Can Predictive Models for Prostate Cancer Patients Derived in the United States of America Be Utilized in European Patients? A Validation Study of the Partin Tables

doi:10.1016/S0302-2838(02)00497-9

European Urology

Volume 43, Issue 1, January 2003, Pages 6-11

https://doi.org/10.1016/S0302-2838(02)00497-9 Get rights and content

Abstract

Objectives: Prostate cancer patients in the US and Europe differ due to selection and treatment differences. Accuracy of predictive tools derived in the US might therefore suffer when applied to European patients. We tested the validity of the widely accepted Partin tables for their ability to predict pathologic stage in German patients.

Methods: Clinical and pathological characteristics were obtained from 1298 consecutive men with clinically localized prostate cancer undergoing radical prostatectomy at the University Hospital Hamburg between January 1992 and February 2000. Receiver operating characteristic (ROC) curve analysis was performed to compare observed and predicted Partin rates for each pathologic stage.

Results: The rate for organ confinement was 56% in Hamburg patients compared to 48% in the Partin study. The rates of Hamburg patients for extracapsular extension without seminal vesicle or lymph node involvement were 25%, for seminal vesicle without lymph node involvement 14% and for lymph node metastases 5%. The corresponding rates of the Partin study were 40, 7 and 5%, respectively. The accuracy of Partin table derived probability was high with an area under the ROC curve of 0.817 (95% CI, 0.757–0.876) for organ confinement and 0.807 (95% CI, 0.781–0.833) for lymph node involvement.

Conclusion: Our study demonstrated that predictive tools for prostate cancer developed in the US could be applied to European patients with comparable accuracy to that reported for validation studies performed with US patients.

Introduction

In recent years we have witnessed the publication of a large number of statistical models and nomograms that predict pathologic stage or outcome in men with clinically localized prostate cancer (PCa) [1], [2], [3], [4], [5], [6], [7]. The vast majority of these predictive tools were derived in the US and it remains uncertain whether these nomograms would be accurate in European patients as detection and treatment strategy may vary in these men [8]. Predictive nomograms are helpful tools in the decision making process, however, validation of these biostatistical models is important since model performance may deteriorate when applied to a new and heterogeneous dataset separate from the training dataset.

The most popular and widespread used of these nomograms are the Partin tables, a predictive tool developed on 4133 patients who were treated in three different, highly reputable institutions in the US [1]. These tables were originally published in 1993 with data from patients treated by a single surgeon at Johns Hopkins University [9]. A validation study published by Kattan and co-workers showed only moderate reliability of these tables and lead to their update based on a multi-institutional dataset [1], [10]. Recently, Blute et al. applied theses tables to 2475 men who underwent radical prostatectomy (RP) at Mayo Clinic and confirmed reliability of this predictive tool in their patients [11].

Despite the good performance of the Partin tables in this separate dataset, concern remains about its general use especially in patients outside the US. Patient selection can substantially differ in Europe compared to the US and it needs to be confirmed whether predictive tools are able to adjust for these differences. The purpose of this study is to provide further validity assessment of the Partin tables using a dataset from an European center to assess whether predictive tools derived in the US would be accurate when applied to these patients.

Section snippets

Study design

Data from 1298 patients who underwent RP for clinically localized PCa between January 1992 and February 2000 at the University Hospital Hamburg-Eppendorf, Germany, were evaluated. Patients with neoadjuvant endocrine therapy, missing information on clinical stage, pretreatment prostate-specific antigen (PSA) level or biopsy Gleason sum were excluded, resulting in a cohort of 1131 consecutive patients available for statistical evaluation.

In all men, serum was obtained for PSA testing before

Results

Clinical and pathological variables from our validation cohort and the Partin study are compared in Table 1. Some differences in the datasets were apparent. T1c cancers were more frequently diagnosed in the Hamburg patients (51%) in comparison to the Partin study (33%). Pretreatment PSA was in general lower in the Partin cohort, with 23% of patients having a PSA level ≤4.0 ng/ml whereas only 9% of patients in our validation cohort showed a PSA level in this range. Overall, 15% of Hamburg

Discussion

In recent years many predictive tools have been published to predict pathologic stage and treatment outcome in patients with clinically localized prostate cancer. However, the only predictive tool published to date that was validated on an external multi-institutional dataset is the prediction of pathologic stage by Partin et al. [1]. These authors combined clinical stage, Gleason sum, and preoperative PSA to predict pathological stage in men treated with RP. We chose the Partin tables

