Development and validation of a risk prediction model for acute chemotherapy-induced nausea and vomiting in pediatric patients with cancers

Introduction

Chemotherapy-induced nausea and vomiting (CINV) is one of the most common and distressing adverse effects of chemotherapy (1). Despite the use of new prophylactic antiemetic drugs, acute CINV still occurs in 60–80% of adults (1), and 80–95% of children (2). CINV is not only an extremely uncomfortable experience but also leads to serious consequences, such as dehydration, electrolyte imbalance, chemotherapy interruption impaired quality of life, and increased medical costs (3,4).

In adults, several risk assessment instruments for CINV have been developed. Identified risk factors include female sex, age younger than 60 years, presence of anxiety, alcohol consumption, history of morning sickness, history of CINV, and Asian ethnicity (5-8). Dranitsaris et al. have developed a widely used CINV prediction model in adults, but it demonstrated low specificity (38.4%) and lacked external validation (9). Hu et al. have developed a nomogram that was only applicable to the first cycle of chemotherapy (10). Such instruments in children are scarce, because the pattern of CINV is different in children from that in adults, depending on age, developmental issues, lifestyle, treatment regimens, and disease acuity (11). In children, chemotherapy-induced nausea (CIN) and chemotherapy-induced vomiting (CIV) have different patterns of risk factors, thus they are treated as two health outcomes, which are different from that in adults. The incidence of CIN is higher than CIV, and active control of CIN can reduce the concurrence of CIV (5,12). In addition, the Multinational Association of Supportive Care in Cancer (MASCC), the European Society of Medical Oncology (ESMO) (13), the American Society of Clinical Oncology (ASCO) (14), and the National Comprehensive Cancer Network (NCCN) (15), have developed guidelines regarding treatment of acute CINV in children. However, there is no international risk assessment tool for CINV in children.

Therefore, this study aimed to develop predictive models for acute CIN and CIV in pediatric cancer patients and validate them externally. Health professionals can then identify children at risk of CINV in advance and provide timely, personalized interventions to mitigate its adverse effects. We present this article in accordance with the TRIPOD reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2024-629/rc).

Methods

Study population

For the training cohort, we recruited 378 children who underwent chemotherapy at the Children’s Hospital, Zhejiang University School of Medicine from August 1, 2022, to March 31, 2023. In the validation cohort, we recruited 230 children in the same hospital from 1 May to 31 August 2023. The inclusion criteria were as follows: (I) hospitalized children diagnosed with tumors requiring chemotherapy; (II) children with ≥1 cycle of intravenous chemotherapy; (III) children aged 1 month to 17 years; and (IV) caregivers who had been caring for the children in the last three months. The exclusion criteria were as follows: children with major physical, psychiatric, or mental disorders.

Ethics approval statement

The study protocol was conducted in accordance with the Declaration of Helsinki and its subsequent amendment (16), and was approved by the Medical Ethics Committee of the Children’s Hospital, Zhejiang University School of Medicine (No. 2022-IRB-223). All caregivers of the patients signed a written informed consent form and voluntarily participated in this study.

Data collection

A self-developed questionnaire was used to collect data. We selected potential risk factors for acute CINV in children based on literature, experts’ opinions, and pediatric CINV prevention guidelines (5,17). We measured demographic factors, disease-related factors, and chemotherapy-related factors. The demographic factors were gender, age, height, weight, body surface area, history of sickness of car, sea and motion, and maternal vomiting during pregnancy. Disease-related factors were history of allergies, negative emotion or mood changes, sleeping time before chemotherapy, tumor type, previous CINV experience, secondhand smoke exposure at household, and health status in the past month. Chemotherapy-related factors were the grade of emetogenicity caused by chemotherapy drugs and the cycle of chemotherapy.

