI research, inflammatory processes are one of the proposed mediating mechanisms [2–4]. Generally, social relationships can be operationalized via structural (e.g., social isolation, living situation), functional (e.g., received/perceived social support), and qualitative aspects (e.g., relationship satisfaction/strain) as well as via assessing the subjective feeling of unmet social needs (loneliness; see Holt-Lunstad [4] for a conceptual overview).

Whilst recent theoretical work recommends a broad operationalization by integrating multiple indicators for social relationships [4], PNI research yet focused on single aspects. For instance, higher social isolation was previously found to be associated with higher levels of the cytokine interleukin 6 (IL-6) and the acute-phase reactant C-reactive protein (CRP) [5, 6]. Better social integration, correspondingly, related to lower concentrations of IL-6, CRP, and the cytokine tumor necrosis factor alpha (TNF-α) [7]. In terms of function, social relationships were reported to associate with these inflammatory parameters in the sense that higher social support related to lower concentrations of IL-6, TNF-α, and CRP [7–9].

Alongside evidence that social relationships can attenuate low-grade inflammation, research has also shown that inflammatory processes shape thinking, feeling, and behaving in social contexts [1, 10, 11]. For instance, social withdrawal or approaching others has been observed in animals and humans under acute inflammatory conditions [10, 12]. Aside from acute inflammation, a withdrawal from peripheral social roles but a maintenance of social contacts in association to IL-6 levels was also observed in the field [13]. Generally, the actual social behavior shown is supposed to vary with the closeness of the other individuals [12–14]. Moreover, gender and the magnitude of the inflammatory response are further aspects discussed to influence the impact acute inflammation has on thinking, feeling, and behavior [12, 15].

Investigating the bidirectional link between inflammation and social relationships in elite athletes is particularly relevant for two main reasons: first, elite athletes’ immune systems are under constant challenge: elite athletes face various sport-related immune stressors with exercise itself being one prominent example [16]. To contribute to recovery and the restoration of homeostasis, the immune system reacts dynamically to each acute bout – ergo sometimes multiple times in an elite athletes’ day [17–19]. Therefore, inflammation of low to high magnitude and especially the range of low-grade inflammation are naturally observable in the sample of elite athletes. Second, social relationships are central in the context of elite sport. Regarding the functional aspect, it is known that the elite athletes’ performance development and their mental health profit from social support [20–22]. Structurally, the environment embeds elite athletes in relationships with various people inside and outside of sport [23]. Nevertheless, at the same time sufficient support from a large and/or high-quality social network cannot be taken for granted in competitive sport: elite sport brings along sociocultural, institutional, inter- and intrapersonal factors, which might enhance the risk for social isolation or loneliness [24]. Examples for these factors include struggling with sport-related norms and ideologies, feeling left behind or misunderstood by the sport institutions and/or close others, and living on one’s own, respectively [24].

This variation in both social relationships and the inflammatory status among elite athletes provides a promising starting point for empirical investigation to gain a more profound understanding of how the association of these aspects varies depending on their respective degrees of manifestation. Yet to date, there are very few studies in which the methodology and statistical analyses allow for conclusions of this kind [25].

There is, moreover, a research gap regarding some inflammatory parameters. For instance, typically pro-inflammatory cytokines and acute-phase reactants, such as IL-6 or CRP, were used as indicators for both acute and low-grade inflammation [9, 13, 26, 27]. Anti-inflammatory cytokines, which play a key role in attenuating inflammatory processes, have been examined far less frequently, and mainly in older studies (for an overview, see Pourriyahi et al. [3] or Unchino et al. [7]). Work investigating the cellular pathway in the context of social relationships is also uncommon. Findings from recent, original articles [28, 29] suggest that less close relationships are associated with a lower functional capacity of the adaptive immune system (for an overview of earlier studies, see Pourriyahi et al. [3]). Notably, both studies were set in the context of aging and thus comprised samples of individuals ranging from middle to late adulthood. Therefore, there remains a gap in research on the association of immune cells and social relationships in a young sample, such as elite athletes.

Given the importance of both inflammation and social relationships for elite athletes and the additional value such sample provides for research, this cross-sectional study is dedicated to investigating the bidirectional link between inflammation and several aspects of social relationships in elite athletes. Through well-considered statistical computations, this study also aims to clarify the directions and magnitudes of effects. Against the background of previous research, it is expected that social relationships negatively predict inflammation. Furthermore, drawing on social behavior during infection [10, 12], it is assumed that these effects may become amplified once inflammatory activity surpasses a certain threshold. Similarly, in line with findings on social isolation and loneliness [3, 6, 25], stronger effects are expected in individuals with poorer social relationships. Overall, effect sizes are anticipated to be small, as immune pathways represent only one of several mechanisms linking health and behavior [2–4].

Methods

Present cross-sectional study was realized within the collaborative, multidisciplinary research project “Individualized performance development in elite sports through holistic and transdisciplinary process optimization” (project acronym: “in:prove”) funded by the German Federal Institute of Sport Science. It refers to demographic, psychosocial, and physiological data collected from February 2022 to February 2025. The study protocol is in accordance with the Declaration of Helsinki, and the university Ethics Committee approved the study (number: AZ 55/22). Each athlete or their legal guardian received written and oral information about data acquisition and privacy policies prior to the study and submitted signed informed consent. It was possible to consent to psychosocial measurements only, physiological measurements only, or both. The acquisition of data was executed during selected training camps of the respective junior and senior national teams. On measurement day, an interdisciplinary research team collected psychosocial and physiological data in permuted order according to the submitted consent. Demographic information (gender, age, athletic discipline) was provided to the research team beforehand.

