Abstract
Purpose
Sepsis causes significant worldwide morbidity and mortality. Inability to clear an infection and secondary infections are known complications in severe sepsis and likely result in worsened outcomes. We sought to characterize risk factors of these complications.
Methods
We performed a secondary analysis of clinical data from 401 subjects enrolled in the PHENOtyping sepsis-induced Multiple organ failure Study. We examined factors associated with prolonged infection, defined as infection that continued to be identified 7 days or more from initial identification, and secondary infection, defined as new infections identified ≥3 days from presentation. Multivariable adjustment was performed to examine laboratory markers of immune depression, with immunocompromised and immunocompetent subjects analyzed separately.
Results
Illness severity, immunocompromised status, invasive procedures, and site of infection were associated with secondary infection and/or prolonged infection. Persistent lymphopenia, defined as an absolute lymphocyte count (ALC)<1000 cells/μL twice in the first five days, and persistent neutropenia, defined as absolute neutrophil count (ANC)<1000 cells/μL twice in the first five days, were associated with secondary and prolonged infections. When adjusted in multivariable analysis, persistent lymphopenia remained associated with secondary infection in both immunocompromised (aOR=14.19, 95% CI [2.69, 262.22] and immunocompetent subjects (aOR=2.09, 95% CI [1.03, 4.17]). Persistent neutropenia was independently associated with secondary infection in immunocompromised subjects (aOR=5.34, 95% CI [1.92, 15.84]). Secondary and prolonged infections were associated with worse outcomes, including death.
Conclusions
Laboratory markers of immune suppression can be used to predict secondary infection. Lymphopenia is an independent risk factor in immunocompromised and immunocompetent patients for secondary infection.
Keywords: sepsis, secondary infection, prolonged infection, lymphopenia, neutropenia
Introduction
Sepsis is a major cause of morbidity and mortality worldwide, with an estimated 19.4 million cases globally each year accounting for 5.3 million deaths yearly.[1] Secondary infections further complicate sepsis admissions, with this problem particularly pronounced in low- and middle-income countries where the risk of developing such an infection is estimated to be twenty times higher than in high-income countries.[2] Part of the mortality burden of sepsis is likely a result of secondary infections, with those developing secondary infections shown to have higher mortality as well as longer intensive care unit (ICU) and hospital length of stay. One study estimated that ICU patients who developed secondary infections had 10.9% higher mortality than those who did not after adjustment for other factors.[3] Even less has been described regarding factors related to persistence of infection. Naturally, it stands to reason that inability to clear an infection would lead to longer hospital stays and potentially greater morbidity and mortality. Indeed, a classic focus of infectious diseases management is source control for this reason.
Prevention of secondary and prolonged infections in sepsis requires identification of targetable risk factors. Though some studies have identified patient characteristics associated with secondary infection, these have primarily been limited to baseline patient characteristics and necessary interventions, and hence not modifiable. It has been hypothesized that secondary infection may, in part, be due to development of a hypoimmune state that can develop during sepsis, and that targeting this by stimulating the immune system may reduce risk of secondary infections. However, the best means of easily identifying these patients at the bedside is unclear. Thus, we sought to identify characteristics associated with secondary and prolonged infection in sepsis patients in the hopes of identifying targetable factors, with the goal of facilitating therapies that might prevent these in future patients.
Methods
We performed a secondary analysis of clinical data from 401 subjects enrolled in the PHENOtyping sepsis-induced Multiple organ failure Study (PHENOMS).[4] This prospective cohort study enrolled children with severe sepsis between 2015 and 2017 across nine pediatric ICUs (PICU) in the Eunice Kennedy Shriver National Institute of Child Health and Human Development Collaborative Pediatric Critical Care Research Network. The goal of the current analysis was to identify risk factors for secondary infection and prolonged infection. Factors examined for associations included demographics, illness severity, immune deficiency, laboratory markers of immune depression, invasive procedures, and infection characteristics. Laboratory markers of immune depression consisted of early neutropenia, defined as at least one measured absolute neutrophil count (ANC) below 1000 cells/μL in Days 0 to 2, persistent neutropenia, defined as at least two measured ANCs below 1000 cells/μL in Days 0 to 5, and early and persistent lymphopenia defined similarly for absolute lymphocyte count (ALC). Associations with immunoparalysis-associated multiorgan failure (IPMOF) by Day 5 and lowest tumor necrosis factor (TNF-α) response in Days 0 to 2 were also examined. IPMOF was defined as having an organ failure index (OFI) of at least 2 on one study day and ex-vivo whole blood endotoxin-stimulated TNF-α response < 200 pg/mL at any time on Day 3 or later.
