Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Feb 1.
Published in final edited form as: Psychoneuroendocrinology. 2024 Dec 25;172:107269. doi: 10.1016/j.psyneuen.2024.107269

Association of systemic inflammation and long-term dysfunction in COVID-19 patients: A prospective cohort

Felipe Dal-Pizzol a,b,*, Bruno Kluwe-Schiavon c, Henrique Ritter Dal-Pizzol d, Gabriele da Silveira Prestes a, Diogo Dominguini a, Carolina Saibro Girardi d, Lucas Santos d, José Cláudio Fonseca Moreira d, Daniel Pens Gelain d, Roger Walz e,f, Tatiana Barichello c, Cristiane Ritter a,b
PMCID: PMC12045274  NIHMSID: NIHMS2071294  PMID: 39778322

Abstract

COVID-19 has significant long-term impacts, including a chronic syndrome known as long-COVID, characterized by persistent symptoms post-recovery. The inflammatory response during acute infection is hypothesized to influence long-term outcomes. This study aimed to identify inflammatory biomarkers predictive of functional outcomes one year after hospital discharge. A prospective cohort study was conducted with 213 COVID-19 patients admitted to ICUs in Southern Brazil between June and November 2020. After exclusions and follow-ups, 109 patients were evaluated for one-year post-discharge. Plasma levels of Th1 (TNF-α, INF-γ, IL-12), Th2 (IL-4, IL-5, IL-6, IL-10, IL-13), and Th17 (IL-17, IL-22) cytokines were measured. Functional outcomes in psychiatric, cognitive, general health, and health perception domains were assessed. Statistical analyses included multivariate regression, regularized partial correlation network analysis, and K-means clustering. We demonstrate that plasma levels of various cytokines, along with demographic and clinical characteristics, can predict four distinct domains of functional outcomes one year following hospital discharge due to COVID-19 and that an hyperinflammatory phenotype was associated with the occurrence of a worse in psychiatric, general health, and health perception domains. The network analysis highlighted complex interconnections among immune markers and clinical variables, elucidating their roles in long-term health. These findings support using biomarkers for patient stratification and indicate potential targets for therapeutic interventions.

Keywords: Long-COVID, inflammation, T-cell response, biomarker

1. Introduction

COVID-19 has caused major impacts worldwide. One of the most important factors in understanding the disease is its inflammatory landscape (de Nooijer et al., 2023). During the pandemic, research indicated that a strong pro-inflammatory response correlated with a more severe clinical manifestation (Hedayati-Ch et al., 2024).

Apart from these acute symptoms, most patients also developed a second and chronic syndrome, called long-COVID (Yang et al., 2024). The most accepted definition is of the persistence of any one of a range of symptoms persisting for at least 3 months after the acute manifestations (WHO, 2021). Somatic and autonomic symptoms, such as malaise, muscle pain, and gastrointestinal problems, as well as neuropsychiatric symptoms, mainly depression, anxiety, chronic fatigue, and post-traumatic stress syndrome, are some of the symptoms that have been ascribed to long-term (Yang et al., 2024).

Long-COVID have also been linked to a strong pro-inflammatory response by the host; plenty of factors are hypothesized to predispose a patient to develop this condition, among them are microglial activation, persistent inflammation, viral load, clotting abnormalities, low blood oxygen, exposure to sedative and analgesic drugs, isolation, and immobility (Altmann et al., 2023). Interestingly the severity of the acute disease does not necessarily correlate with the chronic manifestations, this heterogeneity could be dictated by the host response to the infection (Moyo et al., 2023; Ma et al., 2023; Liew et al., 2024). Additionally, there is some evidence that the degree of acute inflammation is related to long-COVID (Yong et al., 2023), although, a recent study did not find any correlation between inflammation and different aspects of neuropsychiatric function (Chean et al., 2023).

In acute COVID-19 cases, studies show that a heightened Th2 immune response, and a deficient Th1/17, both have been linked to worse patient outcomes (Fouladseresht et al., 2022; Ramakrishnan et al., 2021). Also important is the finding that expansion of dedifferentiated monocytes, exhausted CD4+ T cells, and reduced IFN response are common among COVID-19 patients who developed some kind of neurologic symptoms (Heming et al., 2021). Furthermore, an impaired early antiviral defense (mediated by IFN-γ and IFN-α), associated with high pro-inflammatory cytokines and chemokines levels is also seen in severe COVID-19 patients (Kim and Shin, 2021).

Thus, in this study, we aimed at finding inflammatory biomarkers related to the Th1, Th2, and Th17 responses, that could be clinically relevant to determine four different outcomes domains one year after hospital discharge.

2. Material and methods

2.1. Study design

A prospective cohort was conducted with patients admitted to 6 intensive care units (ICUs) in across 2 hospitals in Southern Brazil between June 2020 and November 2020. During this period, the pre-dominant SARS-CoV-2 variants in Brazil were Alpha, Zeta, and Gamma (Giovanetti et al., 2022). The study was performed according to the Declaration of Helsinki and the Brazilian National Health Council Resolution 466. The Ethics Committee of São José Hospital (31384620.6.1001.5364) and Unimed Joinville Hospital (31384620.6.2002.5362) approved the protocol. All subjects or their surrogates provided written informed consent before inclusion in the study.

2.2. Participants

Patients aged > 18 years who were diagnosed with COVID-19 through reverse-transcription polymerase chain reaction or a rapid antigen test and required supplemental oxygen, non-invasive ventilation, or invasive mechanical ventilation due to COVID-19 pneumonia were included in the study. Exclusion criteria were patients with severe chronic diseases (chronic kidney disease under dialysis, Cirrhosis Child C, severe COPD, severe heart failure) or diseases capable of altering inflammatory response (such as chronic use of immunosuppressants, cancer patients without disease control and HIV without disease control), patients who suffered from neurodegenerative disorders and patients receiving palliative care or with a life expectancy of less than 24 h, as judged by the attending physician, were excluded.

2.3. Procedures

2.3.1. Acute phase

Investigators screened daily all patients admitted to the hospitals, and those who met the inclusion criteria were considered eligible. The patient was invited to participate in the study upon ICU admission, with a maximum inclusion window of 120 h. All necessary information was prospectively collected directly from the patient’s electronic medical record. Prehospital comorbidities were aggregated through the validated Charlson comorbidity index. A detailed description of procedures related to the acute care of these patients was previously published (Corneo et al., 2023; Santana et al., 2024).

