Circulating markers of neutrophil activation and lung injury in pediatric pneumonia in low-resource settings

ABSTRACT

Diagnostic biomarkers for childhood pneumonia could guide management and improve antibiotic stewardship in low-resource settings where chest x-ray (CXR) is not always available. In this cross-sectional study, we measured chitinase 3-like protein 1 (CHI3L1), surfactant protein D (SP-D), lipocalin-2 (LCN2), and tissue inhibitor of metalloproteinases-1 (TIMP-1) in Ugandan children under the age of five hospitalized with acute lower respiratory tract infection. We determined the association between biomarker levels and primary end-point pneumonia, indicated by CXR consolidation. We included 89 children (median age 11 months, 39% female). Primary endpoint pneumonia was present in 22 (25%). Clinical signs were similar in children with and without CXR consolidation. Broad-spectrum antibiotics (ceftriaxone) were administered in 83 (93%). Levels of CHI3L1, SP-D, LCN2 and TIMP-1 were higher in patients with primary end-point pneumonia compared to patients with normal CXR or other infiltrates. All markers were moderately accurate predictors of primary end-point pneumonia, with area under receiver operator characteristic curves of 0.66-0.70 (p<0.05 for all markers). The probability of CXR consolidation increased monotonically with the number of markers above cut-off. Among 28 patients (31%) in whom all four markers were below the cut-off, the likelihood ratio of CXR consolidation was 0.11 (95%CI 0.015 to 0.73). CHI3L1, SP-D, LCN2 and TIMP-1 were associated with CXR consolidation in children with clinical pneumonia in a low-resource setting. Combinations of quantitative biomarkers may be useful to safely withhold antibiotics in children with a low probability of bacterial infection.

Introduction

Pneumonia is the leading infectious cause of childhood mortality worldwide, with the highest burden of disease in low- and middle-income countries of South Asia and Sub-Saharan Africa [1]. Accurate diagnosis is critical because pneumonia requires more resources for management, may require treatment with antibiotics, and is more likely to result in adverse outcomes than milder respiratory infections [2]. Diagnostic imaging (e.g. chest X-ray, CXR) is widely used to distinguish between pneumonia and other respiratory tract infections. Guidelines for the standardization of CXR interpretation have been provided by the WHO, defining ‘primary end-point pneumonia’ by the presence of alveolar consolidation or pleural effusion on CXR [3]. These radiographic findings suggest a bacterial etiology and are associated with more severe disease [4,5]. However, CXR is inaccessible in many low-resource settings due to cost and the need for specialized equipment and trained technologists [5,6]. As a result, clinical criteria are commonly used to diagnose pneumonia and guide treatment. WHO guidelines for the management of childhood illness in low-resource settings classify respiratory tract infections based on the symptoms of fast breathing and chest indrawing as indicators of pneumonia, as well as danger signs for severe pneumonia [7]. While implementation of these diagnostic criteria and corresponding treatment recommendations has reduced pneumonia deaths, they distinguish poorly between infectious etiologies and result in antibiotic overuse [6]. New tools are needed to accurately diagnose pneumonia.

Antibiotic resistance is a growing worldwide concern. Infections caused by drug-resistant organisms lead to higher morbidity and mortality and require more health resources to treat [8]. In low-resource settings, reliance on nonspecific clinical criteria to guide treatment of suspected pneumonia results in an overuse of antibiotics for non-bacterial or self-limiting infections, which likely contributes to antibiotic resistance [9,10]. A randomized controlled trial in Pakistan found that outcomes were no different in children with WHO-defined non-severe pneumonia who received antibiotics compared to placebo [11], suggesting that the currently used clinical criteria are not a good indicator of which patients are most likely to benefit from antibiotics. A highly sensitive test that could be used to accurately rule out bacterial pneumonia would be valuable to identify patients from whom antibiotics could be safely withheld.