Participants

Data were available for 586 elite athletes in total. As there occasionally were repeated measurements for some athletes across the 3 years of data collection, generally the first time point with complete data was considered in this cross-sectional study. Cases were excluded, if either only psychosocial and physiological data were available (n = 143) or if the acquisition of psychosocial data differed more than 180 days from blood collection (n = 1). Documented infection or injury was no reason for exclusion. Consequently, the final sample included in the analyses comprised 442 elite athletes (M_age = 18.29 ± 4.15; 196 female). In this, sample psychosocial and physiological data were collected within a mean timeframe of 3.48 days (SD = 16.61, Mdn = 1.00). All athletes were part of the junior or senior national squad from the disciplines artistic gymnastics (n = 31, all female), 3 × 3 basketball (n = 42, 23 female), ice hockey (n = 116, 19 female), modern pentathlon (n = 37, 18 female), rhythmic gymnastics (n = 26, all female), table tennis (n = 24, 14 female), trampoline gymnastics (n = 29, 19 female), or volleyball (n = 137, 46 female). A dichotomous variable (sport type) was defined to increase group size for following analyses, subsuming athletes from the disciplines 3 × 3 basketball, ice hockey, table tennis, and volleyball to “game sport” and those from the disciplines artistic gymnastics, rhythmic gymnastics, trampoline gymnastics and modern pentathlon to “individual sports.”

Psychosocial Parameters

To collect the psychosocial parameters, athletes filled out a questionnaire either online or by paper and pencil. Among the measures included in the questionnaire, there were the Multidimensional Scale of Perceived Social Support (MSPSS) [30] and the Perceived Available Support in Sport Questionnaire (PASS-Q) [31] to assess a functional aspect of social relationships. The MSPSS focusses on the source of social support and inquires with four items each the general perceived support by family, friends, and “one other significant person.” Scores for each potential source of support and a total social support score were calculated by averaging the four and 12 items, respectively. The PASS-Q focusses on the tangible, informational, emotional, and esteem-related dimension of perceived social support in sports which are inquired with four items per dimension. Scores for each dimension and a total social support score were calculated by averaging the 4 and 16 items, respectively. Both scales are rated by the athletes on a five-point Likert scale (1: strongly disagree – 5: strongly agree), and higher scores generally indicate higher perceived social support.

To assess a structural precondition of social relationships, athletes were asked to indicate their living situation. The response options were: (a) living on their own, (b) living together with their partner, (c) living in a flat-sharing community/boarding home, or (d) living together with their parents/family.

To account for possible confounders, stress was measured by the Perceived Stress Scale 4 (PSS-4) [32]. The scale inquires the athletes’ subjective experience of stress over the last month by four items, which are rated on a five-point Likert-Scale (1: almost never – 5: almost always). A total score was calculated by averaging the four items, and a higher score indicates a higher subjective experience of stress. The Patient Health Questionnaire-4 (PHQ-4) [33] was also included to the analysis. The PHQ-4 is a brief screening instrument, measuring the frequency of anxiety (two items) and depressive symptoms (two items) over the last 4 weeks. Items are rated on a four-point Likert-Scale (1: never – 4: nearly every day). By averaging the respective items, total scores were calculated with higher scores indicating more frequent symptoms of anxiety and depressive mood, respectively.

Physiological Parameters

Blood samples were taken by a medical professional usually before noon. Generally, athletes came in a fed state and followed their usual, individual diet prior to and at the measurement day. Between blood collection and the last bout of exercise, there usually was a minimum time gap of 10 h.

From venous blood (about 25 mL), different immunological and physiological parameters were extracted. A complete blood cell count (in 10⁹/L) was conducted from EDTA blood using a cell counter (Sysmex Hematology Analyzer, Norderstedt, Germany). From serum, the concentrations (each in pg/mL) of the interleukins (IL) IL-1β, IL-6, and IL-10, interferon gamma, and tumor necrosis factor alpha (TNF-α) as well as the hormones insulin and leptin were assessed via the Luminex (LX)-200 instrument from the Luminex^® Multiplex Assay Kit (Bio-Techne, Minneapolis, USA). The concentration of the CRP (in mg/dL) was evaluated using immunoturbidimetric analysis. The hormone level of free Tri-iodothyronine (fT3; pmol/L) and the concentrations of ferritin, vitamin B12, vitamin B9, and 25-OH-vitamin D (each in ng/mL) as well as creatinine kinase (U/L) were analyzed via a chemiluminescent immunoassay. Using photometric methods, the concentration of creatinine (in mg/dL) were analyzed. Serum urea concentrations were determined enzymatically using the urease-glutamate dehydrogenase (GLDH) method on an automated clinical chemistry analyzer (Roche Cobas).

To represent the balance of the inflammatory condition, the systemic inflammation index (SII), the systemic inflammatory response index (SIRI), and the ratio TNF-α:IL-10 were calculated. The formulas used were SII = (absolute neutrophil count) × (absolute platelet count)/(absolute lymphocyte count), and SIRI = (absolute neutrophil count) × (absolute monocyte count)/(absolute lymphocyte count). In these calculated ratios, higher values indicate a more pronounced, and lower ratios a more balanced inflammatory condition.

