External Evaluation of the Briganti Nomogram to Predict Lymph Node Metastases in Intermediate-Risk Prostate Cancer Patients - Beyond the Abstract

Nicolas Peilleron and colleagues performed an evaluation of the precision of the 2012 Briganti nomogram among 285 patients with localized prostate cancer, treated by radical prostatectomy and bilateral lymph node dissection, focusing on 175 intermediate-risk patients.

Predicting the risk of lymph node involvement is still a very active research field. Extended pelvic lymph node dissection adds significant operating time and morbidity to the radical prostatectomy procedure and should, therefore, be offered to selected patients, more likely to benefit from it. While guidelines clearly state that we should use nomograms to estimate the risk and select patients, little is known on the actual applicability of these nomograms in a different setting.

When we started this study, the 2012 Briganti nomogram was the most recently published nomogram relying solely on clinical, biological, and biopsy data (PSA, CT stage, primary and secondary Gleason pattern, and percentage of positive biopsy cores – number of positive cores/total number of cores). It is therefore easy to calculate and the information needed is readily available, making it one of the most widely used nomograms.

Nomograms are usually developed from a global cohort, including patients from all risk groups. However, since lymph node dissection is usually not recommended for low-risk patients (and these patients are now very rarely treated with radical prostatectomy), and strongly recommended for high-risk disease, we believe that intermediate-risk patients are the population of interest.

Interestingly, and probably as expected, the results when applying the nomogram in the global cohort were quite good, with the Briganti nomogram showing an area under the curve (AUC) of 0.75 (with 1 being perfect, and 0.5 close to random). However, when looking at intermediate-risk patients, our population of interest, the AUC decreased to 0.53 (Figure 1). In this specific population, the systematic application of the nomogram with a 5% threshold to select patients for lymph node dissection would have led to the misclassification of 6.6% of patients with a low score but lymph node invasion, while the rate of positive nodes in patients with a higher score was only 8.3%.

ROC and area under the curve

We tried to find predictive factors of lymph node invasion not taken into account in the nomogram. Besides prostate-specific antigen (PSA) and the proportion of positive biopsies, a T3 stage on MRI was associated with a higher risk.

We also had a look at various thresholds that could be applied and whether they would be more relevant in our population. In order to do this, we used both the number of lymph nodes dissections needed to detect one lymph node invasion and the number of positive lymph nodes missed. We found that a threshold of 4% would require 12 dissections to detect 1 lymph node involvement while missing four patients (6%) with positive nodes.

No study is perfect, and ours is no exception! The main limitation lies in the relatively small number of nodes retrieved (median of 10 (7-14) for the entire cohort), and the possible misclassification of patients considered as N0 although lymph node invasion was present in the remaining nodes. To ensure that the contamination of the N0 group was limited, we performed survival analyses, which confirmed very different five-year survival profiles of N0 and N+ patients (Figure 3).

comparative survival curves

Nomograms are created and validated from development and validation cohorts in a defined setting, and it is therefore not surprising that their precision is reduced when used outside this setting. However, the nomogram applied in our intermediate-risk cohort displayed a lower-than-expected precision. This can be explained by the selection of patients different from the development cohort, which gathered patients from all risk groups. Future nomogram development studies will have to focus on intermediate-risk patients, as it is the population of interest for nomogram applications.

Besides, lymph node invasion is a rare event, explaining why the development of precise prediction tools is not straightforward. Even with nomogram use, many patients still get unnecessary procedures.

Let’s hope that the most recent versions of the Briganti nomogram including imaging data, or new imaging modalities such as molecular imaging will allow for a more accurate patient selection in the future!

Written by: Gaelle Fiard, MD, PhD, and Nicolas Peilleron, MD, Department of Urology, Grenoble Alpes University Hospital, Grenoble, France

Read the Abstract