Immune Response Subphenotyping to Predict Mortality in Sepsis: A Prospective Study in Resource-Limited Setting

Abstract

IMPORTANCE:

Sepsis remains a leading cause of death in infectious cases. The heterogeneity of immune responses is a major challenge in the management and prognostication of patients with sepsis. Identifying distinct immune response subphenotypes using parsimonious classifiers may improve outcome prediction, particularly in resource-limited settings.

OBJECTIVES:

This study aimed to evaluate whether classification of the immune response can serve as a predictor of mortality.

DESIGN, SETTING, AND PARTICIPANTS:

This prospective cohort study was conducted in the emergency department, inpatient wards, and ICU of a tertiary hospital. Adult patients diagnosed with sepsis within the previous 24 hours were included. Exclusion criteria were history of RBC transfusion, major thalassemia, decompensated cirrhosis, hematologic malignancy, or use of immunosuppressive or chronic corticosteroid therapy. Demographic, clinical, and laboratory data—including serum ferritin and monocyte human leukocyte antigen-DR/Human Leukocyte Antigen-DR) (mHLA-DR) levels—were collected.

MAIN OUTCOMES AND MEASURES:

Subjects were classified into the following immune subphenotypes: macrophage activation-like syndrome (MALS) (if ferritin > 4420 ng/mL), immunoparalysis (if mHLA-DR < 10,000 receptors/cell and ferritin ≤ 4420 ng/mL), and unclassified (if they did not meet the criteria for either MALS or immunoparalysis). The primary outcome was in-hospital mortality.

RESULTS:

Of the 200 subjects recruited, 54 (27%) were classified into the MALS group, 19 (9.5%) into the immunoparalysis group, and the remainder into the unclassified group. The in-hospital mortality rates for the MALS, immune paralysis, and unclassified groups were 83.3%, 68.4%, and 51.1%, respectively. The proportional hazards assumption was met between the MALS and unclassified groups (crude hazard ratio [HR] 2.3; 95% CI, 1.56–3.35) but not between the immunoparalysis and unclassified groups (crude HR 1.4; 95% CI, 0.76–2.50). After adjusting for confounding variables, MALS’s adjusted HR was 1.7 (95% CI, 1.13–2.49; p = 0.01).

CONCLUSIONS AND RELEVANCE:

The MALS subphenotype is an independent predictor of in-hospital mortality in sepsis.

KEY POINTS

Question: This study aimed to determine whether subphenotypic classification of the immune response could predict mortality in patients with sepsis.

Findings: In this prospective cohort study, 200 adult patients with sepsis were recruited from the emergency department, inpatient wards, and ICU. Patients were classified into macrophage activation-like syndrome (MALS), immunoparalysis, and unclassified groups based on ferritin and monocyte HLA-DR expression. The primary outcome was in-hospital mortality. MALS was associated with significantly higher mortality, with an adjusted hazard ratio indicating its independent predictive value, whereas immunoparalysis did not meet the proportional hazards assumption.

Meaning: The MALS subphenotype is an independent predictor of in-hospital mortality in sepsis.

Sepsis remains a leading cause of mortality, accounting for approximately 20% of global deaths. Despite advances in diagnosis and management, the global sepsis mortality rate remains high at 27% (1). In Indonesia, mortality is significantly higher, with rates reported up to 58.3% in major referral hospitals (2). This is likely underestimated due to underdiagnosis and underreporting in resource-limited settings.

Patients with sepsis exhibit considerable heterogeneity in demographics, infection sources, comorbidities, and immune responses. The current sepsis definition emphasizes a dysregulated host response that contributes to tissue injury, organ dysfunction, and death. Central to this dysregulation is the variability in immune responses, which likely underlies ongoing challenges in diagnosis, treatment, and prognostication (3, 4). Despite growing recognition of distinct immune subphenotypes, current sepsis management remains largely untailored for this biological diversity.