2.3.2. Long-term functional outcomes follow-up

One year after hospital discharge, patients were assessed at the hospital’s outpatient clinic by a senior neuropsychologist to evaluate various functional domains. The evaluation encompassed psychiatric analysis, where symptoms of anxiety and depression were measured using the Hospital Anxiety and Depression Scale (HADS). Cognitive functions were assessed through three specific tests: the Verbal Fluency Test (VFT), the Rey Auditory Verbal Learning Test (RAVLT), and the Attentional Concentration Test (ACT). General health parameters were evaluated using the Pittsburgh Sleep Quality Index (PSQI), the Modified Medical Research Council Dyspnea Scale (mMRC), and the Baecke Physical Activity Questionnaire (BPAQ). Additionally, the impact of health on everyday life was self-reported by the patients through the 12-Item Short Form Survey (SF-12), providing insights into their overall well-being and quality of life. Then, the four long-term functional outcomes were quantified as a z-unit-based composite score by taking the average of each z-transform scale that composes them. To enhance interpretability, all outcomes were adjusted such that higher scores indicate worse outcomes.

2.3.3. Assay procedures

Biological samples were collected in the first 24 h after study inclusion. Blood samples were processed, and plasma was stored at −80 °C until quantification. Plasma protein biomarkers, related to Th1 (TNF-α, IFN-γ, IL-12) and Th2 (IL-4, IL-5, IL-6, IL-10, IL-13), and Th17 (IL-17, IL-22) responses were measured using the Immune Monitoring 65-Plex Human ProcartaPlex™ (Thermo Fisher Scientific Waltham, MA, USA) and a Luminex MAGPIX® system (Luminex Corporation, Austin, TX, USA).

2.4. Statistical analysis

Descriptive data are presented using mean (SD) for normally distributed variables, median (IQR) for skewed or non-normally distributed variables, and count (%) for categorical data. A multivariate stepwise regression analysis was employed to identify the predictors of the four long-term functional outcomes among biomarkers (TNF-α, INF-γ, IL-12, IL-4, IL-5, IL-6, IL-13, IL-17, and IL-22), demographic factors (age, sex, and years of education), and clinical characteristics (highest SOFA score within the first 24 h – as an index of disease severity, Charlson comorbidity index, length of hospital stay, body mass index and corticosteroid use). Additionally, to all outcomes, two regressions were performed, one including and the other excluding years of education, since years of education is described as a major determinant of cognition (Fletcher et al., 2021) sometimes blurring the effect of other associated variables. To further evaluate the relation between cytokines and the severity of illness at the time of assessment. we conducted a correlation analysis involving the biomarker panel, the SOFA score (a severity index recorded at the time of biomarker assessment), and the length of hospital stay (a short-term outcome).

In addition, a regularized partial correlation network was constructed using the EBICglasso algorithm to understand the conditional dependencies among variables. This method identifies associations between pairs of variables after accounting for the influence of other variables in the dataset (Heeren et al., 2020). The network was visualized with nodes representing variables and edges indicating partial correlations between them. Regularization helped refine the network by minimizing the occurrence of false positives and focusing on the most relevant relationships. Grouping was applied to organize variables into categories, such as ’Demographic,’ ’z-scores,’ ’Clinic,’ and subsets related to immune responses (excluding Th17 cytokines since neither IL-17 and IL-22 were identified as predictors of the four long-term functional outcomes in the regression analysis).

Finally, to further understand the relationship between inflammatory biomarkers (excluding Th17 cytokines) and functional outcomes a non-hierarchical K-means clustering analysis was performed to assess for inflammatory markers clustering in different phenotypes, and their association with functional outcomes. A multivariate stepwise regression analysis was used to determine the association between the patient’s outcome and the phenotypes generated by K-means clustering analysis and demographic factors and clinical characteristics. In all analyses, a p-value< 0.05 was adopted as the level for statistical significance.

All analyses were performed using the open-source statistical software R (version 4.2.3).

3. Results

3.1. Sample characteristics

A total of 213 patients were included, with an in-hospital mortality rate of 25 %. Of the 160 remaining patients, there were 5 additional deaths during the one-year follow-up period, and 46 were lost to follow-up. Thus, 109 (68 %) patients were evaluated in the University outpatient clinic one year after hospital discharge. The clinical and demographic characteristics of the 109 evaluated and of the remaining 51 non-evaluated patients were similar (Table 1). From the included patients, seventy-two were male, the mean (SD) age was 53 (12) and the median (IQR) years of education was 12 (8–16). The median (IQR) SOFA score on day 1 was 2 (1–3), the median (IQR) Charlson comorbidity index was 1 (0–2) and the median (IQR) length of hospital stay was 10 (6–19).

Table 1.

Clinical and demographic characteristics of the included and lost follow-up patients.

Included (n = 109) Lost follow-up (n = 51) p-value
Age, years, mean (SD) 53 ± 12 54 ± 14 0.92
Gender, male, n (%) 72 (66) 30 (59) 0.38
YOE, median (IQR) 12 (8–16) 12 (8–16) 0.83
LHOS, median (IQR) 10 (7–19) 10 (6–24) 0.44
SOFA, median (IQR) 2 (1–3) 2 (1–3) 0.88
Charlson, median (IQR) 1 (0–2) 1 (0–2) 0.69
BMI, median (IQR) 28 (26–32) 30 (26–33) 0.41
Corticosteroid use, yes, n (%) 88 (81) 39 (76) 0.54
Open in a new tab

Note. SD, Standard Deviation, IQR, Interquartile Interval, YOE, Years of education, LHOS, Length of hospitalization in day; SOFA, Sequential Organ Failure Assessment Score; Charlson comorbidity index.

3.2. Prediction of the functional outcomes scores by biomarkers

Fig. 1 provides regression estimates indicating the relationship between various predictors and functional outcomes, with additional statistical details given in e-Table 1. For cognitive outcomes one year after discharge, the model including years of education explained 48 % of the variance (r2 = 0.48), with years of education itself presented as a significant negative predictor (p < 0.001), suggesting, as expected, that higher educational level is associated with better cognition after one year. In contrast, both the length of hospital stay (p = 0.002) and the Charlson Comorbidity Index (p = 0.042) were positively associated with worse cognition one year after discharge, suggesting that longer hospitalization time and severe Charlson Comorbidity Index predicted worse cognition after one year. When years of education were removed from the model, both the length of hospital stay (p < 0.001) and the Charlson Comorbidity Index (p = 0.004) remained positively correlated with worse cognition within a model that accounts for 27 % of the variance (r2 = 0.27). Although IL-13 was marginally significant (p < 0.058, 95 % CI [0.225, −0.003]) which could suggest that higher levels of IL-13 during the hospitalizations might be related to worse cognition one year after the discharge, no single biomarker were independently related to cognitive outcomes one year after discharge.