Caregivers recorded the number of occurrences of nausea and vomiting daily using the CINV diary card. The investigators verified and entered the data of the diary into the database daily no later than 24 hours after the end of chemotherapy. CIN was assessed using the BARF scale (18), while CIV was classified into 5 levels based on the United States Department of Health and Human Services 2017 Common Adverse Events Evaluation Criteria (version 5.0) (19). Caregivers reported whether the child had one or more negative emotions or mood changes, such as irritability, frequent crying, restlessness, fear, high alertness, nervousness, etc. The grade of risk for emetogenicity caused by chemotherapeutic drugs was rated as follows:

• Mild risk: <10% of acute chemotherapy emetogenicity;
• Low risk: 10–30%;
• Moderate risk: >30–90%;
• High risk: >90% (11).

In our study, all pediatric patients received standardized antiemetic prophylaxis based on the emetogenic risk of their chemotherapy regimen, according to institutional and international guidelines (e.g., MASCC and ESMO recommendations) (13). Including antiemetics as a covariate could introduce collinearity with chemotherapy regimens, because it was intrinsically correlated to the grade of risk for emetogenicity. Given that our model focused on identifying patient-specific and treatment-specific risk factors beyond standardized prophylaxis, we select the grade of risk for emetogenicity as a potential risk factor.

Adequate sleeping time was defined as 12–16 hours of sleep for infants aged <1 years; 11–14 hours for infants aged 1–2 years; 10–13 hours for children aged 3–5 years; 9–12 hours for children aged 6–12 years; and 8–10 hours for children aged 13–18 years (20).

Statistical analysis

We conducted a descriptive analysis to summarize the data. Normality tests were performed, with means and standard deviations calculated for normally distributed continuous variables and medians with interquartile ranges for non-normally distributed variables. Categorical variables were summarized as numbers and percentages.

The training cohort was used to establish a nomogram for predicting the risk of CIV and CIV, respectively. We applied the two independent samples t-tests (for normally distributed continuous data), Mann-Whitney U test (for non-normally distributed continuous data), and Chi-square tests (for categorical data) to compare differences across groups. Variables with a statistical significance in the last step were selected for the multivariate logistic regression models, and variables that remained statistically significant were selected for establishing the risk prediction model, which was displayed as a nomogram. To assess the model goodness of fit, adjusted R square and the Hosmer-Lemeshow test were used. The receiver operating characteristic (ROC) curve and the area under the curve (AUC) were used to assess the best specificity and sensitivity of the risk prediction model. Calibration curves were employed to evaluate the model’s accuracy and conformity. Both discrimination and calibration were assessed by bootstrapping with a total of 400 resamples. Moreover, decision curve analysis (DCA) reflected the net benefit of the model. ROC curves, calibration curves, and decision curves were illustrated for both the training cohort and the validation cohort.

All statistical analyses were performed using R version 4.2 with the “rms”, “pROC”, “ggplot2” and “dca” packages. Statistical significance was set at P<0.05.

Results

General characteristics

In total, 378 children were included in the training data set to develop the prediction models for acute CIN and CIV. The average age was 6.07 years and the standard deviation was 3.65. Of these children, 143 (37.8%) were girls and 235 (62.2%) were boys; 253 (66.93%) had the acute CIN, and 174 (46.03%) had the acute CIV; 46 out of 174 had grade I vomiting; 29 had grade II vomiting, and 13 had grade III vomiting; 171 (45.24%) had hematological tumors, and 207 (54.76%) had solid tumors.

Selection of predictors for CIN and CIV

Tables S1,S2 present the results of the comparison of the difference between children with and without acute CIN and CIV across subgroups. Eleven variables that had a statistical significance (P<0.05) were selected for the multivariate logistic regression models as potential predictors for acute CIN and CIV, respectively. Tables 1,2 show the results of logistic regression models for CIN and CIV, respectively. Predictors of CIN were nausea and vomiting in previous treatment (OR =3.65, 95% CI: 2.25, 5.92, P<0.001), body weight (OR =1.04; 95% CI: 1.02, 1.06; P=0.001) and anxiety before chemotherapy (OR = 3.21; 95% CI: 1.82, 5.66; P<0.001). Five predictors for CIV were the cycle of chemotherapy (OR =2.07; 95% CI: 1.24, 3.45; P=0.005), classification of the risk of chemotherapy emetogenicity for commonly used chemotherapy medicine (OR =1.97; 95% CI: 1.07, 3.61; P=0.03), having adequate sleeping time during chemotherapy (OR =2.54; 95% CI: 1.43, 4.49; P<0.001), having solid tumor (OR =2.94, P<0.001) and nausea/vomiting in previous treatment (OR =2.93; 95% CI: 1.79, 4.79; P<0.001).

