Relationships between age, soluble triggering receptor expressed on myeloid cells-1 (sTREM-1), and mortality among critically ill adults: a cohort study

Abstract

Background.

Innate immune dysregulation may contribute to age-related differences in outcomes among critically ill adults. Soluble triggering receptor expressed on myeloid cells-1 (sTREM-1) is an important innate immune marker with prognostic value in sepsis, but age-related differences have not been studied.

Methods.

This was a prospective cohort from a large tertiary care hospital enrolling adults from both medical and trauma-surgical ICUs. Plasma sTREM-1 was measured in participants within 24 hours of ICU admission. We analyzed associations between age (≤50 and >50 years) and sTREM-1 using linear regression. We then examined associations between sTREM-1 and both 28-day mortality and persistent organ dysfunction (defined as need for dialysis, vasopressors, or invasive mechanical ventilation) 7 days following admission using relative risk regression.

Results.

Of 231 critically ill adults, older patients (n=122) had higher prevalence of chronic disease and sepsis on enrollment than younger patients, but acute illness severity was similar. Age over 50 was associated with 27% higher sTREM-1 concentrations (95% CI 6%-53%), adjusted for sex and Charlson comorbidity index (CCI). Two-fold higher sTREM-1 was associated with 2.42-fold higher risk for mortality (95% CI 1.57, 3.73) and 1.86-fold higher risk for persistent organ dysfunction (95% CI 1.45, 2.39), adjusted for sex, CCI, and age.

Conclusions.

sTREM-1 was elevated among critically ill older adults, and strongly associated with both death and persistent organ dysfunction. Immune responses associated with sTREM-1 may contribute to age-related differences in ICU outcomes, warranting further study as a potential therapeutic target in older adults.

Introduction

Advanced age is one of the strongest risk factors for death and organ injury in the intensive care unit (ICU) (1-3). A major contributor to these poor outcomes is a dysregulated immune response to critical illness, and an emerging body of evidence reveals important age-related differences in early innate immune mediated inflammation (4-7). Although numbers of key innate immune cells are largely preserved in older adults, phenotype and function are altered. For example, signaling pathways in neutrophils that lead to effective recognition and elimination of pathogens are impaired (8). Aging is also associated with chronic elevations in pro-inflammatory cytokines (9). Some of these immune responses may also be involved in the pathogenesis of chronic age-related diseases, such as cardiovascular disease and cancer, the presence of which further contribute to the risk of ICU mortality (10, 11).

The triggering receptor expressed on myeloid cells-1 (TREM-1) is thought to contribute to mortality in critical illness by potentiating innate immune responses, and its signaling may be altered with aging (12, 13). TREM-1, a member of the immunoglobulin superfamily, is expressed largely on plasma membranes of neutrophils and macrophages/monocytes. TREM-1 activation amplifies oxidative burst, pro-inflammatory cytokine and chemokine production, and neutrophil migration, often synergistically with other pattern recognition receptors (14-18). Best studied in preclinical models of infection, blocking TREM-1 protects against the development of sepsis and shock while improving survival (12). A soluble form of TREM-1 (sTREM-1) may reflect TREM-1 receptor activation and could help identify patients at high risk for related outcomes (19). Prior work demonstrates that sTREM-1 is strongly associated with death among adults with sepsis (19-21). However, its role outside of sepsis is understudied, and differences in sTREM-1 across age groups have not been investigated. Better understanding how sTREM-1 and related immune pathways are altered in older adults during critical illness could clarify why they are more likely to experience poor outcomes, and could inform how to tailor therapies to this high-risk group.

Our primary aim was to investigate the relationship between older age and sTREM-1, as well as sTREM-1 and mortality, in a prospective cohort of critically ill adults admitted for sepsis, trauma, and post-surgical care. We also explored the association between plasma sTREM-1 and a secondary outcome of persistent organ dysfunction, which is a major cause of morbidity and mortality in older critically ill patients but has not been studied in relation to sTREM-1 (22-24). Finally, we explored whether TREM-1 related innate immune dysregulation, as marked by sTREM-1, represented a significant mediator of age-related differences in ICU outcomes.