Pathogen invasion of the lower respiratory tract generates a host inflammatory response, including macrophage activation, increased microvascular permeability, neutrophil recruitment and degranulation, alveolar fluid accumulation, and tissue injury. Measuring markers of inflammation and tissue injury present in the blood during infection may be a promising avenue of research to improve the diagnosis of pneumonia. A recent systematic review of the literature concluded that, while numerous biomarkers have been explored, to date none has been found with sufficient diagnostic accuracy for clinical use [12]. C-reactive protein (CRP) is a marker of inflammation that is commonly used clinically and has been studied extensively as a marker of pediatric pneumonia; however, it has inadequate sensitivity (Sn) and specificity (Sp) to be used in isolation [12]. A combination of biomarkers is likely required to achieve the necessary accuracy.

Chitinase 3-like protein 1 (CHI3L1), also known as YKL-40, is a secreted glycoprotein produced by neutrophils, macrophages and pulmonary epithelial cells. It is upregulated in response to inflammatory cytokines and tissue injury, promotes bacterial clearance, and augments host tolerance to infection [13]. CHI3L1 has been shown to play a role in inflammation in pulmonary conditions including asthma and pulmonary fibrosis [14], and is associated with disease severity in respiratory infections including COVID-19 and community-acquired pneumonia [15–17]. CHI3L1 levels are higher in children with pneumococcal pneumonia than those with viral respiratory infections in Uganda [18].

Surfactant Protein D (SP-D) is a collagen-containing lectin (collectin) that is produced by type II alveolar epithelial cells. SP-D plays an important role in innate immunity in the lungs and is upregulated in response to lung injury or infection [19]. It promotes clearance of pathogens by enhancing phagocytosis via opsonization and stimulating chemotaxis and reactive oxygen species production [19,20]. Lung injury caused by infection and inflammation causes damage to alveolar capillaries, allowing SP-D to leak into the blood, resulting in higher serum SP-D levels found in many acute and chronic lung diseases [21]. SP-D levels are associated with the degree of lung injury, disease severity and poor outcomes in children with respiratory syncytial virus bronchiolitis, influenza, and acute respiratory failure [22–24].

Lipocalin-2 (LCN2) is an acute-phase protein present in neutrophil granules, and is also produced by macrophages, other immune cells and epithelial cells [25]. LCN2 has important antibacterial effects by sequestering iron through chelation of siderophores. It is also involved in inflammation, including recruitment of neutrophils and pro-inflammatory cytokine signaling. In the lungs, LCN2 is induced in response to infections of varied etiologies [26]. Several studies have shown LCN2 to be elevated in patients with pneumonia, with higher levels associated with disease severity and poor outcomes [26,27].

Tissue inhibitor of metalloproteinases-1 (TIMP-1) inhibits matrix metalloproteinases (MMPs), enzymes that degrade the extracellular matrix and play an important role in inflammation and tissue remodeling [28]. TIMPs bind MMPs to inhibit proteolytic function, while also having independent cytokine effects, including modulating cell proliferation, apoptosis, differentiation and angiogenesis [29]. Significant sources of TIMP-1 include macrophages and fibroblasts, and levels are increased during inflammatory processes [30]. In mouse models, pulmonary TIMP-1 expression increases following bleomycin-induced and influenza-induced lung injury [31,32]. Plasma TIMP-1 levels are higher in patients with community acquired pneumonia compared to healthy controls and positively correlate with Pneumonia Severity Index scores [33,34].

In this study, we measured plasma levels of CHI3L1, SP-D, LCN2 and TIMP-1 in children hospitalized for pneumonia in Uganda and determined their association with radiographically-defined primary end-point pneumonia.

Methods

Study design

This was a cross-sectional study, in which we examined the association of biomarkers CHI3L1, SP-D, LCN2 and TIMP-1 with CXR findings among children hospitalized with pneumonia.