Infection classification
Primary infections were any infection identified on day of sepsis onset (defined as Day 1) or the following day. Secondary infections were new infections identified on Day 3 or later, consistent with CDC definition of nosocomial infection.[5] Prolonged infections were any infection that continued to be identified by culture or polymerase chain reaction (PCR) 7 days or more from initial identification. The majority of prolonged infections arose from primary infections, but secondary infections and infections identified prior to sepsis onset could also develop into prolonged infections. Detailed description of infection classification by culture type is described in the Online Resource (eMethods), but in general, positive tests from normally sterile sites were deemed infection, while tests from non-sterile sites depended on density of growth, likelihood of organism pathogenicity at a given body site, and additional information as available. As some sites could be indicative of either infection or colonization / shedding (i.e tracheal aspirates) a sub-analysis was performed including only results that would be considered definitively indicative of infection, specifically abscesses, blood (potential contaminants removed), peritoneal fluid, pleural fluid, spinal fluid, surgical sites, and urine.
Statistical Methods
Tables throughout report medians and interquartile ranges to summarize continuous data while counts and percentages summarize categorical. Non-parametric tests such as Fisher’s exact and Wilcoxon rank-sum test were used to compare demographic, illness severity, immune deficiency markers, and other clinical measures across two outcomes of interest, namely secondary infection and prolonged infection. No adjustment for multiple comparisons was made with alpha < 0.05 used to denote statistical significance.
In multivariable analyses, a step-wise regression approach to model determination was used with entry and stay p-value criteria of 0.20 and 0.05, respectively (Online Resource Tables E1–E2). Immunocompromised status was entered as a forced predictor and was found to be associated with both outcomes. Subsequent analyses investigating associations for primary predictors lymphopenia, neutropenia, IPMOF, and TNF-α response were conducted separately for immunocompetent and immunocompromised patients. Estimates and 95% confidence intervals (CIs) for primary predictors were adjusted for potential confounders as determined by the model determination phase, see footnotes for adjustment specifics.
Microbiology classification was performed by the primary author as described above while blinded to outcomes. All analyses were performed using R version 4.2.1 and SAS 9.4 (SAS Institute; Cary, NC) at the Data Coordinating Center, University of Utah.
Results
Infection characteristics
Of the 401 subjects included, 263 (66%) of subjects had an identified primary infection and the remainder had suspected infection, with 179 (45%) having bacterial infection, 133 (33%) viral, and 4 (1%) fungal; 53 (13%) subjects had primary co-infection (Table 1, E3). There were 68 (17%) who developed secondary infection, and 42 (10%) had prolonged infection; 18 (4%) of these had both secondary and prolonged infection. The most common specimen type for identification of primary infection was nasopharyngeal viral testing (28%), followed by blood (19%), and tracheal aspirate (15%). No associations were observed with type or site of primary infection for secondary infections. For those with prolonged infection, primary bacterial infection, primary viral infection and co-infection were more common than for those without prolonged infection, and they were more likely to have positive tracheal aspirate cultures for primary infectious source as well. When only definitive infections were considered, only identified primary infection and blood as site were associated with prolonged infection. (Online Resource Table E4)
Table 1.
Primary infection characteristics by secondary and prolonged infection
| Secondary Infection | Prolonged Infection | ||||||
|---|---|---|---|---|---|---|---|
