- Email: [email protected]
The Newer the Better? Comparison of the 1997 and 2001 Partin Tables for Pathologic Stage Prediction of Prostate Cancer in China
Oncology The Newer the Better? Comparison of the 1997 and 2001 Partin Tables for Pathologic Stage Prediction of Prostate Cancer in China Xu Gao, Shancheng Ren, Xin Lu, Chuanliang Xu, and Yinghao Sun OBJECTIVES
METHODS
RESULTS
CONCLUSIONS
The use of the Partin tables for pathologic stage prediction of prostate cancer in China has 2 problems. First, the validity of such U.S.-derived predictive tools has not been tested in China. Second, because an updated version of the Partin tables created using a more contemporary cohort was introduced in 2001, it is important to determine whether this renewed version (2001 Partin tables) is superior to the old version (1997 Partin tables). To investigate these problems, we compared these 2 predictive tools for their ability to predict the pathologic stage in Chinese patients. From June 2001 to May 2007, 271 consecutive patients who met the inclusion criteria for Partin table prediction and underwent radical prostatectomy for clinically localized prostate cancer at a single center were evaluated. The predictive accuracy of the 1997 Partin tables and 2001 Partin tables were assessed using receiver operating characteristics analysis. For the 1997 Partin tables, the area under the receiver operating characteristics curve for the prediction of organ-confined disease, established capsular penetration, seminal vesicle involvement, and lymph node involvement was 0.723, 0.674, 0.749, and 0.781, respectively. The corresponding values for the 2001 Partin tables were consistently low at 0.604, 0.539, 0.654, and 0.631. The results of our study have confirmed good predictive accuracy for the 1997 Partin tables for Chinese patients, but not for their updated 2001 version. UROLOGY 72: 1096 –1101, 2008. © 2008 Elsevier Inc.
B
ecause radical prostatectomy (RP) is most effective for organ-confined prostate cancer (PCa), the key question is how to predict the pathologic stage using the preoperative information. The 1997 Partin tables, developed using data from 4133 American patients with PCa, uses 3 preoperative variables, serum prostate-specific antigen (PSA) level, biopsy Gleason score, and clinical stage, to predict 4 pathologic stages: organ-confined disease (OCD), established capsular penetration (ECP), seminal vesicle involvement (SVI), and lymph node involvement (LNI).1 In 2001, this predictive tool was updated to reflect a more contemporary condition of PCa stage at diagnosis.2 After the introduction of the 1997 and 2001 Partin tables, their reliability was tested in different clinical
Xu Gao and Shancheng Ren wish to note that they share first authorship of this article. From the Department of Urology, Changhai Hospital, Second Military Medical University, Shanghai, China Reprint requests: Yinghao Sun, M.D., Department of Urology, Changhai Hospital, Second Military Medical University, 168 Changhai Road, Shanghai 200433 People’s Republic of China. E-mail: [email protected] Submitted: April 15, 2008, accepted (with revisions): July 21, 2008
1096
© 2008 Elsevier Inc. All Rights Reserved
scenarios in the United States and Europe,3-7 but not in China. Because of the different biologic features and the detection procedure of PCa used in China (eg, the obviously lower incidence and no PSA screening for the public), such U.S.-derived predictive tools should be tested in China.