Study population

This study included inpatients at two pediatric facilities in Uganda: Jinja Regional Referral Hospital and Kambuga District Hospital. Children were included based on the following criteria: (1) age under 5 years, (2) hospital admission, and (3) chest indrawing or fast breathing pneumonia. Exclusion criteria were: (1) malaria diagnosis, (2) clinical suspicion of tuberculosis, (3) HIV seropositive (Alere Determine^TM, Abbott, Chicago, IL) and (4) sub-optimal CXR quality. Malaria was diagnosed using light microscopy of field-stained peripheral blood smears or rapid diagnostic tests (mRDT, First Response Malaria Ag. (pLDH/HRP2) Combo Rapid Diagnostic Test, Premier Medical Corporation Limited, India).

Study procedures

Data was collected on the patients’ age and sex, symptoms at presentation, physical exam findings, radiographic findings, recovery times and mortality. A CXR was obtained within 24 hours of admission and was interpreted by a Canadian board-certified radiologist according to the standardized WHO-endorsed definition of primary end-point pneumonia [3]. A standardized case report form was completed for each CXR. The film was classified in terms of quality (uninterpretable, suboptimal, or adequate), findings (end-point consolidation, other (non end-point) infiltrate), presence of pleural effusion, and conclusion (primary endpoint pneumonia, other infiltrate, or no consolidation, infiltrate or effusion) [3]. Tachycardia and tachypnea were defined as heart rate and respiratory rate above the 99^th percentile according to norms in hospitalized children by age [35].

Measurement of CHI3L1, SP-D, LCN2 and TIMP-1 and CRP

Venipuncture was performed at the time of hospital admission, contemporaneous with initial stabilization of the patient and initiation of antibiotics and oxygen. Venous blood was collected in EDTA-coated tubes and centrifuged on site to separate the plasma. Plasma samples were stored at −80°C and then shipped on dry ice to our lab in Edmonton, Canada, where they were again stored at −80°C until analysis. Enzyme-linked immunosorbent assays (ELISAs) for CHI3L1 (dilution factors 1:20 to 1:2000), SP-D (1:5), LCN2 (1:20 to 1:100), TIMP-1 (1:1 to 1:20) and CRP (1:5 to 1:50000) were carried out according to the manufacturer’s instructions (R&D Systems, Minneapolis, MN).

Detection of viruses, bacteria, and fungi from the nasopharymx

A nasopharyngeal (NP) swab was collected at the time of hospital admission, placed in universal transport medium, and stored at −80°C. The sample was shipped to Canada on dry ice for analysis (MTH lab). Commercial kits, KingFisher™ mL Purification System (Thermo Fisher Scientific Inc, Waltham, MA) and the MagaZorb® Total RNA Mini-Prep Kit (Promega, Madison, WI), were used for nucleic acid extraction. The quantitative real-time PCR (qPCR) instrument used was the 7500 Real-Time PCR System (Applied Biosystems®, Foster City, CA). We used a commercial multiplex PCR, FTDResp33 (Fast-Track Diagnostics, Esch-sur-Alzet, Luxembourg), which includes primers for the detection of the following micro-organisms: Streptococcus pneumoniae, Haemophilus influenzae type b, Staphylococcus aureus, rhinovirus, respiratory syncytial virus, human adenovirus, parainfluenza virus, human bocavirus, and Pneumocystis jiroveci. The cycle threshold (Ct) cutoff value for a positive result was set at 38. With respect to S. pneumoniae, a genomic load >log₁₀6.9 in the nasopharynx was used to define clinically significant carriage [36]. To quantify the number of genome copies of S. pneumoniae, we used manufacturer-published standard curves relating the Ct to plasmid DNA concentrations.

Statistical analysis

The primary outcome (dependent variable) was radiographic primary endpoint pneumonia (binary variable). The independent variables of interest were quantitative levels of the four biomarkers (CHI3L1, SP-D, LCN2 and TIMP-1).