Fig. 1.

Fig. 1.

Open in a new tab

Forest Plot of Regression Estimates. Legend: IL, interleukin; INF-γ, interferon-gamma; LHOS, Length of hospitalization in day; TNF-α, Tumor Necrosis Factor-alpha; YOE, Years of Education.

Regarding psychiatric symptoms, the model including years of education (r2 = 0.38) highlights that sex (p < 0.001) and that length of hospital stays (p = 0.001) were significant positive predictors of worse outcome after one year of discharge, while years of education was negatively associated (p = 0.002) with this outcome. In other words, being a woman and longer stay in the hospital predicted more severe psychiatric symptoms one year later, and higher educational level predicted less psychiatric symptoms after a year. Even after adjusting for years of education in the model, TNF-α demonstrated a significant negative association (p = 0.001) with psychiatric symptoms one year after discharge, while IL-4 exhibited a significant positive association (p = 0.017). Specifically, higher TNF-α levels during hospitalization were associated with fewer psychiatric symptoms one year later, whereas higher IL-4 levels predicted an increase in symptoms. Without including years of education, the predictors remained similar, with sex (being a woman, p < 0.001), length of hospital stays (p < 0.001), with the model explaining 31 % of the variance (r2 = 0.31).

In general health models, sex (p = 0.003 in the model with years of education and p = 0.001 in the model without years of education) and length of hospital stays (p = 0.016 in the model with years of education and p = 0.004 in the model without years of education) were consistent positive predictors, suggesting that being a woman and having stayed longer in the hospital is associated with of worse general health one year after discharge. IL-13, on the contrary, showed a negative association (p = 0.005 in the model with years of education and p = 0.004 in the model without years of education) with worse general health, suggesting that higher levels of IL-13 were associated with better general health one year later. Furthermore, in the model without years of education INF-y emerged as a significant positive predictor (p = 0.036) of worse general health. The r2 for these models is 0.31 with years of education and 0.29 without years of education.

Lastly, health perception outcomes were significantly influenced by sex (p < 0.001 for both models), body mass index (p = 0.019 in the model with years of education and p = 0.012 in the model without years of education), and length of hospital stays (p = 0.022 with years of education and p = 0.015 without years of education), suggesting that being a woman, higher BMI, and longer hospitalization periods are associated with worse health perception one year after the discharge. In contrast, both years of education (p = 0.009) and Charlson Comorbidity Index (p = 0.009 in the model with years of education and p = 0.017 in the model without years of education) were negatively associated with worse health perception one year after. Regarding the biomarkers, TNF-α (p = 0.003 with years of education and p = 0.007 without years of education) and IL-10 (p = 0.015 with years of education and p = 0.004 without years of education) were negatively associated with worse health perception, while IL-4 (p < 0.001 with years of education and p < 0.001 without years of education) was positively associated with worse outcome. This means that higher levels of TNF-α and IL-10 predicted better health perception after one year, while higher level of IL-4 predicted worse health perception after one year. The model with years of education explains 47 % of the variance (r2 = 0.47), while the one without years of education explains 45 % (r2 = 0.45).

To further evaluate whether the cytokine panel reflects illness severity at the time of assessment or is truly associated with long-term dysfunction, we conducted a correlation analysis involving the biomarker panel, the SOFA score, and the length of hospital stay (a short-term outcome) (e-Fig. 1). The analysis revealed that only TNF- α and IL-17 significantly and positively correlated with the SOFA score. Additionally, SOFA score, IL-5, IL-6, IL-17, and TNF- α significantly and positively correlated with the length of hospital stay. Interestingly, the majority of cytokines demonstrated positive correlations with one another. These findings suggest a complex interplay between disease severity, cytokine levels, and both short- and long-term outcomes in this patient population.

3.3. Regularized partial correlation network analysis

The regularized partial correlation network analysis in Fig. 2 elucidates conditional dependencies among the demographic, clinical, and inflammatory biomarker variables that were identified using multiple regressions, revealing patterns that shed light on long-term functional outcomes. The cytokine network shows significant interconnections among immune response markers. INF-γ has the strongest positive correlation with IL-12 (partial r = 0.28), IL-4 (partial r = 0.20), and IL-6 (partial r = 0.17). IL-10 is positively correlated with IL-6 (partial r = 0.22). IL-12 and IL-13 showed a positive correlation (partial r = 0.34). Corticosteroid use negatively correlated with IL-4 (partial r = −0.08) and IL-5 (partial r = −0.14). These correlations illustrate the complex interplay among inflammatory markers, reflecting immune response pathways that impact long-term health.

Fig. 2.

Fig. 2.

Open in a new tab

Regularized partial correlation network. Legend: COG, Cognition score; GH, General health score; HP, Health Perception score; IL, interleukin; INF-γ, interferon-gamma; LHOS, Length of hospitalization in day; PSY, Psychiatry score; TNF-α, Tumor Necrosis Factor-alpha; YOE, Years of Education. Blue arrow – Positive correlation; Red arrow – Negative correlation.

Length of hospital stay is associated with multiple outcomes. Specifically, length of hospital stay positively correlates with General Health (partial r = 0.12), Cognitive (partial r = 0.16) and Psychiatric (partial r = 0.01) indicating that longer hospital stays is related to worse outcomes one year after discharge. Importantly, IL-5 precedes the length of hospital stay in the network (partial r = 0.11), suggesting that elevated levels of these cytokines are linked to longer hospital stays. Years of education are negatively correlated with Psychiatric (partial r = −0.08) and Cognitive outcomes (partial r = −0.43), reinforcing what was shown in the multiple regressions, that more years of education contribute to less psychiatric symptoms and better cognition one year later.

To provide a comprehensive analysis of the network structure, we consider three standardized centrality metrics: Strength, Closeness, Betweenness, and Expected Influence (e-Fig. 2). Closeness measures how quickly a node can reach all other nodes in the network through the shortest paths. A higher closeness value indicates that the node can efficiently reach other nodes, implying a central or well-connected position in the network. In this network IL-5, IL-4 and length of hospital stay exhibit high closeness centrality, suggesting they are central nodes with widespread influence across the network. Betweenness indicates how often a node acts as a “bridge” along the shortest paths between two other nodes. Nodes with high betweenness are critical for connecting disparate regions of the network. Cognitive z-scores, IL-5, LHOS have high betweenness centrality, underscoring their importance in maintaining connectivity across different variables. The strength of a node indicates how connected or influential it is within the network. INF-γ, IL-12, IL-10, and Health Perception exhibit high strength, which is comparable to their expected influence. This metric accounts for both direct and indirect connections, offering a comprehensive measure of a node’s overall impact on the network. Overall, the combined analysis reveals that some biomarkers (INF-γ, IL-10, IL-12, and IL-6) and LHOS have the most significant influence on network connectivity. They could be key indicators for predicting long-term patient outcomes.

