Commentary
Article
Author(s):
Emil Lou, MD, PhD, FACP, discusses the predictive and prognostic use of tumor stroma proportion for outcomes and chemoresistance in ovarian cancer.
Tumor stroma proportion (TSP) was found to be a consistent and reproducible marker of survival outcomes and resistance to standard platinum-based chemotherapy in patients with high-grade serous ovarian cancer, supporting its integration into prospective clinical trials and, eventually, clinical practice, according to Emil Lou, MD, PhD, FACP.
Lou and colleagues conducted a confirmatory study that built upon prior research indicating the predictive and prognostic value of TSP in ovarian cancer. Findings published in JAMA Network Open showed that patients in the Tübingen cohort (n = 192) with ovarian cancer who have a higher TSP were more likely to develop resistance to standard-of-care platinum-based chemotherapy. Moreover, high TSP correlated with significantly shorter progression-free survival (PFS; HR, 1.586; 95% CI, 1.093-2.302; P = .02) and overall survival (OS; HR, 1.867; 1.249-2.789; P = .002) in this cohort. There was no significant correlation between TSP levels and chemoresistance, PFS, or OS in The Cancer Genome Atlas cohort (n = 103). Lastly, patients with chemoresistant tumors were twice as likely to have high TSP vs those with chemosensitive tumors (HR, 2.861; 95% CI, 1.256-6.515; P = .01).
“[These] data supported [TSP as] both a predictive biomarker strongly correlating with chemoresistance, and prognostic [of worse outcomes] in those patients,” said Lou, who is a medical oncologist, neuro-oncologist, and gastrointestinal oncologist, as well as an associate professor of medicine in the Division of Hematology, Oncology, and Transplantation at the University of Minnesota Medical School. “We can easily foresee that TSP is a biomarker that could be integrated into any trial anywhere in the world, at a minimal increased cost, if any.”
In an interview with OncLive®, Lou explained the rationale for conducting this study, highlighted key findings that show the correlation between TSP and survival outcomes in ovarian cancer, and discussed plans to not only validate these findings through continued research but incorporate artificial intelligence (AI) into this approach for more accurate predictions.
Lenz: Although I do not treat patients with ovarian cancer myself, ovarian cancer is a major [area of] interest among our research team. In the clinic, I’m a gastrointestinal oncologist, and I treat advanced colorectal or pancreatic cancers. I also have a history of training as a neuro-oncologist, and brain tumors are hard to treat. I put ovarian cancer under that umbrella [of difficult-to-treat, chemoresistant cancers] as well. They’re cancers that are insidious, but also [able] to overcome the [current] standard of care treatments that so far have been standard of care.
In my lab, we’re trying to figure out the cellular and molecular factors that determine which patients [may] have tumors that are more resistant to treatment than others? [We] also [want to know] if there are predictive or prognostic biomarkers that can tell us, even when someone is first diagnosed, that the [tumor is likely to become] resistant to chemotherapy or that this is likely to occur sooner rather than later.
This study has built on those 10 or 12 years [of data] that I’ve been working on [in ovarian] cancer. We did a pilot study consisting of approximately 24 women with diagnosed ovarian cancer and tumor stroma proportion [TSP], which is seen under the microscope and under diagnostic slides. [TSP] emerged as an early candidate [for a predictive biomarker], and we needed to confirm this. That pilot study was published in 2019 in JAMA Oncology, this publication was published in JAMA Network Open in February 2024.
The objective was to look at whether TSP is a potential predictive and prognostic biomarker [in ovarian cancer]. In the world of oncology, prognosis can be related to progression-free survival and overall survival. We hypothesized that TSP could be a prognostic biomarker in that sense. It has been examined in a lot of other cancers like colorectal, pancreatic, and breast cancer, but not as much in ovarian cancer until now. If you want to find a predictive biomarker of chemotherapy outcomes and resistance, you must conduct a pilot study, and if something is supporting that hypothesis, [proceed to a] confirmation study and validation study. This [study] is part of that journey.
When you look under a microscope at a biopsy or a surgical specimen, [we can see] what approximate proportion of a cancer [comprises] actual cancer cells. The rest of it is called stroma, which is a large umbrella [term] to describe a lot of things that are also in the tumor. [This could include] things that compose a physical portion of the tumor, or blood vessels that help cancers become very invasive, or immune cells that might nourish or be part of the composition that makes cancers more aggressive. Based on other studies that have looked at this, a cutoff of 50% has been used. We used that cutoff in the study we published in 2019 and in this study. It’s a standard definition internationally for studies in the field. A TSP greater than 50% means that more than half of someone’s tumor sample is made of stroma and less than 50% [of the sample includes] actual cancer cells, and vice versa. It seems very basic, but it also seems to work in a lot of studies as a cutoff.
When we look at the TSP across these specimens, one of the main things we want to look at is the standard of care that’s used for patients with ovarian cancer. This includes platinum chemotherapy, and sometimes other chemotherapies may be added on to it. Platinum chemoresistance is a known entity in ovarian cancer, and it’s a big challenge we’re trying to overcome. Some of the hardest cases to treat are [patients with] platinum-resistant ovarian cancers.