A sample size calculation indicated that we would need 77 patients to detect a statistically significant difference in biomarker levels between patients with and without CXR consolidation with 80% power at the α = 0.05/4 significance level. To determine this sample size, we used data from another study, showing that log-transformed CHI3L1 levels were mean (standard deviation) 4.8 (1.2) in 30 children with primary endpoint pneumonia versus 3.9 (0.94) in 94 children with normal CXR (pooled standard deviation 1.0, difference of means 0.9, ratio of cases to controls 3.1:1) [37]. We used the Bonferroni correction to adjust for multiple comparisons since we examined 4 biomarkers.

Descriptive statistics were presented as counts (percentages) for binary and categorical variables, and median (interquartile range, IQR) for continuous variables. Data was analyzed non-parametrically, using the Pearson Chi-Square test for binary and categorical variables and the Mann-Whitney U test for continuous variables. Spearman’s rank correlation coefficient was used to assess correlations between continuous variables. Diagnostic test performance characteristics were expressed using sensitivity (Sn), specificity (Sp), positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR(+)), and negative likelihood ratio (LR(-)). To assess the accuracy of the biomarkers in predicting primary end-point pneumonia, we used area under receiver operating characteristic (ROC) curves. Cutoff points were determined using the Youden Index to find the point of maximum discrimination. We used the McNemar test to compare the sensitivities of two diagnostic tests [38].

Ethics statement

This study was approved by the Makerere University School of Biomedical Sciences Research Ethics Committee (REC Protocol # SBS 139), the Uganda National Council on Science and Technology (Ref SS 3331), and the University of Alberta Health Research Ethics Board (Study ID # Pro00106489). Written informed consent was obtained from the parent or legal guardian of each pediatric patient.

Results

A total of 89 children admitted to Jinja Regional Referral Hospital or Kambuga Hospital between February 2014 and July 2015 for fast breathing or chest indrawing pneumonia were included (median age 11 months, 39% female). 22 patients had radiographic primary end-point pneumonia. Patient characteristics are shown in Table 1. There were no significant differences in patient characteristics according to radiographic findings, aside from age and respiratory rate. Of note, the proportion of children with tachypnea based on age-specific norms was similar between groups. Pathogens detected from nasopharyngeal swabs are shown in Supplemental Digital Content 1 (Table).

Table 1.

Patient characteristics, stratified by chest x-ray (CXR) findings.

	CXR findings
	Entire Cohort (N = 89)	Normal (N = 25)	Other infiltrate (N = 42)	Pneumonia (N = 22)	P-value
Demographics
Female sex	35 (39)	9 (36)	18 (43)	8 (36)	0.81
Age [years], median (IQR)	0.92 (0.42–1.7)	0.92 (0.75–2)	0.79 (0.25–1.2)	1.5 (0.46–2.6)	0.028
History
Cough	84 (94)	23 (92)	41 (98)	20 (91)	0.45
Difficulty breathing	78 (89)	21 (88)	37 (88)	20 (91)	0.93
Lethargy	10 (11)	3 (12)	3 (7.1)	4 (18)	0.41
Convulsions	14 (16)	3 (12)	8 (19)	3 (14)	0.74
Vomiting	24 (27)	6 (25)	15 (36)	3 (14)	0.16
Unable to feed/drink	32 (36)	10 (42)	14 (33)	8 (36)	0.8
Physical examination findings
Weight [kg], median (IQR)	8.2 (6.5–10)	9 (7–10)	8 (5.8–9)	9.8 (7–12)	0.11
Underweighta	11 (12)	6 (24)	2 (4.8)	3 (14)	0.067
Heart Rate (bpm), median (IQR)	160 (150–180)	160 (140–180)	160 (160–180)	160 (150–170)	0.34
Tachycardiab	55 (62)	17 (68)	24 (57)	14 (64)	0.66
Respiratory rate (bpm), median (IQR)	68 (52–76)	60 (50–71)	70 (62–78)	62 (50–72)	0.035
Tachypneab	70 (79)	20 (80)	33 (79)	17 (77)	0.97
Oxygen saturation (%), median (IQR)	85 (79–88)	86 (81–88)	85 (78–88)	85 (80–91)	0.89
Hypoxemia (SaO2 <90%)	72 (81)	21 (84)	35 (83)	16 (73)	0.53
Temperature [°C], median (IQR)	38 (37–38)	38 (37–38)	38 (37–38)	38 (37–39)	0.35
Feverc	46 (52)	15 (60)	19 (45)	12 (55)	0.48
Altered level of consciousnessd	14 (16)	5 (20)	6 (15)	3 (14)	0.8
Chest indrawing	74 (86)	20 (83)	34 (85)	20 (91)	0.73
Wheeze	20 (23)	7 (28)	10 (24)	3 (14)	0.47
Stridor	13 (15)	4 (17)	6 (15)	3 (14)	0.97
General danger signe	52 (62)	15 (65)	27 (68)	10 (48)	0.29
Treatment
Ceftriaxone	83 (93)	24 (96)	38 (90)	21 (95)	0.61
Supplemental oxygen	77 (87)	22 (88)	36 (86)	19 (86)	0.97
Outcome
Discharged without disability	79 (89)	22 (88)	40 (95)	17 (77)	0.096
Transferred to another facility	4 (4.5)	2 (8)	0 (0)	2 (9.1)	0.15
Death	4 (4.5)	1 (4)	1 (2.4)	2 (9.1)	0.46