Statistical Analysis

All statistical analyses were performed using IBM SPSS Statistics for Windows Version 29.0.2.0 (IBM Corp., Armonk, NY, USA). Descriptive statistics (mean, standard deviation, minimum, maximum, skewness) were analyzed for each variable, and metric variables were tested for normal distribution (Shapiro-Wilk test). To gain a first insight to the association of the inflammatory markers with the general and sport-related perceived social support as well as to the living situation, nonparametric correlations (Spearman’s rho) or nonparametric group comparisons (Kruskal-Wallis tests; levels of significance were adjusted according to Bonferroni), respectively, were computed. To identify possible third variables data-driven, demographic (age, sport type), physiological (creatinine, urea, vitamin B9, vitamin B12, vitamin D, creatine kinase, fT3, insulin, and leptin) and psychosocial variables (perceived stress, frequency of anxiety, and depressive symptoms) were correlated non-parametrically (Spearman’s rho) with all inflammatory markers and the measures of social support. Moreover, all variables were analyzed for differences regarding gender and the type of sport (Mann-Whitney-U tests; levels of significance were adjusted according to Bonferroni). In all calculations, the levels of significance were set to α ≤ 0.05 (two-sided).

Subsequently, quantile regressions were computed to scrutinize the emerged associations and to check them for robustness. According to the expectation that effects might vary with the degree of inflammation and the social relationships, models were computed once setting indicators for social relationships as the dependent variable and once setting inflammatory markers as the dependent variable. A parameter was considered relevant for quantile regression, if α ≤ 0.001 in previous nonparametric analyses (see results Table 1). Consequently, the SIRI, IL-6, as well as TNF-α (dependent variables) were predicted by the total perceived social support measured by the MSPSS (predictor) and vice versa. A variable was regarded as covariate, either when the level of significance of its correlation to both SIRI and the MSPSS total score was α ≤ 0.05 (see results Table 2), or when level of significance of its group comparisons was α ≤ 0.05 (see results Table 3). First, one model containing only the predictor was calculated, before covariates were added in further regression models. These were executed separately for reasons of model specification (e.g., multicollinearity). Among the covariates were the type of sport, creatinine and urea (analyses having the SIRI as in-/dependent variable) and the frequency of anxiety symptoms (analyses having TNF-α as in-/dependent variable). Moreover, gender was regarded as covariate in all analyses, and when it resulted to have a significant (p ≤ 0.05) impact on the dependent variable in any quartile, the computations were repeated gender-specifically to further investigate possible gender differences. Lastly, models were computed that implemented the living situation as a factor besides the predictor. This was done according to recent recommendations to broadly operationalize social relationships and to combine functional with structural aspects [4]. In all regression models, the parameters were estimated for the quantiles τ = 0.25, τ = 0.50, τ = 0.75, and τ = 0.99 using the simplex-algorithm as estimation method. The statistical method of quantile regression was selected because the estimation of the outcome variable is modelled quantile-specific by median regression, which makes it robust against outliers [34]. Through the quantile-specific estimations, this statistical method is not only beneficial when data are skewed, but also when a distribution’s upper or lower tails are of particular interest [34], as it is the case in this study.

Table 1.

Nonparametric correlations between inflammatory markers and the total perceived social support or its subdimensions

​	SII	SIRI	CRP	IFN-γ	IL-10	IL-1β	IL-6	TNF-α	Ratio TNF-α:IL-10
Family	−0.076	−0.118*	−0.025	−0.012	−0.130*	−0.076	−0.116*	−0.096	0.061
Friends	−0.071	−0.054	−0.034	−0.043	−0.055	−0.086	−0.126*	−0.092	−0.040
Sig. Person	−0.069	−0.136**	−0.083	−0.123*	−0.103	−0.113*	−0.116*	−0.125*	−0.025
Total (MSPSS)	−0.120*	−0.127**	−0.074	−0.082	−0.142*	−0.125*	−0.178**	−0.159**	0.010
Emotional	0.013	0.000	−0.086	−0.035	−0.097	−0.019	−0.051	−0.022	0.072
Esteem	0.047	0.020	−0.073	−0.025	−0.110	−0.043	−0.008	−0.023	0.120*
Informational	0.019	−0.006	−0.068	−0.035	−0.132*	−0.056	−0.009	−0.002	0.126*
Tangible	−0.063	−0.112*	−0.120*	−0.051	−0.082	−0.020	0.006	0.053	0.124*
Total (PASS-Q)	0.002	−0.036	−0.102	−0.042	−0.117*	−0.039	−0.027	−0.002	0.119*

Significance for α ≤ 0.05 and α ≤ 0.001 are marked by * and **, respectively.

MSPSS, Multidimensional Scale of Perceived Social Support; Sig. Person, one other significant person; PASS-Q, Perceived Available Support in Sport Questionnaire; SII, systemic inflammation index; SIRI, systemic inflammatory response index.

Table 2.

Nonparametric correlations (Spearman‘s Rho) between inflammatory markers and social support and psychosocial, physiological, and demographic variables