Subphenotypic classification in sepsis has been introduced, among other purposes, to categorize patients based on observable patterns of immune response. This approach is distinct from endotyping, which defines subtypes by underlying pathobiological mechanisms (5). Patients with sepsis may present a hyperinflammatory subphenotype with systemic inflammatory response syndrome (SIRS), an immunoparalysis subphenotype without overt SIRS, or an intermediate form (6). This immune subphenotyping holds potential for mortality prediction by reflecting differing biological processes. High-income countries increasingly use gene expression profiling for endotype-based prognostication (7, 8), but these methods are impractical in resource-limited settings due to cost and complexity. Thus, there is a need for practical immune response classifications suitable for these environments. Globally, few studies—such as the PROVIDE (Personalized Randomized trial Of Validation and restoration of Immune Dysfunction in severe infections and Sepsis) trial—have used immune subphenotying for mortality prediction in sepsis (9), and none to our knowledge have done so in Indonesia.

This study aimed to assess whether immune response subphenotyping can serve as a predictor of mortality.

MATERIAL AND METHODS

This prospective cohort study recruited patients from the emergency department, hospital wards, and ICU of Dr. Cipto Mangunkusumo National Central Public Hospital between June and December 2024. Ethical approval was obtained from the Health Research Ethics Committee of the Faculty of Medicine, Universitas Indonesia, and Dr. Cipto Mangunkusumo National Central Public Hospital (approval number: KET-672/UN2.F1/ETIK/PPM.00.02/2024; approved May 13, 2024; study title: “Inflammation Status as Predictor of In-hospital Mortality in Sepsis”). All procedures were conducted in accordance with the ethical standards of the institutional review board and the Declaration of Helsinki (1975).

Eligibility criteria consisted of adult patients (18 yr old or older) diagnosed with sepsis according to Sepsis-3 criteria (suspected infection with Sequential Organ Failure Assessment [SOFA] score ≥ 2) (3) within the last 24 hours; with definite or probable source of infection; and consented to participate in the study (patients or the legal guardian). Exclusion criteria included a history of packed red cell transfusion of more than 15 units, those with decompensated liver cirrhosis, hematologic malignancy, on immunosuppressant therapy, or chronic corticosteroid therapy (prednison ≥ 0.4 mg/kg body weight/d or an equivalent dose for a minimum of 15 d). The minimum sample size and justification are provided in Supplementary 1 (https://links.lww.com/CCX/B550).

Data collection from the subjects included demographic data, comorbidities, clinical data, source of infection, and laboratory results. Comorbidity data collected were those included in the Charlson Comorbidity Index (CCI) (10). Clinical parameters were recorded during recruitment and included the Glasgow Coma Scale, mean arterial pressure, administration of vasoactive agents, use of mechanical ventilation, and urine output. The primary source of infection was classified as pneumonia, acute cholangitis, urinary tract infection, skin and soft-tissue infection, intra-abdominal infection, bacteremia, febrile neutropenia, or other types, in accordance with the hospital’s clinical practice guidelines. The final diagnosis of the infection source was determined by the attending physicians. Laboratory data collected included platelet count, bilirubin levels, blood gas analysis, creatinine, serum lactate levels, blood cultures, and cultures from the infection site.

Blood culture samples were obtained from a peripheral site for aerobic culture according to the hospital’s standard of care. In specific cases, such as catheter-related bloodstream infections or infective endocarditis, additional blood samples were collected either from the catheter or from a different peripheral site. Samples were collected before the administration of antibiotics for the infection-related source. Blood culture results were later classified as multidrug-resistant (according to the World Health Organization’s bacterial priority pathogen list, 2024: critical, high, or medium(11) ), non-multidrug-resistant, negative culture (in the absence of bacterial growth), or commensal pathogens (if the bacteria identified were classified as common commensals according to the Centers for Disease Control and Prevention / National Healthcare Safety Network (CDC/NHSN) Common Commensals, 2024 (12)).

Whole blood was also collected for ferritin test and monocyte HLA-DR (mHLA-DR) expression count. Ferritin samples were drawn into serum tubes, whereas mHLA-DR samples were collected in EDTA tubes (BD Biosciences, Franklin Lakes, NJ). Both sample types were sent to an external laboratory for processing. Ferritin was measured using the chemiluminescence microparticle immunoassay principle. The expression of mHLA-DR was measured using monoclonal antibodies and a flow cytometry technique. The reagent was BD Quantibrite anti-HLA-DR/anti-monocyte (catalogue number 340827; BD Biosciences, Franklin Lakes, N J). The HLA-DR molecules on CD14/CD45 monocytes were presented as the number of molecules per monocyte. A hospital phlebotomist performed all blood sampling from the subjects within 24 hours after the subjects were recruited and provided consent.