| Overall (N = 401) | No (N = 333) | Yes (N = 68) | P-value1 | No (N = 359) | Yes (N = 42) | P-value1 | |
| PRIMARY INFECTION (first 48 hours) | |||||||
| Primary infection | 263 (66%) | 217 (65%) | 46 (68%) | 0.7802 | 223 (62%) | 40 (95%) | <.0012 |
| Bacterial | 179 (45%) | 150 (45%) | 29 (43%) | 0.7892 | 154 (43%) | 25 (60%) | 0.0492 |
| Gram positive | 109 (27%) | 89 (27%) | 20 (29%) | 0.6552 | 95 (26%) | 14 (33%) | 0.3612 |
| Gram negative | 98 (24%) | 85 (26%) | 13 (19%) | 0.2832 | 83 (23%) | 15 (36%) | 0.0872 |
| Viral | 133 (33%) | 107 (32%) | 26 (38%) | 0.3272 | 108 (30%) | 25 (60%) | <.0012 |
| Fungal | 4 (1%) | 3 (1%) | 1 (1%) | 0.5262 | 2 (1%) | 2 (5%) | 0.0562 |
| Co-infection | 53 (13%) | 43 (13%) | 10 (15%) | 0.6952 | 41 (11%) | 12 (29%) | 0.0062 |
| SITE OF INFECTION (first 48 hours) | |||||||
| Abscess | 8 (2%) | 7 (2%) | 1 (1%) | 1.0002 | 8 (2%) | 0 (0%) | 1.0002 |
| Blood | 75 (19%) | 60 (18%) | 15 (22%) | 0.4952 | 65 (18%) | 10 (24%) | 0.4022 |
| Bronchial brush | 5 (1%) | 5 (2%) | 0 (0%) | 0.5942 | 4 (1%) | 1 (2%) | 0.4272 |
| Bronchoalveolar lavage | 15 (4%) | 12 (4%) | 3 (4%) | 0.7272 | 11 (3%) | 4 (10%) | 0.0602 |
| Nasopharyngeal | 113 (28%) | 96 (29%) | 17 (25%) | 0.5582 | 96 (27%) | 17 (40%) | 0.0702 |
| Peritoneal fluid | 3 (1%) | 2 (1%) | 1 (1%) | 0.4282 | 1 (0%) | 2 (5%) | 0.0302 |
| Pleural fluid | 9 (2%) | 9 (3%) | 0 (0%) | 0.3682 | 8 (2%) | 1 (2%) | 1.0002 |
| Spinal fluid | 4 (1%) | 4 (1%) | 0 (0%) | 1.0002 | 3 (1%) | 1 (2%) | 0.3592 |
| Sputum | 13 (3%) | 10 (3%) | 3 (4%) | 0.4692 | 10 (3%) | 3 (7%) | 0.1452 |
| Stool / Rectal | 15 (4%) | 13 (4%) | 2 (3%) | 1.0002 | 15 (4%) | 0 (0%) | 0.3842 |
| Surgical site | 4 (1%) | 3 (1%) | 1 (1%) | 0.5262 | 3 (1%) | 1 (2%) | 0.3592 |
| Tracheal Aspirate | 59 (15%) | 47 (14%) | 12 (18%) | 0.4552 | 46 (13%) | 13 (31%) | 0.0042 |
| Urine | 12 (3%) | 9 (3%) | 3 (4%) | 0.4362 | 10 (3%) | 2 (5%) | 0.3642 |
| Vascular catheter | 1 (0%) | 1 (0%) | 0 (0%) | 1.0002 | 1 (0%) | 0 (0%) | 1.0002 |
| Wound (non-surgical) | 6 (1%) | 5 (2%) | 1 (1%) | 1.0002 | 6 (2%) | 0 (0%) | 1.0002 |
Patients with not available (NA) antimicrobial timing or coverage data are not included in p-value calculation.
Fisher’s exact test.
Secondary infections and prolonged infections varied considerably (Online Resource Tables E5–E8). The majority of secondary infections were bacterial (56%), followed by viral (29%). The most common specimen for secondary infections was the blood (35%), followed by tracheal aspirates (24%). Prolonged infections were divided nearly equally between bacterial (49%) and viral (44%). The most common specimens for prolonged infection were from blood (23%) and respiratory (61%), with bronchoalveolar lavage (23%), tracheal aspirates (18%), and nasopharyngeal testing (16%) comprising the majority of these.
Clinical characteristics
Univariable analysis revealed no differences in demographic characteristics across outcomes (Table 2). While pediatric risk of mortality (PRISM) scores did not differ between those with or without secondary or prolonged infection, a higher number of organ failures on Day 0 was observed for those with secondary infection (Table 2). Overall, 104 (26%) of the population were immunocompromised, and subjects with secondary infection were more likely to be immunocompromised. All bedside markers of immune depression as well as IPMOF were more common in those with secondary infection. For those with prolonged infection, malignancy, congenital immunodeficiency, early and persistent neutropenia, persistent lymphopenia, and IPMOF were more common. In terms of resource utilization, those with secondary and prolonged infection were more likely to require multiple procedures, though mechanical ventilation was common across the population and did not differ for either group. When only definitive infections were considered, similar associations between secondary/prolonged infection remained with early/persistent neutropenia and lymphopenia and IPMOF except for early neutropenia and IPMOF with secondary infection (Online Resource Table E9).
Table 2.
Clinical characteristics and markers of immune deficiency by secondary and prolonged infection