In our initial prospective study, we had a valuable resource of collaborators overseas in Europe who had access to [information from 192] women [assessed at Tübingen University Hospital in Germany from 2004 to 2014] who were diagnosed with ovarian cancer. Their slides from biopsies or surgeries that were performed to diagnose ovarian cancer were available in digitized format. Rather than looking under the microscope, we have a digitized image to share with pathology colleagues and within our own team to review. That’s how we can find out what the TSP is.
The other valuable aspect is that we had all the clinical data from that cohort. We had their ages, how old they were at [the time of] diagnosis, what chemotherapy regimens they had received, how long they were [treated with] chemotherapy, and all these other factors that are needed to understand prognosis and how aggressive a cancer might have been, or how long the chemotherapy was successful, if at all.
In addition, we had access to [106] cases from TCGA. The TCGA is a strong and multi-institutional endeavor to create the first genomic fingerprints [that increase our] understanding of a lot of cancers at the genomic and molecular level. That includes ovarian cancer. These are publicly available datasets, so we tapped into that [cohort] for additional confirmation.
We wanted to take these clinical cohorts and determine whether [patients with] a higher TSP, [spent] a shorter amount of time on chemotherapy because the chemotherapy no longer worked. That did play out in the larger [200-patient] dataset. It [wasn’t as apparent] in the TCGA cohort, and we believe [that is because this cohort had a] smaller number of patients. [However], PFS was consistent in both the TCGA and 200 patient German trial cohorts, so a short amount of time before the cancers grew. Most patients who [experienced a shorter PFS] received platinum-based chemotherapy, and then maybe something else after.
Overall, there was a shorter amount of time that patients could be on chemotherapy before there was a progression, whether that be recurrence [of the tumor] despite surgery or [a tumor] that spread to other parts of the body. [These] data supported [TSP as] both a predictive biomarker strongly correlating with chemoresistance, and prognostic [of worse outcomes] in those patients.
In our previous study, we [also] had some indication [that TSP predicted OS], but this study clearly showed a statistically significant, worse OS in the patient that had tumors with lower TSP in addition to more chemoresistance.
We’re fully in the era of molecular oncology. It is common across many types of cancers to get genomic profiling. The chance of genomic profiling uncovering [the likelihood of developing] chemoresistance or targeted drugs that could be helpful is less than 20%. However, TSP is something that we can measure as a diagnosis and it is easily assessed. We also consider it a cheap diagnostic evaluation because it [utilizes] the same slide that’s used for diagnosing [patients] with cancer. [Assessment of TSP] would just take one extra step for any pathologist, and it’s not very complex.
With this confirmation study, we want to perform further validation [of our findings] in larger cohorts. We think that this would be an ideal correlative biomarker. At this stage with these data, the next step is to integrate [TSP] into prospective clinical trials that are treating patients with ovarian cancer either in the first line when they’re newly diagnosed or in subsequent lines. Whether it’s NRG Oncology, The Gynecologic Oncology Group, or other [clinical trial groups] both in North America and internationally. [TSP is] an ideal biomarker that can be further evaluated in a larger cohort and perhaps integrated [into clinical trials.] Along with results from any new therapeutic, [we could] also comment on TSP and perhaps further refine its utility. [We are working] towards understanding if [TSP assessment] can become a part of common clinical practice, and [serve as] useful tool for anyone from the community oncologists all the way to [oncologists at] larger academic centers.
Any manual assessment is always subject to potential human error. When we look at the slides, we gain expertise, but there’s always some margin for human error. That’s something exciting in this era of AI and [we are still learning] how that plays out in healthcare in general, medicine, and clinical care in oncology.
Part of our next move is eventually to merge [AI] with [our identification of TSP] as we expand this into prospective clinical trials. For this to be implemented on a large scale, the vision [is to] leverage AI to reliably identify [TSP]. That would help streamline the process and make it easier to integrate.
This is a confirmatory study utilizing an easy-to-perform evaluation based on material that’s already present at the time of diagnosis. TSP is an assessment that anyone [can perform.] You don’t need to be a pathologist to do it and I’m living proof of that. With experience and familiarity with looking at slides, [anyone can learn to] make that assessment, keeping in mind that this is a potential new prognostic tool that we can use. As we look to validate it further and make it easier to perform, there’s no question that this can move forward in prospective clinical trials. If there are any clinical trialists in the field reading this article, I hope they would consider working together or [using] TSP as a new, easy biomarker to integrate. In the era of global oncology, we have genomic profiling, but that is an expensive test and is not easily accessible worldwide. [However, pathologists] everywhere in the world depend on stained slides to make a diagnosis. Part of our mission is to [promote] equity in cancer care and in clinical trials. Ultimately, if this biomarker is proven to help, it might [allow] oncologists to [assess if] the chemotherapy they were about to give or were giving will work as well biologically, and [they can] start thinking about the next steps for the welfare of this patient.
Lou E, Clemente V, Grube M, et al. Tumor-stroma proportion to predict chemoresistance in patients with ovarian cancer. JAMA Netw Open. 2024;7(2):e240407.2024 Feb 5. doi:10.1001/jamanetworkopen.2024.0407