Subjects were classified into immune response subphenotypes based on predefined biomarker thresholds. Patients were categorized as having macrophage activation-like syndrome (MALS) if their serum ferritin level exceeded 4420 ng/mL. This threshold was adopted based on validation studies conducted by the Hellenic Sepsis Study Group, which demonstrated its diagnostic accuracy in identifying MALS (9, 13). Patients were classified as having immunoparalysis if the expression of mHLA-DR was less than 10,000 receptors per cell and they did not meet the criteria for MALS. The mHLA-DR threshold, though more variable across studies, was selected based on evidence linking low expression levels with poor sepsis outcomes (14). Subjects who did not fulfill the criteria for either MALS or immunoparalysis were categorized as unclassified. An event was defined as death on hospitalization from any cause. Subjects discharged for outpatient follow-up and lost to follow-up, including those discharged against medical advice or transferred to another hospital, were included in the censor group.

Statistical analysis of the research data was performed using SPSS, Version 27 (Armonk, NY) and Stata, Version 18.5 (StataCorp, College Station, TX). The latter was specifically used for the -ln ln survival plots and the global test. Age, CCI, and SOFA score data were presented as means with sds. Data on sex, comorbidities, septic shock, use of mechanical ventilation, source of infection, outcomes, and positive blood cultures were presented as frequencies with corresponding proportions. Kaplan-Meier curve analysis was used to assess the assumption of proportional hazards between immune response groups. This assumption was further evaluated using the -ln ln survival plots and the global test. The proportional hazards assumption was met if the survival curves for the different immune response groups were parallel on the Kaplan-Meier curve and -ln ln survival scale, with a p value of greater than or equal to 0.05 from the global test.

Cox regression was used for both bivariate and multivariate analyses to determine survival differences between immune response groups. Results were presented as crude and adjusted hazard ratios (HRs) with 95% CIs and p values. Bivariate analysis was conducted on the MALS and immunoparalysis groups, with the unclassified group as the reference or comparison group. If a significant crude HR was found in the bivariate analysis (i.e., the CI of the HR did not include the value of 1, with p < 0.05), then the bivariate analysis of that group was continued with multivariate analysis.

Multivariate analysis was performed to determine whether the specific inflammatory response identified in the bivariate analysis was an independent predictor of in-hospital mortality, after adjusting for covariates known to be associated with mortality in sepsis. Covariates were selected a priori based on clinical relevance and existing literature, following the principles outlined by Harrell for regression modeling (15). To maintain model parsimony in light of the number of events, three covariates were included: SOFA score, CCI, and age. These variables were chosen to account for the primary established contributors to mortality in sepsis (16). The adjusted HR for significant immune response subphenotypes was calculated by incorporating these covariates into the multivariate Cox regression model.

The delta HR was calculated to determine whether any covariate was a confounder in the multivariate analysis. The delta HR was computed by determining the difference in HR before and after adding a given variable, divided by the HR after the variable was added. A covariate was considered a confounder if the delta HR was greater than or equal to 10%.

Before incorporating the three covariates in the Cox regression model, a bivariate analysis was conducted to assess the significance of each variable in relation to mortality using univariate Cox regression testing. These variables were then entered in the multivariate analysis using a stepwise approach, starting with the variable with the smallest p value in the bivariate analysis and gradually adding those with larger p values. This process continued until the final adjusted HR was obtained after incorporating all eligible variables.

In both bivariate and subsequent multivariate analyses, age, CCI, and SOFA score were treated as continuous variables. To assess potential multicollinearity among the covariates included in the multivariate model, Pearson correlation coefficients were calculated.

RESULTS

We screened 280 potentially eligible patients between June and December 2024. Eighty patients were excluded due to exclusion criteria, technical reasons, or because they declined to participate. At the end of the study period, 200 subjects were recruited (Fig. 1).

Figure 1.

