Original Research

Evaluation of risk stratification for predicting adverse outcomes in paediatric febrile neutropenia

Motunrayo O. Adekunle, Alan Davidson, Marc Hendricks

Received: 14 May 2025; Accepted: 11 Sept. 2025; Published: 14 Nov. 2025

Copyright: © 2025. The Author(s). Licensee: AOSIS.
This work is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license (https://creativecommons.org/licenses/by/4.0/).

Abstract

Background: Adverse outcomes from febrile neutropenia (FN) are a common cause of morbidity and mortality in childhood cancer management. Outpatient management of individuals at low risk of adverse outcomes has been reported to reduce the cost of care, improve health-related quality of life (HRQL), and reduce the risk of hospital-acquired infections.

Aim: To validate a tool for risk stratification of adverse outcomes in paediatric oncology patients at the Red Cross War Memorial Children’s Hospital, Cape Town, South Africa.

Setting: A retrospective data collection from 01 January 2017 to 31 December 2019 of children with a cancer diagnosis who had chemotherapy-induced febrile neutropenia.

Methods: The study population comprised children with confirmed cancer diagnoses on chemotherapy with FN. Each episode of FN was evaluated.

Results: Analysis was performed on 254 (99.2%) out of 256 oncology cases seen over the study period. In all, there were 267 chemotherapy-induced FN episodes in 179 patients. Ninety-nine (37.1%) adverse outcomes occurred in 267 FN episodes. Validation of a risk stratification tool of adverse outcome demonstrated a sensitivity and specificity of 52.3% and 62.0%, respectively. Positive and negative predictive values were 56.3% and 58.2%, respectively. The area under the curve translated to a 57.1% accuracy (p-value of 0.064). In our cohort, the coordinates of the curve’s best predictive values were between 7.5 and 8.5.

Conclusion: A lower cut-off using the Swiss Paediatric Oncology Group (SPOG) FN risk index best predicted adverse outcomes in our cohort, although the tool could not be validated.

Contribution: The SPOG FN tool is useful in our setting with easy accessibility of the parameters, however, a lower cut-off value is required to determine patients at risk of adverse events.

Keywords: febrile neutropenia; cancer; chemotherapy; children; risk factors; adverse outcomes.

Introduction

Febrile neutropenia (FN) is a life-threatening treatment-related complication in children with cancer. It is a major cause of prolonged hospital admissions and treatment alteration. Consequently, serial admissions for FN significantly impact health-related quality of life (HRQL) in children with cancer and their families.^1,2

The risk stratification of adverse outcomes in FN helps to identify individuals requiring hospital admission from those who could be treated as outpatients. Outpatient management of individuals at low risk of adverse outcome helps to reduce the rate of hospitalisation, the cost of treatment, and hospital-acquired infections and improves HRQL.^2,3

Unfortunately, inconsistencies between studies make adopting a risk stratification tool for FN difficult. In addition, valid concerns related to tools developed in high-income settings and the transfer of these applications to low- and middle-income environments have been tabled.⁴ This is related to the stark differences in healthcare systems, patients’ nutritional status and varying pathogenic organisms in well-resourced versus heavily under-resourced countries.⁴ These difficulties prompted a call for more studies in low- and middle-income countries (LMICs) to identify risk stratifiers for FN that may be setting-specific and would contribute to the development of guidelines which match the needs of children with cancer in less resourced settings.⁴ The International Febrile Neutropenia Guideline made recommendations regarding the validation of risk stratification for various regions before adaptation given the impact of geographical and temporal differences,⁵ but this is yet to be adopted at the paediatric oncology unit of Red Cross War Memorial Children’s Hospital (RCWMCH). This study aimed to validate the Swiss Paediatric Oncology Group (SPOG) FN index, which is a tool for risk stratification.⁶ These findings, we hoped, would aid in determining those at risk of adverse outcomes requiring hospital admissions compared to those who would benefit from outpatient care. Additionally, this may also impact infection-related morbidity and mortality.

Methods

This retrospective data collection was conducted at the oncology unit of Red Cross War Memorial Children’s Hospital (RCWMCH), Cape Town, South Africa. The duration of the study was 3 years (01 January 2017 to 31 December 2019).⁷

Study population

The study participants were children newly diagnosed with either haematologic or biopsy-proven solid cancers on chemotherapy. Febrile neutropenia was defined as axillary temperature (≥ 38.0 °C) and neutropenia (absolute neutrophil count [ANC] < 1.0 × 10⁹ cells/L),⁸ and each episode of FN was evaluated.

Data collection method

All clinical and laboratory parameters at admission for each episode of FN were evaluated. The intensity of chemotherapy received prior to the FN was classified.⁹ Adverse outcomes were classified according to Ammann et al.⁶ Swiss Paediatric Oncology Group FN Risk Index was used as a predictor of adverse outcome as shown in Table 1.⁶ The cut-off value for high risk of adverse outcomes was a score of at least nine and individuals with a score less than nine were classified as low risk for adverse outcomes.