MATERIAL AND METHODS From June 2001 to May 2007, 302 patients underwent RP for localized PCa (confirmed by 10-core biopsy) at our department. Patients who had undergone neoadjuvant endocrine therapy, had missing clinical stage, pretreatment PSA level, or Gleason score information were excluded, resulting in a cohort of 271 consecutive patients available for validation. However, patients with clinical Stage T1a-T1b (n ⫽ 8) and T3a (n ⫽ 9) were also excluded from the test of the 2001 Partin tables, because predictive values were no longer provided for patients with these disease stages in the 2001 version. When the final pathologic stage was predicted using the 1997 Partin tables, the nomograms provided each patient with a probability. However, when the SVI and LNI were predicted using the 2001 Partin tables, the nomograms could not provide a probability for 29 and 31 0090-4295/08/$34.00 doi:10.1016/j.urology.2008.07.047
Table 1. Clinical and pathologic characteristics of present cohort and Partin cohort Characteristic
Present Cohort (n ⫽ 271)
Partin Cohort (n ⫽ 4133)
5 (1.8) 3 (1.1) 76 (28.0) 85 (31.4) 83 (30.6) 10 (3.7) 9 (3.3)
1 3 33 29 21 10 3
7 (2.6) 92 (33.9) 95 (35.1) 77 (28.4)
23 48 21 8
30 (11.1) 31 (11.4) 104 (38.4) 70 (25.8) 36 (13.3)
5 17 51 22 5
148 (54.6) 66 (24.4) 30 (11.1) 27 (10.0)
48 40 7 5
Clinical stage T1a T1b T1c T2a T2b T2c T3a PSA (ng/mL) ⬍4 4-10 10-20 ⬎20 Gleason score ⬍5 5 6 7 ⬎7 Pathologic stage OCD ECP SVI LNI
Abbreviations: PSA, prostate-specific antigen; OCD, organ-confined disease; ECP, established capsular penetration; SVI, seminal vesicle involvement; LNI, lymph node involvement. Data in parentheses are percentages.
patients, respectively, because the data were insufficient data to calculate the probability. All pretreatment PSA levels were measured before prostatic manipulation. The clinical stage was defined according to the 1992 TNM staging system. The pathologic staging into OCD, ECP, SVI, and LNI categories was determined using the criteria described by Partin et al.1,2 and performed by a single pathologist. Statistical calculations were performed using the Statistical Package for Social Sciences software, version 11.0 (SPSS, Chicago, IL) and Statistical Analysis Systems, version 9.0 (SAS Institute, Cary, NC). Two-sided tests with significance at P ⫽ .05 were used. The predictive accuracies of the 1997 and 2001 Partin tables were quantified using receiver operating characteristics (ROC) analysis and area under the ROC curve (AUC) assessment. The extent of overestimation or underestimation of the observed pathologic stage by the 1997 Partin tables was explored using the nonparametric, local regression Loess smoothing technique.
RESULTS All patients in our cohort were Chinese, and the median age was 67 (ranged 53-77). The comparison of the distribution of clinical and pathologic variables for the 1997 Partin tables study and the present study is presented in Table 1. The comparison of the distribution for the 2001 Partin tables study and the present study is presented in Table 2. Figure 1 shows the AUCs for a given pathologic stage when these 2 nomograms were applied to the present cohort. UROLOGY 72 (5), 2008
Table 2. Clinical and pathologic characteristics of present cohort and Partin update cohort Characteristic Clinical stage T1c T2a T2b T2c PSA (ng/mL): 0-2.5 2.6-4 4.1-6 6.1-10 ⬎10 Gleason score 2-4 5-6 3⫹4 4⫹3 8-10 Pathologic stage OCD ECP SVI LNI
Present Cohort* (n ⫽ 254)
Parting Update Cohort (n ⫽ 5079)
76 (29.9) 85 (33.5) 83 (32.7) 10 (3.9)
63 32 11 3
0 (0.0) 5 (2.0) 25 (9.8) 61 (24.0) 163 (64.2)
7 10 27 35 21
28 (11.0) 127 (50.0) 29 (11.4) 37 (14.6) 33 (13.0)
0.6 79 13 4.4 3
140 (55.1) 64 (25.2) 26 (10.2) 24 (9.4)
64 30 4 2
Abbreviations as in Table 1. Data in parentheses are percentages. * Seventeen patients with clinical Stage T1a-T1b or T3a excluded during validation of 2001 Partin tables because new tables no longer provide predictive values for them.