Data represent n (%) unless otherwise specified; IQR, interquartile range.

^aWeight-for-age below −3SD [49]

^bVital sign >99^th percentile for age [35].

^cAxillary temperature >37.5°C.

^dOne or more of the following: fails to watch or follow with eyes, fails to localize to painful stimulus, fails to cry or verbalize appropriately with pain.

^eOne or more of the following: vomiting, convulsions, unable to feed/drink, or altered consciousness [50].

Admission levels of CHI3L1, SP-D, LCN2 and TIMP-1 were elevated in patients with primary end-point pneumonia compared to patients with normal CXR or other infiltrates (Figure 1). Biomarker levels stratified by CXR findings are summarized in Supplemental Digital Content 2 (Table). Among patients with primary end-point pneumonia, pleural effusion was associated with higher levels of LCN2 (210 pg/mL [IQR 180-240] vs 120 pg/mL [IQR 69-130], p = 0.017). CHI3L1, LCN2 and TIMP-1 levels were positively correlated with CRP, a marker of systemic inflammation used in clinical practice (see Figure, Supplemental Digital Content 3). Figure 2 shows ROC curve analysis for each biomarker as a predictor of primary end-point pneumonia. Optimal cutoff points, areas under the ROC (AUROC) curves, and test performance characteristics of each biomarker in predicting primary end-point pneumonia are shown in Table 2, including clinical reference biomarker, CRP.

Figure 1.

Association of biomarker levels with chest x-ray (CXR) findings in 89 children hospitalized for clinical pneumonia in Uganda. CXR findings are indicated as primary end-point pneumonia (P), other infiltrates (O), or normal CXR (N). a. Levels of chitinase 3-like protein 1 (CHI3L1) were higher in P than O. b. Levels of surfactant protein D (SP-D) were higher in P than N. c. Levels of lipocalin-2 were higher in P than both N and O. d. Levels of tissue inhibitor of metalloproteinases-1 (TIMP-1) were higher in patients with P than O. *, p < 0.05; **, p < 0.01.

Figure 2.

Receiver operating characteristic curves for biomarkers CHI3L1, SP-D, LCN2 and TIMP-1 as predictors of chest x-ray. Biomarker cutoff points were determined based on Youden index: CHI3L1 > 140 pg/mL, SP-D > 5.2 pg/mL, LCN2 > 110 pg/mL, and TIMP-1 > 150 pg/ml.

Table 2.

