Predicting Treatment-Related Fatigue in the Context of Lymphoma and Chronic Lymphocytic Leukemia
Authors: Steve Kalloger¹, Amanda Watson¹, Shawn Sajkowski¹, Natacha Bolaños², Lorna Warwick²
¹ Department of Research & Information, Lymphoma Coalition, Mississauga, ON, Canada
² Management, Lymphoma Coalition, Mississauga, ON, Canada
Introduction
Cancer-related fatigue (CRF) is highly prevalent, significantly impacts patient well-being, and is a hugely under-reported and poorly understood phenomenon.
This study aimed to build a predictive model to help identify patients at risk for CRF based on routinely collected clinical information in the context of lymphoma and chronic lymphocytic leukemia (CLL).
Methods
Study Design
The Lymphoma Coalition conducted the 2022 Global Patient Survey on Lymphomas & CLL to capture patient experiences.
- Total respondents: 8,637
- Patients receiving treatment: 5,135
- Patients reporting incidence of treatment-related side effects (N): 32
Respondents
- Age range with fatigue: 18–87
- Age range without fatigue: 20–90
- Average age (fatigue cohort): 56
- Average age (no fatigue cohort): 56
Sex distribution:
- Female: 38%
- Male: 62%
Statistical Analysis
- Side effects, biological sex, and age were analyzed using multinomial regression with the least absolute shrinkage and selection operator (LASSO) method.
- Model validation used the corrected Akaike Information Criterion (AICc).
- Accuracy was quantified with the receiver operating characteristic (ROC) and a confusion matrix.
Results
- Fatigue identified as a side effect: 67% (N = 3,427) of patients treated for lymphoma or CLL.
- Model accuracy: Correctly identified 85% of patients with CRF.
- Misclassification rate: 24%.
Key Findings
- Significant predictors of CRF included:
- Inability to multitask
- Lack of concentration
- Headaches
- Respiratory problems
- Peripheral neuropathy
- Sexual/intimacy problems
Receiver Operating Characteristic (ROC):
- Area under curve (AUC): 0.880 (very good to excellent classification).
Confusion Matrix
| Actual: Not Selected | Actual: Selected | |
|---|---|---|
| Predicted: Not Selected | 0.558 | 0.145 |
| Predicted: Selected | 0.442 | 0.855 |
- Misclassification Rate: 0.244
Conclusion
Fatigue is an overlooked side effect of many cancer treatments due to its subjective nature.
This analysis demonstrates that the majority of patients at risk for fatigue can be identified using routinely collected clinical data.
The use of machine learning techniques such as LASSO allows integration of many parameters, enabling prediction of risk profiles for patients undergoing therapy.
Limitations
- The model has a high rate of misclassification.
Implications
- This model could be used as an indicator for physicians to initiate conversations with patients about fatigue, ensuring prompt diagnosis and delivery of optimal, timely treatment.
Conflict of Interest
The study was sponsored by AbbVie, BMS, Pharmacyclics, and Roche. None of the authors benefited personally from the research.
Contact Information
- For further details on the LC 2022 GPS: Lymphoma Coalition Global Patient Survey
- Email: amanda@lymphomacoalition.org
Summary
This study by the Lymphoma Coalition analyzed data from the 2022 Global Patient Survey on Lymphomas & CLL to predict cancer-related fatigue (CRF) in patients receiving treatment.
- Prevalence: Fatigue was reported by 67% (3,427 patients).
- Model performance: A predictive model using LASSO regression correctly identified 85% of patients with CRF but had a 24% misclassification rate.
- Key predictors: Inability to multitask, lack of concentration, headaches, respiratory problems, peripheral neuropathy, and sexual/intimacy problems.
- Accuracy: The model showed strong performance with an AUC of 0.880 (very good to excellent).
Key Takeaway
Fatigue remains an overlooked side effect of lymphoma and CLL treatments. This model demonstrates that routinely collected clinical data can help identify patients at risk, supporting earlier conversations and timely interventions by physicians.
Limitations
- High misclassification rate limits precision.
- Further refinement is needed before clinical application.