The AUC of the 1997 Partin tables was 0.723 (95% confidence interval [CI] 0.664-0.782), 0.674 (95% CI 0.603-0.744), 0.749 (95% CI 0.667-0.830), and 0.781 (95% CI 0.688-0.874) for OCD, ECP, SVI, and LNI, respectively. The corresponding AUCs of the 2001 Partin tables were consistently low at 0.604 (95% CI 0.5350.673), 0.539 (95% CI 0.463-0.616), 0.654 (95% CI 0.553-0.755), and 0.631 (95% CI 0.522-0.740). These statistical data have confirmed that the 1997 Partin tables had reasonably predictive accuracy for OCD, SVI, and LNI (AUC ⬎0.7) and that the 2001 Partin tables had lower predictive accuracy in the present cohort. Figure 2 shows the Loess smoothing plots of the prediction of the observed pathologic stage using the 1997 Partin tables. The nomogram tended to underestimate the probability of OCD and overestimate the probability of SVI and LNI. For the prediction of ECP, an important overestimation was noted for 0%-60%, but for ⬎60%, an underestimation was noted.
COMMENT Of the patients diagnosed with clinically localized PCa, ⬎40% are found to have pathologic extraprostatic extension or pelvic LNI. For these patients, RP alone cannot cure the cancer, and about 50% will develop recurrence within 5 years after surgery.8,9 Thus, the prediction of the pathologic PCa stage is of great importance in 1097
Figure 1. Receiver operating characteristic curves of (Left) 1997 Partin tables and (Right) 2001 Partin tables for predicting different pathologic features: (A) Organ-confined disease, (B) established capsular penetration, (C) seminal vesicle involvement, and (D) lymph node involvement.
Figure 2. Nonparametric, local regression Loess smoothing plots of pathologic stage prediction using 1997 Partin tables. Table-predicted probabilities shown on x-axes; y-axes represent observed pathologic stage. Perfect predictions correspond to slope of 1. (A) Organ-confined disease (OCD), (B) established capsular penetration (ECP), (C) seminal vesicle involvement (SVI), and (D) lymph node involvement (LNI).
1098
UROLOGY 72 (5), 2008
treatment decision making. Because imaging studies are not accurate for staging PCa, preoperative clinical and pathologic parameters are often used to predict the pathologic stage and thus identify the patients most likely to benefit from the RP. Many predictive models have been published to predict the pathologic outcomes of patients with clinically localized PCa.10-14 These predictive tools can help counsel individual patients in choosing the most appropriate management option. Among these predictive tools, the most widely used is the 1997 Partin tables. This nomogram was constructed in the 1990s and had been validated externally and interracially several times, reflecting its good performance in different scenarios.3,4,6,7 As the only predictive tool of the pathologic stage that has been extensively validated, the 1997 Partin tables have been recommended for their clinical application in the seventh and eighth editions of Campbell’s Urology.15,16 This nomogram has also gained acceptance as a useful guide in clinical practice in the United States and other countries. However, in the PSA era, a tremendous change has occurred in the stage of PCa at presentation, especially in the United States and European countries, with more PCa detected in men with a lower serum PSA level and biopsy Gleason score. In accordance with this change, the 1997 Partin tables were updated in 20012 to provide a more accurate prediction of the pathologic stage under this “migration” phenomenon. The 2001 Partin tables have not been as thoroughly evaluated for their predictive accuracy5-7; however, they have been substituted for the older Partin tables in the ninth edition of CampbellWalsh Urology.17 Unlike in Western countries, PCa is not a common malignancy in China. However, several reports have shown a trend in China toward an increasing incidence of PCa. Sim and Cheng18 reviewed the registered cases of PCa in some Asian countries, including China, from 1978 to 1997 and found that the incidence had increased by 5%-118%. In addition, the wide spread of the PSA test, which promotes the early diagnosis of PCa, has resulted in RP being performed more widely in China than before.18-20 Therefore, Chinese urologists urgently need to adopt a predictive pathologic outcome tool in their clinical practice, such as the Partin tables. Thus, 2 key questions caused our concern. First, the 1997 and 2001 Partin tables were derived in the United States, and it remains uncertain whether these nomograms would be accurate for Chinese patients, because the biologic features and disease detection could vary among these patients. Second, the 2001 Partin tables were constructed from a more contemporary cohort of U.S. patients to reflect the tremendous change in the stage of PCa at presentation in the United States. The intent of this new version is to provide more accurate information about the pathologic stage for patients durUROLOGY 72 (5), 2008
