This cross-sectional study was conducted at the Guangdong Provincial People’s Hospital and People’s Hospital of Guangdong Province Ganzhou Hospital (Ganzhou Municipal Hospital), both of which are tertiary referral hospitals. Ethical approval for the use of clinical data for research purposes was granted by the Ethics Committee of Guangdong Provincial People’s Hospital. Between November 2022 and August 2024, we totally enrolled 470 patients from Guangdong Provincial People’s Hospital and People’s Hospital of Guangdong Province Ganzhou Hospital (Ganzhou Municipal Hospital). The flow diagram of patient screening process is shown in Figure 1 . The study objective and data confidentiality were communicated to the participants before participating in the study. According to the declaration of Helsinki, written informed consents were acquired.

The inclusion criteria for the patients were as follows (1): age ≥ 18 years, (2) glioma confirmed through CT, MRI, and pathological examination, (3) diagnosed with glioma for the first time, (4) complete clinical data available, (5) patients with clear consciousness and literacy, capable of participating in scale assessment, (6) patients who voluntarily signed the informed consent form. Exclusion criteria were as follows: (1) psychiatric disorders, impaired mental status, or communication difficulties, (2) patients in pregnancy or lactation period, (3) concurrent diagnosis of other malignant tumors, (4) patients with acute infections or immune system disorders, (5) patients with an estimated postoperative survival time of less than 3 months, (6) history of brain inflammation or neurological deficits, (7) received psychotherapy or intervention within one year.

The basic sociodemographic data of participant was collected using a self-prepared sociodemographic questionnaire in Chinese. The information including age, gender, body mass index (BMI), occupation, marital status, educational level, and personal monthly income. Clinical data were retrieved from the hospital information and management system, including cancer stage, chemotherapy, self-care ability, insomnia, neuroticism, pain, depression, anxiety, and physical exercise. All the information was collected one month after the patient received treatment.

In this study, the Numerical Rating Scale (NRS) was used to measure the levels of cancer-related fatigue and the intensity of pain experienced by patients with glioma (17). The NRS typically consists of a horizontal line with numbers ranging from 0 to 10. The endpoints of the scale represented the levels of cancer-related fatigue and pain, with 0 indicating no fatigue or no pain and 10 indicating the worst possible fatigue or the worst possible pain.

In order to assess sleep quality and disturbances of glioma patients over a one-month time interval, the Pittsburgh Sleep Quality Index (PSQI) was used in this study (18). Each component score of the PSQI ranges from 0 to 3, with 3 indicating the greatest dysfunction or disturbance. The seven component scores are then summed to obtain a global PSQI score, which ranges from 0 to 21. Higher scores indicate poorer sleep quality.

The Perceived Social Support Scale (PSSS) was used to measure the patients’ perceptions of the support they received from their social network (19, 20). It includes items prompting individuals to rate their agreement with statements on social support. Patients rated their responses on a Likert scale of 1 to 7 or 1 to 5, where higher scores indicating stronger perceived social support.

Health-related quality-of-life was measured by the 36-item Short Form Health Survey (SF-36) (21, 22). This is a patient-reported questionnaire that covers eight health domains: physical functioning (PF), role physical (RP), bodily pain (BP), general health (GH), vitality (VT), social functioning (SF), role emotional (RE), and mental health (MH). Different options for each item have different score weightings. The final score ranges from 0 (worst general health status) to 100 (best health status) for each domain, and higher scores indicate better quality of life.

In this study, all patients were categorized into two groups based on their NRS scores of cancer-related fatigue: a no or mild fatigue group (scores 0-3) and a moderate to severe fatigue group (scores 4-10). We have included 26 potential risk factors contributing to cancer related fatigue in glioma patients. There were 10 demographic and clinical factors, including age, BMI, occupation, marital status, education level, personal monthly income, cancer stage, chemotherapy, self-care ability, and physical exercise. There were 16 mental and psychological factors, including sleep quality, neuroticism, intensity of pain, depression, anxiety, perceived social support, quality-of-life, and so on.

