For the hundreds of thousands of people treated for colorectal cancer each year, the surgery is only half the battle. The adjuvant chemotherapy that follows—given to mop up residual disease—is notorious for producing a tangled web of side effects: fatigue, nausea, neuropathy, anxiety, depression, pain, and sleep disturbance that rarely arrive one at a time. A new prospective longitudinal study published in Supportive Care in Cancer has now applied an unusual tool borrowed from statistical physics to map how this symptom burden reorganizes itself over the course of treatment, and the results reveal a striking asymmetry in how different symptom groups behave as chemotherapy progresses.
The research team, led by investigators at Soochow University in Suzhou, China, together with collaborators at NYU Rory Meyers College of Nursing and the University of Nevada, Reno, followed 380 patients undergoing adjuvant chemotherapy after colorectal cancer surgery. Rather than simply counting symptoms at each visit, the researchers modeled the entire symptom landscape as a network—a web of interacting nodes in which each symptom influences the others. They then estimated a quantity called network temperature, a measure of how freely the system fluctuates among different possible configurations of symptom states.
The concept comes from the Ising model, a mathematical framework originally developed to describe magnetism in physics, in which units flip between two states—in this case, the presence or absence of a symptom—under the influence of their neighbors. In a hot network, symptoms behave erratically: a patient’s overall symptom configuration can shift dramatically from one moment to the next, with different combinations of problems appearing and disappearing. In a cold network, the system is locked into a small number of stable configurations, meaning the symptom pattern a patient experiences tends to persist. Temperature, in this sense, is a direct readout of the stability of a person’s symptom experience.
To carry out the analysis, the team assessed patients at three stages of chemotherapy and included only symptoms with a prevalence of at least 20 percent at every assessment, ensuring the network contained problems that were genuinely common across the treatment course. They fitted multigroup Ising network models to the data and used the Spinglass community detection algorithm—another import from statistical physics—to identify clusters of symptoms that hang together more strongly with each other than with the rest of the network. Confidence intervals for the temperature estimates were generated through 1,000 patient-level nonparametric bootstrap resamples, a resampling technique that quantifies how much uncertainty surrounds each measurement.
Three symptom clusters emerged from the analysis: an emotional distress cluster, a somatic–gastrointestinal cluster, and a fatigue–neuropathy cluster. This grouping mirrors what clinicians observe at the bedside, where physical complaints such as nausea, appetite loss, and bowel symptoms tend to travel together, while fatigue and chemotherapy-induced peripheral neuropathy form their own distinct axis of toxicity.
The headline finding was that the overall network temperature of the symptom system decreased across the three chemotherapy stages. In plain terms, as treatment went on, patients’ symptom configurations became more constrained—less prone to wholesale rearrangement from one assessment to the next. The symptom experience, chaotic at the outset of therapy, gradually settled into more predictable patterns. This is the first time network temperature has been applied to symptom science in oncology, adapting a metric previously used to study the stability of depression symptoms across adolescence.
But the cluster-level results told two very different stories. The fatigue–neuropathy cluster showed a marked and steady cooling: its relative temperature fell from a normalized reference value of 1.00 at the first assessment to 0.60 (95 percent confidence interval 0.49–0.74) at the second stage and 0.45 (0.36–0.55) at the third. That means the way fatigue and neuropathy co-occurred in individual patients became progressively more locked-in over the course of chemotherapy. The authors suggest this reflects the biology of these toxicities—both are driven by cumulative, treatment-related mechanisms such as the metabolic disruptions underlying cancer-related fatigue and the nerve damage produced by oxaliplatin, a standard adjuvant drug for stage III colon cancer whose neurotoxicity is well documented in recent trials.
The emotional distress cluster moved in the opposite direction. Its temperature rose over time, from the reference value of 1.00 to 1.06 (0.96–1.19) at stage two and 1.12 (1.01–1.25) at stage three. Although the confidence intervals overlap the reference value at the second stage, the upper trajectory indicates that anxiety and depression maintained—or even increased—their configurational variability throughout treatment. Unlike the physical toxicities, which become entrenched, the emotional dimension of the patient experience remained fluid, with different patterns of distress surfacing at different points. This is consistent with a broader literature showing that fear of cancer recurrence, coping strategies, and psychological adjustment continue to evolve long after the acute physical side effects have stabilized, and that anxiety and depression among colorectal cancer survivors are common and closely tied to quality of life.
The clinical implications of this asymmetry are significant. If physical symptom clusters become stable and predictable as chemotherapy proceeds, clinicians can, in principle, anticipate them: a patient showing fatigue and neuropathy early in treatment is likely to retain that configuration, making it a sensible target for early, sustained interventions such as dose modification, exercise programs, or neuropathy management protocols. Emotional distress, by contrast, demands repeated, longitudinal screening, because its presentation is not fixed—what a patient reports at week six may bear little resemblance to what they report at week eighteen. The authors argue that their findings highlight the importance of longitudinal, cluster-informed symptom assessment during chemotherapy, rather than one-off symptom surveys that may miss the dynamic structure of the burden.
The study also demonstrates the practical value of importing tools from network science into supportive cancer care. Network analysis of multivariate data has matured rapidly in psychological science over the past decade, and metrics such as network temperature offer a way to quantify not just which symptoms a patient has, but how stable and interdependent their symptom system is—a dimension that conventional severity scores cannot capture. For colorectal cancer, a disease whose global incidence continues to climb according to the latest GLOBOCAN estimates, and whose survivors face months of adjuvant therapy, understanding the architecture and dynamics of symptom networks may prove as important as understanding the drugs that cause them. The work was supported by the Major Project of Philosophy and Social Science Research in Colleges and Universities in Jiangsu Province, and the datasets are available from the corresponding author on reasonable request.
Subject of Research: Temporal stability of symptom clusters measured by network temperature in colorectal cancer patients receiving adjuvant chemotherapy
Article Title: Network temperature and stability of symptom clusters in colorectal cancer patients undergoing adjuvant chemotherapy: a prospective longitudinal study
Article References: Cao, C., Liu, S., Shan, S., Qi, X., Mao, W., Yang, Z., & Tian, L. (2026). Network temperature and stability of symptom clusters in colorectal cancer patients undergoing adjuvant chemotherapy: a prospective longitudinal study. Supportive Care in Cancer, 34(10), Article 951. https://doi.org/10.1007/s00520-026-11188-7
Image Credits: AI Generated
DOI: 10.1007/s00520-026-11188-7
Keywords: colorectal cancer, adjuvant chemotherapy, symptom clusters, network analysis, network temperature, Ising model, fatigue, peripheral neuropathy, emotional distress, supportive care, longitudinal study, patient-reported outcomes
News Source: Nathaniel Bowman. (October 18, 2026). Symptom Networks Cool Down During Chemotherapy, but Emotional Distress Refuses to Settle. Scienmag.