Materials and Methods

Study Population

The Critical Illness Translational Research Cohort (CITRC) is an ongoing prospective observational cohort of patients admitted to the medical or trauma-surgical ICUs at Harborview Medical Center (Seattle, WA) (25). Patients were enrolled within 24 hours of ICU admission if they had at least 2 criteria for the systemic inflammatory response syndrome (23). For this analysis we included patients recruited between 2015 and 2018 who had plasma obtained on enrollment. Exclusion criteria included the inability to provide informed consent, non-English speaking, metastatic cancer, CD4 cell count <200, immunosuppressive medication prior to admission (including prednisone >20 mg/day or equivalent glucocorticoid; antiproliferative agents; mTOR inhibitors; calcineurin inhibitors; and tumor necrosis factor-alpha inhibitors), and do-not-resuscitate orders within the first 24 hours of ICU admission. This study was approved by the Institutional Review Board at the University of Washington.

Measurement of sTREM-1

Blood was collected into EDTA anticoagulant from patients within 24 hours of enrollment. Plasma was isolated by centrifugation and stored at −80°C. We measured sTREM-1 concentrations using an electrochemiluminescent immunoassay (Meso Scale Discovery, Rockville, MD). Samples were diluted 5-fold to fit within the dynamic range of the assay (3.93–3,000 pg/ml). A reference plasma sample and duplicates for a subset of samples were included on each plate for quality control purposes.

Clinical Data and Outcomes

Clinical data, including demographics, chronic comorbidities, and acute hospitalization data, were captured in standardized case report forms through chart review and patient interviews. Our primary outcome was mortality 28 days from ICU admission. We also examined a secondary composite outcome of persistent organ dysfunction, where persistent organ dysfunction included need for dialysis, mechanical ventilation, or vasopressors 7 days after admission among those surviving to that point (26). We included this outcome because persistent organ dysfunction is thought to be on the causal pathway to death among critically ill patients (22, 23). Sepsis was defined by a modified Sequential Organ Failure Assessment (SOFA) score of two or more, with either suspected or microbiologically confirmed infection within the first 48 hours of ICU admission (23). Shock was defined by need for vasopressor therapy.

Statistical Analysis

We summarized baseline characteristics by age. We stratified by age groups of ≤ 50 years and > 50 years based on the mean age in this cohort for our primary analyses. This age threshold is also consistent with epidemiologic data showing outcomes related to sepsis and critical illness rise sharply after 50 years of age (2, 3, 27, 28). We reported mean and standard deviation for normally distributed continuous variables; median and interquartile range for non-normal continuous variables; and number and proportion for categorical variables. We compared these across age groups using t-tests, Wilcoxon rank sum tests, and Fisher’s exact tests respectively.

We first examined the association between age group and sTREM-1 concentrations using linear regression. We adjusted for sex and Charlson comorbidity index (CCI) as pre-specified confounders (11, 29).

We then examined the associations between sTREM-1 and both 28-day mortality and persistent organ dysfunction using a relative risk (RR) regression, with Poisson distribution and robust standard errors (30). These analyses were adjusted for sex, CCI, and for age as a continuous variable. Since sTREM-1 has primarily been studied in infection and septic shock, we stratified this analysis by sepsis and by shock to see whether these conditions modified the effect of sTREM-1 on outcomes.

In these models we used log₂-transformed sTREM-1 values as a continuous variable due to the right-skewed nature of these measurements. We chose not to adjust for baseline physiologic illness severity because this would incorporate factors proposed to be on the causal pathway from TREM-1 activation to death and organ dysfunction. We considered a P value < 0.05 as statistically significant and provided 95% confidence intervals (CI). Analyses were conducted in STATA (version 16.0).