Data analysis

Data analysis was conducted using the SPOG risk index as the primary stratification tool. The predictive performance of the index for adverse outcomes was assessed by calculating sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). Sensitivity represented the proportion of patients with adverse outcomes correctly identified as high risk, while specificity represented the proportion without adverse outcomes correctly identified as low risk. Positive predictive value and NPV were calculated to evaluate the clinical utility of the index in predicting true risk status. Receiver operating characteristic (ROC) analysis was used to determine the area under the curve (AUC) and the optimal cut-off for predicting adverse outcomes. Data were presented using bar charts, pie charts and box-and-whisker plots, as appropriate. Statistical significance was set at p < 0.05, and 95% confidence intervals were calculated for all estimates.

Ethical considerations

Ethical approval was obtained from the Health Research Ethics Committee of the University of Cape Town (reference number: HREC 351/2020) and approved on 02 July 2020.

Results

Demographic characteristics of all cancer patients

In all, 267 FN episodes occurred in 179 patients who received chemotherapy. In all, 99 adverse events were seen in all FN episodes (37.1%). Microbiologically defined infection was the commonest adverse event (n = 71; 71.7%), as shown in Figure 1.

FIGURE 1: Adverse outcomes in study subjects.

Based on the SPOG FN Risk Index (cut-off ≥ 9 for adverse outcomes), the median score was 7 (Figure 2).

FIGURE 2: Median score of study subjects.

Risk stratification in febrile neutropenia (n = 267)

In our cohort, 146 patients (54.7%) were classified as low risk (score < 9) using the SPOG tool. A higher proportion of adverse events occurred in patients with scores ≥ 9, although this association was not statistically significant (p = 0.159) (Table 2).

As shown in Table 3, validating the SPOG risk index tool, sensitivity and specificity were 50.6% and 57.1%, respectively.

Determining accuracy of the adopted tool

An ROC curve was developed to predict the best cut-off for adverse outcomes using the SPOG tool in our cohort (Figure 3). The area under the curve was 0.571, which translates to 57.1% accuracy (p = 0.064) (Table 4).

FIGURE 3: Receiver operating characteristic curve.

With the exclusion of haemoglobin as a parameter for adverse risk stratification, there was a significantly higher accuracy of SPOG tool predicting adverse outcome in our cohort (p = 0.02).

Point coordinates on the curve

As shown in Table 5, curve coordinates were generated to evaluate the sensitivity and specificity values that best predict the risk of adverse outcomes. Given the inverse relationship between sensitivity and specificity, the optimal coordinate was defined as the point with the highest combined values of both measures. In our cohort, these points corresponded to a score of 8.5, and to 6.5 when haemoglobin was excluded.

Discussion

Adverse events from chemotherapy-induced febrile neutropenia are a common complication in paediatric oncology. As seen in our cohort, about one-third of adverse events occurred in all FN episodes and is supported by similar findings from other high income countries (HICs) (France 23.4%) and upper middle-income countries (UMICs) (Brazil 32%, India 34%).^10,11,12 This is not surprising considering that there is considerable overlap between the intensity of the chemotherapy regimens used in our unit compared to these centres, and so one would expect a similar risk profile.

Microbiologically defined infection (MDI) was the commonest adverse event in our cohort and again is consistent with earlier reports.¹² Microbiologically defined infection is associated with a significant risk of prolonged hospital admission and death.¹³ It has also been reported to be a predictor of adverse events such as bacteraemia and serious medical conditions by some authors.¹⁴ Death constituted 4% of adverse outcomes (i.e. 2.2% of patients) and considering the reportedly wide range of mortality as a result of FN in children (0.4% – 12%), places us in an advantageous position compared to other LMICs, where insufficient supportive care, among others, accounts for the high death rate from FN.^{6,13,15,16,17}

We selected the SPOG FN risk index as a validation tool for adverse events occurrence. This validation tool was selected in our study because the parameters evaluated are routinely done in all patients with FN in our institution. A median score of seven was observed in our cohort, suggesting that most patients fell below the SPOG high-risk threshold for adverse outcomes. This reflects on the higher proportion of subjects falling into the category of low risk for adverse events in our cohort. The inclusion of haemoglobin, which had the highest score by Ammann et al.⁶ perhaps, explains this finding. The triggering of FN post blood transfusion is not a common occurrence in our practice. This would have resulted in lower scores and risk stratification in our cohorts who had low haemoglobin levels despite marked risk for adverse events. However, we did not set out to evaluate the proportion of patients that developed chemotherapy induced neutropenia (CIN) with fever following blood transfusion.

In our study, a higher though not significant proportion of patients that had adverse events were categorised as high risk using the SPOG FN risk index. About one-third of those that had adverse events were stratified as low risk using the validation tool. This indicates that the risk for adverse events in our institution requires a lower score index compared to cut-off value of nine extrapolated by SPOG. In middle income countries, adaptation of lower score will help to escalate care promptly and reduce the mortality rate from FN, which remains high.