​	SII	SIRI	CRP	IFN-γ	IL-10	IL-1β	IL-6	TNF-α	Ratio TNF-α:IL-10	Family	Friends	Significant person	Total (MSPSS)	Emotional	Esteem	Informational	Tangible	Total (PASS-Q)
Perc. Stress	0.058	0.003	−0.106	0.019	0.105	0.132*	−0.016	0.039	−0.133*	−0.301**	−0.112*	−0.145**	−0.245**	−0.155**	−0.245**	−0.154**	−0.101*	−0.179**
Anx. Sympt.	0.085	0.062	0.003	−0.059	0.132*	0.171**	0.051	0.117*	−0.094	−0.219**	−0.027	−0.058	−0.157**	−0.056	−0.170**	−0.164**	−0.154**	−0.154**
Depr. Sympt.	0.028	0.023	−0.127*	−0.001	0.160**	0.141*	0.025	0.092	−0.138*	−0.298**	−0.080	−0.129**	−0.213**	−0.152**	−0.230**	−0.211**	−0.173**	−0.219**
Creatinine	0.137**	0.173**	0.155**	0.035	−0.031	−0.016	0.093	0.081	0.162**	−0.127*	−0.132*	−0.096	−0.124*	−0.202**	−0.086	−0.024	−0.169**	−0.137**
Urea	0.166**	0.149**	0.004	0.035	0.018	0.016	0.079	0.086	0.067	−0.083	−0.139**	−0.134*	−0.152**	−0.124*	−0.066	−0.047	−0.146**	−0.117*
Vitamin B9	−0.102*	−0.093	−0.025	−0.043	−0.034	−0.103	−0.068	−0.020	0.014	0.100*	0.030	0.094*	0.098*	−0.044	−0.050	0.011	0.017	−0.021
Vitamin D	−0.051	−0.085	0.062	0.033	−0.084	−0.069	−0.002	−0.031	0.107	0.017	−0.010	0.043	0.021	−0.048	−0.013	−0.078	−0.089	−0.066
Vitamin B12	0.032	−0.009	0.122*	0.045	−0.027	−0.055	0.083	0.030	0.077	0.001	−0.115*	−0.014	−0.085	0.018	0.046	0.080	0.038	0.053
CK	−0.060	−0.014	0.126*	0.001	−0.121*	−0.203**	−0.089	−0.097	0.099	0.080	0.001	0.028	0.065	−0.070	0.040	0.103*	−0.046	0.003
fT3	0.079	0.066	−0.213**	0.035	0.033	0.112*	0.178**	0.144**	0.074	0.013	−0.143**	0.043	−0.070	0.040	0.037	0.063	0.055	0.061
Insulin	0.142*	0.153**	−0.161**	−0.003	0.064	0.103	0.168**	0.143*	0.059	−0.075	−0.080	−0.076	−0.102	0.050	0.026	0.020	−0.006	0.024
Leptin	0.139**	0.081	−0.141*	0.081	0.187**	0.210**	0.141*	0.129*	−0.161**	−0.035	0.119*	−0.042	0.030	0.094	0.042	−0.063	0.038	0.046
Age	0.001	−0.033	0.130*	0.052	0.054	−0.083	−0.142*	−0.078	−0.106	−0.149**	0.076	−0.128**	−0.032	−0.125**	−0.138**	−0.192**	−0.232**	−0.201**

Significance for α ≤ 0.05 and α ≤ 0.001 are marked by * and **, respectively.

Anx. Sympt., frequency of anxiety symptoms; CK, creatine Kinase; Depr. Sympt., frequency of depressive symptoms; MSPSS, Multidimensional Scale of Perceived Social Support; PASS-Q, Perceived Available Support in Sport Questionnaire; Perc. Stress, perceived stress; SII, systemic inflammation index; SIRI, systemic inflammatory response index.

Table 3.

Results for the nonparametric group comparisons (Kruskal-Wallis- and Mann-Whitney-U test)

​	Living situation		Gender		Type of sports
	H (df = 3)	p value	z	p value	z	p value
DV: perceived social support
Family	1.98	0.576	0.92	0.360	−0.89	0.372
Friends	0.39	0.942	4.02	<0.001**	−1.96	0.050
Sig. Person	9.18	0.027	1.89	0.059	−0.49	0.625
Total (MSPSS)	1.66	0.646	2.86	0.004*	−1.24	0.216
Emotional	3.39	0.336	3.25	0.001*	−1.47	0.140
Esteem	0.85	0.837	1.93	0.054	−1.62	0.106
Informational	3.21	0.360	0.39	0.690	−1.50	0.133
Tangible	4.23	0.238	3.48	<0.001**	−3.70	<0.001**
Total (PASS-Q)	3.80	0.284	2.80	0.005*	−2.45	0.014
DV: inflammation
Leucocytes	12.70	0.005	−2.22	0.027	1.59	0.111
Erythrocytes	4.85	0.183	−12.58	<0.001**	3.83	<0.001**
Thrombocytes	9.73	0.021	4.26	<0.001**	−4.37	<0.001**
Neutrophils	4.52	0.211	−2.90	0.004	3.38	<0.001**
Lymphocytes	23.45	<0.001**	2.39	0.017*	−3.19	0.001*
Monocytes	22.11	<0.001**	−2.51	0.012*	−0.24	0.981
SII	0.31	0.958	−1.77	0.076	2.37	0.018*
SIRI	6.01	0.111	−4.21	<0.001**	3.84	<0.001**
IL-1β in pg/mL	2.51	0.473	1.70	0.089	1.33	0.183
IL-6 in pg/mL	3.39	0.336	−1.14	0.255	1.81	0.071
IL-10 in pg/mL	2.95	0.399	1.43	0.154	0.03	0.973
IFN-γ in pg/mL	3.28	0.350	−1.48	0.138	0.11	0.910
TNF-α in pg/mL	1.83	0.608	−0.79	0.428	0.04	0.686
Ratio TNF-α:IL-10	1.71	0.634	−3.54	<0.001**	0.82	0.414
CRP in mg/dL	0.73	0.866	−4.92	<0.001**	3.79	<0.001**

Significant values are marked by * and ** and highlighted by bold print. According to the Bonferroni-adjustment, levels of significance were α ≤ 0.006 and α ≤ 0.0001 (DV: social support) and α ≤ 0.0029 and α ≤ 0.0001 (DV: inflammation), respectively.