Test performance characteristics of CHI3L1, SP-D, LCN2, and TIMP-1, individually and in combination to predict radiographic primary end-point pneumonia.

	CHI3L1	SP-D	LCN2	TIMP-1	All 4 below thresholdb	CRPc
Thresholda	140 pg/mL	5.2 pg/mL	110 g/mL	150 pg/mL		59 μg/mL
AUROC curve	0.69 (0.54–0.84)	0.67 (0.55–0.80)	0.70 (0.56–0.84)	0.66 (0.51–0.80)		0.75 (0.60–0.89)
Sensitivity (%)	50 (28–72)	61 (36–83)	73 (50–89)	50 (28–72)	95 (76–100)	64 (41–83)
Specificity (%)	91 (81–97)	71 (59–82)	72 (59–82)	81 (70–90)	44 (31–57)	88 (78–95)
Positive predictive value (%)	65 (38–86)	38 (21–58)	46 (29–63)	48 (27–69)	36 (24–50)	64 (41–83)
Negative predictive value (%)	85 (74–92)	87 (74–94)	89 (77–96)	83 (71–91)	96 (82–100)	88 (78–95)
Positive likelihood ratio	5.5 (2.3–13)	2.1 (1.3–3.7)	2.6 (1.6–4.1)	2.7 (1.4–5.2)	1.7 (1.3–2.1)	5.3 (2.6–11)
Negative likelihood ratio	0.55 (0.36–0.84)	0.54 (0.3–0.99)	0.38 (0.19–0.77)	0.62 (0.4–0.95)	0.11 (0.016–0.76)	0.41 (0.24–0.72)

CHI3L1, chitinase-3-like-1; SP-D, surfactant protein D; LCN2, lipocalin-2; TIMP-1, tissue inhibitor of metalloproteinase-1; CRP, C-reactive protein; AUROC, area under receiver-operator characteristic curve.

^aBased on receiver operator characteristic curve analysis for maximum discrimination between primary end-point pneumonia and normal chest radiograph or other infiltrates (Youden index).

^bIndicates patients with CHI3L1 < 140 pg/mL, SP-D <5.2 pg/mL, LCN2 < 110 pg/mL, and TIMP-1 < 150 pg/mL.

^cCRP provided as a comparator biomarker, used in clinical practice.

We next explored the potential clinical utility of these four markers for the management of children with suspected pneumonia (Table 2). Twenty-eight patients (31%) had levels of CHI3L1, SP-D, LCN2, and TIMP-1 which were all below their respective cutoff points. Only one of these patients (3.4%) had primary end-point consolidation on CXR, compared to 20/61 patients (36%) with one or more elevated biomarkers (Figure 3). The LR(-) (i.e. all four markers below the cutoff) was 0.11 (95%CI 0.015 to 0.73). This corresponds to a Sn of 95% (95%CI 76% to 100%) and an NPV of 96% (95%CI 82% to 100%) to rule out radiographic pneumonia. Relative to a comparator biomarker used in clinical practice, CRP (LR(-) 0.41, Sn 64%, NPV 88%), the combination of CHI3L1, SP-D, LCN2, and TIMP-1 provided superior ability to predict those children without radiographic primary end-point pneumonia (p = 0.041, Table 2).

Figure 3.

Probability of chest x-ray (CXR) consolidation according to the number of pneumonia biomarkers (CHI3L1, SP-D, LCN2 and TIMP-1) elevated above the threshold. Among 89 patients admitted for pneumonia in Uganda, 22 (25%) had CXR consolidation (dotted line). Of the 28 patients with no biomarkers elevated, only one (3.4%) had CXR consolidation. A greater number of elevated biomarkers was associated with a higher probability of CXR consolidation.

Discussion

In low-resource settings, pneumonia is a significant cause of childhood morbidity and mortality, but diagnostics such as chest radiography are not always available. Here we show that CHI3L1, SP-D, LCN2 and TIMP-1 are associated with the presence of CXR consolidation and together can be used to identify children at low risk of primary endpoint pneumonia (LR(-) 0.11). These results are encouraging for the development of a biomarker panel to replace the CXR for pneumonia diagnosis in low-resource settings.