Lastly, we sought to explore whether innate immune dysregulation, as reflected by sTREM-1, mediated the relationship between older age and ICU outcomes. We performed a causal mediation analysis using the “mediation” package in R, using linear regression to model the relationship between age and sTREM-1, and logistic regression to model relationships between age, sTREM-1 and ICU outcomes (31). This approach allowed us to decompose the total effect of age on outcomes into the indirect effect of age on outcomes (i.e. the effect mediated through sTREM-1, our primary parameter of interest) and a direct effect through mechanisms other than those reflected by sTREM-1. Bootstrapping was used to calculate 95% confidence intervals (32). Primarily we chose to adjust only for sex, assuming that the effect of age on sTREM-1 and outcomes is partially through chronic disease and increased risk for infection (33). However since causal mediation models can be vulnerable to confounders of the mediator-outcome relationship, we also built models adjusting for CCI and the presence of sepsis on enrollment.

Results

Cohort characteristics

Of 231 critically ill patients, 53% were older than 50 years. Overall, the cohort was predominantly male, and 63% were admitted to a trauma-surgical ICU while 37% were admitted to a medical ICU (Table 1). Relative to younger adults, adults over 50 had a high proportion of pre-existing chronic diseases. Sepsis was more common among older adults, with skin, bloodstream, and lung infections being most common overall (Table S1). Although there were no differences in shock and invasive mechanical ventilation on enrollment by age, there were trends towards higher 28-day mortality and persistent organ dysfunction among older adults (Table 1).

Table 1:

Cohort description by age^a

	Total	Age ≤ 50 years	Age > 50 years	P value
	N=231	N=109	N=122
Demographics
Age (years)	50 (14)	38 (9)	61 (8)	<0.001
Female	76 (33%)	39 (36%)	37 (30%)	0.40
Race				0.83
White	179 (77%)	81 (74%)	98 (80%)
Asian	8 (3%)	5 (5%)	3 (2%)
Black	23 (10%)	12 (11%)	11 (9%)
Native American	17 (7%)	9 (8%)	8 (7%)
Pacific Islander	4 (2%)	2 (2%)	2 (2%)
Chronic Disease
Coronary artery disease	23 (10%)	2 (2%)	21 (17%)	<0.001
Chronic obstructive pulmonary disease	33 (14%)	3 (3%)	30 (25%)	<0.001
Heart failure	19 (8%)	2 (2%)	17 (14%)	<0.001
Chronic kidney disease	23 (10%)	6 (6%)	17 (14%)	0.046
Diabetes	70 (30%)	21 (19%)	49 (40%)	<0.001
Charlson comorbidity index	1 (0-1)	0 (0-1)	1 (1-2)	<0.001
Characteristics on enrollment
ICU type				0.34
Medical	86 (37%)	37 (34%)	49 (40%)
Trauma-Surgical	145 (63%)	72 (66%)	73 (60%)
Sepsis	177 (77%)	77 (71%)	100 (82%)	0.045
Shock	92 (40%)	41 (38%)	51 (42%)	0.59
Mechanical ventilation	109 (47%)	49 (45%)	60 (49%)	0.60
SOFAb score	6 (3-10)	6 (3-9)	7 (3-10)	0.15
WBC (103 cells per μL)	17 (12-23)	17 (13-22)	18 (12-24)	0.51
Absolute PMNs (103 cells per μL)	14 (8-20)	12 (8-16)	17 (8-21)	0.17
Absolute monocytes (103 cells per μL)	0.90 (0.49-1.48)	0.68 (0.49-1.08)	0.99 (0.43-1.79)	0.41
sTREM-1 (pg/mL)	353 (243-553)	275 (204-451)	385 (285-585)	<0.001
Outcomes
28-day mortality	19 (8%)	5 (5%)	14 (11%)	0.091
Persistent organ dysfunction at day 7c	41 (18%)	16 (15%)	25 (21%)	0.25

^a.Age expressed as mean (standard deviation); other continuous variables expressed as median (interquartile range). Categorical variables expressed as number (%). P values reflect comparison of these variables between groups using t-tests, Wilcoxon rank sum and Fisher’s exact tests where appropriate.

