Prostate MRI and Clinicopathologic Risk Calculator to Predict Laterality of Extraprostatic Extension at Radical Prostatectomy

Eric V. Li, Sai Kumar, Jonathan Aguiar, Mohammad Adnan Siddiqui, Zequn Sun, Clayton Neill, Edward M. Schaeffer, Ashley E. Ross, Hiten D. Patel

TL;DR

Researchers have developed a side-specific risk calculator that predicts exactly where prostate cancer is likely to break through the prostatic capsule (extraprostatic extension). By combining multiparametric MRI (mpMRI) data with clinical pathology, this tool helps surgeons decide when it is safe to spare nerves on one side of the prostate while performing a wider excision on the other.

What Problem Does This Paper Address?

When a surgeon performs a radical prostatectomy (removing the prostate), they face a high-stakes balancing act. If they cut too close to the prostate, they might leave behind cancer cells that have spread just outside the gland—a condition known as extraprostatic extension (EPE). If they cut too wide, they risk damaging the delicate neurovascular bundles responsible for erectile function.

While existing "nomograms" (statistical models) can predict the overall likelihood of EPE, they rarely tell the surgeon which side of the prostate is at risk. Furthermore, while MRI has become a standard tool, it often misses microscopic EPE. This paper addresses the need for a "side-specific" map to guide surgical planning, allowing for aggressive cancer removal where needed and nerve preservation where possible.

Methods

The authors conducted a retrospective study of 601 patients across an eleven-hospital healthcare system who underwent mpMRI followed by surgery.

The core of their approach was a tiered side-definition system. They categorized each side of the prostate as "dominant" or "non-dominant" based on:

  1. The highest biopsy Gleason Grade Group (GG).
  2. The highest PIRADS score (an MRI-based risk scale).
  3. The number and size of lesions.
  4. The percentage of positive biopsy cores.

They then used multivariable logistic regression to identify which clinical and imaging factors actually predicted EPE on each specific side. Finally, they packaged these findings into an online, open-access risk calculator.

Datasets

The study utilized a clinical dataset from the Northwestern University healthcare system, covering patients treated between July 2018 and November 2022.

  • Sample Size: 601 men.
  • Imaging: 3 Tesla mpMRI evaluated by specialized genitourinary radiologists.
  • Pathology: Biopsy data (targeted and systematic) and final surgical pathology reports.

Key Insights and Ideas

  • Laterality is Independent: One of the most important findings was that factors on one side of the prostate (like a high-grade tumor or a suspicious MRI lesion) were only associated with EPE on that specific side. They did not influence the risk on the contralateral side.
  • The "Abutment" Limitation: Radiologists often look for "capsular abutment"—where the tumor physically touches the edge of the prostate on an MRI. While this was common (81% of patients), it had very poor specificity (27%). In other words, just because a tumor touches the edge doesn't mean it has broken through. Consequently, abutment did not improve the final predictive model.
  • PIRADS 5 and High Gleason Grades: These remained the strongest predictors. Specifically, a PIRADS 5 lesion or a Gleason Grade Group of 3–5 on a particular side significantly increased the odds of EPE on that side.

Key Results

The resulting models showed strong discriminative power, outperforming or matching traditional whole-gland models while adding the benefit of side-specific prediction:

  • Prevalence: EPE was found in 53% of the surgical cases.
  • Model Accuracy: The Area Under the Curve (AUC) was 0.77 for the dominant side and 0.79 for the non-dominant side.
  • Clinical Utility: The researchers successfully integrated age, PSA density, side-specific PIRADS, side-specific Gleason Grade, and systematic core positivity into a functional online calculator.

Limitations and Future Work

The authors acknowledge several hurdles before this becomes a universal standard:

  • Retrospective Nature: The data was collected after the fact; a prospective trial would be the gold standard.
  • Single System Study: The model was developed within one healthcare system and requires "external validation" (testing on patients from different regions/hospitals) to ensure it generalizes well.
  • Interrater Variability: MRI interpretation can vary significantly between radiologists. The authors suggest that future iterations incorporating Artificial Intelligence (AI) and radiomics could provide more objective and standardized imaging inputs to further refine the calculator's accuracy.

Summary updated

Abstract

<title>Abstract</title> <bold>Background</bold> Traditional nomograms can inform the presence of extraprostatic extension (EPE) but not laterality, which remains important for surgical planning, and have not fully incorporated multiparametric MRI data. We evaluated predictors of side-specific EPE on surgical pathology including MRI characteristics and developed side-specific EPE risk calculators. <bold>Methods</bold> This was a retrospective cohort of patients evaluated with mpMRI prior to radical prostatectomy (RP) in our eleven hospital healthcare system from July 2018-November 2022. The dominant side was defined pre-operatively using a tiered system based on laterality of highest biopsy Gleason Grade Group (GG), highest PIRADS lesion, number of lesions, and cancer volume. Univariable and multivariable logistic regression were performed for overall EPE, dominant side EPE, and non-dominant side EPE. Internal validation with leave one out and calibration curves were completed. <bold>Results</bold> EPE was identified in 53% (317/601) of patients at RP. Side-specific factors (PIRADS, GG, abutment) were only associated with EPE on their respective side. Final variables in the model associated with EPE on the dominant and non-dominant sides included age, log PSA density (PSAD), side-specific PIRADS 5, side-specific GG3-5, and percentage positivity of systematic cores. AUCs for dominant and non-dominant side EPE were 0.77 (95% CI 0.73-0.80) and 0.79 (95% CI 0.74-0.84), respectively. MRI-identified abutment and prostate health index (PHI) did not improve model discrimination. Risk calculators available online at https://rossnm1.shinyapps.io/PredictionOfEPELaterality/. <bold>Conclusions</bold> PSA, side-specific PIRADS, side-specific GG, and percentage positivity of systematic cores were associated with side-specific EPE at RP and incorporated into a risk calculator to assist in surgical planning and nerve-sparing decisions at time of RP.
BibTeX
@inproceedings{li2025prostate,
  author = {Eric V. Li and Sai Kumar and Jonathan Aguiar and Mohammad Adnan Siddiqui and Zequn Sun and Clayton Neill and Edward M. Schaeffer and Ashley E. Ross and Hiten D. Patel},
  title = {{Prostate MRI and Clinicopathologic Risk Calculator to Predict Laterality of Extraprostatic Extension at Radical Prostatectomy}},
  booktitle = {Prostate},
  year = {2025},
  doi = {10.1038/s41391-024-00928-7},
}
Figure