MSPSS, Multidimensional Scale of Perceived Social Support; Sig. Person, one other significant person; PASS-Q, Perceived Available Support in Sport Questionnaire; SII, systemic inflammation index; SIRI, systemic inflammatory response index; IFN-γ, interferon gamma.

Results

Generally, no variable followed normal distribution (p ≤ 0.001) except for the erythrocytes (Shapiro-Wilk: W(442) = 0.996, p = 0.299) and fT3 (W(441) = 0.995, p = 0.124). The immunological and physiological variables as well as the perceived stress, the frequency of anxiety and depressive symptoms had a positive (right) skew. The variables of perceived social support and their subdimensions were negatively (left) skewed, with medians reaching from 4.00 (PASS-Q total score and its subdimensions) to 5.00 (MSPSS subdimension: significant person). Details about the descriptive statistics for each variable and for the tests on normal distribution can be found in online supplementary Table A (for all online suppl. material, see https://doi.org/10.1159/000550042).

Correlations and Group Comparisons

Results for the nonparametric correlations between inflammatory markers and social support can be found in Table 1. Notably, for the associations of the SIRI, IL-6, and TNF-α to the total, general perceived social support (MSPSS), the p values lay below 0.001. Effect sizes were small and except for the ratio of TNF-α:IL-10 all effects had the expected direction.

Small effect sizes were also visible for correlations between the perceived social support or inflammation and some other psychosocial, physiological, and demographic variables (see Table 2). Among the variables correlating significantly to both the total perceived social support (MSPSS) and the associated inflammatory parameters (the SIRI, IL-6, TNF-α) were the frequency of anxiety symptoms, creatinine, urea, and age.

For some aspects of the perceived social support and some inflammatory markers, there were significant small differences between female and male athletes as well as between athletes from individual and game sport (see Table 3). For instance, female athletes reported significantly higher perceived social support in total (MSPSS: Mdn_female = 4.67; Mdn_male = 4.58; PASS-Q: Mdn_female = 4.13; Mdn_male = 3.94) and from friends (Mdn_female = 4.5; Mdn_male = 4.25) than male athletes. Male athletes had a higher ratio TNF-α:IL-10 (Mdn_female = 2.39; Mdn_male = 2.97) and SIRI (Mdn_female = 0.79; Mdn_male = 0.99) than female athletes, amongst others. Moreover, the SIRI (Mdn_individual = 0.79; Mdn_game = 0.93) was significantly higher for athletes from game sports.

Regarding the living situation, the lymphocyte and monocyte count, but not the SIRI differed significantly between the groups (see Table 3). Details for the post hoc pair-wise group comparisons are provided in online supplementary Table B.

Quantile Regression

To further scrutinize these identified relationships between social support and inflammation, quantile regressions were calculated and checked for robustness. Figure 1a–c depicts the results of the quantile regressions, where the quantiles were set according to inflammation indicated by the SIRI, IL-6, and TNF-α. In more detail, a higher general social support significantly predicted a lower SIRI in the lowest quartile. This indicates that social support is negatively associated to systemic inflammatory response in the group of athletes with balanced inflammatory response (green). This effect or its trend was robustly found across almost all models containing the covariates. Details for these models, such as the Pseudo-R² or the main effect of the covariates on the dependent variable, are provided in online supplementary Table C. Gender-specific models revealed that among female athletes, the SIRI was significantly predicted by the general social support in the lowest two quartiles (b_Q1 = −0.21, T(194) = −3.69, p < 0.001, Pseudo-R² = 0.030, MAE = 0.49; b_Q2 = −0.22, T(194) = −2.43, p = 0.016, Pseudo-R² = 0.022, MAE = 0.44). In male athletes, the general social support predicted the SIRI in the highest quartile (b_Q4 = −7.45, T(244) = −2.21, p = 0.028, Pseudo-R² = 0.125, MAE = 7.24).

Fig. 1.

Quantile-specific results inclusive confidence intervals for the prediction of the SIRI (a), the concentrations of IL-6 (b), and the concentrations of TNF-α through the perceived social support (c). The scatter plots include quantile-specific regression lines. IL-6, interleukin 6; MAE, mean absolute error; Q1, quantile one; Q2, quantile two; Q3, quantile three; Q4, quantile four; SIRI, systemic inflammatory response index; TNF-α, tumor necrosis factor alpha.

Regarding IL-6 and TNF-α, a higher general, total perceived social support (MSPSS) significantly predicted lower concentrations of these cytokines across all quartiles (blue, purple, red) except for the lowest quartile (green). Notably, the regression coefficients and the Pseudo-R² augmented across quartiles (see Fig. 1b, c), indicating that the strength of the negative association and the explanation of variance increased with rising cytokine levels. When covariates were considered in the model, the significance of the effect of IL-6 and TNF-α mostly only remained for the upper two and the highest quartile, respectively (see online suppl. Table B). While the frequency of anxiety symptoms did not predict the concentrations of TNF-α in any quartile, leptin, the type of sport, and gender contributed significantly to the explanation of variance of both IL-6 and TNF-α in the highest quartile. When the models were calculated per gender, results regarding IL-6 were found to be similar in female athletes to those across all athletes (b_Q2 = −1.41, T(153) = −2.29, p = 0.023, Pseudo-R² = 0.016, MAE = 4.69; b_Q3 = - 5.67, T(153) = −3.25, p = 0.001, Pseudo-R² = 0.051, MAE = 5.58; b_Q4 = −47.82, T(153) = −59.47, p < 0.001, Pseudo-R² could not be estimated, MAE = 30.03). Regarding TNF-α, gender-specific calculations found that in female athletes the general social support predicted the cytokine concentrations significantly in the second and the fourth quartile (b_Q2 = −1.34, T(153) = - 2.70, p = 0.008, Pseudo-R² = 0.010; MAE = 5.29; b_Q4 = −46.63, T(153) = −31.72, p < 0.001, Pseudo-R² could not be estimated, MAE = 43.07). In male athletes, the general social support (MSPSS) did neither predict IL-6 nor TNF-α.