^b.SOFA sequential organ failure assessment

^c.Persistent organ dysfunction includes need for vasopressors, dialysis, or invasive mechanical ventilation on day 7, among 106 patients ≤ 50 and 119 patients >50 who survived to day 7

The median concentration of sTREM-1 in the cohort overall was 353 pg/mL (interquartile range 243-553 pg/mL). To evaluate sTREM-1 as a general inflammatory marker, we examined its relationship to leukocyte counts on enrollment (Figure S1). sTREM-1 was somewhat correlated with white blood cell count (Spearman coefficient 0.375, P < 0.001), with median white blood cells being highest in the upper tertile of sTREM-1 (Figure S1). sTREM-1 also had some correlation with absolute PMNs (Spearman coefficient 0.363, P = 0.008) but not with absolute monocyte count.

sTREM-1 by age and age-related chronic disease

Adults over 50 years of age had significantly higher sTREM-1 concentrations on enrollment compared to younger adults (Figure 1). In contrast, white blood cell count and SOFA scores were not significantly higher among older adults. After adjusting for sex and CCI, age > 50 was still associated with 27% higher (Coef. 1.27, 95% CI 1.06, 1.53) sTREM-1 concentrations (Table 2). After noting higher proportion of sepsis among older adults, we adjusted for this and observed age was still significantly associated with higher sTREM-1 (Table S2).

Figure 1: Illness severity markers by age.

Log₂ transformed sTREM-1 concentrations, white blood cell count, and sequential organ failure assessment (SOFA) score by age group. Only sTREM-1 was significantly different by age group using Wilcoxon rank sum tests. *** P < 0.001.

Table 2:

Fold-change in soluble triggering receptor expressed on myeloid cells-1 (sTREM-1) associated with age greater than 50 years

Age group	Unadjusted		Adjusted for sex		Adjusted for sex and CCIa
	Coef. (95% CI)	P	Coef. (95% CI)	P	Coef. (95% CI)	P
Age ≤ 50 years	1.00 (reference)		1.00 (reference)		1.00 (reference)
Age > 50 years	1.33 (1.13, 1.57)	0.001	1.34 (1.13, 1.57)	0.001	1.27 (1.06, 1.53)	0.009

^a.CCI Charlson comorbidity index

We also examined differences in sTREM-1 by specific age-related chronic diseases in our cohort to see whether these explained higher sTREM-1 in older adults. sTREM-1 was higher in individuals with diabetes, coronary artery disease, and chronic kidney disease (Figure S2). However, further adjustment for these comorbidities showed similar relationships between older age and higher sTREM-1 (Coef. 1.27, 95% CI 1.08, 1.48, Table S2).

Unlike with groups stratified by age and age-related diseases, we did not see significant differences in sTREM-1 concentrations or in key clinical features by sex (Table S3).

sTREM-1 and ICU outcomes

We then examined the relationship between sTREM-1 and short-term ICU outcomes that could result from dysregulated early innate immune activation. Nonsurvivors were older in age, with more chronic disease and higher severity of illness by SOFA score (Table S4).

A two-fold increase in circulating sTREM-1 level was associated with a 2.42-fold higher risk of 28-day mortality (95% CI 1.57, 3.73), adjusting for age, sex, and CCI (Table 3). This association was similar in patients with and without sepsis. The strength of the association was modified by presence of shock, with sTREM-1 associated with over 4 times the risk death (RR 4.56, 95% CI 2.30, 9.02) among those with shock on ICU admission.

