Multiparametric Modelling of Survival in Pancreatic Ductal Adenocarcinoma Using Clinical, Histomorphological, Genetic and Image-Derived Parameters

Georgios Kaissis, Friederike Jungmann, Sebastian Ziegelmayer, Fabian Lohöfer, F Harder, Anna Melissa Schlitter, Alexander Muckenhuber, Katja Steiger, Rebekka Schirren, Helmut Frieß, Roland M. Schmid, Wilko Weichert, Marcus R. Makowski, Rickmer BrarenJournal of Clinical Medicine 2020 · DOI

TL;DR

This study introduces a multiparametric approach to predict survival in pancreatic ductal adenocarcinoma (PDAC) by integrating clinical, histopathological, genetic, and novel image-derived features. The findings highlight that imaging-derived features are independent predictors of survival and, when combined with other data, achieve a high performance in predicting patient outcomes, suggesting their crucial role in future risk stratification.

What Problem Does This Paper Address?

Pancreatic ductal adenocarcinoma (PDAC) is notoriously aggressive, with a grim prognosis and limited treatment options. Accurately predicting patient survival is crucial for effective treatment planning and patient stratification. While various biomarkers exist, including clinical factors, histopathology, and genetic markers, their individual predictive power is often limited. Furthermore, many studies focus on novel imaging biomarkers (radiomics) without rigorously comparing their value against established parameters or exploring their combined potential. This paper aims to bridge this gap by conducting a comprehensive multiparametric analysis to assess the predictive value of different data types and their synergistic effect on PDAC survival prediction.

Methods

The researchers employed a retrospective cohort study design involving 103 patients with resectable PDAC. They collected a rich dataset encompassing:

  • Clinical and Histopathological Data: Age, sex, chemotherapy regimens, tumor size, lymph node status, grading, and resection status.
  • Morpho-molecular and Genetic Data: Tumor morphology, molecular subtype, and mutations in key genes like TP53, KRAS, SMAD4, and CDKN2A/p16.
  • Image-Derived Features (Radiomics): Quantitative features extracted from CT scans using the PyRadiomics software.

To handle the high dimensionality of the radiomic features, they used unsupervised machine learning techniques, specifically Principal Component Analysis (PCA), to reduce dimensionality while retaining 99% of the variance. Feature selection was performed using a log-rank test. The core of their analysis involved multivariate Cox proportional hazards modeling to assess the prognostic strength of each parameter category and their combinations. A five-fold cross-validation was used to evaluate predictive performance, with the concordance index (CI) serving as the primary metric.

Datasets

The study utilized a dataset of 103 patients with resectable PDAC, with clinical, histopathological, genetic, and CT imaging data collected retrospectively. No new public benchmark datasets were introduced.

Key Insights and Ideas

  • Imaging Features Shine: The study strongly suggests that image-derived features (radiomics) are not just supplementary but independent predictors of survival in PDAC. This is significant because it offers a non-invasive way to glean prognostic information.
  • The Power of Combination: While individual parameter groups showed moderate predictive power, the combined model, integrating all data types, achieved the highest performance. This underscores the complexity of PDAC and the need for a holistic approach to prediction.
  • Limitations of Traditional Markers: Morpho-molecular and genetic parameters, while important, showed the lowest predictive power in this study. This could be due to sampling errors in biopsies or inherent tumor heterogeneity not fully captured by these markers alone.
  • Non-invasive Advantage: The paper emphasizes the value of non-invasive imaging biomarkers, especially in palliative and neo-adjuvant settings where invasive procedures might be less feasible or desirable.

Key Results

The study reported the following concordance indices (CI) for out-of-sample data:

  • Clinical and Histopathological Features: CI = 0.63
  • Morpho-molecular and Genetic Features: CI = 0.53
  • Image-Derived Features: CI = 0.65
  • Combined Model (all parameters): CI = 0.65

These results indicate that imaging features alone performed comparably to clinical and histopathological data and that the combined model achieved the highest predictive performance. Specifically, in the combined model, three image-derived feature groups, tumor subtype, tumor size (pT), and nodal status (pN) were found to be significant and independent predictors of overall survival.

Limitations and Future Work

The authors acknowledge several limitations:

  • Sample Size: The study included 103 patients, which they note is a common limitation in current imaging studies and can affect external validation.
  • External Validation: The results were not tested on an independent external validation cohort, which is crucial for confirming generalizability.
  • Cost and Accessibility: The extensive morpho-molecular and genetic workup is costly and personnel-intensive, limiting its widespread clinical adoption.

Future work should focus on:

  • Direct Comparison: Conducting side-by-side comparisons of whole-tumor tissue analysis with imaging-derived features to better understand their true predictive performance.
  • Multi-centric Cooperation: Establishing multi-centric collaborations to gather larger sample sizes for more robust multiparametric data integration and clinical validation.
  • Prospective Testing: Testing the developed models prospectively to ascertain their external validity and clinical utility.

Summary updated

Abstract

Rationale: Pancreatic ductal adenocarcinoma (PDAC) remains a tumor entity of exceptionally poor prognosis, and several biomarkers are under current investigation for the prediction of patient prognosis. Many studies focus on promoting newly developed imaging biomarkers without a rigorous comparison to other established parameters. To assess the true value and leverage the potential of all efforts in this field, a multi-parametric evaluation of the available biomarkers for PDAC survival prediction is warranted. Here we present a multiparametric analysis to assess the predictive value of established parameters and the added contribution of newly developed imaging features such as biomarkers for overall PDAC patient survival. Methods: 103 patients with resectable PDAC were retrospectively enrolled. Clinical and histopathological data (age, sex, chemotherapy regimens, tumor size, lymph node status, grading and resection status), morpho-molecular and genetic data (tumor morphology, molecular subtype, tp53, kras, smad4 and p16 genetics), image-derived features and the combination of all parameters were tested for their prognostic strength based on the concordance index (CI) of multivariate Cox proportional hazards survival modelling after unsupervised machine learning preprocessing. Results: The average CIs of the out-of-sample data were: 0.63 for the clinical and histopathological features, 0.53 for the morpho-molecular and genetic features, 0.65 for the imaging features and 0.65 for the combined model including all parameters. Conclusions: Imaging-derived features represent an independent survival predictor in PDAC and enable the multiparametric, machine learning-assisted modelling of postoperative overall survival with a high performance compared to clinical and morpho-molecular/genetic parameters. We propose that future studies systematically include imaging-derived features to benchmark their additive value when evaluating biomarker-based model performance.
BibTeX
@article{kaissis2020multiparametric,
  author = {Georgios Kaissis and Friederike Jungmann and Sebastian Ziegelmayer and Fabian Lohöfer and F Harder and Anna Melissa Schlitter and Alexander Muckenhuber and Katja Steiger and Rebekka Schirren and Helmut Frieß and Roland M. Schmid and Wilko Weichert and Marcus R. Makowski and Rickmer Braren},
  title = {{Multiparametric Modelling of Survival in Pancreatic Ductal Adenocarcinoma Using Clinical, Histomorphological, Genetic and Image-Derived Parameters}},
  journal = {Journal of Clinical Medicine},
  year = {2020},
  doi = {10.3390/jcm9051250},
}
Figure