In Figure 2a–c, the results of the regressions with quantiles set according to the general, total perceived social support (MSPSS) are depicted. In more detail, the SIRI resulted as significant predictor in the quartiles of moderate general social support but not in the lowest and highest quartiles. When the models containing the covariates were computed, the effect of the SIRI on the general social support was no longer observable. Details for all models, such as the Pseudo-R² or the main effect of the covariates on the dependent variable, are provided in online supplementary Table D. Gender-specific computations of the quantile regressions showed no relation between the SIRI and the general social support in the lowest three quartiles for both male and female. Regression analysis in the highest quartile led to an infinite solution in female athletes and to a significant prediction in men, but the estimate’s coefficient and Pseudo-R² were zero (b_Q1 = 0.00, T(244) = −19.05, p < 0.001, Pseudo-R² = 0.000, MAE = 0.56).

Fig. 2.

Quantile-specific results inclusive confidence intervals for the prediction of the perceived social support through the SIRI (a), the concentrations of IL-6 (b), and the concentrations of TNF-α (c). The scatter plots include quantile-specific regression lines. IL-6, interleukin 6; MAE, mean absolute error; Q1, quantile one; Q2, quantile two; Q3, quantile three; Q4, quantile four; SIRI, systemic inflammatory response index; TNF-α, tumor necrosis factor alpha.

Regarding the prediction of the general, total perceived social support (MSPSS) through the concentrations of the cytokine IL-6, a significant, negative relationship was found for the three lowest quartiles (red, purple, blue in Fig. 2b). This indicates that when social support is perceived as relatively low or moderate, higher concentrations of IL-6 were related to lower levels of social support and vice versa. This effect was robust for the third quartile (for details, see online suppl. Table D). According to gender-specific calculations a higher IL-6 related to lower general social support in both male (b_Q3 = −0.02, T(153) = −7.00, p < 0.001, Pseudo-R² = 0.091, MAE = 0.43) and female athletes (b_Q3 = −0.02, T(158) = −5.04, p < 0.001, Pseudo-R² = 0.039, MAE = 0.38).

When TNF-α was the predictor for the general, total perceived social support (MSPSS), very small, significant effects were observable for all quartiles indicating that higher levels of the cytokine related to lower general social support and vice versa (see Fig. 2c). Yet, these effects did not persist in the models containing the covariates (for details, see online suppl. Table D). The gender-specific quantile regressions revealed that in male athletes the concentrations of TNF-α significantly predicted the general social support in the highest three quartiles (b_Q2 = −0.01, T(153) = −3.69, p < 0.001, Pseudo-R² = 0.069, MAE = 0.38; b_Q3 = −0.01, T(153) = −6.86, p < 0.001, Pseudo-R² = 0.065, MAE = 0.43; b_Q4 = −0.01, T(153) = −9.92, p < 0.001, Pseudo-R² could not be calculated, MAE = 0.58), whereas in female athletes this was solely the case for the third and fourth quartile (b_Q3 = −0.01, T(161) = −2.05, p = 0.042, Pseudo-R² = 0.007, MAE = 0.40; b_Q4 = 0.00, T(161) = 18.50, p < 0.001, Pseudo-R² could not be calculated, MAE = 0.43).

Discussion

This cross-sectional study examined the bidirectional link between inflammation and social relationships in a sample of elite athletes. Quantile regression analyses, which allow to provide a fine-grained picture of an association in relation to the degree of the dependent variable, revealed a differentiated pattern: in one direction the total, general perceived social support was a robust predictor of a lower SIRI in the lowest quartile, of lower concentrations of TNF-α in the highest three, and of IL-6 in all quartiles. In the other direction, the inflammatory markers predicted a slightly lower perception of the total, general social support when social support was relatively moderate (both middle quartiles). Here, only the effect of IL-6 in the third quartile was found to be robust. While in the former quantile regressions the results were mainly driven by elite athletes of female gender, effects in the latter quantile regressions did not differ between genders.

The present observation of social relationships predicting inflammation is largely in line with and extends previous research on several points: first, like the results presented here, there is evidence for small but significant effects of social support [7], social isolation, or loneliness [5, 6] on low-grade inflammation reflected by the concentrations of cytokines, such as IL-6 and TNF-α in the general population. When discussing these results, it is important to distinguish between physiological low-grade inflammation, which reflects subtle immune activation linked to behavioral and metabolic processes, and infection-associated high-grade inflammation, characterized by strong acute-phase responses. The current findings relate to the former. An expected yet novel insight from this study is that the beneficial effect of social support on IL-6 and TNF-α was more pronounced at higher circulating cytokine levels; a finding that could be revealed using the quantile regressions. This specific link may reflect a buffering role of social support during phases of acute inflammation, when pro-inflammatory cytokine concentrations are typically elevated [1]. For elite athletes, whose immune systems regularly respond to immune challenges such as acute bouts of exercise, a lower cytokine release may indicate a faster recovery and an earlier readiness for subsequent high-intensity exercise. As this in turn is intertwined with performance [35], the cytokines thus could be one part of the explanation of how the elsewhere reported advantage of social support for athletic performance (for an overview see, e.g., Freeman [36]) is mediated at the physiological level. Interestingly, no robust effects of social relationships on the acute-phase reactant CRP were found in this or in other studies with young samples [5, 9, 37, 38]. This may relate to the more indirect communication of CRP with the brain compared to cytokines, which act via humoral, neural, or cellular pathways [1]. Such indirect mechanisms might be too rigid to rapidly reflect social influences.