In this study, patient signs and symptoms did not differ according to CXR findings (Table 1), suggesting that clinical features alone may be unsatisfactory to guide antibiotic prescription. This is supported by a recent meta-analysis, in which WHO-approved signs of fast breathing (Sn 62%) and chest indrawing (Sn 48%) showed poor diagnostic performance in identifying radiographic pneumonia in children under 5. [39] Similarly, pathogen detection was not strongly associated with CXR consolidation (Table, Supplemental Digital Content 1). Although respiratory syncytial virus was more common in the group with other infiltrates (consistent with its role in causing bronchiolitis), identification of bacterial pathogens was not associated with CXR consolidation, reflecting the inability of NP swabs to distinguish true infection from bacterial colonization of the upper respiratory tract. Similarly, multiple, often overlapping etiologies were only weakly associated with CXR findings in the large multi-country Pneumonia Etiology Research for Child Health (PERCH) case-control study [39].

In contrast, each biomarker was statistically significantly elevated in children with CXR consolidation (p < 0.05 for all markers). However, the clinical utility of individual biomarkers may be limited since the AUROCs of 0.66 to 0.70 suggest poor discrimination [40–42]. When considered in combination, a greater number of elevated biomarkers was associated with a higher probability of CXR consolidation (Figure 3), and patients with no elevated biomarkers had approximately a ten-fold lower likelihood of CXR consolidation compared to the pretest likelihood (LR(-) 0.11).

Selection of these biomarkers was based on their known role in the pathophysiology of lung diseases, and our results confirm and extend previous findings on the role of these biomarkers in pediatric pneumonia in low-resource settings. CHI3L1 is secreted by activated macrophages and neutrophils during lung inflammation [13], and was higher in children with primary end-point pneumonia compared to children with normal CXR in Tanzania [37]. SP-D is produced by alveolar cells during tissue injury and infection. Higher levels were associated with increased disease severity in children with community-acquired pneumonia in Egypt [20]. LCN2 is a marker of neutrophil degranulation elevated in lung infection. LCN2 levels were associated with disease severity in pediatric pneumonia in Africa and it has been proposed as a marker to distinguish bacterial pneumonia from viral infections [9,43]. TIMP-1 is involved in tissue remodeling during lung injury [30], and has been identified as a biomarker that could be used in combination with other markers to distinguish between pneumonia etiologies in children in Mozambique [44]. Therefore, the association of these four biomarkers with primary end-point pneumonia in children in low-resource settings is biologically plausible and supported by previous studies. Because each of these biomarkers is involved in distinct but interconnected pathways relating to neutrophil activation and tissue injury during lung infection, using them in combination is likely to provide higher diagnostic utility than any one biomarker in isolation.

We explored the implication of these findings for the development of a diagnostic test to rule out primary end-point pneumonia for improved antimicrobial stewardship. Our results suggest that low levels of all four markers can be used to confidently rule out primary end-point pneumonia with a LR(-) of 0.11 and Sn of 95%. This appears to be more useful than CRP, another biomarker used in clinical practice, which we found to have a LR(-) of 0.41, and which has a Sn reported in the literature of 70% in distinguishing bacterial from viral pneumonia [12]. Similarly, procalcitonin, another biomarker widely studied in pneumonia, also has a lower Sn (69%) [12]. CHI3L1 was previously shown to predict primary end-point pneumonia in Tanzania (AUROC 0.79, Sn 93% and Sp 63%) [37], which is similar to our findings (AUROC 0.69, Sn 50%, Sp 91%). LCN2 has similarly been shown to predict probable bacterial pneumonia, as defined by positive blood cultures or CXR consolidation (AUROC 0.91, Sn 77%, Sp 94%) [9], which supports our findings in primary end-point pneumonia (AUROC 0.70, Sn 73%, Sp 72%). Diagnostic performance of these four markers in combination was similar to the combination of CHI3L1 and CRP proposed in another study (Sn 93%, LR(-) 0.083) [37]. Other biomarkers associated with radiographic pneumonia include proadrenomedullin [45] and procalcitonin [37]. In another study, biomarkers (white blood cell count, absolute neutrophil count, CRP, and procalcitonin) were less useful [46]. Therefore, these four biomarkers show promising clinical utility for ruling out primary end-point pneumonia and are equivalent or superior to other markers identified in the literature.