Development, predictive accuracy, and net benefit of the risk prediction nomogram

Results from supplementary Table S3 indicate that the training and validation data are comparable. Figure 1 presents the nomograms for predicting acute CIN, ROC, calibration, and decision curves. Three predictors identified through logistic regression analysis were incorporated into the nomogram (Figure 1A):

• Body weight (kg);
• History of nausea and vomiting in previous therapy; and
• Negative emotions or mood changes during chemotherapy.

Figure 1 Development, calibration, and evaluation of prediction model for the acute CIN. (A) Nomogram predicting the CIN; (B,C) calibration curves for the nomogram with the training and validation cohorts, respectively; (D,E) ROC for the nomogram, with the training and validation cohorts, respectively; (F,G) decision curves for the nomogram, with the training and validation cohorts, respectively. CIN, chemotherapy-induced nausea; ROC, receiver operating characteristic.

The model demonstrated an R² value of 0.217 and a C-index of 0.736. Higher total points on the nomogram corresponded to an increased risk of acute CIN. For instance, a body weight of 30 kg corresponded to a score of 26 on the total score scale. If a child with a body weight of 30 kg, no nausea and vomiting in previous therapy, and had negative emotion or mood changes during chemotherapy, his/her total score is around 70 (i.e., 26+0+44), which corresponded to the risk score of 0.85. In this case, we can say that the probability of such a child having acute CIN was about 85%. In the training dataset, the AUC was 0.744 (Figure 1B); while in the validation dataset (n=230), the AUC was 0.568 (Figure 1C). Figure 1D shows the calibration curve in the training dataset which was close to the ideal diagonal line, and the Hosmer-Lemeshow goodness of fit (GOF) test demonstrated that the model had a good fit (P=0.81). In the validation dataset, the calibration curve is not very close to the ideal diagonal line (Figure 1E), and the P value of the Hosmer-Lemeshow test is 0.02. Moreover, the decision curve in the training cohort showed a significantly better net benefit, while it showed a worse net benefit of the predictive model compared with that of the training cohort (Figure 1F,1G).

Figure 2 presents the nomogram for predicting acute CIV, ROC, calibration and decision curves. Five predictors identified by the logistic regression analysis were included to develop the nomogram (Figure 2A). In the training dataset, the AUC was 0.736 (Figure 2B). The calibration curve was close to the ideal diagonal line (Figure 2D), and the Hosmer-Lemeshow test demonstrated that the model had a good fit (p=0.31). Furthermore, the decision curve shows a significantly better net benefit in the predictive model (Figure 2F). In the validation cohort, the AUC was 0.719 (Figure 2C). The calibration curve is very close to the ideal diagonal line (Figure 2E), and the P value of the Hosmer-Lemeshow test is 0.14. Moreover, the decision curve also shows a significant net benefit of the predictive model (Figure 2G).

Figure 2 Development, calibration, and evaluation of prediction model for the acute CIV. (A) Nomogram predicting the CIV; (B,C) calibration curves for the nomogram with the training and validation cohorts, respectively; (D,E) ROC for the nomogram, with the training and validation cohorts, respectively; (F,G) decision curves for the nomogram, with the training and validation cohorts, respectively. CIN, chemotherapy-induced nausea; ROC, receiver operating characteristic.

Discussion

In this study, we identified previous CINV experience, negative emotion or mood changes during chemotherapy, and body weight as predictors for acute CIN, while previous CINV experience, inadequate sleeping time before chemotherapy, three or more chemotherapy cycles, high emetogenicity of chemotherapy regimens, and solid tumors were predictors for acute CIV.