Table 3:

Relative risk of 28-day mortality associated with two-fold higher soluble triggering receptor expressed on myeloid cells-1 (sTREM-1), in total sample and stratified by key subgroups

Group	N	# deaths (%)	Unadjusted		Adjusted for sex		Adjusted for sex, CCIa		Adjusted for sex, CCIa, age
			RR (95% CI)	P	RR (95% CI)	P	RR (95% CI)	P	RR (95% CI)	P
All	231	19 (8%)	2.48 (1.73, 3.56)	<0.001	2.48 (1.70, 3.63)	<0.001	2.38 (1.58, 3.58)	<0.001	2.42 (1.57, 3.73)	<0.001
Sepsis
Sepsis	177	15 (8%)	2.60 (1.67, 4.06)	<0.001	2.66 (1.64, 4.31)	<0.001	2.55 (1.54, 4.24)	<0.001	2.54 (1.51, 4.30)	<0.001
No Sepsis	54	4 (8%)	2.39 (1.31, 4.34)	0.004	2.34 (1.31, 4.19)	0.004	2.43 (1.29, 4.57)	0.006	2.77 (1.17, 6.54)	0.020
Shock
Shock	92	11 (12%)	4.08 (2.36, 7.07)	<0.001	4.59 (2.26, 9.33)	<0.001	4.49 (2.18, 9.26)	<0.001	4.56 (2.30, 9.02)	<0.001
No Shock	139	8 (6%)	1.66 (1.00, 2.75)	0.049	1.61 (0.99, 2.62)	0.053	1.64 (0.94, 2.88)	0.084	1.58 (0.81, 3.11)	0.181

^a.CCI Charlson comorbidity index

Consistent with the aforementioned relationships, we saw that a two-fold increase in sTREM-1 was associated with a greater risk of persistent organ dysfunction (1.86 fold, 95% CI 1.45, 2.39) among those who survived at least 7 days after ICU admission (Table 4). This relationship was similar across most subgroups, except among patients without sepsis where there was no significant association.

Table 4:

Relative risk of persistent organ dysfunction^a associated with two-fold higher soluble triggering receptor expressed on myeloid cells-1 (sTREM-1), in total sample and stratified by key subgroups

Group	N	# events (%)	Unadjusted		Adjusted for sex		Adjusted for sex, CCIb		Adjusted for sex, CCIb, age
			RR (95% CI)	P	RR (95% CI)	P	RR (95% CI)	P	RR (95% CI)	P
All	225	41 (18%)	1.85 (1.44, 2.38)	<0.001	1.85 (1.44, 2.38)	<0.001	1.87 (1.46, 2.40)	<0.001	1.86 (1.45, 2.39)	<0.001
Sepsis
Sepsis	172	34 (20%)	2.00 (1.52, 2.64)	<0.001	2.00 (1.51, 2.64)	<0.001	2.01 (1.53, 2.65)	<0.001	2.01 (1.53, 2.66)	<0.001
No Sepsis	53	7 (13%)	1.10 (0.51, 2.40)	0.807	1.12 (0.55, 2.29)	0.755	1.20 (0.57, 2.55)	0.633	1.03 (0.43, 2.43)	0.952
Shock
Shock	86	24 (28%)	1.59 (1.05, 2.41)	0.030	1.58 (1.03, 2.43)	0.036	1.59 (1.01, 2.51)	0.043	1.61 (1.02, 2.56)	0.043
No Shock	139	16 (12%)	1.90 (1.32, 2.73)	<0.001	1.84 (1.31, 2.59)	<0.001	1.83 (1.34, 2.50)	<0.001	1.82 (1.34, 2.48)	<0.001

^a.persistent organ dysfunction includes need for vasopressors, dialysis, or invasive mechanical ventilation on day 7, among patients who survived to day 7

^b.CCI Charlson comorbidity index

sTREM-1 as a mediator of age-related differences in ICU outcomes

After observing that age was associated with higher sTREM-1 concentrations, and that sTREM-1 concentrations were associated with 28-day mortality and persistent organ dysfunction at 7 days, we had a theoretical basis to assess whether the inflammation reflected by sTREM-1 mediated age-related differences in ICU outcomes. The average indirect effect of age on death mediated through sTREM-1 was statistically significant (β = 0.028, 95% CI 0.009, 0.054, p = 0.002), while the remaining direct effect (i.e. through mechanisms other than sTREM-1) was not significant, albeit similar in effect size (Figure 2). The indirect effect of age through sTREM-1 on persistent organ dysfunction was also significant (β = 0.058, 95% CI 0.024, 0.108, p <0.001, Figure S3). These indirect effects remained significant in models that also adjusted for CCI and sepsis (Figures S4-5).

