Researchers developed models for predicting morning fatigue following administration of chemotherapy in patients with cancer. They presented their findings in the journal Supportive Care in Cancer.

The study included adult outpatients who were receiving chemotherapy for cancer related to the breast, gastrointestinal system, gynecologic system, or the lung. Participants completed questionnaires 1 week before receiving chemotherapy and 1 week after receiving it. The level of morning fatigue was assessed with the Lee Fatigue Scale (LFS).

The researchers conducting the study developed 7 different models based on regression and machine learning (ML) algorithms for predicting morning fatigue. These models incorporated multiple patient characteristics, in addition to pretreatment morning fatigue LFS total scores or scores on LFS scale items for morning fatigue. Models using pretreatment LFS total scores included 145 variables, and models using pretreatment LFS scale items included 157 variables. The research team evaluated these models to identify the best predictors of morning fatigue severity 1 week after receiving chemotherapy.

Continue Reading

Based on available data from both pretreatment and posttreatment time points, there were 1235 patients, mean age 57.0 years, included in the dataset. The patient-reported mean score for morning fatigue 1 week after chemotherapy was 3.63, based on patients’ LFS total scores and when using scores for LFS scale items. The researchers noted that this was higher than a clinically meaningful threshold score of 3.2 on the LFS for morning fatigue.

A type of ML-based model called an elastic net model was found to perform the best overall at fitting data related to morning fatigue. This model predicted a mean score for morning fatigue of 3.63 after chemotherapy when either the pretreatment LFS total scores or LFS scale items were included in the model.

When pretreatment LFS total scores for morning fatigue were included as a variable in the elastic net model, this variable appeared to be the strongest predictor of morning fatigue after chemotherapy. The second most important predictor in this particular analysis was the Karnofsky Performance Status score.

When the elastic net model included pretreatment LFS scale item scores instead of the LFS total scores, the variables that appeared to be most important for predicting morning fatigue after chemotherapy were, in order, the Karnofksy Performance Status score, the Center for Epidemiological Studies–Depression Somatic Subscale score, and scores for 2 individual LFS scale items. The individual scale items were the terms “exhausted” and “worn out”; the remaining terms related to fatigue showed less predictive importance in this model.

“This is the first study to use ML techniques to accurately predict the severity of morning fatigue using both total and individual items from the LFS,” the researchers wrote in their report.


Kober KM, Roy R, Conley Y, et al. Prediction of morning fatigue severity in outpatients receiving chemotherapy: less may still be more. Support Care Cancer. 2023;31(5):253.  doi:10.1007/s00520-023-07723-5