ing this stage migration. However, whether this newer version is more applicable than the older version when applied in Chinese clinical practice remains unknown. The present study was conducted to explore these questions. Although the possible limitations of this study were the relatively small sample size and the single-center study, we believe that our results are valid. Considering the low incidence of PCa in China, and the consecutive enrollment of the present cohort, 271 patients could serve as a representative sample for validation. Another limitation of our study was that it was a single-center study. If the present study had been conducted as a multicenter study, although the number of the cases could have been greater, the validation results might have been less convincing. The use of a single center meant that all surgery was performed exclusively by 2 experienced urologists (X.G. and Y.H.S.), nearly all patients (only 9 patients [3%] were excluded from the 302 cases because of missing parameters) had the clinical data required for the validation, and central pathologic review could be easily achieved. Our results have demonstrated that the 1997 Partin tables could be applied to Chinese patients, although this predictive tool was derived in the United States. The AUC for OCD (0.723), SVI (0.749), and LNI (0.781) showed the accurate predictive ability of this nomograms when applied to our cohort. Similar results could be seen from several previous validation studies performed in and outside the United States, confirming that the 1997 Partin tables are able to adjust for patient selection differences across different clinical scenarios.3,4,6,7 The Loess plots in Figure 2 showed that the nomogram tended to underestimate OCD and generally overestimate ECP, SVI, and LNI. This should be noted during clinical application because patients with predicted late-stage disease might select alternative, potentially less curable, treatment modalities. However, we have concluded that the 2001 Partin tables are not accurate enough to apply to urologic practice in China. The AUC of the 2001 Partin tables was 0.604, 0.539, 0.654, and 0.631 for OCD, ECP, SVI, and LNI, respectively, consistently inferior to those of the 1997 Partin tables. Because an AUC of ⬍0.700 could be considered a poor result in ROC analysis, we believe that the 2001 Partin tables do not have a reasonable predictive value for the final pathologic features in Chinese patients with PCa. The decreased accuracy of the 2001 Partin tables was likely because of the following. First, similarly lower pretreatment variables was seen in the 2001 Partin cohort compared with both to our cohort and the 1997 Partin cohort (Tables 1 and 2). For example, 62.3% of our patients had Stage T2a PCa or less and a comparable ratio of 66% was seen in the Partin 1997 cohort. However, a significantly greater ratio of 95% was seen in the development cohort of the 2001 Partin tables. Also, the percentage of 1099
patients with a Gleason score of 2-4 and ⱖ7 was 11.1% and 25.8% in our study, respectively. The corresponding ratios had decreased to 5% and 22% in the 1997 Partin cohort and had decreased further to 0.6% and 17.4% in the 2001 Partin cohort. Again, the corresponding ratio in the 1997 Partin cohort was between that of our study and the 2001 Partin cohort. According to Tables 1 and 2, the other ratios in the 1997 Partin cohort could also be seen between that of our patients and that of the 2001 Partin cohort, such as the percentage of patients with a PSA of 4-10 ng/mL or ⱖ10 ng/mL and with SVI and LNI, except for the percentage of patients with OCD, which was lowest in the 1997 Partin cohort (48%) and the greatest in the 2001 Partin cohort (64%). Differing population characteristics between the validation cohort and the development cohort are quite common and also have been well described in other validation studies.3-7 To a certain extent, these differences could be compensated for by the transportability of a well-performed predictive tool, such as the 1997 Partin tables in the present study. However, as the differences show a trend toward more significance, the accuracy of a predictive tool will be compromised, even to the point of no longer qualifying for clinical practice. In relation to our validation study, this is one of the main reasons accounting for the lower accuracy of the 2001 Partin tables. Second, the basic structural changes of the 2001 Partin tables have reduced their predictive accuracy for the present cohort. During the past decades, a tremendous PCa stage migration has occurred in western countries. In accordance with this change, Partin et al.1,2 revised their former nomogram with a more contemporary cohort in 2001. In this revised nomogram, some basic structural changes were made. For example, the PSA groups were stratified into narrower intervals when ⬍10 ng/mL. However, the new tables put those patients with a PSA of ⬎10 ng/mL into the same group. This might be more suitable to a validation setting with a low PSA level. Nonetheless, when applied to the present validation cohort, in which nearly 65% of the patients had a PSA level of ⬎10 ng/mL, the theoretical advantage related to narrower divisions in the pretreatment PSA level has not yielded an appreciable gain in predictive accuracy. Finally, the 2001 Partin tables provided low coverage for the present validation cohort. In our cohort, 17 patients (6.3%) in our validation cohort had clinical Stage T1a-T1b or T3a. These patients were excluded from the validation of the 2001 tables because the new tables no longer provided predictive values for individuals with these stages. Furthermore, the probabilities were not available for another 29 and 31 patients (10.7% and 11.4%) for the predicted SVI and LNI using the newer tables according to their clinical variables. These data were considered as missing during the final ROC analysis. Even though the missing data do not necessarily have 1100
negative effect on the accuracy of the predictive tool, a more comprehension one (ie, the 1997 Partin tables), would be more suitable for clinical use from intuitive and practical considerations.