Having a history of CINV in previous chemotherapy was a predictor for both acute CIN and CIV in children. This may be explained by the past experience of nausea and vomiting, making the reflex pathway more active, which may lead to a declined threshold of the CINV response (21). An early study also showed that the previous CINV experience increased at least sixfold in subsequent cycles if it was poorly controlled in the first cycle (22). Negative emotions or mood changes were significant predictors of acute CIN, consistent with findings in adults. Vanbockstael et al. identified anxiety as an independent risk factor for CINV (23). The potential explanation was that negative emotions may active the hypothalamic-pituitary-adrenal axis, leading to elevated levels of adrenocorticotropic hormones, cortisol, and growth hormone-releasing peptide, which can cause autonomic symptoms, thus inducing nausea (24). We recommend that caregivers and healthcare professionals closely monitor children’s emotional well-being during chemotherapy. Collaboration with pediatric psychologists and medical social workers may be necessary to address negative emotions effectively.

Body weight was also associated with acute CIN in children. Furlanetto et al. suggested that body surface is an independent risk factor for CINV in adults (22). However, children’s weight is more dynamic due to growth, development, and disease-related factors, making it a more sensitive index than body mass index (BMI) and body surface area. Clinicians and nurses need to adjust the dose of chemotherapy drugs according to the daily monitored weight. Our study suggested that having solid tumors was associated with a higher risk of acute CIN. There have been no earlier reports on tumor types and the risk of CINV in children. It might be connected to factors such as surgery, discomfort from surgical incisions thereafter, and different chemotherapy treatments given to children with solid tumors.

We found that inadequate sleeping time was a significant risk factor for acute CIV. Several studies in adults have shown that repeated chemotherapy predisposes may disturb the circadian rhythm and activate the autonomic nervous system (9,25-27). In our study, undergoing three or more cycles of chemotherapy was significantly associated with acute CIV. Children may be affected by factors such as drugs, environment, and discomfort during chemotherapy, and as chemotherapy cycles increase, these factors have an accumulated effect and promote the development of acute CIV in children. In contrast, Dranitsaris et al. reported a negative correlation between the number of chemotherapy cycles and CINV occurrence in adults (9). Chemotherapeutic drugs with high emetogenicity were also identified as key predictors for acute CIV in children, which is consistent with the results of several studies in adults (8). Patients receiving highly emetogenic chemotherapy regimens are more likely to develop CINV. Our study suggested that having solid tumors was associated with a higher risk of acute CIN. There have been no earlier reports on tumor types and the risk of CINV in children. It might be connected to factors such as surgery, discomfort from surgical incisions thereafter, and different chemotherapy treatments given to children with solid tumors.

Strengths and limitations

To the best of our knowledge, the present study is the first to develop a prediction model for assessing the risk of acute CIN and CIV in pediatric cancer patients. Our findings highlight that risk factors identified in adults—such as age, gender, history of motion sickness, or exposure to household secondhand smoke—are not directly applicable to children. This model effectively differentiates between children with or without CIN/CIV, enabling healthcare professionals to implement pharmacologic and nonpharmacologic interventions to prevent adverse outcomes. In addition, the prediction model also provides separate predictive assessments of the risk of CIN and CIV, helping clinicians and nurses to effectively identify children at high risk of CIN and actively control the occurrence of CIN through psychological interventions.

However, there are several limitations in this study. First, the study population was derived from a single children’s hospital, limiting the generalizability of the findings. Second, the present prediction model applies to acute CINV and does not account for anticipatory or delayed CINV. Nevertheless, addressing and intervening in acute CINV during chemotherapy may help mitigate the occurrence of delayed or anticipatory CINV in subsequent cycles. Dupuis et al. demonstrated that the acute CINV control significantly impacts delayed-phase symptoms (17), which provided critical methodological and conceptual guidance for our subsequent research endeavors on delayed CINV in pediatric patients with cancers.