The second way in which this study extends current knowledge is by considering other inflammatory markers yet understudied: anti-inflammatory cytokines have been little researched in the context of social relationships to date, and these few available findings additionally suggest an uncorrelatedness of IL-10 [39, 40] or the ratio TNF-α:IL-10 [41] with aspects of social relationships. Remarkably, the present study revealed significant correlations of social support and IL-10 as well as with the ratio TNF-α:IL-10 in elite athletes. As IL-10 typically rises after acute exercise as a counter-regulator of inflammation [18, 19], this anti-inflammatory cytokine may reflect how exercise strain moderates or mediates this association in elite athletes. Also, psycho-neuroinflammatory changes concomitant to depression [42, 43] could possibly explain the observed association. Despite the novelty of the correlative patterns of the present study, these indeed were not scrutinized because their level of significance did not exceed the criterium of α = 0.01 set a priori for the deeper quantile regression analysis. Moreover, the present study is the first to transfer the systemic inflammatory index (SII) and the systemic inflammatory response index (SIRI) from the medical [44, 45] to the psychosocial context in elite sports. These two blood cell-based indices are composed by the ratio of three immune cells and therefore more comprehensively reflect the systemic inflammatory status than single-cell parameters [44]. The finding that both indices correlated with the total, general perceived social support in elite athletes is a further indication of lower social support relating to an imbalance toward innate immunity [28]. Such imbalance occurs when the circulating levels of innate immune cells, such as neutrophils, are elevated and/or when the lymphocyte counts are low. This pattern of systemic inflammation can be observed in initial stages of post-exercise recovery processes, among others [17, 45]. Interestingly, this study’s quantile regressions suggest that social support may contribute to maintaining the dynamic equilibrium between pro- and anti-inflammatory processes under conditions of low-grade immune activation. For elite athletes, restoring such equilibrium is critical to maintaining physiological capacity, to becoming more resilient to training and thus to optimizing performance. Yet, in some cases athletes experience a prolonged inflammatory imbalance, which can even become a persistent (subclinical) inflammatory state. A status of so-called chronic, low-grade inflammation can pose a risk for health, hinder adaptation to training, and lead to performance stagnation or decline [46, 47]. Even though the quantile regression suggests a beneficial effect on inflammation when the systemic inflammatory response is low, limitingly no statements about its chronicity can be derived. Nevertheless, the present results provide grounds for future research investigating the role of (perceived) social support in immune management in elite athletes.

In terms of immune management of elite athletes, the present study further demonstrates the need to individualize approaches based on the athletes’ gender. In particular, female elite athletes should be focused on because the effects of social support on inflammation were mainly present in females. Gender differences in research on inflammation are not novel [12, 28, 38, 47] and sex hormones [12, 48], microglial biology (for details, see the review from Smith and Bilbo [48]) or sensitivity to social stress [28, 48] are among the suggested explanations. However, it should be noted that the present study was not designed to offer an in-depth analysis of gender-specific mechanisms. Appropriate parameters such as sex hormones were not collected because the investigation of gender differences was not the primary goal of this work. Although not causal, the correlational findings nevertheless underline the beneficial effect general social support has on the immune response especially in female athletes.

It is also worth noting that the structural aspect “living situation” did not provide any profound insight in predicting the elite athletes’ inflammation. Integrating several aspects of social relationships into the operationalization of the concept is recommended, as this broader view minimizes assessment errors [4]. It can be speculated that the living situation was not an important predictor for inflammation as elite athletes lead an inherently active lifestyle [49] and throughout the day meet various people from inside and outside of sport. The findings on functional and structural aspects of social relationships predicting inflammation thus underline the importance of the general, total social support the elite athlete perceives.

Results investigating the link between social relationships and inflammation considering the degree of the perceived, total general social support were not robust and less conclusive. Generally, the effects of inflammation on social support were small. This was expected given previous observation that the immune pathways are one of several ways linking health and behavior [2–4]. Interestingly, the link was only present, when social support was perceived as relatively moderate, but not when it was comparatively low or high. Especially in cases with poorer social relationships this study anticipated stronger effects due to previous findings on social isolation and loneliness [3, 6]. One possible reason for the absence of this awaited association might lay in the skewed distribution of the dependent variable: even the lowest quartile at average “rather agreed” to have received general social support. A finer breakdown, e.g., by singling out those athletes who “partly agreed” or less, statistically would have been inappropriate due to a small subgroup size (n = 6). The absence of an association at the upper end of the skewed distribution could be attributed to a ceiling effect: since the total, general perceived social support was maximal throughout the highest quartile, inflammation could not contribute to explanation of variance.

Why inflammation negatively predicts the perception of social support in the middle quartiles cannot be finally answered based on the cross-sectional, associative design of the study. Transferring the knowledge on social behavior during inflammation [10–12] to this study, one conceivable reason is that elite athletes actively seek social support from close others during acute inflammation. Being simply co-present with them [26, 50] or receiving emotional, informational or tangible support maybe enhancing recovery in the broader sense and subsequently could lower or balance inflammatory responses. Content wise, the different effects found across quartiles could be explained by motivational changes concomitant to acute inflammation, which have been discussed as enhancing recovery or rest [15]. It could be speculated that such motivational changes are higher in those athletes, who in general perceive relatively moderate social support, as they might be more aware of their (social) resources. Elite athletes with relatively low perceived social support accordingly would less expect benefits from others, which again might shift the motivation toward other promising sources for providing recovery. As the findings of quantile regressions explaining the social support through inflammatory markers, however, were not robust, the findings and the explanatory hypotheses should not be overrated.