CONCLUSIONS Our results have indicated that the 1997 Partin tables adequately predict the pathologic outcomes in Chinese men who undergo RP for PCa. They are particularly useful for the prediction of OCD, SVI, and LNI. However, the 2001 version of the Partin tables, which were updated in 2001 to “better represent the contemporary presentation of PCa with stage shift phenomena,” might not be applicable to general urologic practice in China because their predictive values for each pathologic feature were consistently lower than those of the 1997 version. References 1. Partin AW, Kattan MW, Subong EN, 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;277:1445-1451. 2. Partin AW, Mangold LA, Lamm DM, et al. Contemporary update of prostate cancer staging nomograms (Partin tables) for the new millennium. Urology. 2001;58:843-848. 3. Blute ML, Bergstralh EJ, Partin AW, et al. Validation of Partin tables for predicting pathological stage of clinically localized prostate cancer. J Urol. 2000;164:1591-1595. 4. Graefen M, Augustin H, Karakiewicz PI, et al. 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. Eur Urol. 2003;43:6-11. 5. Eskicorapci SY, Karabulut E, Turkeri L, et al. Validation of 2001 Partin tables in Turkey: A multicenter study. Eur Urol. 2005;47: 185-189. 6. Ayyathurai R, Ananthakrishnan K, Rajasundaram R, et al. Predictive ability of Partin tables 2001 in a Welsh population. Urol Int. 2006;76:217-222. 7. Augustin H, Eggert T, Wenske S, et al. Comparison of accuracy between the Partin tables of 1997 and 2001 to predict final pathological stage in clinically localized prostate cancer. J Urol. 2004; 171:177-181. 8. Hull GW, Rabbani F, Abbas F, et al. Cancer control with radical prostatectomy alone in 1,000 consecutive patients. J Urol. 2002; 167:528-534. 9. Eastham JA, Scardino PT. Radical prostatectomy. In: Walsh PC, Retik AB, Vaughan ED, et al. eds. Campbell’s Urology, Vol. 4. Philadelphia: WB Saunders; 2002:3080-3106. 10. Peller PA, Young DC, Marmaduke DP, et al. Sextant prostate biopsies: A histopathologic correlation with radical prostatectomy specimens. Cancer. 1995;75:530-538. 11. Bostwick DG, Qian J, Bergstralh E, et al. Prediction of capsular perforation and seminal vesicle invasion in prostate cancer. J Urol. 1996;155:1361-1367. 12. Gilliland FD, Hoffman RM, Hamilton A, et al. Predicting extracapsular extension of prostate cancer in men treated with radical prostatectomy: Results from the population based prostate cancer outcomes study. J Urol. 1999;162:1341-1345. 13. Rabbani F, Bastar A, Fair WR. Site specific predictors of positive margins at radical prostatectomy: An argument for risk based modification of technique. J Urol. 1998;160:1727-1733. 14. Badalament RA, Miller MC, Peller PA, et al. An algorithm for predicting organ confined prostate cancer using the results obtained
UROLOGY 72 (5), 2008
from sextant core biopsies with prostate specific antigen level. J Urol. 1996;156:1375-1380. 15. Carter HB, Partin AW. Diagnosis and staging of prostate cancer. In: Walsh PC, Retik AB, Vaughan ED, et al., eds. Campbell’s Urology. Philadelphia: WB Saunders; 1998:2519-2537. 16. Carter HB, Partin AW. Diagnosis and staging of prostate cancer. In: Walsh PC, Retik AB, Vaughan ED, et al., eds. Campbell’s Urology. Philadelphia: WB Saunders; 2002:3055-3079. 17. Carter HB, Allaf ME, Partin AW. Diagnosis and staging of prostate cancer. In: Wein AJ, Kavoussi LR, Novick AC, et al.,
UROLOGY 72 (5), 2008
eds. Campbell-Walsh Urology. Philadelphia: Saunders Elsevier; 2007:2912-2931. 18. Sim HG, Cheng CW. Changing demography of prostate cancer in Asia. Eur J Cancer. 2005;41:834-845. 19. Peyromaure M, Debre B, Mao K, et al. Management of prostate cancer in China: A multicenter report of 6 institutions. J Urol. 2005;174:1794-1797. 20. Tsuji X, Li I, Kuwahara M, et al. Mass screening of prostate cancer in Changchun city of China. Int Urol Nephrol. 2004;36: 541-548.