| Secondary Infection | Prolonged Infection | ||||||
|---|---|---|---|---|---|---|---|
| Overall (N = 401) | No (N = 333) | Yes (N = 68) | P-value1 | No (N = 359) | Yes (N = 42) | P-value1 | |
| DEMOGRAPHICS | |||||||
| Male | 222 (55%) | 185 (56%) | 37 (54%) | 0.8942 | 194 (54%) | 28 (67%) | 0.1412 |
| Race | 0.0902 | 0.5422 | |||||
| White | 270 (67%) | 229 (69%) | 41 (60%) | 241 (67%) | 29 (69%) | ||
| Black or African American | 83 (21%) | 66 (20%) | 17 (25%) | 75 (21%) | 8 (19%) | ||
| Multiracial | 3 (1%) | 1 (0%) | 2 (3%) | 2 (1%) | 1 (2%) | ||
| Other | 19 (5%) | 15 (5%) | 4 (6%) | 17 (5%) | 2 (5%) | ||
| Ethnicity | 0.4762 | 1.0002 | |||||
| Unknown or Not Reported | 14 (3%) | 13 (4%) | 1 (1%) | 13 (4%) | 1 (2%) | ||
| Hispanic or Latino | 66 (16%) | 57 (17%) | 9 (13%) | 59 (16%) | 7 (17%) | ||
| Not Hispanic or Latino | 321 (80%) | 263 (79%) | 58 (85%) | 287 (80%) | 34 (81%) | ||
| ILLNESS SEVERITY MEASURES | |||||||
| PRISM2 | 8 [3, 14] | 8 [3, 14] | 9 [3, 16] | 0.1943 | 8 [3, 14] | 10 [5, 17] | 0.1983 |
| Organ failures on Day 0 | 0.0142 | 0.0562 | |||||
| 0 | 3 (1%) | 3 (1%) | 0 (0%) | 3 (1%) | 0 (0%) | ||
| 1 | 175 (44%) | 157 (47%) | 18 (26%) | 161 (45%) | 14 (33%) | ||
| 2 | 152 (38%) | 119 (36%) | 33 (49%) | 134 (37%) | 18 (43%) | ||
| 3 | 57 (14%) | 45 (14%) | 12 (18%) | 52 (14%) | 5 (12%) | ||
| 4 | 8 (2%) | 5 (2%) | 3 (4%) | 5 (1%) | 3 (7%) | ||
| 5 | 6 (1%) | 4 (1%) | 2 (3%) | 4 (1%) | 2 (5%) | ||
| IMMUNE DEFICIENCY MARKERS | |||||||
| Immunocompromised | 104 (26%) | 79 (24%) | 25 (37%) | 0.0332 | 88 (25%) | 16 (38%) | 0.0642 |
| Bone marrow or stem cell transplantation | 23 (6%) | 14 (4%) | 9 (13%) | 0.0082 | 18 (5%) | 5 (12%) | 0.0802 |
| Solid organ transplant | 8 (2%) | 7 (2%) | 1 (1%) | 1.0002 | 8 (2%) | 0 (0%) | 1.0002 |
| Malignancy at ICU admission | 49 (12%) | 39 (12%) | 10 (15%) | 0.5422 | 39 (11%) | 10 (24%) | 0.0232 |
| Congenital immunodeficiency | 12 (3%) | 7 (2%) | 5 (7%) | 0.0362 | 7 (2%) | 5 (12%) | 0.0052 |
| Rheumatologic disease | 7 (2%) | 4 (1%) | 3 (4%) | 0.0982 | 5 (1%) | 2 (5%) | 0.1602 |
| Sickle cell disease | 3 (1%) | 3 (1%) | 0 (0%) | 1.0002 | 3 (1%) | 0 (0%) | 1.0002 |
| Steroid use (chronic or acute) | 59 (15%) | 45 (14%) | 14 (21%) | 0.1362 | 52 (14%) | 7 (17%) | 0.6502 |
| Neutropenia (ANC<1000 on Days 0–2) | 64 (16%) | 47 (14%) | 17 (25%) | 0.0302 | 52 (14%) | 12 (29%) | 0.0252 |
| Neutropenia (twice ANC<1000 on Days 0–5) | 38 (9%) | 23 (7%) | 15 (22%) | <.0012 | 28 (8%) | 10 (24%) | 0.0032 |
| Lymphopenia (ALC<1000 on Days 0–2) | 209 (52%) | 164 (49%) | 45 (66%) | 0.0122 | 185 (52%) | 24 (57%) | 0.5182 |
| Lymphopenia (twice ALC<1000 on Days 0–5) | 152 (38%) | 110 (33%) | 42 (62%) | <.0012 | 130 (36%) | 22 (52%) | 0.0452 |
| Confirmed IPMOF (Days 0–5) | 61 (15%) | 45 (14%) | 16 (24%) | 0.0422 | 49 (14%) | 12 (29%) | 0.0202 |
| Lowest TNF–alpha (Days 0–2) n, median [q1, q3] | 251, 364 [86, 928] | 215, 364 [93, 857] | 36, 340 [50, 1049] | 0.9993 | 230, 368 [86, 944] | 21, 290 [107, 623] | 0.7653 |
| INVASIVE PROCEDURES (first 48 hours) | |||||||
| CRRT | 32 (8%) | 21 (6%) | 11 (16%) | 0.0122 | 27 (8%) | 5 (12%) | 0.3602 |
| Plasma exchange | 9 (2%) | 4 (1%) | 5 (7%) | 0.0092 | 9 (3%) | 0 (0%) | 0.6062 |
| Mechanical ventilation (invasive or non-invasive) | 357 (89%) | 293 (88%) | 64 (94%) | 0.2002 | 316 (88%) | 41 (98%) | 0.0672 |
| ECMO | 22 (5%) | 15 (5%) | 7 (10%) | 0.0752 | 16 (4%) | 6 (14%) | 0.0192 |
| Transfusion, RBC | 150 (37%) | 118 (35%) | 32 (47%) | 0.0752 | 128 (36%) | 22 (52%) | 0.0432 |
| Transfusion, platelets | 95 (24%) | 70 (21%) | 25 (37%) | 0.0082 | 72 (20%) | 23 (55%) | <.0012 |
PRISM: pediatric risk of mortality; ICU: intensive care unit; ANC: absolute neutrophil count, cells/μL; ALC: absolute lymphocyte count, cells/μL; IPMOF: immunoparalysis-associated multiorgan failure; TNF: tumor necrosis factor; CRRT: continuous renal replacement therapy; ECMO: extracorporeal membrane oxygenation; RBC: red blood cells.
Patients with not available (NA) antimicrobial timing or coverage data are not included in p-value calculation.
Fisher’s exact test.
Wilcoxon rank-sum test.