3.4. Determination of endophenotypes based on all biomarkers and their association with functional outcomes

To determine endophenotype classes we used a K-means cluster analysis using all measured biomarkers, and it was found two endophenotypes comprising 77 (cluster 1 - hypoinflammatory) and 32 (cluster 2 - hyperinflammatory) patients. Table 2 shows the sociodemographic, clinic, and biomarker data of the two clusters. There were no significant differences in the socio-demographic and clinical characteristics of the two clusters, except for lower corticosteroid use in the hypoinflammatory cluster. Cluster 1 patients showed lower levels of all inflammatory biomarkers measured.

Table 2.

Biomarkers phenotyping and clinical characteristics.

Phenotype 1 – Hypoinflammatory (n = 77) Phenotype 2 – Hyperinflammatory (n = 32) p-value*
Gender, male, n (%) 51 (66) 21 (66)  0.95
Age, mean (SD) 52 (12) 53 (12)  0.67
Charlson comorbidity index, median (IQR) 1.0 (0–2) 1 (0–2)  0.37
SOFA, median (IQR) 2 (1–3) 2 (1–3.8)  0.75
Length of hospital stay, days, median (IQR) 10 (6–19) 11 (7–23)  0.21
Years of education, mean (SD) 12.4 (4.8) 11.3 (4.7)  0.30
BMI, median (IQR) 29 (26–32) 27 (25–31)  0.19
Corticosteroid use, yes, n (%) 67 (87) 21 (66)  0.010
TNF–α 3.3 (1.8–9.4) 11.5 (8.1–17.4) < 0.001
IFN-γ 10.3 (8–15.3) 21.3 (9.7–25.0) < 0.001
IL–4 101 (67–189) 444 (385–527) < 0.001
IL–5 5 (2.2–8.2) 10.9 (5.6–17.3) < 0.001
IL–6 9.1 (4.4–12.8) 13.8 (8.3–43.6)  0.004
IL–10 3.8 (2.2–5.3) 7.9 (5.1–10.0) < 0.001
IL–12 4.5 (3.1–6.3) 6.2 (4.9–8.2)  0.004
IL–13 5.0 (1.0–8.1) 7.9 (4.0–10.4)  0.011
Open in a new tab

Note. SD, Standard Deviation, IQR, Interquartile Interval, IL, interleukin; INF-γ, interferon-gamma; LHOS, Length of hospitalization in day; TNF-α, Tumor Necrosis Factor-alpha; SOFA, Sequential Organ Failure Assessment Score

*

non-ajusted p-values.

Table 3 shows the results of regression analysis examining the associations between endophenotypes and functional outcomes while controlling for the effects of demographic and clinical variables. All outcomes were worse in the hyperinflammatory phenotype and only the cognition score did not reach statistical significance when adjusted for demographic and clinical characteristics (p = 0.08) when compared to the hypoinflammatory cluster (Table 3).

Table 3.

Biomarkers phenotyping and functional outcomes.

Functional outcomes Phenotype 1 Hypoinflammatory (n = 77) Phenotype 2 Hyperinflammatory (n = 32) p-value*
Cognition score, mean (SD) − 318 (1.9) 0.534 (1.8) 0.08
General Health score, mean (SD) −0.306 (1.7) 0.245 (1.5) 0.03
Health perception score, mean (SD) −.297 (1.5) 0.467 (1.6) 0.007
Psychiatry score, mean (SD) −0.311 (1.7) 0.629 (1.8) 0.03
Open in a new tab

Note.

*

Adjusted for demographic (age, sex, years of education) and clinical characteristics (worst SOFA score in the first 24 h, Charlson comorbidity index, length of hospital stay, BMI and corticosteroid use).

4. Discussion

Here, we demonstrate that plasma levels of various cytokines, alongside demographic and clinical characteristics, can predict outcomes across four distinct domains of functional recovery one year after hospital discharge from COVID-19. We demonstrate that plasma levels of various cytokines, when collected early after hospital admission, along with demographic and clinical characteristics, can predict four distinct domains of functional outcomes one year following hospital discharge due to COVID-19 and that an hyperinflammatory phenotype was associated with the occurrence of a worse in psychiatric, general health, and health perception domains. Notably, TNF-α, IL-4, IL-10, IL-13 and INF-y were identified as central biomarkers influencing long-term outcomes, and the regularized partial correlation network analysis suggested a complex interplay among inflammatory markers. These data add to the literature since most of published studies measured only a few domains of the impairments observed in long-COVID and did not measure different inflammatory mediators together to relevant clinical characteristics.

During SARS-CoV-2 infection Th1, Th2, and Th17 responses are activated and often associated with disease severity (Fouladseresht et al., 2022; Ramakrishnan et al., 2021). One of the hypotheses to explain the development of long-Covid is the magnitude of the acute inflammatory response to SARS-CoV2 (Altmann et al., 2023). The outcomes reported here are very similar to those observed in survivors of non-COVID acute respiratory distress syndrome (ARDS) or sepsis (Hughes et al., 2018; Darden et al., 2021; Mankowski et al., 2022; Maciel et al., 2019; Rocha et al., 2023). In these entities, inflammatory biomarkers, collected early after hospital or ICU admission, were also associated with long-term outcomes. Biomarkers of endothelial injury and inflammation collected at admission are associated with long-term cognitive impairment in critically ill patients (Hughes et al., 2018). Elevated circulating IL-6 and IL-10 concentrations at hospital discharge were associated with long-term cognitive dysfunction in ICU survivors (Maciel et al., 2019). In addition to that, some studies on long-COVID tried to detect similar patterns. Recently, IL-12 (Th1), IL-4 IL-13 (Th2), and IL-17 (Th17) levels were measured at hospital admission and 29 patients were followed until one year, but there was no significant association between their levels and the development of long-COVID symptoms (Cezar et al., 2024). In a recent meta-analysis that included 24 different inflammatory biomarkers collected early after SARS-CoV-2 infection, only the levels of lymphocytes and IL-6 were significantly higher in long-COVID than in non-Long-Covid cases (Yong et al., 2023). Non-specific CRP and procalcitonin when measured during acute infection are poorly correlated with the further occurrence of symptoms of anxiety, depression, or PTSD in survivors of COVID-19 (Chean et al., 2023). Patients who went on to develop long-COVID demonstrated significantly higher levels of TNF-α and IP-10 from 07 measured biomarkers. An intriguing aspect of our findings is that, among the four evaluated domains, only cognition was not associated with any of the measured cytokines. Furthermore, the relationship between cytokines and clinical outcomes was not uniformly positive; for example, TNF-α exhibited a negative association with various outcomes, whereas IL-4 showed a positive association with the same outcomes. This was further demonstrated by the regularized partial correlation network analysis, in which certain cytokines, such as IL-5, emerged as central nodes despite not appearing as significant in the regression analysis. These observations may be attributed to the design of our study. Blood samples were collected only once at the time of admission, thereby limiting our ability to capture the complex interactions that might occur between cytokines as the disease progresses. The biological activities of cytokines are frequently synergistic, occasionally summative, and at times, one cytokine can modulate the actions of another. However, we were unable to assess these temporal correlations.