Conclusions

A combination of pretreatment PSA, clinical stage and Gleason score as used by the Partin tables provide accurate prediction of pathologic stage. This could be demonstrated in a patient cohort where diagnostic and treatment strategies might differ compared to the US. While other clinical features have the potential to improve predictive accuracy of statistical models, the restriction to routinely obtained features as predictor variables allows general applicability.
Editorial Comment
Bob Djavan,

Acknowledgements

This research was supported by grants from the Deutsche Krebshilfe and Deutsche Forschungsgemeinschaft (GR 1866/1-1). Dr. Karakiewicz was partially supported by the American Foundation for Urologic Diseases, the National Cancer Institute of Canada and by the Medical Research Council of Canada.

References (20)

M.T. Gettman et al.
Prediction of patient outcome in pathologic stage T2 adenocarcinoma of the prostate: lack of significance for microvessel density analysis
Urology
(1998)
J.J. Bauer et al.
Biostatistical modeling using traditional preoperative and pathological prognostic variables in the selection of men at high risk for disease recurrence after radical prostatectomy for prostate cancer
J. Urol.
(1998)
A.W. Partin et al.
The use of prostate-specific antigen, clinical stage and Gleason score to predict pathological stage in men with localized prostate cancer
J. Urol.
(1993)
M.L. Blute et al.
Validation of Partin tables for predicting pathological stage of clinically localized prostate cancer
J. Urol.
(2000)
H. Huland et al.
Systematic biopsies and digital rectal examination to identify the nerve-sparing side for radical prostatectomy without risk of positive margin in patients with clinical stage T2, N0 prostatic carcinoma
Urology
(1994)
M. Graefen et al.
A validated strategy for side-specific prediction of organ confined prostate cancer: a tool to select for nerve-sparing radical prostatectomy
J. Urol.
(2001)
S. Soh et al.
Has there been a recent shift in the pathological features and prognosis of patients treated with radical prostatectomy?
J. Urol.
(1997)
S. Conrad et al.
Systematic sextant biopsies improve preoperative prediction of pelvic lymph nodemetastases in patients with clinically localized prostatic carcinoma
J. Urol.
(1998)
A.W. Partin et al.
Combination of prostate-specific antigen, clinical stage and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update
JAMA
(1997)
M. Graefen et al.
Early prostate-specific antigen relapse after radical retropubic prostatectomy: prediction on the basis of preoperative and postoperative tumor characteristics
Eur. Urol.
(1999)

There are more references available in the full text version of this article.

Cited by (63)