Figure 2: Mediation analysis of age, sTREM-1, and 28-day mortality.

Total effect of age on death decomposed into an indirect effect (i.e. effect mediated through sTREM-1) and direct effect (i.e. effect through mechanisms other than those reflected by sTREM-1). Age modeled as a binary variable, by ≤ 50 years and > 50 years. Models adjusted for sex.

Sensitivity analysis

We performed a sensitivity analysis using an alternative age threshold, stratifying patients by <60 and ≥60 years. Comorbidity burden, median sTREM-1, and mortality were higher among adults aged ≥60 years compared to those <60 years (Table S5). The association between age and sTREM-1, as well as the average indirect effects of age on both death and persistent organ dysfunction remained significant after adjustment for sex (Tables S6-S8). These effects were attenuated however when adjusting for CCI.

Discussion

In this cohort study we demonstrated that older age is associated with higher plasma sTREM-1 concentrations during critical illness. We then showed that sTREM-1 is a strong predictor of mortality and persistent organ dysfunction across groups reflecting varying illness severity in the ICU. Finally, we provided evidence that the relationship between older age and ICU outcomes may be mediated by innate immune dysregulation as reflected by sTREM-1. Altogether, these findings justify further investigation of how the TREM-1 pathway contributes to the mortality and organ dysfunction in older patients with critical illness.

Our study adds to the growing understanding of age-related defects in early innate immune responses during critical illness, and to our knowledge is the first description of how sTREM-1 varies by age among critically ill adults (33, 34). Possible reasons we observed higher sTREM-1 concentrations among older adults include the higher burden of chronic disease and sepsis in this age group. In addition to extensive evidence of TREM-1 activation and sTREM-1 release in sepsis, there is emerging research linking the pathway to multiple chronic age-related diseases. sTREM-1 and TREM-1 have been implicated in cancers, coronary artery disease, stroke, and kidney disease in prior work, and we similarly observed higher sTREM-1 concentrations among patients with chronic kidney disease, coronary artery disease, as well as diabetes (14, 35). On the other hand, our primary analysis showed age greater than 50 years was associated with significantly higher sTREM-1 even after adjusting for multiple chronic diseases and sepsis. This raises the possibility that sTREM-1 may also reflect age-related innate immune defects during critical illness that are independent of these factors. Supportive of this idea, one prior study compared TREM-1 mediated neutrophil function in cells isolated from elderly otherwise healthy participants to those from young adults (13). The study found that cells from elderly participants displayed alterations in TREM-1 mediated respiratory burst, TREM-1 signal transduction, and cell surface TREM-1 recruitment upon stimulation. The authors hypothesized that this altered recruitment in particular could contribute to higher sTREM-1 concentrations in older adults during times of TREM-1 activation. Apart from this work we are unaware of other studies directly investigating age-related changes in sTREM-1/TREM-1 in health or in illness. Additional mechanistic studies are needed to examine exactly how aging influences the sTREM-1/TREM-1 pathway.

We also demonstrated strong associations between sTREM-1 and ICU outcomes, which combined with the results of our mediation analysis, support closer investigation of the sTREM-1/TREM-1 pathway as a potential therapeutic target in critical illness among older adults. While the observational nature and limited sample size of our study can bias the mediation effect estimates, the results from our primary analysis remained robust to adjustment for chronic disease and sepsis, and consistent with the surrogate outcome of persistent organ dysfunction (37). Interestingly, previous work from a large ICU cohort found that other biomarkers of the innate immune response (e.g. interleukin-6, interleukin-8, interleukin-1β, and tumor necrosis factor) did not explain age-related differences in ICU mortality (38). Although their findings may have differed from ours because they focused on patients and plasma samples at the time of acute respiratory distress syndrome, we hypothesize sTREM-1 and TREM-1 reflect a distinct aspect of the innate immune response with greater contribution to age-related differences in ICU outcomes. A phase IIb trial testing the efficacy of a TREM-1 inhibitor among patients with septic shock is currently ongoing, and our work supports examining heterogeneity of treatment effect by age (19, 39). Altogether, our findings corroborate the value of sTREM-1 as a marker of deleterious outcomes, and suggest that older adults comprise an important subgroup that could benefit from investigational immune-modulating therapies for critical illness.