Association with outcomes and resource utilization
Development of secondary infection and prolonged infection was associated with several clinical outcomes (Online Resource Table E10). Those with secondary infection or prolonged infection were more likely to die in the PICU (24% vs 8%, 29% vs 9% respectively); have fewer organ failure-free days (median 11 vs 22, 8 vs 21), PICU-free days (3 vs 17, 0 vs 17), and vent-free days (6 vs 19, 0 vs 19); and experience prolonged multiorgan dysfunction syndrome (MODS) (53% vs 20%, 64% vs 21%) [p<0.001 for all comparisons]. Both groups were more likely to require continuous renal replacement therapy (CRRT) (32% vs 9%, 31% vs 11% [p<0.001 for both]) and extracorporeal membrane oxygenation (ECMO) (15% vs 6% [p=0.021], 21% vs 6% [p=0.002]). Additionally, they were more likely to exhibit macrophage activation syndrome (16% vs 4% [p<0.001]; 14% vs 5% [p=0.029]) and develop thrombocytopenia-associated multiple organ failure (TAMOF) (28% vs 5% [p<0.001], 31% vs 7% [p<0.001]). All associations remained in the sub-analysis when only definitive infections were considered except for macrophage activation syndrome (MAS) and sequential liver failure associated multiorgan failure (SMOF) with secondary infection (Online Resource Table E11).
Multivariable models
When multivariable analysis was performed, secondary infection was found to be independently associated with plasma exchange, immunocompromised status, and number of organ failures on Day 0 (Online Resource Table E12). Prolonged infection was found to be independently associated with identified primary infection, platelet transfusion, mechanical ventilation, malignancy at ICU admission, positive tracheal aspirate culture, immunocompromised status, and congenital immunodeficiency (Online Resource Table E13).
Association of markers of immune depression with outcomes
Given that immunocompromised status was independently associated with both outcomes of interest, and an assumption that cytopenias would be more common in immunocompromised patients, subjects were divided into immunocompetent and immunocompromised for examining markers of immune depression related to outcomes. In univariable analysis (Online Resource Table E14), immunocompetent subjects with persistent neutropenia were more likely to have prolonged infection (OR=7.65, 95% CI [1.85, 28.78]). Those with persistent lymphopenia were more likely to develop secondary infection (OR=2.14, 95% CI [1.08, 4.16]). When adjusted in multivariable analysis, persistent lymphopenia remained associated with development of secondary infection (OR=2.09, 95% CI [1.03, 4.17]) (Table 3).
Table 3.
Adjusted estimates for primary predictors
| Immunocompetent | Immunocompromised | |||||||
|---|---|---|---|---|---|---|---|---|
| Secondary Infection1 | Prolonged Infection2 | Secondary Infection1 | Prolonged Infection2 | |||||
| Odds ratio (95% CI) | P-value | Odds ratio (95% CI) | P-value | Odds ratio (95% CI) | P-value | Odds ratio (95% CI) | P-value | |
| Neutropenia (ANC<1000 on Days 0–2) | 0.78 (0.17, 2.53) | 0.705 | 2.39 (0.63, 7.77) | 0.188 | 2.99 (1.14, 8.19) | 0.026 | 0.93 (0.28, 3.00) | 0.898 |
| Neutropenia (ANC<1000 on Days 0–5) | 1.16 (0.15, 5.34) | 0.866 | 4.50 (0.85, 20.98) | 0.075 | 5.34 (1.92, 15.84) | 0.001 | 1.02 (0.29, 3.52) | 0.980 |
| Lymphopenia (ALC<1000 on Days 0–2) | 1.55 (0.79, 3.05) | 0.198 | 0.72 (0.28, 1.76) | 0.473 | 2.57 (0.75, 12.09) | 0.142 | 0.37 (0.08, 1.75) | 0.201 |
| Lymphopenia (ALC<1000 on Days 0–5) | 2.09 (1.03, 4.17) | 0.042 | 1.15 (0.44, 2.81) | 0.768 | 14.19 (2.69, 262.22) | <.001 | 0.64 (0.13, 3.13) | 0.568 |
| Confirmed IPMOF (Days 0–5) | 1.85 (0.69, 4.55) | 0.213 | 1.84 (0.54, 5.49) | 0.313 | 0.91 (0.30, 2.57) | 0.862 | 2.14 (0.64, 7.04) | 0.211 |
| Lowest TNF–alpha (Days 0–2) 3 | 1.00 (1.00, 1.00) | 0.402 | 1.00 (1.00, 1.00) | 0.809 | -- | -- | 1.00 (1.00, 1.00) | 0.063 |
Results are based on models adjusting for plasma exchange and the number of organ failures on Day 0.
Results are based on models adjusting for primary infection, platelets transfusion, and tracheal aspirate sample collection.
For immunocompromised patients, there were insufficient counts to support the multivariable logistic regression model with secondary infection as outcome.
In univariable analysis (Online Resource Table E15), immunocompromised subjects with persistent neutropenia were more likely to develop secondary infection (OR=4.62, 95% CI [1.77, 12.38]). Similar trends were seen for persistent lymphopenia (OR=17.22, 95% CI [3.37, 315.26]. In multivariable analysis, persistent neutropenia and persistent lymphopenia remained significantly associated with secondary infection (OR=5.34, 95% CI [1.92, 15.84], OR=14.19, 95% CI [2.69, 262.22], respectively) (Table 3).
