Weakness, fatigue, poor balance, and trouble walking and standing are all symptoms of sarcopenia, a type of muscle loss associated with aging common in patients with cancer, including multiple myeloma (MM). In a recent study, Mayo Clinic researchers set out to determine if heterogeneous overall survival (OS) rates in patients with MM are influenced by sarcopenia and whether its presence has prognostic value.
Detailing their findings in Cancer, the researchers described how computed tomography (CT) utilizing a deep learning-based segmentation approach affected OS rates in older patients diagnosed with newly diagnosed MM (NDMM) (2022; https://doi.org/10.1002/cncr.34545). Led by Bharat Nandakumar, MBBS, a research fellow at Mayo Clinic, the study investigators developed a machine learning-based convolutional neural network algorithm that identified sarcopenia by reviewing CT images of the abdomens of patients.
To detect sarcopenia, researchers used precise segmentation and measured the skeletal muscle at the L3 vertebrae level. These measurements were further normalized by the patient's height to obtain the skeletal muscle index and classify if they were sarcopenic or not. The Mayo Clinic study included 322 patients with NDMM, with 67 (28%) categorized with high-risk fluorescence in situ hybridization (FISH) cytogenetics. A total of 53 percent were sarcopenic based on their peri-diagnosis standard-dose CT scan.
In the study, the OS and 2-year mortality rates for patients with sarcopenia and MM were 44 months and 40 percent, respectively. By comparison, patients with NDMM who did not have sarcopenia had an OS rate of 90 months, or more than twice as long as those with sarcopenia, and a mortality rate of 18 percent.
"The prevalence of sarcopenia in patients with multiple myeloma has been described in the literature, but I wanted to investigate whether the presence of sarcopenia impacted the survival experience and had any prognostic value," Nandakumar told Oncology Times. "Utilizing a novel AI platform based on a deep learning algorithm, we were able to detect sarcopenia from patient's baseline CT images with high levels of accuracy in a timely fashion, which was instrumental to the study."
A New Prognostic Tool
Several recent studies have evaluated the role of machine learning, CT, and combinations of the respective forms of artificial intelligence and medical imaging in the detection, diagnosis, and treatment of sarcopenia in patients with various cancers, including MM. For example, recent research analyzed a deep learning algorithm segmentation approach to determine the skeletal muscle index and sarcopenia in metastatic renal cancer (Eur Radiol 2022; doi: 10.1007/s00330-022-08579-9).
In another study, researchers using a novel CT imaging approach showed sarcopenia is prevalent in MM and also associated with increased cardiovascular events. Unlike the Mayo Clinic study, however, an association between sarcopenia and OS was not found (Bone Marrow Transplant 2021; doi: 10.1038/s41409-020-01008-9). Still another analysis in this research employed a novel CT method that suggested sarcopenia is prevalent in MM and also associated with increased early post-transplant cardiovascular complications in patients undergoing autologous hematopoietic cell transplantation.
In the new Mayo Clinic study, the researchers demonstrated that, although various disease and patient factors are responsible for the heterogeneity of OS in patients with MM, the presence of sarcopenia is associated with poorer survival rates, Nandakumar noted.
"This study is unique, as it is the first of its kind to employ a deep learning-based platform on cross-sectional imaging to identify sarcopenia in patients with MM and derive useful prognostic information," he said. "The presence of sarcopenia conferred worse prognosis independent of other risk factors like older age, high-risk cytogenetic abnormalities, and ISS Stage II or III."
Potential Clinical Implications
Sarcopenia increasingly is viewed as a risk factor in geriatric patients. Several studies have associated sarcopenia with falls, loss of function, malnutrition, and high rates of mortality. An estimated 23-68 percent of patients share the aging syndrome in clinical settings (Int J Environ Res Public Health 2021; https://doi.org/10.3390/ijerph18168299). For patients with sarcopenia and NDMM, the findings in the new Mayo Clinic study come with many clinical implications, Nandakumar said.
"The novel AI-based platform used for detecting sarcopenia from CT images can potentially be integrated into clinical practice providing the basis for real-time decision-making at the time of diagnosis, such as incorporating pre-habilitation or rehabilitation services, increased supportive care, or considering treatments with a different toxicity profile, etc. This would allow provisions for optimizing management plans and mitigate risk based on the presence of sarcopenia," he explained.
"The adverse impact of sarcopenia identified by this methodology may also provide some explanation to the heterogeneous nature of survival outcomes in patients with MM despite accounting for traditional prognostic features like FISH cytogenetics and ISS stage," added Nandakumar.
The ability of the machine learning approach applied in combination with CT imaging to predict and improve survival outcomes in patients with MM may extend to other cancers and diseases.
"This study demonstrates the potential and feasibility of using AI-based platforms in clinical practice that can provide real-time prognostic information," said Nandakumar.
Meanwhile, one possibility for further research into AI and CT will be looking for the genesis of sarcopenia in patients with MM.
"It would be worthwhile to explore whether there are any complex pathophysiological mechanisms or interactions between the tumor microenvironment and skeletal muscle that results in sarcopenia," Nandakumar said.
Chuck Holt is a contributing writer.
Top 10 Best Hospitals for Cancer