Our study has several limitations. It took place in a low-income country, where factors such as socioeconomic status, nutrition, co-infection and access to health resources may affect underlying patient health, disease course and outcomes. These factors limit the generalizability of our findings to high-income settings. Our selection of biomarkers for this study was based on biological plausibility supported by existing literature, but other markers may be similarly informative. An alternative approach could be to identify potential biomarkers from an unbiased high-throughput screening method (e.g. proteomics-based analysis). Exploration of other potential biomarkers of pneumonia is warranted. Furthermore, some statistical analyses were limited by our sample size. Further studies are needed to validate these findings and confirm the clinical utility of these biomarkers in pediatric pneumonia in low-resource settings. Chest radiography was chosen as the comparator in this study because it is the current acceptable method for diagnosing bacterial pneumonia in clinical practice. The gold standard diagnostic test for childhood pneumonia is lung biopsy. Although CXR is commonly used in clinical practice, it has several limitations. First, variability in CXR interpretation may occur between observers. Only one radiologist provided the CXR interpretation in our study, whereas independent agreement between two or more blinded readers would be desirable to account for inter-observer variability. Second, there may be overlap in the CXR findings between different infectious etiologies (e.g. viral, bacterial, atypical bacterial, tuberculous, and fungal) [6]. Other radiographic and microbiologic methods of detecting bacterial infection, such as computed tomography or bronchoalveolar lavage, expose patients to radiation, are invasive, and were not feasible in this study. Nevertheless, the association between these four biomarkers and CXR findings suggests that CHI3L1, TIMP-1, SP-D and LCN2 may be clinically useful markers of primary end-point pneumonia that warrant further study.

Potential clinical applications of these findings include the development of a fingerprick blood test to replace the CXR in low-resource settings. For example, these abundant circulating proteins could be adapted to a multiplex point-of-care lateral flow immunochromatographic platform, similar to the malaria rapid diagnostic test (mRDT) [47]. A fingerprick bedside diagnostic tool would have the advantage of reduced radiation exposure and no need for expensive CXR equipment. A second application could be to improve antibiotic stewardship in low-resource settings. Using biomarkers as a surrogate marker of CXR consolidation may help to identify children in whom it would be safe to withhold antibiotics, thus reducing antibiotic overuse for viral infections. The LR(-) of 0.11 of these four markers may give the healthcare provider confidence in withholding antibiotics by minimizing the risk of missing a bacterial infection that may result in severe outcomes [48]. In this study, 93% of children received broad-spectrum antibiotics (ceftriaxone); however, only 25% had consolidation on CXR. If we had managed patients with low levels of all four biomarkers without antibiotics (31% of patients in our cohort), we could have spared antibiotic exposure in a third of patients. Because antibiotic resistance results in increased use of health resources and worse outcomes for patients with drug-resistant infections, a reduction in antibiotic use for common childhood respiratory infections would be highly advantageous. In conclusion, due to the high burden of pediatric pneumonia and current diagnostic challenges in low-resource settings, CHI3L1, SP-D, LCN2 and TIMP-1 warrant further study as markers of bacterial pneumonia to aid in clinical decision making and reduce antibiotic overuse.

Funding Statement

The work was supported by the Grand Challenges Canada [S4 0206-01]; Women and Children’s Health Research Institute [3584]

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

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