Evolution in time
Lastly, longitudinal examination of these groups revealed that differences in ALC and outcomes developed early and persisted. Figure 1 demonstrates that those who developed secondary infection presented with lower lymphocyte counts than those that did not (p=0.008). Additionally, while those without secondary infection have rising counts over the following five days, those with secondary infection continued with significantly lower counts, (p=0.015 to <0.001). These differences remained when only definitive infections were considered (Online Resource Figure E1). Conversely, this trend was less pronounced for neutrophil counts, with only some days showing significant differences. When examined according to immunocompromised status, the trends for lymphopenia were present in both groups, while differences in ANC were only present in the immunocompromised group. (Online Resource Figures E2 and E3)
Fig. 1. ALC/ANC over time by secondary infection.
Comparing the lowest lymphocyte counts (ALCs) and lowest neutrophil counts (ANCs) over the first six calendar days since sepsis onset among patients with (light gray) and without (dark gray) secondary infection. Differences in lowest lymphocyte and neutrophil counts at each day were compared using the Wilcoxon rank-sum test. Resulting p-values are reported. Red lines indicate the threshold used to designate early and persistent lymphopenia/neutropenia.
Moreover, those with persistent lymphopenia developed secondary infection throughout the first 4 weeks, compared to those without persistent lymphopenia who rarely developed secondary infection and primarily in the first 2 weeks, if ever (Figure 2A), a difference that was maintained when only definitive infections were considered (Figure E4). A similar trend was seen for those with persistent neutropenia (Figure 2B).
Fig. 2. Time to secondary infection by persistent lymphopenia status.
Panel A and B show Kaplan-Meir curves of the percentage of patients that do not have secondary infection over time. Panel A compares rates by persistent lymphopenia and Panel B by persistent neutropenia. The hazard ratio (95% CI) for early and persistent lymphopenia compared to those without was 2.64 (1.62, 4.31), p-value <0.0001 from the log-rank test. Similarly, persistent neutropenia showed 2.77 (1.56, 4.91), p-value 0.0003.
Discussion
This analysis demonstrated that routinely available markers of immune depression can be used to predict secondary infection and prolonged infection in children with sepsis. Moreover, we observed that persistent lymphopenia, defined as an ALC < 1000 twice in the first 5 days, was independently associated with secondary infection in both immunocompromised and immunocompetent patients, while persistent neutropenia, defined as an ANC < 1000 twice in the first 5 days, was independently associated with secondary infection in immunocompromised patients, after controlling for other factors. Patients who developed secondary infections were observed to have significantly lower lymphocyte counts throughout the first 5 days of admission compared to those who did not develop secondary infections. This trend remained even when patients were subdivided by immunocompromised status. Moreover, those with persistent lymphopenia in the first five days appeared to be at risk of developing secondary infections throughout the first 4 weeks, compared to those without persistent lymphopenia who primarily developed secondary infections within the first two weeks and at a much lower rate. To our knowledge, this is the largest study in pediatric patients with sepsis identifying lymphopenia as an independent risk factor related to development of new, secondary infections in an acute setting, and is also notable in demonstrating it as such in both immunocompromised and immunocompetent patients separately. As a secondary analysis of a multicenter study with limited exclusion criteria, its results should be widely generalizable.
That neutropenia is a risk factor for secondary infection is unsurprising. There has been a longstanding recognition that neutropenic patients are at significantly increased risk of infection, with the variety of potential pathogens increasing with the duration of neutropenia. It is for this reason that oncology patients with fever and neutropenia are advised to present to the nearest emergency room, and the workup expands considerably if no source is identified on initial investigations. Prolonged lymphopenia has been recognized since early in the AIDS epidemic as placing patients at risk of certain opportunistic infections, and similar concerns plague transplant recipients. However, the role of shorter periods of lymphopenia in contributing to infection risk has remained in question. Even so, lymphopenia has long been recognized as a frequent hematologic abnormality in patients with sepsis [6–11], and this appears to be a result of lymphocyte apoptosis [8,12–14]. Moreover, there has been a clear association with worse outcomes. For instance, a large retrospective study of adults with community-acquired pneumonia observed that 39% had lymphopenia, and that those patients were more likely to require ICU admission, had longer hospital length of stay, and died more often [11]. Another study of adult sepsis patients with bacteremia also found that lymphopenia was an independent predictor of mortality [7]. It is certainly possible that lymphopenia in these situations is merely a marker of disease severity, rather than playing a direct role in outcomes. It has at times even been proposed that lymphopenia in sepsis may be protective, preventing excessive inflammation and secondary injury [13]. However, this has been called into question by animal model studies demonstrating that prevention of lymphopenia actually results in improved outcomes in sepsis [12,15].
Whether lymphopenia contributes directly to worse outcomes is not definitive at this point, and the mechanisms by which it may are even less certain and potentially multifactorial. Some consideration has been given to secondary damage simply caused by immune dysregulation itself that may be related to lymphopenia. However, as our results show, persistent lymphopenia was itself an independent risk factor for secondary infection, regardless of immune status. Additionally, the majority of secondary infections were bacterial, which is perhaps surprising given that lymphocytes are not classically associated with combatting bacterial infections. The apparent effect of lymphopenia develops early and is sustained as well, given that these patients developed secondary infections at rates higher than those without lymphopenia from about a week after admission onward.