Some studies addressed the issue of persistent inflammation and the occurrence of long-COVID. When these markers were collected late after recovery from acute infection, only IL-6 levels were associated with persistent symptoms (Peluso et al., 2021). This was also noted when blood was collected 8 months after recovery from COVID-19, from a panel of 21 cytokines only IL-1β, IL-6, and TNF showed a significant correlation with long-term symptoms (Schultheiß et al., 2022). At this same time-point, despite increased plasma levels of IL-5, IL-9, IL-17 F, IL-18, IL-22, IL-23, IL-33, CCL2/MCP-1, and sCD163 in post-COVID-19 disease as compared to individuals who never had COVID-19, only CCL2/MCP-1 and IL-8 were correlated to persistent symptoms (Schultheiß et al., 2023). Moreover, recently, it was found that increased complement activation and complement-mediated tissue injury could be a major component of sustained inflammation after COVID-19 (Cervia-Hasler et al., 2024). Long-Covid is associated with increased frequencies of tissue-migrating CD4 + T cells and exhausted SARS-CoV-2-specific CD8 + T cells up to 8 months after COVID-19 (Yin et al., 2024). Using whole-body positron emission tomography imaging it was demonstrated higher levels of activated T lymphocytes in the brain stem, spinal cord, bone marrow, nasopharyngeal and hilar lymphoid tissue, cardiopulmonary tissues, and gut wall when compared to pre-pandemic controls (Peluso et al., 2024). T lymphocytes in the spinal cord and gut wall this was associated with the presence of long-Covid symptoms (Peluso et al., 2024). It was suggested that this could be driven by antigen-antigen complexes, involving both autoantibodies and antibodies against herpesviruses, clotting dysfunction and reactivation of Epstein Barr Virus (Cervia-Hasler et al., 2024; Ryu et al., 2024). (Cheung et al., 2022; Stein et al., 2022; Swank et al., 2023). Intracellular SARS-CoV-2 single-stranded Spike protein-encoding RNA and double-stranded Spike protein-encoding RNA were demonstrated in the rectosigmoid lamina propria tissue up to 600 days following initial COVID-19, suggesting that tissue viral persistence could be associated with long-term immunologic perturbations (Peluso et al., 2024).

These studies helped to advance our understanding of the mechanisms of long-COVID but had some limitations. Long-COVID is defined by the persistence of some symptoms after a variable timepoint after recovery, thus a heterogeneous group of patients is mixed during analysis. Furthermore, several studies did not control the effect of biomarkers on relevant clinical characteristics of the patients, or patients were retrospectively recruited based on the presence or not of long-COVID symptoms. Additionally, there remains no universally accepted definition of long-COVID (ITT episode 37, 2024). We here determined four different functional domains, using well-validated tools, and it is relevant to notice that there is no single pattern of biomarkers that was associated with outcomes, and only cognitive score was not related to any one of the measured cytokines, nor to the different phenotypes. Different outcomes were related to a range of different biomarkers reflecting the complexity of the transition from the acute phase of the disease to long-term functional outcomes. Of the 10 measured cytokines only 4 (Th2 – IL-4, IL-10, and IL-13 and Th1 – INF-γ and TNF-α) were associated with the different functional outcomes. The regularized partial correlation network also emphasizes the role of specific biomarkers in the overall network connectivity. It is important to emphasize that the time limitation for blood collection may have impaired the detection of some interactions between cytokines. Consequently, one of the biomarkers identified in this study as being associated with long-COVID could potentially influence a cascade of different molecules, which may be the primary contributors to the observed outcomes. Furthermore, the biomarkers linked to long-term outcomes might differ if we were able to assess the effect of persistent inflammation on long-term symptoms.

Besides cytokines some clinical characteristics emerges as relevant predictors of worse functional outcomes. In alignment with previous studies on long-COVID, female sex emerged as a significant predictor in the final model across three of the four domains (Antony et al., 2023; Azambuja et al., 2024; Elias et al., 2024; Prestes et al., 2022). Another strong predictor identified was length of hospital stay, which was associated with all four domains assessed in this study. It is possible that length of hospital stay may serve as a more comprehensive marker of disease severity and baseline patient characteristics, such as age and comorbidities (Alimohamadi et al., 2022), compared to individual variables, thus providing valuable prognostic information for long-term outcomes. This factor is particularly relevant when determining priorities for rehabilitation programs. Targeting interventions towards female patients and those with prolonged hospital stays could potentially lead to a more efficient allocation of healthcare resources. Additionally, these findings reinforce the external validity of our results and supports the use of distinct functional domains as outcome measures, rather than relying on the non-standardized definition of Long-COVID.