1101
METHODS
RESULTS
CONCLUSIONS
The use of the Partin tables for pathologic stage prediction of prostate cancer in China has 2 problems. First, the validity of such U.S.-derived predictive tools has not been tested in China. Second, because an updated version of the Partin tables created using a more contemporary cohort was introduced in 2001, it is important to determine whether this renewed version (2001 Partin tables) is superior to the old version (1997 Partin tables). To investigate these problems, we compared these 2 predictive tools for their ability to predict the pathologic stage in Chinese patients. From June 2001 to May 2007, 271 consecutive patients who met the inclusion criteria for Partin table prediction and underwent radical prostatectomy for clinically localized prostate cancer at a single center were evaluated. The predictive accuracy of the 1997 Partin tables and 2001 Partin tables were assessed using receiver operating characteristics analysis. For the 1997 Partin tables, the area under the receiver operating characteristics curve for the prediction of organ-confined disease, established capsular penetration, seminal vesicle involvement, and lymph node involvement was 0.723, 0.674, 0.749, and 0.781, respectively. The corresponding values for the 2001 Partin tables were consistently low at 0.604, 0.539, 0.654, and 0.631. The results of our study have confirmed good predictive accuracy for the 1997 Partin tables for Chinese patients, but not for their updated 2001 version. UROLOGY 72: 1096 –1101, 2008. © 2008 Elsevier Inc.
B
ecause radical prostatectomy (RP) is most effective for organ-confined prostate cancer (PCa), the key question is how to predict the pathologic stage using the preoperative information. The 1997 Partin tables, developed using data from 4133 American patients with PCa, uses 3 preoperative variables, serum prostate-specific antigen (PSA) level, biopsy Gleason score, and clinical stage, to predict 4 pathologic stages: organ-confined disease (OCD), established capsular penetration (ECP), seminal vesicle involvement (SVI), and lymph node involvement (LNI).1 In 2001, this predictive tool was updated to reflect a more contemporary condition of PCa stage at diagnosis.2 After the introduction of the 1997 and 2001 Partin tables, their reliability was tested in different clinical
Xu Gao and Shancheng Ren wish to note that they share first authorship of this article. From the Department of Urology, Changhai Hospital, Second Military Medical University, Shanghai, China Reprint requests: Yinghao Sun, M.D., Department of Urology, Changhai Hospital, Second Military Medical University, 168 Changhai Road, Shanghai 200433 People’s Republic of China. E-mail: [email protected] Submitted: April 15, 2008, accepted (with revisions): July 21, 2008
1096
© 2008 Elsevier Inc. All Rights Reserved
scenarios in the United States and Europe,3-7 but not in China. Because of the different biologic features and the detection procedure of PCa used in China (eg, the obviously lower incidence and no PSA screening for the public), such U.S.-derived predictive tools should be tested in China.