The best means of applying this information in clinical trials is a matter of great interest. Some investigators have demonstrated improved outcomes in animal models of sepsis by blocking mechanisms of lymphocyte apoptosis, specifically caspase-mediated apoptosis [12]. However, the investigators themselves note that there are a multitude of caspases, and it is not clear which one should be targeted to improve outcomes, if any at all. Additionally, as we and many others have observed, a significant proportion of patients with sepsis present with lymphopenia, potentially limiting the usefulness of anti-apoptotic agents. The role of agents to restore lymphocyte counts, such as IL-7 [16] or IL-2 [17], remains uncertain at this point and requires further study. Part of this uncertainty lies in the need to ensure that potential therapies are targeted appropriately. Many promising sepsis therapeutics have failed to show efficacy in large trials; there has been growing suspicion that many of these failures were due to the heterogeneity of sepsis and the need to individualize these treatments.
Until such investigations can be completed, our newfound demonstration that the presence of ongoing lymphopenia places patients at risk of secondary infection can be used to inform management of patients with sepsis. Specifically, this easily identifiable finding should prompt a lower threshold for evaluating for new infection and potentially beginning empiric coverage in the setting of new fever. It is noteworthy that the ALCs of patients who developed secondary infections were observed to be lower than those who did not from the day of sepsis onset, making these patients identifiable early. These results also demonstrate that patients with persistent lymphopenia are at higher risk than otherwise might be expected and should bolster adherence to hygiene prevention measures to reduce risk of introducing pathogens.
Application of these results to addressing prolonged infection is perhaps less straightforward. Lymphopenia and neutropenia were both associated with prolonged infection, but these associations did not hold in multivariable analysis. Moreover, when restricted to only definitive infections in the sub-analysis, there were only 13 subjects with prolonged infections, significantly limiting any conclusions. Furthermore, approximately half of these were comprised of viremias, which are already known to improve more slowly in some cases. Thus, in this scenario, these cytopenias can at best be viewed as markers of severity and perhaps as an indication for more aggressive therapies aimed at the initial infection itself, such as source control or, if applicable, reduction of immune suppression.
There are several limitations to our results. As a secondary analysis of a prior study, data collection was not designed for the purposes of our analysis, leading to some imprecision. For instance, we attempted to assess the effect of timely and appropriate antibiotic treatment by comparing sepsis onset to first administration of antibiotics, as well as comparing susceptibility profiles of organisms to empiric regimens, where appropriate. Unfortunately, only dates without times were available for these comparisons, severely limiting the analysis. There was also a lack of clinical information surrounding tracheal aspirate cultures, significantly impacting the adjudication of these results as infections or not, while determination of blood culture contaminants and ongoing viral shedding can be challenging. To address this concern, a sub-analysis excluding these types of results was performed in order to examine only definitive infections. Although this also likely eliminated true infections as well, the associations observed remained largely unchanged. Additionally, it is possible that some infections classified as early secondary infections may have simply been primary infections diagnosed belatedly due to focus on potential bacterial etiologies. As most secondary infections occurred at least several days into the hospital course though, the effects of this are likely to be small. Lastly, the observational nature of this secondary analysis precludes attribution of true causality.
Conclusion
Illness severity, immunocompromised status, invasive procedures, and site of infection were associated with secondary infection and/or prolonged infection in pediatric sepsis. Bedside markers of immune suppression can be used to predict secondary infection, specifically lymphopenia (ALC < 1000) in immunocompromised and immunocompetent patients and neutropenia (ANC < 1000) in immunocompromised patients. Strategies targeting lymphopenic and neutropenic patients warrant investigation to prevent secondary infection in pediatric sepsis.
Supplementary Material
Acknowledgments:
Supported, in part, by grant R01GM108618 (to Dr. Carcillo) from the National Institute of General Medical Sciences, and by 5U01HD049934-10S1 (to Dr Carcillo), from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services and the following cooperative agreements: U10HD049983, U10HD050096, U10HD049981, U10HD063108, U10HD063106, U10HD063114, U10HD050012, and U01HD049934.
The authors would like to thank Dr. Gregory Storch for careful reading and feedback on the manuscript.
Statements and Declarations
Zachary Aldewereld and Brendan Connolly have no conflicts of interest. Drs Carcillo, Berg, Wessel, Pollack, Meert, Hall, Newth, Doctor, Cornell, Harrison, Zuppa, Reeder, Banks, Holubkov, and Dean received support for article research from the NIH. Dr. Carcillo’s institution also received funding from the National Institute of General Medical Sciences. Dr. Pollack disclosed that his research is supported by philanthropy from Mallinckrodt Pharmaceuticals. Dr. Hall received funding from LaJolla Pharmaceuticals (service as a consultant), unrelated to the current submission. Dr. Newth received funding from Philips Research North America. Dr. Doctor’s institution received funding from the Department of Defense and Kalocyte. Dr. Cornell disclosed he is co-founder of Pre-Dixon Bio. Dr. Holubkov received funding from Pfizer (Data Safety Monitoring Board [DSMB] member), Medimmune (DSMB member), Physicians Committee for Responsible Medicine (biostatistical consulting), DURECT Corporation (biostatistical consulting), Armaron Bio (DSMB past member), and St Jude Medical (DSMB past member). The remaining authors have disclosed that they do not have any potential conflicts of interest.
