Predicting chemotherapy-induced febrile neutropenia outcomes in adult cancer patients: an evidence-based prognostic model.

Abstract

Aims: This thesis explored and examined the clinical factors associated with the outcomes of chemotherapy-induced febrile neutropenia for adult cancer patients and confirms the independent predictive value of these factors. Established as predictors, the factors were used to formulate a multivariable prognostic model to stratify patients according to their risk groupings (high- or low-risk) for adverse outcomes for febrile neutropenia. Newly developed models underwent preliminary validation for their performance as prognostic models for febrile neutropenia outcomes. Background: Accuracy in risk stratification for cancer patients presenting with chemotherapy-induced febrile neutropenia is of critical importance. Serious morbidity may result when treatment is tailored according to misclassified levels of risk. New predictors and prediction tools used for risk stratification have been reported in the recent years. A systematic review was conducted on this topic as part of the thesis and the findings showed a lack of conclusive information on predictive values for some factors identified as predictors, and limitations in prognostic research studies’ methodologies which affect the internal and external validity of the risk prediction tools. Methods: Clinical factors identified through the systematic review contributed to the candidate factors investigated. Additional factors were also included based on other primary studies not included in the systematic review. A retrospective review of patients’ medical records was conducted. Tests of association using univariate analysis were conducted on these variables. Significant variables were tested and adjusted for confounders in a multivariate logistic regression analysis to formulate a multivariable tool for risk stratification of patients presenting with febrile neutropenia. Results: Predictive values for some variables were re-established while some variables failed to demonstrate their predictive values in a univariate analysis. After statistically adjusting to the current factors used in existing prognostic models, a new risk prediction tool was developed predict the risk of adverse outcomes. This tool has been subjected to preliminary validation that confirmed its potential utility. Limitations of the study included single-centre data and the small sample size. Conclusions: Application of a risk prediction tool has its benefits and limitations. However, enhancement of the methodological rigor and comprehensiveness of reporting of results in prognosis research needs to be emphasised for clarity in interpretation and implementation of the studies’ findings. Despite the promising initial validation of the tool developed in this thesis, further extensive validation and evaluation of the tool’s performance are needed to show the true impact of the tool on clinical practice.