Consolidated Standards of Reporting Trials diagram of participants’ recruitment. *A history of packed red cell transfusion of more than 15 units, those with decompensated liver cirrhosis, hematologic malignancy, on immunosuppressant therapy, or chronic corticosteroid therapy (prednisone ≥ 0.4 mg/kg body weight/d or an equivalent dose for a minimum of 15 d). ED = emergency department.

Of 200 subjects, 54 were classified into the MALS group (27%), 19 into the immunoparalysis (9.5%), and the remaining 127 subjects (63.5%) into the unclassified group. The proportion of subjects in the immunoparalysis group was significantly lower than anticipated.

The characteristics of the subjects are presented in Table 1. The mean age of the subjects was comparable across all three immune response groups. The number of male and female subjects was similar; however, the proportion of males was higher in the immunoparalysis group compared with the other two groups. The comorbidity index, proportion of septic shock, SOFA score, and in-hospital mortality were the highest in the MALS group, followed by the immunoparalysis and unclassified groups. The most common comorbidities observed among the study subjects were solid tumors, followed by diabetes mellitus, chronic kidney disease, chronic liver disease, and cerebrovascular accident or transient ischemic attack. The highest proportions of solid tumors and chronic liver disease were found in the MALS group, whereas the highest proportion of diabetes mellitus was observed in the immunoparalysis group. The most common sources of infection were community-acquired pneumonia, followed by acute cholangitis, nosocomial pneumonia, and intra-abdominal infections. Subjects with immunoparalysis had the highest proportions of nosocomial pneumonia and intra-abdominal infections compared with the other groups. Blood cultures were positive in 24% of subjects, with most blood pathogens identified falling under the category of non-priority pathogen. The distributions of ferritin and mHLA-DR expression levels across immune response groups are presented in Supplementary 2 (https://links.lww.com/CCX/B550).

TABLE 1.

Characteristics of Subjects Based on Immune Response Profile

Characteristics	Total (n = 200)	Macrophage Activation-Like Syndrome (n = 54)	Imunoparalysis (n = 19)	Unclassified (n = 127)
Demography
Age, yr, $\overline{\mathrm{x}}$ ± sd	52.0 ± 14.66	51.2 ± 15.18	55.1 ± 14.29	51.9 ± 14.55
Male, n (%)	100 (50)	28 (51.9)	13 (68.4)	59 (46.5)
Charlson Comorbidity Index, $\overline{\mathrm{x}}$ ± sd	4.0 ± 2.80	4.8 ± 2.99	3.9 ± 2.79	3.8 ± 2.70
Comorbidities
Solid tumor	82 (41.0)	32 (59.3)	9 (47.4)	41 (32.2)
Diabetes mellitus	55 (27.5)	7 (13.0)	6 (31.6)	42 (33.1)
Chronic kidney disease	23 (11.5)	4 (7.4)	1 (5.3)	18 (14.2)
Chronic liver disease	21 (10.5)	9 (16.7)	0 (0)	12 (9.5)
Cerebrovascular accident	21 (10.5)	6 (11.1)	1 (5.3)	14 (11.0)
Congestive heart failure	18 (9.0)	3 (5.6)	1 (5.3)	14 (11.0)
Hemiplegia	11 (5.5)	3 (5.6)	0 (0)	8 (6.3)
Myocardial infarct	11 (5.5)	2 (3.7)	0 (0)	9 (7.1)
Other comorbiditiesa	18 (9.0)	6 (11.1)	0 (0)	12 (9.5)
Severity
Septic shock, n (%)	130 (65.0)	40 (74.1)	13 (68.4)	77 (60.6)
Sequential Organ Failure Assessment score, x̄ ± sd	7.3 ± 3.37	8.7 ± 3.58	7.4 ± 2.87	6.6 ± 3.17
Outcome
Death, n (%)	123 (61.5)	45 (83.3)	13 (68.4)	65 (51.1)
Intensive care admission, n (%)	44 (22.0)	7 (13.0)	6 (31.6)	31 (24.4)
Source of infection
Pneumonia (community-acquired and nosocomial)	102 (51.0)	29 (53.8)	10 (52.6)	63 (49.6)
Acute cholangitis	31 (15.5)	8 (14.8)	2 (10.5)	21 (16.5)
Intra-abdominal	23 (11.5)	6 (11.1)	5 (26.3)	12 (9.4)
Skin and soft tissue	20 (10.0)	3 (5.6)	2 (10.5)	15 (11.8)
Urinary tract	15 (7.5)	5 (9.3)	0 (0)	10 (7.9)
Bacteremiab	3 (1.5)	1 (1.9)	0 (0)	2 (1.6)
Febrile neutropenia	3 (1.5)	1 (1.9)	0 (0)	2 (1.6)
Bone and joint	2 (1.0)	1 (1.9)	0 (0)	1 (0.8)
Ear, nose, and throat	1 (0.5)	0 (0)	0 (0)	1 (0.8)
Culture results
Positive blood culture, n (%)	48 (24.0)	14 (25.9)	8 (42.1)	26 (20.5)
Priority classificationc, n (%)
Critical	15 (31.3)	4 (28.6)	2 (25.0)	9 (34.6)
High	2 (4.2)	0 (0)	1 (12.5)	1 (3.8)
Medium	0 (0)	0 (0)	0 (12.5)	0 (0)
Non-priority	31 (64.6)	10 (71.4)	5 (62.5)	16 (61.5)