Validation of the Partin tables for prostate cancer in the Mexican population
2015, Revista Mexicana de Urologia
Las tablas de Partin ayudan a predecir el resultado de la pieza de la prostatectomía radical y decidir si el paciente se beneficiará del procedimiento. Dichos resultados se generaron en población estadounidense y se han validado en otros países. En México no existe ningún estudio que valide su uso en nuestra población.
Validar la predicción realizada mediante tablas de Partin en pacientes mexicanos con cáncer de próstata localizado sometidos a prostatectomía radical en un hospital de la Ciudad de México.
Se revisaron 106 expedientes de pacientes a quienes se les realizó prostatectomía radical como tratamiento para el cáncer de próstata localizado en el periodo de enero de 2004 a junio de 2014 y se compararon mediante curvas ROC los resultados obtenidos con la cohorte de Partin de 2007 con las que se validaron las tablas actualizadas.
La distribución de resultado patológico fue extensión extraprostática 22.6%, vesículas seminales 12.3%, ganglios linfáticos 2.8% mientras que en la cohorte de Partin de 2007 la distribución fue extensión extraprostática 22%, vesículas seminales 3%, ganglios linfáticos 1%. Al hacer el análisis con curvas ROC la precisión para predecir los resultados fue de 64.6, 81.5 y 89% para extensión extraprostática, vesículas seminales y ganglios linfáticos respectivamente.
Los resultados obtenidos corroboran la utilidad de las tablas de Partin de 2007 para predecir resultados patológicos en los pacientes de nuestra institución, con cifras similares e incluso mayores a los estudios de validación en Norteamérica y Europa.
En pacientes sometidos a prostatectomía radical en nuestra institución la precisión global de las tablas de Partin es similar a la cohorte original de 2007.
The Partin tables aid in predicting radical prostatectomy specimen results and in deciding whether a patient will benefit from the procedure. The results justifying the tables were from a United States population and have been validated in other countries. There are no studies in Mexico validating the use of these tables in the Mexican population.
To validate the prediction of results through the use of Partin tables in Mexican patients with localized prostate cancer that underwent radical prostatectomy at a single hospital in Mexico City.
One hundred and six case records were reviewed of patients that underwent radical prostatectomy as treatment for localized prostate cancer within the time frame of January 2004 to June 2014. For validation purposes, ROC curve results were compared with those of the 2007 Partin cohort.
The pathologic result distribution was extraprostatic extension 22.6%, seminal vesicle invasion 12.3%, and lymph node invasion 2.8%, whereas the distribution for the 2007 Partin cohort was extraprostatic extension 22%, seminal vesicle invasion 3%, and lymph node invasion 1%. The analysis of the ROC curves showed that the accuracy for predicting the results was 64.6, 81.5, and 89% for extraprostatic extension, seminal vesicle invasion, and lymph node invasion, respectively.
The results obtained corroborate the usefulness of the 2007 Partin tables for predicting the pathologic results in patients at our institution; our figures were similar and even higher than those of the validation studies in North America and Europe.
The overall accuracy of the Partin tables in patients that underwent radical prostatectomy in our institution was similar to that of the original 2007 cohort.
Predicting the risk of non-organ-confined prostate cancer when perineural invasion is found on biopsy
2014, Urology
To more precisely define the risk of non–organ-confined (non-OC) prostate cancer among men with perineural invasion (PNI) identified on prostate biopsy.
The Johns Hopkins radical prostatectomy database was queried for men with PNI reported on prostate biopsy. Patients with and without non-OC disease were compared for differences in preoperative clinical and pathologic characteristics, including three biopsy-based measures of tumor volume (number of cores with cancer, percentage of cores with cancer, and maximum percent core involvement with cancer). After evaluating the different preoperative variables in univariate analyses, a multivariable logistic regression model was generated, and bootstrap estimates of the risk of non-OC disease were calculated.
In total, 556 patients with PNI were analyzed, 279 (50.2%) of whom were found to have non-OC prostate cancer. In univariate analyses, preoperative prostate-specific antigen, clinical T stage, biopsy Gleason sum, and the three biopsy-based measures of tumor volume were significantly associated with non-OC disease. Of the three measures of tumor volume, the best fit to the data and highest degree of model discrimination were obtained using maximum percent core involvement with cancer. Incorporating this variable, preoperative prostate-specific antigen, clinical T stage, and biopsy Gleason sum into a multivariable model, the estimated risk of non-OC disease was found to range from 13.8% to 94.4% (bootstrap corrected c-index = 0.735).
Men with PNI on prostate biopsy are at a wide range of risk for non-OC disease. Preoperative estimation of this risk is improved by considering readily available biopsy estimates of tumor volume.
Assessing the most accurate formula to predict the risk of lymph node metastases from prostate cancer in contemporary patients treated with radical prostatectomy and extended pelvic lymph node dissection
2013, Radiotherapy and Oncology
The aim of this study was to perform a head-to-head comparison of the Roach formula vs. two other newly developed prediction tools for lymph node invasion (LNI) in prostate cancer, namely the Nguyen and the Yu formulas.
We included 3115 patients treated with radical prostatectomy and extended pelvic lymph node dissection (ePLND), between 2000 and 2010 at a single center. The predictive accuracy of the three formulas was assessed and compared using the area-under-curve (AUC) and calibration methods. Moreover, decision curve analysis compared the net-benefit of the three formulas in a head-to-head fashion.