The mechanisms of persistent inflammation during COVID-19 recovery have yet to be fully determined (Brightling et al., 2023) but were also noted in non-COVID ARDS and sepsis (Yende et al., 2019). Residual inflammation persists among certain individuals, either those with the most severe illness or inefficient clearing of SARS-CoV-2 or related to other unknown factors (Liew et al., 2024; Michael et al., 2023). Thus, defining the endotypes of patients could help in the understanding of this concept. This is a concept that has already been studied in non-COVID ARDS and sepsis (Calfee et al., 2014; Calfee et al., 2018; Puthucheary et al., 2020; Soussi et al., 2022; Sinha et al., 2023; Neyton et al., 2024). Classic studies identified a hyperinflammatory phenotype (higher plasma concentrations of inflammatory cytokines, lower serum bicarbonate, and higher vasopressor requirements) and a hyperinflammatory phenotype (Puthucheary et al., 2020; Sinha et al., 2023; Neyton et al., 2024). A similar approach was used for COVID-19 to determine short-term outcomes and response to corticosteroids (Sinha et al., 2021). When referring to long-term outcomes, few studies have focused on mortality. At hospital discharge septic patients assigned to “phenotype B” had more impaired cardiovascular and kidney functions, hematological disorders, and inflammation and had higher one-year mortality (Soussi et al., 2022). Based on the six-month trajectories of CRP and sPDL1 it was identified a phenotype in sepsis survivors with high-CRP and sPDL-1 (hyperinflammatory/immunosuppressed phenotype) had higher one-year mortality (Yende et al., 2019). These phenotypes probably reflect different gene expression signatures as recently demonstrated (Neyton et al., 2024). In sepsis, the hyperinflammatory phenotype was associated with elevated expression of innate immune response genes. The hipoinflammatory phenotype had elevated expression of adaptative response genes and T-cell response genes (Neyton et al., 2024). In a small cohort (n = 24), patients without post-COVID symptoms are more frequently in the “resolved” endotype – those with robust inflammatory and hemostatic in the hospital that resolved after discharge (An et al., 2023). Patients with persistently dampened (suppressive endotype) or persistently activated (unresolved endotype) had more frequent post-COVID symptoms (An et al., 2023). Based on oxidative damage, antioxidant, and inflammatory biomarkers an endotype with higher oxidative damage and inflammation and lowered antioxidant defenses are related to somatic and mental symptoms in Long-Covid (Al-Hakeim et al., 2023). Our hypoinflammatory phenotype exhibited a significantly higher use of corticosteroids compared to the hyperinflammatory phenotype. However, the impact of the phenotype remained significant in the regression analysis, independent of corticosteroid use. Interest in the effect of COVID-19 treatments on the incidence of Long COVID is growing. A recent meta-analysis demonstrated that antiviral treatment offers protective effects against Long COVID; however, neither corticosteroids nor monoclonal antibodies showed a similar impact. Therefore, a deeper understanding of these phenotypes could inform future precision treatment strategies.

The major strengths of our findings are the prospective study design, the evaluation of different functional domains late after hospital discharge and the evaluation of several biomarkers together to relevant clinical and demographic characteristics. Despite this, we had a loss of follow-up of 32 % during the one-year evaluation. This bias could be diminished by the fact that those patientsś clinical and demographic characteristics are like those patients that were evaluated. In addition, we do not have a temporal analysis of the biomarkers, and this would add relevant information to a better understanding of the pathogenesis of long-covid. Moreover, due to the nature of our data, we were unable to obtain a robust diagnosis regarding the presence of neuropsychiatric disorders prior to hospital admission. This limitation could influence our analysis and should be considered when interpreting the results. In this context, the study does not include a control group of non-COVID-19 patients, making it unclear whether the observed inflammatory responses are specific to COVID-19 or are generally associated to the general critically ill patients or general poor health outcomes. However, to the best of our knowledge, no published data exist that have contemporaneously collected data from COVID-19 patients and another critically ill population, which would serve as the ideal control group for a direct comparison. Using populations from different time periods would introduce bias due to the unique environmental stressors associated with the COVID-19 pandemic. Due to the single time point of blood collection, it is not possible to determine whether the measured cytokines are intrinsically related to long-term outcomes or merely serve as markers of inflammatory events triggered during the acute phase of COVID-19 and sustained throughout recovery and long COVID. Taking into account our results and published data, it is likely that a complex interplay exists between individual cytokines, clinical characteristics, and long-term outcomes. Inflammation probably plays a role in long COVID, both during the acute phase and throughout the recovery period, although the specific contribution of individual cytokines to long COVID symptoms remains unclear.

5. Conclusion

Biomarkers of inflammation are associated with different functional domains of long-COVID one year after hospital discharge and could be used to aggregate patients in different phenotypes to a better stratification of the risk of long-term impairments. Whether anti-inflammatory strategies could influence long-term symptoms should be further studied but a rational to this is supported by our results.

Supplementary Material

Supplementary material

Acknowledgements

FD-P is the guarantor of the content of the manuscript, including the data and analysis.

FD-P made substantial contributions to the conception and design of the work, analysis, and interpretation of data for the work; drafting the work; and final approved the version to be published.

BK-S, HRD-P, GSP, DD, CSG, LS, JCFM, DPG, RW, TB made substantial contributions to the design of the work the acquisition and analysis of data for the work and final approved the version to be published.

CR made substantial contributions to the conception and design of the work, analysis, and interpretation of data for the work; revised the work for important intellectual content, and final approved the version to be published.

The Texas Alzheimer’s Research and Care Consortium – TARCC 2022–26, The National Football League Players Association - NFLPA, National Institute of Health (NIH)/National Institute on Aging (NIA) grant (R01 AG072491) to TB and FDP

Funding

MCTIC/CNPq/FNDCT/MS/SCTIE/DECIT, 07/2020, grant number 401263/2020–7. BRF S.A. Hub unrestricted donation. FAPERGS-PPSUS #21/2551–0000073–2 The funding sources did not have any role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Data Availability

The datasets generated during the current study are available from the corresponding author on reasonable request.

Abbreviations:

ACT

Attentional Concentration Test

BPAQ

Baecke Physical Activity Questionnaire

COGz

Cognitive z-scores

COVID

Coronavirus Disease

COVID-19

Coronavirus Disease 2019

HADS

Hospital Anxiety and Depression Scale

HIV

Human Immunodeficiency Virus

ICU

Intensive Care Unit

IFN-α

Interferon alpha

IFN-γ

Interferon gamma

IL

Interleukin

IQR

Interquartile Range

MCP-1

Monocyte Chemoattractant Protein-1

MMRC

Modified Medical Research Council Dyspnea Scale

PASC

Post-Acute Sequelae of SARS-CoV-2 infection

PSQI

Pittsburgh Sleep Quality Index

PTSD

Post-Traumatic Stress Disorder

RAVLT

Rey Auditory Verbal Learning Test

SARS-CoV-2

Severe Acute Respiratory Syndrome Coronavirus 2

SD

Standard Deviation

COPD

Chronic Obstructive Pulmonary Disease

SF-12

12-Item Short Form Survey

SOFA

Sequential Organ Failure Assessment

Th1

T Helper Type 1 cells

Th1/17

T Helper Type 1/17 cells

Th2

T Helper Type 2 cells

TNF-α

Tumor Necrosis Factor alpha

VFT

Verbal Fluency Test

Footnotes

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Lucas Santos: Methodology, Investigation, Formal analysis. Carolina Saibro Girardi: Methodology, Investigation, Formal analysis. Diogo Dominguini: Methodology, Investigation, Formal analysis. Gabriele da Silveira Prestes: Methodology, Investigation, Data curation, Conceptualization. Henrique Ritter Dal-Pizzol: Writing – original draft, Methodology, Investigation, Formal analysis. Bruno Kluwe-Schiavon: Methodology, Investigation, Formal analysis. Cristiane Ritter: Writing – original draft, Funding acquisition, Formal analysis, Conceptualization. Felipe Dal Pizzol: Writing – review & editing, Project administration, Formal analysis, Data curation, Conceptualization. Tatiana Barichello: Writing –review & editing, Methodology, Investigation, Formal analysis. Roger Walz: Writing – review & editing, Project administration, Funding acquisition, Conceptualization. Daniel Pens Gelain: Writing – review & editing, Funding acquisition, Conceptualization. José Cláudio Fonseca Moreira: Writing – review & editing, Project administration, Funding acquisition, Conceptualization.

