AI-Powered Map Reveals Cancer Cachexia Research Is Growing Faster Than Science Itself

Cancer cachexia, the devastating wasting syndrome that strips patients of muscle and body weight, has long lived in the shadow of cancer research itself. Now, a sweeping 25-year analysis of the scientific literature has quantified just how quickly the field is catching up, and where it is heading next. The study, published in Cancer Reports, examined 5,775 peer-reviewed publications indexed in the Web of Science Core Collection between 2001 and 2025, and combined classic bibliometric methods with artificial intelligence tools to chart the intellectual landscape of a syndrome that affects roughly a third of all cancer patients.

The headline finding is one of accelerating growth. Publication output in cancer cachexia fits an exponential curve with a coefficient of determination of 0.98, corresponding to an annual growth rate of 6.7 percent. That figure comfortably outpaces the estimated 4.1 percent growth rate of scientific publishing as a whole. Using the statistical rule of 70, the researchers calculated that the number of cachexia papers doubles roughly every 10.4 years, a pace that marks the field as one of the more dynamic corners of oncology research. The authors suggest that the founding of the Society for Sarcopenia, Cachexia and Wasting Disorders in 2008, the launch of the Journal of Cachexia, Sarcopenia and Muscle in 2010, and the creation of the Cancer Cachexia Society in 2017 all helped push the field onto this steep trajectory.

Yet the analysis also underscores how small the field remains relative to the problem it addresses. Over a quarter century, fewer than 6,000 papers met the study’s inclusion criteria, while the same database search parameters returned nearly 2.3 million papers on cancer overall and roughly 54,000 on pancreatic cancer alone. The researchers argue that cachexia remains one of the most underrecognized consequences of malignancy, partly because early diagnosis is difficult and a standardized clinical definition has been slow to settle. This is despite cachexia being a fundamental driver of poor prognosis, undermining treatment tolerance, physical function, quality of life, and ultimately survival.

The journal landscape is strikingly concentrated. The Journal of Cachexia, Sarcopenia and Muscle dominates every metric the team measured, including total papers, locally cited sources, and the local H-index, publishing around eight percent of all cachexia-related papers since its founding. Other leading outlets span both basic science and clinical nutrition, a reflection of the syndrome’s close ties to malnutrition. The authors note that while conventional nutritional support cannot fully reverse skeletal muscle loss in cachexia, early nutritional intervention remains essential for preventing mild disorders from progressing into irreversible problems.

Collaboration patterns reveal a field that is global but geographically clustered. The 5,775 documents were written by 31,545 authors, averaging 8.1 co-authors per paper, with 21.4 percent involving international co-authorship. Network analysis using the Louvain community-detection algorithm identified six author communities with distinct regional flavors: one dominated by Italian researchers, one mainly Asian, a small German group, and a more diverse cluster organized around the prolific and highly cited Professor Vickie Baracos. At the country level, the United States leads in total papers and citations, followed by China and Japan, though China and Japan fall outside the top ten in average citations per paper. The United Kingdom and Canada, despite lower output, punch above their weight in citation impact, and the late Professor Kenneth Fearon of Edinburgh ranked third in both local citations and H-index despite not reaching the top 25 by paper count.

The most methodologically novel part of the study lies in its use of artificial intelligence to interpret keyword networks. The team built a co-occurrence network of 88 keywords linked by 1,597 edges and used the Louvain algorithm to identify four communities. Rather than naming these clusters by hand, they fed the keyword lists to the GPT-3.5-turbo large language model with a standardized prompt template, fixed parameters including a temperature of zero, and instructions to generate concise two-to-five-word labels. The AI named the four communities Cancer Prognosis and Body Composition, Cancer Care Management, Muscle Metabolic Regulation, and Cancer Appetite Modulators, descriptions that the authors say aptly summarize the overall landscape of cachexia research.

Trend analysis added a temporal dimension. After converting keywords into dense vector embeddings using SapBERT, a transformer model trained on biomedical terminology, the researchers applied k-means clustering to group related terms and again used the language model to annotate the results. The data show a clear generational succession in inflammation research: work on interferon-gamma peaked around 2006, tumor necrosis factor and TNF-alpha around 2008 and 2012, and interleukin-6, the highest-frequency cytokine keyword, centered around 2015. The most recent cytokine activity belongs to GDF-15, a member of the TGF-beta superfamily, which peaked in 2025. Meanwhile, older mechanistic themes such as muscle protein degradation, tissue protein turnover, and ubiquitin-dependent proteolysis all have median years before 2010, indicating waning interest as the field has shifted toward clinical and translational questions.

Burst analysis, which flags keywords whose annual frequency rises at least five-fold above baseline, revealed shorter-term hotspots that barely overlap with the long-term trends; only eight keywords appeared in both analyses. Recent bursts include dexamethasone in 2023, the immunotherapy drug pembrolizumab in 2022 and 2025, and the broader term immune-checkpoint inhibitor in 2024 and 2025, suggesting that the intersection of cachexia and cancer immunotherapy is currently one of the hottest areas in the field. Metabolic keywords also flared, with PGC-1alpha bursting in 2019, thermogenesis in 2022, and hypermetabolism in 2021 and 2022. On the therapeutic front, the appetite-stimulating drug anamorelin and its formulation Ono-7643 showed peaks around 2023 and 2024, though meta-analyses of trial data report significant gains in body weight and lean mass without improvements in survival, grip strength, or appetite scores.

Citation analysis highlighted which papers have shaped the field and where new ideas may be entering from outside. Ten of the fifteen most-cited papers relate to the clinical definition, prediction, and management of cachexia, several published within the last five years, indicating that guideline development remains an active frontier. Among the top three is the 2024 New England Journal of Medicine trial of ponsegromab, a humanized monoclonal antibody that inhibits GDF-15, a striking example of translation moving from the first GDF-15 cachexia papers in 2016 to a clinical therapeutic eight years later. Intriguingly, two highly globally cited papers, one on tumor extracellular vesicles inducing liver metabolic dysfunction and another on the systemic manifestations of COPD, have almost no local citations, suggesting that knowledge from adjacent fields may represent promising, untapped avenues for cachexia research.

The authors are candid about the limitations of their approach. The simple Boolean search for cancer and cachexia may miss relevant papers, author keywords were absent from 21.1 percent of records, and language-model-generated labels are nondeterministic interpretive summaries rather than rigorous classifications, though the underlying network, trend, and burst computations were performed independently of the AI annotation. Even so, the study offers the first comprehensive longitudinal map of cancer cachexia research and a clear message: the field is growing fast, collaboration is strong but geographically siloed, and the next wave of activity is likely to come from immunometabolism, biomarkers, and AI-accelerated bibliometrics that dig deeper into the full text of the literature than keywords alone ever could.

Subject of Research: Bibliometric and AI-guided analysis of 25 years of cancer-induced cachexia research trends

Article Title: The Growth and Trends in Cancer‐Induced Cachexia Research: A Bibliometric and AI‐Guided Visualization Analysis

Article References: O'Connell, T. M., Gedela, R., & Pin, F. (2026). The Growth and Trends in Cancer‐Induced Cachexia Research: A Bibliometric and AI ‐Guided Visualization Analysis. Cancer Reports, 9(10), Article e70690. https://doi.org/10.1002/cnr2.70690

Image Credits: AI Generated

DOI: 10.1002/cnr2.70690

Keywords: cancer cachexia, bibliometrics, artificial intelligence, GDF-15, muscle wasting, immunotherapy, anamorelin, ponsegromab, interleukin-6, research trends, Web of Science, keyword analysis