Conclusions

The prediction models developed in the present study had relatively good fitness. It provides clinical healthcare professionals with an effective and easy-to-use tool for predicting the risk of having CINV; thus, they could provide timely and personalized interventions to prevent CINV and to reduce adverse events associated with CINV before chemotherapy.

Acknowledgments

We gratefully thank all the patients and caregivers who were willing to participate in this study. In addition, we thank the Pediatric Evidence-based Medical and Clinical Research Laboratory, for the great support for methodology.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2024-629/rc

Data Sharing Statement: Available at https://tp.amegroups.com/article/view/10.21037/tp-2024-629/dss

Peer Review File: Available at https://tp.amegroups.com/article/view/10.21037/tp-2024-629/prf

Funding: This study was funded by Zhejiang Medical and Health Science and Technology Project (grant No. 2024647912) and the Zhejiang Provincial Basic Public Welfare Research Program Project (grant No. LGF22F030003).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2024-629/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study protocol was conducted in accordance with the Declaration of Helsinki and its subsequent amendment, and was approved by the Medical Ethics Committee of the Children’s Hospital, Zhejiang University School of Medicine (No. 2022-IRB-223). All caregivers of the patients signed a written informed consent form and voluntarily participated in this study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sommariva S, Pongiglione B, Tarricone R. Impact of chemotherapy-induced nausea and vomiting on health-related quality of life and resource utilization: A systematic review. Crit Rev Oncol Hematol 2016;99:13-36. [Crossref] [PubMed]
2. Flank J, Sparavalo J, Vol H, et al. The burden of chemotherapy-induced nausea and vomiting in children receiving hematopoietic stem cell transplantation conditioning: a prospective study. Bone Marrow Transplant 2017;52:1294-9. [Crossref] [PubMed]
3. Ilyas AB, Bahaj RK, Shaikh AA, et al. Breast Cancer Patients’ Perceptions of Their Experience With Chemotherapy-Induced Nausea and Vomiting and Its Impact on Quality of Life in Jeddah, Saudi Arabia. Cureus 2020;12:e12038. [Crossref] [PubMed]
4. Adel N. Overview of chemotherapy-induced nausea and vomiting and evidence-based therapies. Am J Manag Care 2017;23:S259-65.
5. Patel P, Robinson PD, Cohen M, et al. Prevention of acute and delayed chemotherapy-induced nausea and vomiting in pediatric cancer patients: A clinical practice guideline. Pediatr Blood Cancer 2022;69:e30001. [Crossref] [PubMed]
6. Di Mattei VE, Carnelli L, Carrara L, et al. Chemotherapy-Induced Nausea and Vomiting in Women With Gynecological Cancer: A Preliminary Single-Center Study Investigating Medical and Psychosocial Risk Factors. Cancer Nurs 2016;39:E52-9. [Crossref] [PubMed]
7. Mosa ASM, Hossain AM, Lavoie BJ, et al. Patient-Related Risk Factors for Chemotherapy-Induced Nausea and Vomiting: A Systematic Review. Front Pharmacol 2020;11:329. [Crossref] [PubMed]
8. Puri S, Hyland KA, Weiss KC, et al. Prediction of chemotherapy-induced nausea and vomiting from patient-reported and genetic risk factors. Support Care Cancer 2018;26:2911-8. [Crossref] [PubMed]
9. Dranitsaris G, Molassiotis A, Clemons M, et al. The development of a prediction tool to identify cancer patients at high risk for chemotherapy-induced nausea and vomiting. Ann Oncol 2017;28:1260-7. [Crossref] [PubMed]