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.psyneuen.2024.107269.

References

  1. Al-Hakeim HK, Al-Rubaye HT, Al-Hadrawi DS, Almulla AF, Maes M, 2023. Long-COVID post-viral chronic fatigue and affective symptoms are associated with oxidative damage, lowered antioxidant defenses and inflammation: a proof of concept and mechanism study. Mol. Psy 28, 564–578. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Alimohamadi Y, Yekta EM, Sepandi M, et al. , 2022. Hospital length of stay for COVID-19 patients: a systematic review and meta-analysis. Multidiscip. Respir. Med 17 (1), 856. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Altmann DM, Whettlock EM, Liu S, Arachchillage DJ, Boyton RJ, 2023. The immunology of long COVID. Nat. Rev. Immunol 23, 618–634. [DOI] [PubMed] [Google Scholar]
  4. An AY, Baghela A, Zhang P.Get,et al. , 2023. Post-COVID symptoms are associated with endotypes reflecting poor inflammatory and hemostatic modulation. Front. Immunol 14, 1243689. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Antony B, Blau H, Casiraghi E, et al. , 2023. Predictive models of long COVID. EBioMedicine 96, 104777. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Azambuja P, Bastos LSL, Batista-da-Silva AA, et al. , 2024. Prevalence, risk factors, and impact of long COVID in a socially vulnerable community in Brazil: a prospective cohort study. Lancet Reg. Health Am 37, 100839. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Brightling CE, Evans RA, Singapuri A, et al. , 2023. Long COVID research: an update from the PHOSP-COVID Scientific Summit. Lancet Respir. Med 11, e93–e94. [DOI] [PubMed] [Google Scholar]
  8. Calfee CS, Delucchi K, Parsons PE, Thompson BT, Ware LB, Matthay MA, NHLBI ARDS Network, 2014. Subphenotypes in acute respiratory distress syndrome: latent class analysis of data from two randomised controlled trials. Lancet Respir. Med 2, 611–620. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Calfee CS, Delucchi KL, Sinha P, Matthay MA, Hackett J, Shankar-Hari M, McDowell C, Laffey JG, O’Kane CM, McAuley DF, Irish Critical Care Trials Group, 2018. Acute respiratory distress syndrome subphenotypes and differential response to simvastatin: secondary analysis of a randomised controlled trial. Lancet Respir. Med 6, 691–698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Cervia-Hasler C, Brüningk SC, Hoch T, et al. , 2024. Persistent complement dysregulation with signs of thromboinflammation in active Long Covid. Science 383, 7942. [DOI] [PubMed] [Google Scholar]
  11. Cezar R, Kundura L, André S, et al. , 2024. T4 apoptosis in the acute phase of SARS-CoV-2 infection predicts long COVID. Front. Immunol 14, 1335352. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Chean D, Benarroch S, Pochard F, Kentish-Barnes N, Azoulay E, 2023. Mental health symptoms are not correlated with peripheral inflammatory biomarkers concentrations in COVID-19-ARDS survivors. Intensive Care Med. 49, 109–111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Cheung CCL, Goh D, Lim X, et al. , 2022. Residual SARS-CoV-2 viral antigens detected in GI and hepatic tissues from five recovered patients with COVID-19. Gut 71, 226–229. [DOI] [PubMed] [Google Scholar]
  14. Corneo E, Garbelotto R, Prestes G, et al. , 2023. Coagulation biomarkers and coronavirus disease 2019 phenotyping: a prospective cohort study. Thromb. J 21, 80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Darden DB, Brakenridge SC, Efron PA, et al. , 2021. Biomarker evidence of the persistent inflammation, immunosuppression and catabolism syndrome (PICS) in chronic critical illness (CCI) after surgical sepsis. Ann. Surg 274, 664–673. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Elias TP, Gebreamlak TW, Gebremeskel TT, et al. , 2024. Determinants of post-acute COVID-19 syndrome among hospitalized severe COVID-19 patients: a 2-year follow-up study. PLoS One 19 (5), e0298409. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fletcher J, Topping M, Zheng F, Lu Q, 2021. The effects of education on cognition in older age: evidence from genotyped Siblings. Soc. Sci. Med 280, 114044. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Fouladseresht H, Ghamar Talepoor A, Eskandari N, et al. , 2022. Potential immune indicators for predicting the prognosis of COVID-19 and trauma: similarities and disparities. Front. Immunol 12, 785946. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Giovanetti M, Fonseca V, Wilkinson E, et al. , 2022. Replacement of the Gamma by the Delta variant in Brazil: impact of lineage displacement on the ongoing pandemic. Virus Evol. 8, veac024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hedayati-Ch M, Ebrahim-Saraie HS, Bakhshi A, 2024. Clinical and immunological comparison of COVID-19 disease between critical and non-critical courses: a systematic review and meta-analysis. Front. Immunol 15, 1341168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Heeren A, Bernstein EE, McNally RJ, 2020. Bridging maladaptive social self-beliefs and social anxiety: a network perspective. J. Anxiety Dis 74, 102267. [DOI] [PubMed] [Google Scholar]
  22. Heming M, Li X, Räuber S, et al. , 2021. Neurological manifestations of COVID-19 feature T cell exhaustion and dedifferentiated monocytes in cerebrospinal fluid. Immunity 54, 164–175. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Hughes CG, Patel MB, Brummel NE, et al. , 2018. Relationships between markers of neurologic and endothelial injury during critical illness and long-term cognitive impairment and disability. Intensive Care Med. 44, 345–355. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. ITT Episode 37, 2024. What exactly is Long-Covid. N. Engl. J. Med 391, e17. [DOI] [PubMed] [Google Scholar]
  25. Kim YM, Shin EC, 2021. Type I and III interferon responses in SARS-CoV-2 infection. Exp. Mol. Med 53, 750–760. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Liew F, Efstathiou C, Fontanella S, et al. , 2024. Large-scale phenotyping of patients with long COVID post-hospitalization reveals mechanistic subtypes of disease. Nat. Immunol 25, 607–621. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Ma Y, Deng J, Liu Q, Du M, Liu M, Liu J, 2023. Long-term consequences of asymptomatic SARS-CoV-2 infection: a systematic review and meta-analysis. Int. J. Environ. Res. Public Health 20, 1613. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Maciel M, Benedet SR, Lunardelli EB, et al. , 2019. Predicting long-term cognitive dysfunction in survivors of critical illness with plasma inflammatory markers: a retrospective cohort study. Mol. Neurobiol 56, 763–767. [DOI] [PubMed] [Google Scholar]
