Predictive Risk Model for Endometrial Cancer Based on Preoperative Examination of Those With Endometrial Intraepithelial Neoplasia
Gynecologic oncologists currently lack risk models that can accurately predict an endometrial cancer diagnosis at the time of hysterectomy for endometrial intraepithelial neoplasia. To fill this gap, Michelle Ertel, MD, and her colleagues set out to develop a calculated risk model based on specific preoperative risk factors that can accurately predict endometrial cancer at the time of hysterectomy. Dr Ertel presented the team's findings at the Society of Gynecologic Oncology (SGO) 2024 Annual Meeting in San Diego, CA.
Additional Resource:
Ertel MD, Bell S, Warshafsky MD, Skjoldager K, Ridgley J, Mysona DP, Lesnock JL. A predictive nomogram for endometrial cancer based on preoperative assessment among patients with endometrial intraepithelial neoplasia. Poster presented at: The Society of Gynecologic Oncology (SGO) 2024 Annual Meeting on Women’s Cancer; March 16-18, 2024; San Diego, CA. Accessed March 7, 2024. www.sgo.org/events/annual-meeting/
For more SGO 2024 coverage, visit the Newsroom.
TRANSCRIPTION:
Michelle Ertel, MD: My name is Michelle Ertel, and I am a current gynecologic oncology fellow at McGee Women's Hospital with the University of Pittsburgh Medical Center.
Consultant360: What prompted this study?
Dr Ertel: Yeah, so, you know, we as gynecologic oncologists are seeing more and more patients with endometrial intraepithelial neoplasia or CAH (complex atypical hyperplasia) because the risk of occult malignancy at the time of hysterectomy for these patients is about 40%. And there's a lot of debate amongst us about what the best practice for lymph node assessment is in these patients - whether that is to perform a routine sentinel lymph node dissection or basing it off of your frozen section from the hysterectomy specimen and using Mayo criteria to determine their risk of lymph node involvement.
And so really the goal of this project was to try to see if we could better identify those who are at the highest risk for cancer in order to stratify who we should actually be doing sentinel lymph node dissections on.
C360: Can you please summarize the results of your study?
Dr Ertel: So, what we found is we were actually able to build two different clinical prediction models. One was for the overall cohort and one was in a pre-menopausal cohort. And so, in the overall cohort, what we found is the two most important factors for predicting whether or not somebody had endometrial cancer was the biopsy method, whether or not they had a D&C (dilation and curettage) versus an endometrial biopsy, as well as the endometrial stripe thickness. And then interestingly in sort of the premenopausal cohort, which we defined as patients less than the age of 52, there were about 13 factors that we took into account after we had reviewed more than 27 sort of hypothesized and known clinical risk factors. We found 13 that seemed to be the most predictive and were able to build a model using elastic net, multinomial logistic regression modeling to give an accurate prediction of who is at risk of cancer.
C360: Is this model close to implementation in clinical practice, or is more validation needed?
Dr Ertel: That's a good question. That's ultimately the goal is to be able to put this into clinical practice. I think the next to the very next step that we need to be able to do that is actually to apply it to an external validation set. And so, what I mean by that is we based all of our modeling off of our patient population in Pittsburgh and the patients that we see and treat here. And we don't know whether or not it would be valid in a patient population that is different than ours here. And so the next step really will be to validate that in an external cohort.
C360: What are the gaps in our knowledge that still remain?
Dr Ertel: It would be the external validation. You know, one thing I will highlight is that our population in Pittsburgh had very little ethnic and racial diversity. And so it's a little bit unknown whether or not this model would be applicable to populations that have a higher incidence of racial and ethnic diversity. And so to address this, kind of our very next step is to partner with an institution that has a different patient population from ours. ours to use as our external validation set.
C360: What were the main takeaways from your study?
Dr Ertel: Yeah so I think you know initially we had set out to find a model that would predict who does have cancer and what we found is that it's actually easier to predict or to rule out patients that are low risk of having cancer, which is also as clinically useful because what that means is that we're able to kind of identify a subset of patients who really don't need a lymph node assessment and can safely omit that at the time of hysterectomy. But I think what we found is that it is possible to accurately predict this based off of preoperative assessment. And so hopefully down the road, we'll be able to apply this and further kind of tailor our treatment for patients who have this pre-cancerous diagnosis.
C360: Did any of your study results surprise you?
Dr Ertel: I think one thing that did surprise us is that when we looked at all of these 27, risk factors, some of them are known risk factors for the development of endometrial cancer, it's things like BMI. We know that the higher the BMI, the more likely you are at risk. And in our study, really BMI didn't seem to pan out in terms of its ability to predict whether or not somebody had a cancer diagnosis or not.
Certainly, this is based on our cohort alone, and so needs to be externally validated like we talked about. about. But it seems to be that when you sort of factor in many other different risk factors the importance of BMI is not as high as we maybe thought.