References
- 1.Fleischmann C, Scherag A, Adhikari NKJ, Hartog CS, Tsaganos T, Schlattmann P, et al. Assessment of global incidence and mortality of hospital-treated sepsis current estimates and limitations. Am J Respir Crit Care Med. 2016;193:259–72. [DOI] [PubMed] [Google Scholar]
- 2.Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D, Kievlan DR, et al. Global, regional, and national sepsis incidence and mortality, 1990–2017: analysis for the Global Burden of Disease Study. The Lancet. 2020;395:200–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.van Vught LAV, Klouwenberg PMCK, Spitoni C, Scicluna BP, Wiewel MA, Horn J, et al. Incidence, risk factors, and attributable mortality of secondary infections in the intensive care unit after admission for sepsis. JAMA - Journal of the American Medical Association. 2016;315:1469–79. [DOI] [PubMed] [Google Scholar]
- 4.Carcillo JA, Berg RA, Wessel D, Pollack M, Meert K, Hall M, et al. A Multicenter Network Assessment of Three Inflammation Phenotypes in Pediatric Sepsis-Induced Multiple Organ Failure. Pediatric Critical Care Medicine. 2019;20:1137–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.CDC, Ncezid, DHQP. Identifying Healthcare-associated Infections (HAI) for NHSN Surveillance. 2023. [Google Scholar]
- 6.Felmet KA, Hall MW, Clark RSB, Jaffe R, Carcillo JA. Prolonged Lymphopenia, Lymphoid Depletion, and Hypoprolactinemia in Children with Nosocomial Sepsis and Multiple Organ Failure. The Journal of Immunology. 2005;174:3765–72. [DOI] [PubMed] [Google Scholar]
- 7.Drewry A, Samra N, Skrupky L, Fuller B, Compton S, Hotchkiss R. Persistent lymphopenia after diagnosis of sepsis predicts mortality. Shock. 2014;42:383–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Girardot T, Rimmelé T, Venet F, Monneret G. Apoptosis-induced lymphopenia in sepsis and other severe injuries. Apoptosis. Springer New York LLC; 2017. p. 295–305. [DOI] [PubMed] [Google Scholar]
- 9.Remy S, Kolev-Descamps K, Gossez M, Venet F, Demaret J, Javouhey E, et al. Occurrence of marked sepsis-induced immunosuppression in pediatric septic shock: a pilot study. Ann Intensive Care. 2018;8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sheikh H, Vahedi M, Bagheri A, Jahanshir A, Seyedhosseini J, Vahidi E. Association of Lymphopenia with Short Term Outcomes of Sepsis Patients; a Brief Report [Internet]. Arch Acad Emerg Med. 2019. Available from: www.jemerg.com [PMC free article] [PubMed] [Google Scholar]
- 11.Cilloniz C, Peroni HJ, Gabarrús A, García-Vidal C, Pericàs JM, Bermejo-Martin J, et al. Lymphopenia Is Associated With Poor Outcomes of Patients With Community-Acquired Pneumonia and Sepsis. Open Forum Infect Dis. 2021;8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hotchkiss RS, Tinsley KW, Swanson PE, Chang KC, Cobb JP, Buchman TG, et al. Prevention of lymphocyte cell death in sepsis improves survival in mice [Internet]. 1999. Available from: www.pnas.org [DOI] [PMC free article] [PubMed]
- 13.Hotchkiss RS, Coopersmith CM, Karl IE. Prevention of Lymphocyte Apoptosis-A Potential Treatment of Sepsis? 2005. [DOI] [PubMed] [Google Scholar]
- 14.Lai Y, Li X, Li T, Li X, Nyunoya T, Chen K, et al. Protein arginine N-methyltransferase 4 (PRMT4) contributes to lymphopenia in experimental sepsis. Thorax. 2022;thoraxjnl-2021-217526. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lv ZY, Shi YL, Bassi GS, Chen YJ, Yin LM, Wang Y, et al. Electroacupuncture at ST36 (Zusanli) Prevents T-Cell Lymphopenia and Improves Survival in Septic Mice. J Inflamm Res. 2022;15:2819–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Francois B, Jeannet R, Daix T, Walton AH, Shotwell MS, Unsinger J, et al. Interleukin-7 restores lymphocytes in septic shock: the IRIS-7 randomized clinical trial. JCI Insight. 2018;3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lord JD, McIntosh BC, Greenberg PD, Nelson BH. The IL-2 receptor promotes proliferation, bcl-2 and bcl-x induction, but not cell viability through the adapter molecule Shc. J Immunol [Internet]. 1998;161:4627–33. Available from: http://www.ncbi.nlm.nih.gov/pubmed/9794391 [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.