  29. Mankowski RT, Anton SD, Ghita GL, Brumback B, Darden DB, Bihorac A, Moore FA, 2022. Older adults demonstrate biomarker evidence of the persistent inflammation, immunosuppression, and catabolism syndrome (PICS) after sepsis. J. Gerontol.: Ser. A 77 (1), 188–196. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Michael BD, Dunai C, Needham EJ, et al. , 2023. Para-infectious brain injury in COVID-19 persists at follow-up despite attenuated cytokine and autoantibody responses. Nat. Commun 14, 8487. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Moyo E, Chimene M, Moyo P, Musuka G, Mangoya D, Murewanhema G, Dzinamarira T, 2023. Risk factors and clinical presentations of long COVID in Africa: a scoping review. J. Infect. Public Health 16, 1982–1988. [DOI] [PubMed] [Google Scholar]
  32. Neyton LP, Sinha P, Sarma A, et al. , 2024. Host and microbe blood metagenomics reveals key pathways characterizing critical illness phenotypes. Am. J. Respir. Crit. Care Med 209, 805–815. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. de Nooijer AH, Pickkers P, Netea MG, Kox M, 2023. Inflammatory biomarkers to predict the prognosis of acute bacterial and viral infections. J. Crit. Care 78, 154360. [DOI] [PubMed] [Google Scholar]
  34. Peluso MJ, Lu S, Tang AF, et al. , 2021. Markers of immune activation and inflammation in individuals with postacute sequelae of severe acute respiratory syndrome coronavirus 2 infection. J. Infect. Dis 224, 1839–1848. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Peluso MJ, Ryder D, Flavell RR, et al. , 2024. Tissue-based T cell activation and viral RNA persist for up to 2 years after SARS-CoV-2 infection. Sci. Transl. Med 16, eadk3295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Prestes GDS, Simon CS, Walz R, et al. , 2022. Respiratory outcomes after 6 months of hospital discharge in patients affected by COVID-19: a prospective cohort. Front. Med 9, 795074. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Puthucheary ZA, Gensichen JS, Cakiroglu AS, et al. , 2020. Implications for post critical illness trial design: sub-phenotyping trajectories of functional recovery among sepsis survivors. Crit. Care 24, 1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Ramakrishnan RK, Al Heialy S, Hamid Q, 2021. Implications of preexisting asthma on COVID-19 pathogenesis. Am. J. Physiol. Lung Cell. Mol. Physiol 320, L880–L891. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Rocha FRD, Gonçalves RC, Prestes GDS, et al. , 2023. Biomarkers of neuropsychiatric dysfunction in intensive care unit survivors: a prospective cohort study. Crit. Care Sci 35, 147–155. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Ryu JK, Yan Z, Montano M, et al. , 2024. Fibrin drives thromboinflammation and neuropathology in COVID-19. Nature 633, 905–913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Santana A, Prestes GDS, Silva MDD, et al. , 2024. Identification of distinct phenotypes and improving prognosis using metabolic biomarkers in COVID-19 patients. Crit. Care Sci 36, e20240028en. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Schultheiß C, Willscher E, Paschold L, et al. , 2022. The IL-1β, IL-6, and TNF cytokine triad is associated with post-acute sequelae of COVID-19. Cell Rep. Med 3, 100633. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Schultheiß C, Willscher E, Paschold L, et al. , 2023. Liquid biomarkers of macrophage dysregulation and circulating spike protein illustrate the biological heterogeneity in patients with post-acute sequelae of COVID-19. J. Med. Virol 95, e28364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Sinha P, Furfaro D, Cummings MJ, 2021. Latent class analysis reveals COVID-19–related acute respiratory distress syndrome subgroups with differential responses to corticosteroids. Am. J. Respir. Crit. Care Med 204, 1274–1285. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Sinha P, Kerchberger VE, Willmore A, et al. , 2023. Identifying molecular phenotypes in sepsis: an analysis of two prospective observational cohorts and secondary analysis of two randomised controlled trials. Lancet Respir. Med 11, 965–974. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Soussi S, Sharma D, Jüni P, et al. , 2022. Identifying clinical subtypes in sepsis survivors with different one-year outcomes: a secondary latent class analysis of the FROG-ICU cohort. Crit. Care 26, 114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Stein SR, Ramelli SC, Grazioli A, et al. , 2022. SARS-CoV-2 infection and persistence in the human body and brain at autopsy. Nature 612, 758–763. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Swank Z, Senussi Y, Manickas-Hill Z, et al. , 2023. Persistent circulating severe acute respiratory syndrome coronavirus 2 spike is associated with post-acute coronavirus disease 2019 sequelae. Clin. Infect. Dis 76, e487–e490. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. World Health Organization. A clinical case definition of post COVID-19 condition by a Delphi consensus, 6 October 2021. No. WHO/2019-nCoV/Post_COVID 19_condition/Clinical_case_definition/2021.1. World Health Organization, 2021. [Google Scholar]
  50. Yang J, Markus K, Andersen KM, et al. , 2024. Definition and measurement of post-COVID-19 conditions in real-world practice: a global systematic literature review. BMJ Open 14, e077886. [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Yende S, Kellum JA, Talisa VB, et al. , 2019. Long-term host immune response trajectories among hospitalized patients with sepsis. JAMA Netw. Open 2 e198686–e198686. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Yin K, Peluso MJ, Luo X, et al. , 2024. Long COVID manifests with T cell dysregulation, inflammation and an uncoordinated adaptive immune response to SARS-CoV-2. Nat. Immunol 25, 218–225. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Yong SJ, Halim A, Halim M, et al. , 2023. Inflammatory and vascular biomarkers in post-COVID-19 syndrome: a systematic review and meta-analysis of over 20 biomarkers. Rev. Med. Virol 33, e2424. [DOI] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials

Supplementary material
NIHMS2071294-supplement-Supplementary_material.pdf (228.1KB, pdf)

Data Availability Statement

The datasets generated during the current study are available from the corresponding author on reasonable request.

RESOURCES