Overall, 10.8% of patients had LNI. The LNI-predicted risk was >15% in 25.5%, 3.4%, and 10.2% of patients according to the Roach, Nguyen and Yu formula, respectively. The AUC was 80.5%, 80.5% and 79%, respectively (all p > 0.05). However, the Roach formula demonstrated more favorable calibration and generated the highest net-benefit relative to the other examined formulas in decision curve analysis.
All formulas demonstrated high and comparable discrimination accuracy in predicting LNI, when externally validated on ePLND treated patients. However, the Roach formula showed the most favorable characteristics. Therefore, its use should be preferred over the two other tools.
Development and validation of a UK-specific prostate cancer staging predictive model: UK prostate cancer tables
2012, British Journal of Medical and Surgical Urology
To construct new prostate cancer staging lookup tables based on a dataset collated by the British Association of Urological Surgeons (BAUS) and to validate them and compare their predictive power with Partin tables.
Complete data on 1701 patients was collated between 1999 and 2008 across 57 UK centres. Lookup tables were created for prediction of pathological stage (PS) using PSA level, biopsy Gleason score (GS) and clinical stage, replicating Partin's original approach.
Tables were generated using logistic regression (LR) and bootstrap resampling methods and were internally validated and externally validated using concordance indices (CI) and area under the receiver operating characteristic curve (AUROC) respectively.
The CI and AUROC analyses indicate that Partin tables performed poorly on UK data in comparison with US data.
The UK prostate cancer tables performed better than Partin tables but the predictive power of all models was relatively poor.
The study shows that the predictive power of Partin tables is reduced when applied to the UK population.
Models generated using LR methodology have fundamental limitations, and we suggest alternative modelling methods such as Bayesian networks.
National Comprehensive Cancer Network practice guidelines 2011: Need for more accurate recommendations for pelvic lymph node dissection in prostate cancer
2012, Journal of Urology
Citation Excerpt :
Specifically our current sample is based on a European cohort while our previous report sample was based on a North American cohort. Previous data suggest that European patients might harbor less favorable PCa variants than their North American counterparts.19,20 In particular, patients treated in Europe have shown a higher LNI risk compared to those treated in the United States.20
The 2011 NCCN (National Comprehensive Cancer Network) guidelines for prostate cancer recommend pelvic lymph node dissection at radical prostatectomy in all individuals with a nomogram predicted lymph node invasion probability of 2% or greater. We examined the ability of these guidelines to correctly predict lymph node invasion in patients treated with extended pelvic lymph node dissection.
We examined 3,064 consecutive patients treated with radical prostatectomy and extended pelvic lymph node dissection between 2000 and 2010. We formally validated the NCCN guideline nomogram using discrimination, calibration and decision curve analysis as benchmarks. Moreover the performance characteristics of the 2% nomogram cutoff as well as other cutoff values (range 1% to 10%) were tested.
Overall 10.0% of patients had lymph node invasion. The discrimination accuracy of the NCCN guideline nomogram was 79.8%, with a maximum underestimation of the lymph node invasion risk of 41.2%. On decision curve analysis the NCCN nomogram fared better than not performing pelvic lymph node dissection in all patients. However, in the prediction range between 0% and 9% the nomogram did not fare better than performing pelvic lymph node dissection in all patients. The use of the 2% cutoff would allow the avoidance of 49.3% of pelvic lymph node dissections, at the cost of missing 20.3% of patients with lymph node invasion.
The NCCN nomogram tends to significantly underestimate the real lymph node invasion rate. Moreover the use of the currently recommended cutoff of 2% to trigger pelvic lymph node dissection might not be appropriate.
Decision curve analysis to compare 3 versions of Partin Tables to predict final pathologic stage
2012, Urologic Oncology: Seminars and Original Investigations
To perform a decision curve analysis (DCA) to compare the Partin Tables 1997, 2001, and 2007 for their clinical applicability.
Clinical and pathologic data of 687 consecutive patients treated with open radical prostatectomy for clinically localized prostate cancer between 2003 and 2008 at a single institution were used. DCA quantified the net benefit relating to specific threshold probabilities of extraprostatic extension (EPE), seminal vesicle involvement (SVI), and lymph node involvement (LNI).
Overall, EPE, SVI, and LNI were recorded in 17.8, 6.0, and 1.2%, respectively. For EPE predictions, the DCA favored the 2007 version vs. 1997 for SVI vs. none of the versions for LNI.
DCA indicate that for very low prevalence conditions such as LNI (1.2%), decision models are not useful. For low prevalence rates such as SVI, the use of different versions of the Partin Tables does not translate into meaningful net gains differences. Finally, for intermediate prevalence conditions such as EPE (18%), despite apparent performance differences, the net benefit differences were also marginal. In consequence, the current analysis could not confirm an important benefit from the use of the Partin Tables and it could not identify a clearly better version of any of the 3 available iterations.

View all citing articles on Scopus

View full text

Can Predictive Models for Prostate Cancer Patients Derived in the United States of America Be Utilized in European Patients? A Validation Study of the Partin Tables

Abstract

Introduction

Section snippets

Study design

Results

Discussion

Conclusions

Acknowledgements

Urology

J. Urol.

J. Urol.

J. Urol.

Urology

J. Urol.

J. Urol.

J. Urol.

Combination of prostate-specific antigen, clinical stage and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update

JAMA

Early prostate-specific antigen relapse after radical retropubic prostatectomy: prediction on the basis of preoperative and postoperative tumor characteristics

Eur. Urol.