^aAIDS, peripheral vascular disease, peptic ulcer, chronic obstructive pulmonary disease, connective tissue disease, and dementia.

^bIncluded endocarditis and catheter-related bloodstream infection.

^cProportion to all positive results of blood culture.

At follow-up, 123 subjects experienced in-hospital death, while 77 subjects were censored from the analysis. Of those 77, 66 were discharged for outpatient follow-up, and 11 were lost to follow-up. The median survival was 10 days (95% CI, 7.4–12.6) (Supplementary 3, https://links.lww.com/CCX/B550).

The difference between in-hospital mortality among the three immune response groups is shown in the Kaplan-Meier curve in Figure 2.

Figure 2.

Kaplan-Meier survival curve of subjects based on immune response. MALS = macrophage activation-like syndrome.

The median survival of MALS, immunoparalysis, and unclassified groups was 4 days (95% CI, 2.6–5.2), 8 days (95% CI, 3.0–13.0), and 15 days (95% CI, 10.1–19.9), respectively. The proportional hazard assumption was met between the MALS and the unclassified group, while the assumption was not met between the immunoparalysis and the unclassified group (Supplementary 4, https://links.lww.com/CCX/B550).

The HR between immune response groups are presented in Table 2. The HR for the MALS group was statistically significant, while the HR for the immunoparalysis group was not statistically significant. Therefore, multivariate analysis for survival was performed only for the MALS group. These bivariate analysis findings confirm the results of the Kaplan-Meier curve analysis, which showed a significant difference in survival between the MALS and the unclassified groups but no significant difference between the immunoparalysis and the unclassified groups.

TABLE 2.

Bivariate Analysis Between Immune Responses

Immune Response, n (%)	Death		Hazard Ratio	95% CI	p
	Yes (n = 123)	No (n = 77)
Macrophage activation-like syndrome	45 (83.3)	9 (16.7)	2.3	1.56–3.35	< 0.001
Immunoparalysis	13 (68.4)	6 (31.6)	1.4	0.76–2.50	0.292
Unclassified	65 (51.2)	62 (48.8)	Reff

Reff = the “unclassified” group was used as the reference group in this analysis.

In the assessment of multicollinearity, a statistically significant moderate correlations were observed between age and CCI (r = 0.45; p < 0.001). Despite this correlation, all covariates—SOFA score, CCI, and age—were retained in the multivariate model, given their established roles as independent predictors of sepsis-related outcomes.

Multivariate analysis to assess the MALS as a predictor of mortality during hospitalization was conducted using Cox regression, as shown in Table 3.

TABLE 3.

Multivariate Analysis on Macrophage Activation-Like Syndrome

Confounders	HR (95% CI)	p	Delta HR (%)
Crude
Macrophage activation-like syndrome	2.3 (1.56–3.35)	< 0.001
Adjusted
Sequential Organ Failure Assessment score	1.8 (1.25–2.73)	0.002	23.8
Charlson Comorbidity Index	1.7 (1.13–2.48)	0.01	10.3
Age	1.7 (1.13–2.49)	0.01	0

HR = hazard ratio.