The SPOG FN risk tool yielded low predictive values in our cohort. Green et al. reported low scores in another institution in South Africa and also in the Netherlands using the same validation tool.^18,19 Green et al. reported a high incidence of bacteraemia that had a score below nine using SPOG risk stratification, a documented possibility for the low yield in obtained results.¹⁸ The latter authors, however, reported a higher specificity with re-assessment 8 h – 24 h after admission.¹⁹ The authors concluded that the low predictive values were possibly due to the inclusion of haemoglobin of at least 9 g/dL as a scoring parameter. Ammann et al.⁶ included high haemoglobin in its tool postulating that fever occurred in FN following blood transfusion, a finding which is not constantly reported in clinical practice. We did not set out to do a reassessment validation of the adopted tool in our cohort; but in line with the report of Miedema et al.,¹⁹ onset of fever to be termed as FN post-transfusion was not a common occurrence in our practice and this would have accounted for the low score below nine in patients that had adverse events in our cohort. We decided to remove haemoglobin value as a parameter for risk stratification and observed a significantly higher accuracy. This further supported the fact that fever in FN does not necessary occur following blood transfusion and that high haemoglobin level is not a consistent parameter that predicts adverse events in FN. There will be a need for more studies to revalidate this tool with the exclusion of haemoglobin as a scoring parameter for further comparison of its prediction and adaptation for middle-income countries. In addition, the wide disparity in adverse outcomes in our study compared to Ammann et al. may have contributed to the conflicting result.⁶ We recorded a high proportion of patients with adverse outcomes (37.1% vs. 29%), MDI on (71.7% vs. 22%) and ICU admissions (12% vs. 4.3%) compared to Ammann et al.⁶

Using the validation tool by Ammann et al.,⁶ a lower cut-off will be required in our setting to categorise patients as high risk for adverse events. The adaptation will reduce false positive results and untoward mortality. In LMICs where supportive care is not optimal, it is more beneficial to provide wider inclusion of more patients at risk to initiate prompt and early treatment and to reduce morbidity and mortality.

In conclusion, this current study failed to validate SPOG tool to correctly classify CIN patients. The retrospective study design may have contributed to this. Also, a lower cut-off value better predicts adverse outcomes in our setting. Adaptation of the tool with lower cut-off as well as exclusion of haemoglobin as a screening parameter will help us reduce the rate of false negatives of adverse events.

Despite the possible disadvantages of retrospective data collection, the effect of missing data was not significant considering we exceeded 99% data recruitment. However, we were unable to reassess patients’ parameters after admission for re-evaluation for risk assessment, which has been shown to be a better predictor in previous studies. Despite these drawbacks, we have shown that the tool is useful in our own setting with easy accessibility of the parameters; this can be used to help prioritise patients at risk of adverse outcomes. A lower cut-off value is required to determine patients at risk of adverse events; this has not been explored in middle-income countries. In addition, haemoglobin as a parameter requires review as stated by Meidema et al.¹⁹ We have documented this important finding for the first time in sub-Saharan Africa. However, a question still remains about the feasibility of reassessment in low-income countries as well as the possibility of developing tools that better predict adverse outcomes without the need for reassessment, as the time lag may constitute a clinical risk for patients.

Acknowledgements

The authors would like to acknowledge all staff at the Red Cross War Memorial Children’s Hospital haematology oncology service.

This article includes content that overlaps with research originally conducted as part of Motunrayo O. Adekunle’s master’s thesis titled ‘Prevalence, Risk Factors and Predictors of Adverse Outcomes of Febrile Neutropenia in Oncology Patients on Chemotherapy at the Red Cross War Memorial Children’s Hospital: A Three Year Retrospective Study’, submitted to the Faculty of Medicine, University of Cape Town in 2021. The thesis was supervised by Marc Hendricks and Alan Davidson. Portions of the data, analysis, and/or discussion have been revised, updated, and adapted for journal publication. The original thesis is publicly available at: https://hdl.handle.net/11427/36504 content. The author affirms that this submission complies with ethical standards for secondary publication, and appropriate acknowledgement has been made of the original work.

A related article, focusing on the risk factors and predictors of outcome in febrile neutropenia, has been published in the SA Journal of Oncology, 2023; 7(0), a232. The present article addresses a distinct research question, focusing on the validation of the febrile neutropenia tool, taking the predictors into consideration.

The authors declare that they have no financial or personal relationships that may have inappropriately influenced them in writing this article.

M.O.A. contributed to the conceptualisation, manuscript development and data analysis. A.D. and M.H. supervised the study and contributed to the manuscript development. All authors contributed to the article, discussed the results, and approved the final version for submission and publication.

This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

The data that support the findings of this study are available from the corresponding author, M.O.A., upon reasonable request.

The views and opinions expressed in this article are those of the authors and are the product of professional research. The article does not necessarily reflect the official policy or position of any affiliated institution, funder, agency or that of the publisher. The authors are responsible for this article’s results, findings and content.

References