Summary
A longitudinal study linked routine clinical records from 52,236 adults with cancer to identify recurring episodes of significant weight loss. The episodes clustered around disease progression and were associated with consistent blood-marker changes and several tumor mutations.
Researchers have assembled a longitudinal, multimodal analysis of cancer-related weight loss using routine clinical records from 52,236 adults treated at Memorial Sloan Kettering Cancer Center. Published in Nature Medicine on 12 October 2026, the study linked weight measurements with blood tests, tumor sequencing and records of disease progression to examine when significant weight loss occurs and how it relates to other features of cancer.
Cancer cachexia is a wasting syndrome involving progressive, unintentional weight loss and muscle loss. For this analysis, the researchers identified weight-loss episodes from changes in patients’ body mass index (BMI): at least 5% over six months, or at least 2% for people whose baseline BMI was below 20 kg/m². The definition allowed the team to track episodes over time, but it did not capture every aspect of cachexia.
Episodes clustered around progression
The researchers analysed 1,741,505 longitudinal BMI measurements and identified 30,938 qualifying episodes. At least one episode occurred in 46.63% of the MSK cohort; among patients who had an episode, 22.56% experienced more than one. The pattern varied by cancer type: episodes were particularly common in pancreatic and stomach cancers.
Across 18 cancer types, radiographically recorded progression events fell within weight-loss episodes more often than expected under the study’s timing-based comparison in 17. Episodes were also associated with poorer overall survival across these cancer types. These are observational associations: the analysis shows that weight loss and progression often coincided, but cannot establish that one caused the other.
The researchers found a similar episode pattern in an independent cohort of 5,785 patients from Dana-Farber Cancer Institute. That replication supports the usefulness of the episode-detection approach across two health systems, while the cohorts still reflect patients whose care and data were captured at specialist cancer centres.
Blood markers and tumor mutations add context
During weight-loss episodes, routine blood tests showed a broadly consistent pattern across cancers. Albumin and hemoglobin tended to be lower, while markers including alkaline phosphatase and aspartate aminotransferase tended to be higher. For example, 28.4% of albumin measurements during episodes were below the study’s reference range, compared with 12.0% outside episodes. The authors note that some blood changes could also be related to chemotherapy, rather than cachexia alone.
Tumor mutations were associated with episode risk in several cancers. TP53 mutations were linked to increased risk across multiple cancer types, including lung cancer. Such results identify associations between tumor genetics and weight-loss patterns; they do not by themselves show that a mutation directly causes cachexia.
The team also developed models to estimate the risk of future weight-loss episodes in lung, colorectal and prostate cancer. In an external test using a reduced lung-cancer model and 1,159 patients, the model’s concordance index was 0.625, indicating modest ability to distinguish patients who experienced an episode earlier from those who did not. The study presents these models as a potential basis for identifying patients who may need closer monitoring, not as an established clinical decision tool.
A key qualification is that the researchers used weight change as a practical marker, rather than measuring muscle loss across the full cohort. BMI can miss muscle depletion, including in people with higher body weight, and the study lacked enough imaging data to apply the muscle-based component of cachexia definitions broadly. The analysis maps clinically recorded weight-loss episodes at scale; fuller body-composition and prospective clinical studies would be needed to clarify the biology and guide interventions.