Multivariate analysis identified MALS as a significant predictor of in-hospital mortality in sepsis (adjusted HR 1.7; 95% CI, 1.13–2.49; p = 0.01). Subjects with MALS had nearly a two-fold increased risk of in-hospital death compared with subjects in the unclassified group. Additionally, the analysis revealed that the SOFA score and CCI acted as confounding factors in the model.

DISCUSSION

This study found that most subjects were classified into the unclassified group, while about a quarter were assigned to the MALS group. Only a small proportion were categorized as immunoparalysis. Sepsis patients with MALS had a significantly higher risk of in-hospital death compared with the reference group. No significant survival difference was observed between immunoparalysis and unclassified groups.

The proportion of subjects classified as MALS (27%) was comparable to the PROVIDE clinical trial (20%). However, the immunoparalysis group was substantially smaller than expected and much lower than in PROVIDE (9.5% vs. 42.9%). This lower-than-anticipated sample size may limit the interpretability of some findings.

Macrophage activation-like syndrome is a subset of sepsis marked by excessive hyperinflammation due to pathologic macrophage activation (17, 18). It is characterized by “ultra” hypercytokinemia. The term hyperferritinemia syndrome is sometimes used interchangeably with MALS to highlight that elevated ferritin levels are not just a nonspecific acute-phase response. Ferritin can induce the expression of Toll-like receptors, triggering a cascade that includes more interleukin (IL)-1β production, which exacerbates the inflammation more. The Hellenic Sepsis Study Group proposed ferritin as a simplified marker of MALS, replacing more complex criteria like the HScore or liver dysfunction with coagulopathy (13).

Conversely, immunoparalysis in sepsis is typically diagnosed using a range of markers, with Myla-DR expression being the most established. Low Myla-DR levels are associated with more severe immune dysfunction and worse outcomes. HLA-DR is a class II major histocompatibility complex molecule involved in antigen presentation. It enables antigen-presenting cells to activate CD4+ T cells, which in turn stimulate cytotoxic T cells and B cells. Reduced Myla-DR expression impairs this process and disrupts adaptive immune activation (14, 19). In addition to these immune subsets, other immunologic sub phenotypes likely exist but have not been characterized to date.

Importantly, this sub phenotypic classification is not only prognostic but may support the future application of personalized immunotherapy. The PROVIDE trial applied sub phenotypic classification to stratify patients and guide treatment. Subjects were categorized into MALS, immunoparalysis, or unclassified groups using ferritin levels and mHLA-DR expression, and were then administered tailored immunotherapies. In the MALS group, the immunotherapy arm demonstrated a reduction in organ dysfunction and improved 7-day survival, although this benefit did not persist at 28 days. Unfortunately, the efficacy of immunotherapy in the immunoparalysis group could not be reliably assessed due to the limited number of enrolled subjects (9).

In our study, the overall in-hospital mortality rate was 61.5%, which is considerably higher than in many other sepsis studies. This may be explained by the recruitment setting—a national referral hospital—where patients often present with more advanced disease and complex comorbidities (20, 21).

The MALS group experienced the highest mortality (83.3%), followed by the immunoparalysis (68.4%) and unclassified (51.1%) groups. Multivariate analysis confirmed MALS as an independent predictor of mortality, even after adjusting for confounders. Although mortality in the immunoparalysis group was higher than in the unclassified group, the survival difference was not statistically significant. The MALS group also showed the highest SOFA scores, comorbidity indices, and prevalence of septic shock, followed by the immunoparalysis and unclassified groups. These findings align with previous studies linking elevated SOFA scores and multiple comorbidities with poor sepsis outcomes (22–24). These clinical characteristics may contribute to the elevated mortality observed in both MALS and immunoparalysis groups.

Several studies support the association between the MALS subphenotype and increased mortality. Research in elderly ICU patients and in those with community-acquired pneumonia showed that MALS was linked to significantly worse outcomes (25, 26). A nested case-control study from the ProCESS trial (Protocol-Based Care for Early Septic Shock), which analyzed data from 1,341 patients, found that MALS patients had higher organ dysfunction scores and comorbidity indices. MALS was an independent predictor of 90-day mortality (odds ratio 3.1; 95% CI, 1.4–7.5; p = 0.008) (27). Due to its unique pathobiology, MALS is often unresponsive to conventional sepsis therapy, which may explain its association with poor outcomes (28).