Prior to risk factors selection, 26 potential risk factors were normalized using a z-score approach to eliminate index dimension differences of the data. In the training cohort, risk factor selection and model development were carried out. The least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation was employed to identify the risk factors with non-zero coefficients. Subsequently, a multivariable logistic regression analysis was performed on the risk factors identified by the LASSO model to determine significant risk factors for nomogram development.

A nomogram was constructed with the significant risk factors extracted by LASSO model and multivariable logistic regression analyses. The receiver operating characteristic (ROC) curve and area under the curve (AUC) illustrated the predictive accuracy and conformity of this nomogram in the training cohort and validation cohort. The C-index and calibration plots visually revealed the predictive accuracy of nomogram. Decision curve analysis (DCA) was used to evaluate the clinical net benefit of nomogram.

All statistical analyses were performed by SPSS software version 25.0 (IBM Corporation; United States) and R software (version 3.4.1, Boston, MA, USA). The ages, BMI, and the scores of NRS, PSQI, PSSS and SF-36 were continuous variables approximately normally distributed, which were described using means and standard deviation (SD). The descriptive statistics for categorical variables were used frequencies and percentages. The differences between the training and validation cohorts in general characteristics were analyzed by Student’s t test or Pearson’s χ2 test. The multivariable logistic regression analyses were conducted to determine the hazard ratios (HR) and 95% confidence intervals (CI) for events. The variables with p<0.05 in multivariate logistic regression analyses were considered as the significant risk factors leading to cancer-related fatigue in glioma patients. We considered p<0.05 to be the statistically significance.

A total of 470 glioma patients were included in this study, with 284 patients as the training cohort and 186 patients as the validation cohort. The general characteristics of two cohorts were summarized in Table 1 . No statistical difference was observed between two cohorts, indicating the grouping was randomized and reasonable. The Numerical Rating Scale (NRS) score of cancer-related fatigue was 4.18 ± 2.77 for training cohort and 4.09 ± 2.84 for validation cohort. In the training cohort, 129 glioma patients experienced no or mild cancer-related fatigue, whereas 155 patients reported moderate or severe cancer-related fatigue. In the validation set, 104 patients were divided into the no or mild fatigue group, and 82 patients were divided into the moderate or severe fatigue group.

After LASSO algorithm, 12 risk factors were retained based on non-zero coefficients for the prediction of cancer-related fatigue ( Figure 2 ). These factors included occupation, cancer stage, sleep quality, anxiety, the PSSS score, and the seven health domains of the SF-36, excluding role physical.

To determine the independent risk factors leading to cancer-related fatigue of glioma patients, multivariable logistic regression analyses were performed in the training cohort. As shown in Table 2 , the cancer stage (p=0.014), and the scores of PSSS (p<0.001), physical functioning (PF) (p<0.001), bodily pain (BP) (p=0.031), general health (GH) (p<0.001), and mental health (MH) (p=0.009) were the independent risk factors for cancer-related fatigue of glioma patients.

A clinically quantitative predictive model nomogram was developed based on the independent risk factors extracted above ( Figure 3 ). The nomogram demonstrated excellent forecasting ability, with the concordance-index was 0.964 (0.935-0.993). For each glioma patient, higher total points indicated as a greater risk of developing cancer-related fatigue. We further examined the predictive ability of this nomogram on both the training and validation cohorts. The AUC values of nomogram were 0.916 (CI: 0.879-0.953) in the training cohort ( Figure 4A ) and 0.885 (CI: 0.829-0.941) in the validation cohort ( Figure 4B ). The calibration curves of this nomogram exhibited a notable concordance with the ideal diagonal line ( Figures 4C, D ). Moreover, the decision curve analyses (DCA) illuminated that the optimal utility of this nomogram which achieved the high clinical net benefit across the entire range of reasonable threshold probabilities in both training and validation cohorts ( Figures 4E, F ).