One strength of this work is that it extends the evidence base of sTREM-1 as a prognostic marker in sepsis to a more general ICU population. Although several prior investigations have showed that higher sTREM-1 levels are associated with higher risk of death, most had small sample sizes, limited characterization of comorbidities, and were restricted to sepsis and pneumonia (20, 21, 40). Our cohort included patients admitted for acute medical diagnoses, trauma, and post-surgical care, and we also found that sTREM-1 is associated with death among patients without sepsis or suspected infection. The prognostic value of sTREM-1 in critically ill patients without infection has not been clearly studied before, although one small study showed patients after elective cardiac surgery and out-of-hospital cardiac arrest without infection had similar sTREM-1 concentrations as patients with severe sepsis (41). Interestingly, whereas our subgroup of patients without sepsis represented the lower end of illness severity, the subgroup of shock was among the highest risk for death/organ dysfunction and here we observed an even stronger association between sTREM-1 and mortality. This relationship is consistent with the strong body of preclinical evidence illustrating that TREM-1 actually amplifies responses to inflammatory cytokines that are elevated in early septic shock (12, 42). Overall, we believe these findings are important in showing that sTREM-1 offers prognostic value across a wide range of diagnoses and illness severity.

Finally, we showed sTREM-1 was also associated with persistent organ dysfunction at 7 days. Prior work of sTREM-1 in organ dysfunction focused on mortality prediction among patients with either acute kidney injury or acute respiratory distress syndrome (43, 44). Traditionally research agendas have focused on prediction of mortality, but there is increasing support to develop other outcome measures (45). Organ dysfunction is a proximal surrogate for mortality risk, causes significant morbidity, and has been identified as a meaningful outcome in some patient populations (26, 46, 47).

Our study has several limitations. First, this study is only moderate in size and single-center, though it stands as one of the largest investigations of sTREM-1 among ICU patients. In particular, we had relatively few mortality events when stratified by subgroups and not all of the findings were replicated in the subgroup analysis of persistent organ dysfunction, so the finding of effect modification by shock should be validated in other cohorts. Second, our sensitivity analysis was underpowered with relatively few patients greater than 60 years, although the general effect of older age on sTREM-1 remained consistent. Third, our assessment of age-related comorbidities was largely based on patient/surrogate interview and chart review and can vary in completeness, though represent standard, common approaches (48). Finally, our study focuses only on plasma measurements of sTREM-1 as a marker of TREM-1 mediated inflammation, instead of direct measures of TREM-1 expression and activation, since such measurements are easier to translate into trials and clinical practice (19). It remains unclear which aspect of immune dysregulation downstream of TREM-1 contributes to poor outcomes, although we note sTREM-1 itself is unlikely to be a contributor; prior research suggests it may have a role as a decoy receptor to attenuate TREM-1’s effects (14, 19). Despite these limitations, we observed strong relationships we observed between older age and higher sTREM-1, as well as sTREM-1 and mortality and intend this work to be hypothesis-generating.

Conclusions

In conclusion, we found that older age is strongly associated with higher sTREM-1 concentrations during critical illness. sTREM-1 is also a powerful predictor of 28-day mortality and persistent organ dysfunction at day 7, which includes need for vasopressors, dialysis, and/or invasive mechanical ventilation. These data support further investigation of how sTREM-1/TREM-1 may influence age-related differences in outcomes during critical illness.