Although MALS remains an independent predictor of mortality, the potential confounding effects of the SOFA score and comorbidities in this relationship cannot be disregarded. Patients with MALS in this study exhibited both high SOFA scores and a greater burden of chronic illness. These factors may contribute to or even predispose patients to the MALS subphenotype, reinforcing its association with increased mortality.

The lack of survival difference between the immunoparalysis and unclassified groups may be explained by several factors. First, the immunoparalysis sample size was insufficient to detect meaningful differences. This was despite using a higher mHLA-DR cutoff (< 10,000 receptors/cell) compared with the PROVIDE study (< 5,000 receptors/cell) (9). Second, the study evaluated short-term (in-hospital) mortality; long-term outcomes may be more relevant for assessing the impact of immunoparalysis. Third, the majority of patients (81%) were recruited from the emergency department. Immunoparalysis tends to occur later in the disease course and is more common in patients with prolonged hospitalization (29). Fourth, prior studies show that immunoparalysis leads to worse outcomes in older adults (30), yet this study population had a balanced mix of elderly and non-elderly patients. Lastly, patients receiving immunosuppressive therapy or long-term corticosteroids—who are at increased risk for immunoparalysis—were excluded from this study, which may have further reduced detection rates.

As intended, this study demonstrates that immune subphenotyping is a feasible and meaningful prognostic tool in sepsis, particularly in resource-limited settings. Ferritin and mHLA-DR are not merely classification markers; they reflect underlying immune dysregulation with clinical implications. Although the immunoparalysis group did not show a survival difference, understanding immune heterogeneity remains crucial for sepsis care.

There are several limitations in this study. First, the small number of immunoparalysis subjects limits statistical power. Second, the predominance of emergency department recruitment likely contributed to the underrepresentation of late-stage immune dysfunction. Third, exclusion criteria related to conditions affecting ferritin and mHLA-DR may restrict generalizability. Fourth, the study used single-timepoint measurements of biomarkers. Longitudinal monitoring, particularly of mHLA-DR expression, could provide improved prognostic accuracy (31). Finally, the elevated mortality rate—reflecting the severity of illness at a national referral hospital—may influence the study’s external validity.

Nevertheless, this study has several notable strengths. It is among the few from resource-limited settings to apply immune subphenotyping in sepsis prognostication. This approach could also guide immunomodulatory therapies. For example, patients with a MALS subphenotype may benefit from cytokine-targeted treatments such as IL-1 receptor antagonists, glucocorticoids, IV immunoglobulin, or extracorporeal blood purification (32). In contrast, those with immunoparalysis may be candidates for immune-stimulatory interventions, including interferon-γ, granulocyte-macrophage colony-stimulating factor, IL-7, or anti-programed death-ligand 1 antibodies (33). Furthermore, the diverse patient cohort also supports generalizability to a broader sepsis population.

CONCLUSIONS

The MALS subphenotype is an independent predictor of in-hospital mortality in patients with sepsis, with a mortality risk more than twice that of the unclassified subphenotype. No significant difference in in-hospital mortality was observed between sepsis patients with immunoparalysis and those with an unclassified immune response, although this finding may have been affected by an underpowered sample.

ACKNOWLEDGMENTS

The authors express their sincere gratitude to Ms. Utami Susilowati for her statistical expertise in the data analysis, Dr. Gilbert Sterling Octavius for his invaluable assistance in preparing the article, and to Dr. Getty Innash Nandika and Dr. Chyntia for their efforts in recruiting patients. We also extend our thanks to the Emergency Department staff at Dr. Cipto Mangunkusumo National Central Public Hospital for their support in the recruitment process. Our heartfelt appreciation goes to Professor Erni J. Nelwan, along with the faculty and fellows of the Division of Tropical Medicine and Infectious Diseases, Department of Internal Medicine, Dr. Cipto Mangunkusumo National General Hospital, Faculty of Medicine Universitas Indonesia, for their unwavering support of our research.

Supplementary Material

cc9-7-e1315-s001.pdf ^{(588.2KB, pdf)}