The research team enrolled 202 patients beginning their first postoperative adjuvant chemotherapy at a tertiary hospital in Haikou, China, between August 2025 and June 2026. After accounting for withdrawals and incomplete data, 110 patients entered the primary analysis, contributing 4,134 valid assessments out of 4,620 planned, an overall completion rate of 89.5%. Participants rated nausea, vomiting, fatigue, loss of appetite, diarrhea, constipation, abdominal distension, abdominal pain, sleep disturbance, anxiety, depression, dry mouth, and numbness or tingling in the hands and feet on 0-to-10 scales during three daily windows: morning, afternoon, and evening.

The study's central methodological innovation was its use of a multilevel vector autoregressive (mlVAR) model, which treats each symptom as a node in a network and estimates directed, time-lagged associations between symptoms measured hours apart. Unlike cross-sectional surveys, this approach can reveal whether a symptom at one assessment predicts changes in another symptom at the next one, while separately estimating contemporaneous associations among symptoms occurring at the same moment. The researchers stress that these are predictive associations within short observation intervals, not proof of causal chains.

The trajectory findings were strikingly dynamic. All 13 symptoms showed statistically significant time effects after false-discovery-rate correction. Fatigue, loss of appetite, abdominal distension, sleep disturbance, and dry mouth tended to peak in the middle of the post-chemotherapy week and then decline, while nausea and anxiety generally decreased over time.

Depression and numbness or tingling in the hands and feet showed cumulative trends, rising during the later portion of the observation window, a pattern the authors suggest may reflect cumulative neurotoxicity and the psychological weight of ongoing treatment, though these mechanisms were not directly tested. Several symptoms were also more severe in the second chemotherapy cycle than in the first. Fatigue scores rose by 0.567 points, sleep disturbance by 0.394, loss of appetite by 0.338, numbness and tingling by 0.504, and dry mouth by 0.132, all with corrected P values below 0.001, while anxiety declined slightly.

Despite these severity shifts, the overall shape of the short-term symptom trajectories remained broadly similar between cycles, and no statistically significant time-by-cycle interactions emerged. The network analysis identified loss of appetite as the most influential symptom in the temporal network, with the highest out-strength score of 0.629, followed by dry mouth at 0.491 and nausea at 0.457. The strongest individual temporal edge ran from loss of appetite to subsequent fatigue, with a standardized coefficient of approximately 0.123, followed by links from loss of appetite to abdominal distension and sleep disturbance, from sleep disturbance to fatigue, and from nausea to loss of appetite.

In the contemporaneous network, the strongest simultaneous associations appeared between fatigue and sleep disturbance, abdominal distension and abdominal pain, and anxiety and depression. Importantly, the authors emphasize that the strongest single temporal association was modest in magnitude, at roughly 0.12 on a standardized scale. Patient-level bootstrap analysis showed that loss of appetite ranked first in out-strength in 95.5% of resamples, and case-dropping analyses indicated comparatively high stability for temporal centrality estimates, but the researchers caution that these bootstrap intervals reflect precision in the direction of individual edge estimates rather than overall network stability.

Centrality rankings, they note, measure relative prominence within the observed 13-symptom network and should not be read as direct indicators of clinical effect size. Sensitivity analyses tested the robustness of these patterns from multiple angles. When the researchers adjusted for post-chemotherapy time, time of day, supportive medications, demographic and clinical factors, and chemotherapy regimen, the loss-of-appetite-to-fatigue association persisted, though several other edges were attenuated and nausea moved to the top of the out-strength ranking.

Restricting the sample to the 91 patients receiving oxaliplatin-containing regimens (CAPOX or mFOLFOX6) preserved the network structure with correlations of 0.976 and 0.982 against the primary model, and the top three symptoms by out-strength remained unchanged. Comparing the two cycles, the temporal networks showed a structural correlation of 0.663 and the contemporaneous networks 0.697, with permutation tests finding no significant between-cycle differences in global strength or maximum edge weight. The clinical implication, according to the study, is that symptom monitoring after chemotherapy should be continuous and phase-sensitive rather than reliant on isolated assessments.

Because loss of appetite, dry mouth, and nausea showed comparatively prominent outgoing temporal associations, the authors propose them as candidate symptoms for enhanced monitoring and future hypothesis-driven intervention research. The network also suggests a linked monitoring strategy for gastrointestinal symptoms, appetite, sleep, and fatigue, in which nausea and dry mouth preceded changes in appetite, which in turn preceded changes in fatigue and sleep disturbance. However, the authors are explicit that these observational findings do not demonstrate that intervening on any one symptom will reduce subsequent symptom burden, and that unmeasured time-varying factors such as medication exposure, oral intake, hydration, and daily activity may have influenced multiple symptoms simultaneously.

The study carries important limitations. Participants were recruited from a single hospital, and 92 of the 202 enrolled patients did not enter the primary analysis, raising the possibility of selection bias related to digital literacy, health status, or adherence. The three-times-daily schedule may have missed symptom dynamics on shorter timescales, only the first two chemotherapy cycles and 13 prespecified symptoms were assessed, and boundary singular fits occurred in the random-effects structures, prompting the authors to urge cautious interpretation of individual parameter estimates.

The sample-size target was pragmatic and literature-informed rather than simulation-based.

Subject of Research: Temporal symptom network dynamics in early post-chemotherapy colorectal cancer patients assessed by ecological momentary assessment

Article Title: A study on the temporal network characteristics of early post-chemotherapy core symptoms in colorectal cancer patients based on ecological momentary assessment

Article References: Lv, Y., Xu, L., Li, M., Cai, X., Ye, L., Wang, X., Wang, X., Chen, Y., & Lu, Y. (2026). A study on the temporal network characteristics of early post-chemotherapy core symptoms in colorectal cancer patients based on ecological momentary assessment. Frontiers in Oncology, 16, Article 1942076. https://doi.org/10.3389/fonc.2026.1942076

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1942076

Keywords: colorectal cancer, chemotherapy, ecological momentary assessment, symptom networks, mlVAR, fatigue, loss of appetite, nausea, sleep disturbance, patient-reported outcomes, nursing, longitudinal study