Researchers Develop Dynamic Prognostic Model for Patients With Intrahepatic Cholangiocarcinoma Undergoing Curative Resection

Timothy Pawlik, MD, PhD, MPH, Wexner Medical Center, The Ohio State University, Columbus, Ohio, discusses results from a study in which he and his colleagues used landmark analysis to assess patients undergoing curative intent surgery for intrahepatic cholangiocarcinoma with the goal of developing a dynamic prognostic tool.

Dr Pawlik et al report that this tool provides a more dynamic prognostic model to assess long-term outcomes, by both incorporating traditional factors associated with survival, as well as adjusting for the overall survival time a patient has already experienced in each subsequent year.

Transcript:

Hello, my name's Tim Pawlik. I'm the Chair of Surgery at the Ohio State University Wexner Medical Center, and also a surgical oncologist that treats patients with liver and pancreatic tumors. I wanted to talk with you today about our recent paper in the Annals of Surgical Oncology on landmarking to predict long-term survival among patients with intrahepatic cholangiocarcinoma.

Intrahepatic cholangiocarcinoma is a relatively rare disease that has a grave prognosis with a 5-year survival in the range of only about 30% among patients, even who have a resectable disease. In fact, the overall cure rate is probably more in the order of about 10% to 15%, yet there is significant heterogeneity with regards to long-term prognosis based on patient, tumor, and hospital-specific factors. Survival estimation is an important aspect of many research studies among surgeons. Typically, most surgeons and most statisticians use commonly-used techniques such as the Kaplan-Meier estimator in log-rank tests or Cox proportional hazard models. But these survival models can have several shortcomings, and don't take into consideration some of the dynamic factors that may be associated with long-term survival.

In addition, these survival estimates provide data in aggregate fashion for prognostic groups, which may be helpful for general predictions, but may not be applicable to individual patients who have very specific factors with regards to their demographics, or their tumor, or their surgical care, and can't really allow for the tailoring of prognosis for individual patients in your clinic. Landmark analysis is a type of survival analysis that splits follow-up time at a common pre-specified time point, the so-called landmark. In turn, landmark analysis allows for a more dynamic prediction of survival that accounts for the time that's elapsed since the initial surgery, as well as the occurrence of any intervening events such as recurrence.

The purpose of the current study was to use landmark analysis as a dynamic prognostic model to assess patients who had undergone curative intent resection for an intrahepatic cholangiocarcinoma, and then to develop a tool that may assist patients and providers to predict prognosis that may help tailor treatment, follow-up strategy, and also inform discussions around prognosis for patients with intrahepatic cholangiocarcinoma.

We identified patients from a multi-institutional international database that involves centers in North America, Europe, and Asia, as well as Australia, and had a cohort of roughly 600 or 700 patients who had undergone a curative intent operation for intrahepatic cholangiocarcinoma. Then we performed a landmark analysis to estimate overall survival prediction. We assessed the performance of the model, both internally with bootstrapping, resampling, and then assessed the discrimination using a standard Harrell's concordance index, and the accuracy with calibration plots. Not surprisingly, on multivariate analysis, there were several factors that were associated with overall survival. These included age, tumor size, margin status, morphologic subtype of the intrahepatic cholangiocarcinoma, as well as factors that you find in the [American Join Committee on Cancer] AJCC Staging Manual, such as T and N category, as well as histologic grade. Not surprisingly, also, tumor recurrence was also strongly associated with overall survival.

Interestingly, we noted that the effect of these variables on overall survival changed over time, and results were provided as a survival plot, and predicted probability of overall survival at any specific time in the patient that saw follow-up. What we were able to demonstrate was the dynamic impact that each of these prognostic factors had on overall survival was dependent on the time at which they were being assessed in the patient's follow-up period.

For example, a 65-year-old patient with an introductory T1 grade 3 or 4 intrahepatic cholangiocarcinoma that measured 3 centimeters, who had undergone an R0 resection, had a calculated estimated 3-year survival using our proposed tool of 76%. This overall survival estimate increased if the patient had already survived 1 year, to 79%, with a good discrimination ability of about 0.8 on the model. Looking at other patients, depending on the factors associated with their tumor, their grade, their T and N category, and the amount of time that they had survived from the time of operation, we could estimate the varying prognostic 5-year survival for these patients based on these factors, thereby providing a much more dynamic model and tool for providers to use in the clinic to assess long-term outcomes for patients undergoing curative intense surgery for intrahepatic cholangiocarcinoma.

These types of studies are important for several reasons. One, as noted, patients really want detailed data about their specific survival risk or probability, rather than aggregate data. Standard models, like a Kaplan-Meier curve, estimate that maybe, in general, survival at 5 years is maybe 30% or 40%. But that specific patient isn't 30% or 40% alive at 5 years. They're either alive or deceased. We need to develop tools that are specifically tailored to individual patients.

In addition, patients’ overall prognosis changes dynamically with time. The longer that a patient survives, from the time of their initial curative intent operation, probably, that's the most important factor which predicts future further survival. Survival models not only need to incorporate traditional clinical pathologic features like tumor grade, tumor number, nodal status, it also needs to incorporate, in a much more dynamic fashion, the overall survival time that the patient has already experienced when they come back to your clinic at year 2, or year 3, or year 4. Therefore, the type of analyses that we performed in this paper in Annals of Surgical Oncology that incorporates a landmark analysis into an easy-to-use tool for physicians in the clinic, may be something that is helpful not only to providers in assessing the overall risk of recurrence or death, which can inform perhaps conversations around the need for adjuvant treatment or the need for more aggressive or less aggressive surveillance, but also can provide patients with more detailed individual information about their own specific prognosis.

While we use this tool and develop this tool in intrahepatic cholangiocarcinoma, our groups and others have utilized landmark analysis and conditional survival analysis to look at a whole host of other tumors. In particular, these types of analyses can be helpful, especially among patients who are diagnosed with cancers who have a particularly poor prognosis because, as patients survive longer, the factors that may have been initially associated with their prognosis, like tumor factors, become less and less important. What becomes more important is the fact that patient has survived 1, 2, 3, 4 years from the time of surgery.

Again, I appreciate the opportunity to share with you a little bit about our research around intrahepatic cholangiocarcinoma, and perhaps some more novel methods to assess survival, analyze survival, and provide more meaningful data to providers and patients in the clinical setting. Once again, thank you very much.

Source:

Spolverato G, Capelli G, Lorenzoni G, et al. Dynamic prediction of survival after curative resection of intrahepatic cholangiocarcinoma: A landmarking-based analysis. Ann Surg Oncol. 2022;29:7634-7641. doi:10.1245/s10434-022-12156-1