10. Hu Z, Liang W, Yang Y, et al. Personalized Estimate of Chemotherapy-Induced Nausea and Vomiting: Development and External Validation of a Nomogram in Cancer Patients Receiving Highly/Moderately Emetogenic Chemotherapy. Medicine (Baltimore) 2016;95:e2476. [Crossref] [PubMed]
11. Cope DG. Clinical Updates in Nausea and Vomiting. Semin Oncol Nurs 2022;38:151249. [Crossref] [PubMed]
12. Sherani F, Boston C, Mba N. Latest Update on Prevention of Acute Chemotherapy-Induced Nausea and Vomiting in Pediatric Cancer Patients. Curr Oncol Rep 2019;21:89. [Crossref] [PubMed]
13. Roila F, Molassiotis A, Herrstedt J, et al. 2016 MASCC and ESMO guideline update for the prevention of chemotherapy- and radiotherapy-induced nausea and vomiting and of nausea and vomiting in advanced cancer patients. Ann Oncol 2016;27:v119-33. [Crossref] [PubMed]
14. Hesketh PJ, Bohlke K, Lyman GH, et al. Antiemetics: American Society of Clinical Oncology Focused Guideline Update. J Clin Oncol 2016;34:381-6. [Crossref] [PubMed]
15. Network NCC. NCCN Clinical Practice Guidelines in Oncology (NCCN GuidelinesVR). Antichemotherapy emetogenicity. Version 2 2016. Available online: https://www.nccn.org/professionals/physician_gls/pdf/antichemotherapy emetogenicity.pdf
16. World Medical Association Declaration of Helsinki. ethical principles for medical research involving human subjects. JAMA 2013;310:2191-4. [Crossref] [PubMed]
17. Dupuis LL, Tamura RN, Kelly KM, et al. Risk factors for chemotherapy-induced nausea in pediatric patients receiving highly emetogenic chemotherapy. Pediatr Blood Cancer 2019;66:e27584. [Crossref] [PubMed]
18. Watcha MF, Lee AD, Medellin E, et al. Clinical Use of the Pictorial Baxter Retching Faces Scale for the Measurement of Postoperative Nausea in Children. Anesth Analg 2019;128:1249-55. [Crossref] [PubMed]
19. U.S. Department of Health and Human Services. Common Terminology Criteria for Adverse Events (CTCAE) Version 5.0. Available online: https://www.hhs.gov/search-results/index.html?query=Common+Terminology+Criteria+for+Adverse+Events+%28CTCAE%29+Version+5.0&referrer=https%3A%2F%2Fwww.hhs.gov%2Findex.html
20. Paruthi S, Brooks LJ, D’Ambrosio C, et al. Recommended Amount of Sleep for Pediatric Populations: A Consensus Statement of the American Academy of Sleep Medicine. J Clin Sleep Med 2016;12:785-6. [Crossref] [PubMed]
21. Lee KM, Jung DY, Hwang H, et al. Late chronotypes are associated with neoadjuvant chemotherapy-induced nausea and vomiting in women with breast cancer. Chronobiol Int 2017;34:480-91. [Crossref] [PubMed]
22. Furlanetto J, Eiermann W, Marmé F, et al. Higher rate of severe toxicities in obese patients receiving dose-dense (dd) chemotherapy according to unadjusted body surface area: results of the prospectively randomized GAIN study. Ann Oncol 2016;27:2053-9. [Crossref] [PubMed]
23. Vanbockstael J, Coquan E, Gouerant S, et al. How to improve the prevention of chemotherapy-induced nausea and vomiting? The French NAVI study. Support Care Cancer 2016;24:1131-8. [Crossref] [PubMed]
24. DeVuono MV, Parker LA. Cannabinoid Hyperemesis Syndrome: A Review of Potential Mechanisms. Cannabis Cannabinoid Res 2020;5:132-44. [Crossref] [PubMed]
25. Vaid AK, Gupta S, Doval DC, et al. Expert Consensus on Effective Management of Chemotherapy-Induced Nausea and Vomiting: An Indian Perspective. Front Oncol 2020;10:400. [Crossref] [PubMed]
26. Perez-Pozuelo I, Zhai B, Palotti J, et al. The future of sleep health: a data-driven revolution in sleep science and medicine. NPJ Digit Med 2020;3:42. [Crossref] [PubMed]
27. Jung D, Lee KM, Kim WH, et al. Longitudinal Association of Poor Sleep Quality With Chemotherapy-Induced Nausea and Vomiting in Patients With Breast Cancer. Psychosom Med 2016;78:959-65. [Crossref] [PubMed]