Despite a considerable increase in publications in recent years, liquid biopsy is not yet routinely used in cancer diagnostics and tumor monitoring. In contrast to direct biopsies, liquid biopsies use body fluids that are taken far away from the primary tumor. It can therefore overlook a relevant part of the tumor, but conversely be representative of the more aggressive counterpart.
(red) Venous blood or cerebrospinal fluid is usually used for the liquid biopsy (LB). Tumor-derived material can be present either in free form (circulating tumor nucleic acids and circulating tumor cells [CTCs]) or in membrane-bound vesicles (microvesicles [MVs] and exosomes [EXs]). Recently, platelets have shown their enormous potential for liquid biopsy as tumor-adapted platelets (TEPs). Against this background, however, intrinsic brain tumors pose an additional challenge for several reasons, such as their low incidence and the cost-effectiveness of early screening, the lack of evidence for early treatment options, and the presence of the blood-brain barrier (BBB) as a potential suppressor of migrating tumor cells and their low ability to metastasize via the blood. One of the major challenges in integrating brain tumor LB into clinical routine is to evaluate reliable standards and to increase the rather low and variable sensitivity, which is currently around 10-60%, as the excretion of tumor DNA into the CSF does not seem to be a universal feature of diffuse glioma. In addition, there is still insufficient data on the impact of tumor type (glioblastoma vs. IDH-mutated astrocytoma vs. lower grade gliomas), location, extent of BBB disruption and disease stage on the sensitivity, specificity and clinical utility of individual liquid biopsy biomarkers and their combination, as well as on the best type of CSF sampling (either lumbar or cisternal CSF).
Machine learning supports LB
A systematic review focused on the application of machine learning (ML) to LB in brain tumors with the aim of providing neurosurgeons with a practical guide to understanding state-of-the-art practices and open challenges. The study was conducted in accordance with the PRISMA-P guidelines (preferred reporting items for systematic review and meta-analysis protocols). An online literature search was conducted in the PubMed/Medline, Scopus and Web of Science databases using the following search query: “([Flüssigbiopsie] AND [Glioblastom OR Hirntumor] AND [machine learning OR artificial intelligence))”.
The search in the literature yielded a total of 55 results. Duplicate entries were removed (n=24). 31 records were screened and four records were excluded by title and abstract screening; 27 reports were searched for retrieval and four studies were not accessible and were therefore discarded. Twenty-three reports were found to be relevant to our research question and assessed for suitability. Nine reports were excluded because they did not meet the inclusion criteria. Ultimately, 14 articles were included in the study.
Currently not used for brain tumors
The main advantages of LB are that they are non-invasive, repeatable and can be evaluated in real time. Intrinsic brain tumors represent a difficult group of tumors for which liquid biopsy can be routinely used. Only sparse and early evidence has been published in the literature that different phases of gliomagenesis are characterized by different secreted proteomic biomarkers. LB plays an important role in diagnosis when tissue biopsy is not feasible due to the risk of excessive morbidity, e.g. in deep-seated or multicentric lesions or in advanced age and a variety of comorbidities. However, this does not mean that there is no potential for LB in gliomas at other stages, e.g. for monitoring treatment outcomes and prognostic stratification of affected patients. The level of an ideal, hypothetical biomarker should be elevated at the time of diagnostic imaging, then drop significantly after surgical removal of the tumor and remain low during further treatment, helping to distinguish progression from pseudoprogression and increasing the number of cases eligible for potentially curative options or more successful therapies. This is all the more true as the WHO classification of CNS tumors for 2021 calls for a characterization of brain tumors not only on the basis of their histology, but also independently on the basis of their molecular characteristics.
However, the small number of papers in the literature on brain tumors shows that liquid biopsy is not currently used for CNS tumors. However, ML can play a role in LB in brain tumors – especially in early detection, as it is more effective than conventional statistical methods in identifying the early signs of brain tumors by analyzing the genetic and proteomic markers in liquid biopsies. Precision medicine can also benefit, as it can quickly help predict the effectiveness of different therapies by analyzing the molecular signature of tumors. It also delivers faster results. ML algorithms can quickly and accurately analyze LB results and provide physicians with actionable insights that can help them make treatment decisions.
Source: Menna G, et al: Is There a Role for Machine Learning in Liquid Biopsy for Brain Tumors? A Systematic Review. Int J Mol Sci 2023; 24(11): 9723.
InFo NEUROLOGY & PSYCHIATRY 2024; 22(3): 33