When considering potential explanations for the variation in effects across quantiles, the role of fatigue and mood alterations should also be considered. These psychological conditions are often concomitant to inflammatory alterations and may exhibit symptomatic reductions in social participation or motivation amongst others [51–53]. That they mediate or moderate the association between inflammation and social relationships in this study yet seems unlikely against its findings: only the frequency of anxiety symptoms resulted to significantly predict the total, general perceived social support and this was only true for one quartile of one model. Nevertheless, potential contributions from these psychological conditions cannot entirely be excluded. An additional plausible interpretation is that, in elite athletes, inflammation generally does not explain the perception of social support. Given that these athletes often train multiple times per day and that the immune system responds dynamically to each exercise bout [17–19], habitual exposure to transient immunological challenges might settle the effect. Yet, to finally answer the question to which extend inflammation predicts the perception or the seeking of social support, further research is required.

Limitations and Strengths

This study pioneers in research on inflammation and social relationships through its methodological approach. It shed important light on the cellular pathway of the immune-to-brain-communication and helped converge on the important question of directions and magnitudes of effects. Limitingly, the design and the methodology of this study do not permit causal conclusions, also due to a lack of a control group. Moreover, a longitudinal analysis of within-subject differences has not yet been realized because of insufficient statistical power. Hence, this study indeed provides important first insights into the question whether there is an association between inflammation and social support, but further research is needed to uncover the underlying mechanisms explaining why and how these processes are linked.

Another characteristic distinguishing the study from previous research is the young sample of elite athletes, which generally can be seen as extraordinarily healthy (for a counter opinion see Maffetone and Laursen [54]). Unexpectedly, the informative value of the highest and lowest quartiles regarding social relationships was limited due to the distribution’s negative skewness and its shift toward strong perceived social support. Precisely the poor social relationships were expected to bring a knowledge gain to research on social isolation or loneliness, which in the end did not materialize. Regarding inflammation, on the other hand, the quartile-specific examination captures very well the variability in the sample of elite athletes and consequently provided important, new insights. The small degree of standardization and additionally insufficient information about the athletes’ daily and training routine the days immediately prior to measurement, however, makes it difficult to trace the inflammatory status back to a possible reason; next to infection or injury, it is very likely that in this study’s sample of elite athletes inflammation was influenced by previous exercise as transitory elevations in response to exercise can persist longer than the prescribed, minimum time gap of 10 h [18, 46]. Although the reason for inflammation was irrelevant for the research question and this aspect therefore does not limit the validity of this study’s results, considering the possibility of an exercise-induced inflammation could be important for research.

Conclusion

Existing articles on immune management in athletes (see, e.g., Palmowski et al. [16]) have largely neglected psychosocial aspects to date. The present findings, however, demonstrate that particularly the general functional aspect of social relationships negatively and robustly predicted the concentrations of the cytokines IL-6 and TNF-α as well as the SIRI. In this, results differed quantile-specifically with the magnitude of inflammation. Specifically, the effect of the total, general perceived social support on the cytokine levels increased at higher inflammation, whereas its beneficial association with the SIRI instead emerged primarily when innate and adaptive immune activity were most balanced. Effects of inflammation on social support were not robust across the degree of the general, total perceived social support in elite athletes. Social relationships, therefore, seem to be affected by inflammation only under specific conditions, which have yet to be more precisely identified.

Statement of Ethics

The study protocol is in accordance with the Declaration of Helsinki, and the Ethics Committee of the Justus Liebig University Giessen approved the study (ethical Approval No. AZ 55/22; approval date: May 10, 2022). Each athlete, and in cases of underage participants, their legal guardian, received written and oral information about data acquisition and privacy policies prior to the study and submitted signed informed consent.

Conflict of Interest Statement

Prof. Dr. Karsten Krüger was a member of the journal’s Editorial Board at the time of submission. The remaining authors have no conflicts of interest to declare.

Funding Sources

Research was funded by the German Federal Institute for Sport Sciences in 2021–2025 (Grant No. 081901/21-25, English title: Individual performance development in elite sports by holistic and transdisciplinary process optimization). The data analyzed for the purposes of this study was part of a multidisciplinary large-scale data set, which included multiple points of measurement and a cross-sectional and longitudinal perspective. The subset of data included in this study covered the cross-sectional data collected in the period February 2022 to February 2025. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Author Contributions

Kati Wiedenbrüg: conceptualization, formal analysis, investigation, methodology, visualization, and writing – original draft and editing; Lisa Musculus: investigation, project administration, validation, and writing – review; Celine Hilpisch and Sebastian Hacker: investigation, project administration, visualization, and writing – review; Karsten Krüger: conceptualization, funding acquisition, investigation, methodology, project administration, resources, supervision, and writing – review.

Funding Statement

Research was funded by the German Federal Institute for Sport Sciences in 2021–2025 (Grant No. 081901/21-25, English title: Individual performance development in elite sports by holistic and transdisciplinary process optimization). The data analyzed for the purposes of this study was part of a multidisciplinary large-scale data set, which included multiple points of measurement and a cross-sectional and longitudinal perspective. The subset of data included in this study covered the cross-sectional data collected in the period February 2022 to February 2025. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its supplementary material files. Further inquiries can be directed to the corresponding author.