MATERIAL AND METHODS From June 2001 to May 2007, 302 patients underwent RP for localized PCa (confirmed by 10-core biopsy) at our department. Patients who had undergone neoadjuvant endocrine therapy, had missing clinical stage, pretreatment PSA level, or Gleason score information were excluded, resulting in a cohort of 271 consecutive patients available for validation. However, patients with clinical Stage T1a-T1b (n ⫽ 8) and T3a (n ⫽ 9) were also excluded from the test of the 2001 Partin tables, because predictive values were no longer provided for patients with these disease stages in the 2001 version. When the final pathologic stage was predicted using the 1997 Partin tables, the nomograms provided each patient with a probability. However, when the SVI and LNI were predicted using the 2001 Partin tables, the nomograms could not provide a probability for 29 and 31 0090-4295/08/$34.00 doi:10.1016/j.urology.2008.07.047
Table 1. Clinical and pathologic characteristics of present cohort and Partin cohort Characteristic
Present Cohort (n ⫽ 271)
Partin Cohort (n ⫽ 4133)
5 (1.8) 3 (1.1) 76 (28.0) 85 (31.4) 83 (30.6) 10 (3.7) 9 (3.3)
1 3 33 29 21 10 3
7 (2.6) 92 (33.9) 95 (35.1) 77 (28.4)
23 48 21 8
30 (11.1) 31 (11.4) 104 (38.4) 70 (25.8) 36 (13.3)
5 17 51 22 5
148 (54.6) 66 (24.4) 30 (11.1) 27 (10.0)
48 40 7 5
Clinical stage T1a T1b T1c T2a T2b T2c T3a PSA (ng/mL) ⬍4 4-10 10-20 ⬎20 Gleason score ⬍5 5 6 7 ⬎7 Pathologic stage OCD ECP SVI LNI
Abbreviations: PSA, prostate-specific antigen; OCD, organ-confined disease; ECP, established capsular penetration; SVI, seminal vesicle involvement; LNI, lymph node involvement. Data in parentheses are percentages.
patients, respectively, because the data were insufficient data to calculate the probability. All pretreatment PSA levels were measured before prostatic manipulation. The clinical stage was defined according to the 1992 TNM staging system. The pathologic staging into OCD, ECP, SVI, and LNI categories was determined using the criteria described by Partin et al.1,2 and performed by a single pathologist. Statistical calculations were performed using the Statistical Package for Social Sciences software, version 11.0 (SPSS, Chicago, IL) and Statistical Analysis Systems, version 9.0 (SAS Institute, Cary, NC). Two-sided tests with significance at P ⫽ .05 were used. The predictive accuracies of the 1997 and 2001 Partin tables were quantified using receiver operating characteristics (ROC) analysis and area under the ROC curve (AUC) assessment. The extent of overestimation or underestimation of the observed pathologic stage by the 1997 Partin tables was explored using the nonparametric, local regression Loess smoothing technique.
RESULTS All patients in our cohort were Chinese, and the median age was 67 (ranged 53-77). The comparison of the distribution of clinical and pathologic variables for the 1997 Partin tables study and the present study is presented in Table 1. The comparison of the distribution for the 2001 Partin tables study and the present study is presented in Table 2. Figure 1 shows the AUCs for a given pathologic stage when these 2 nomograms were applied to the present cohort. UROLOGY 72 (5), 2008
Table 2. Clinical and pathologic characteristics of present cohort and Partin update cohort Characteristic Clinical stage T1c T2a T2b T2c PSA (ng/mL): 0-2.5 2.6-4 4.1-6 6.1-10 ⬎10 Gleason score 2-4 5-6 3⫹4 4⫹3 8-10 Pathologic stage OCD ECP SVI LNI
Present Cohort* (n ⫽ 254)
Parting Update Cohort (n ⫽ 5079)
76 (29.9) 85 (33.5) 83 (32.7) 10 (3.9)
63 32 11 3
0 (0.0) 5 (2.0) 25 (9.8) 61 (24.0) 163 (64.2)
7 10 27 35 21
28 (11.0) 127 (50.0) 29 (11.4) 37 (14.6) 33 (13.0)
0.6 79 13 4.4 3
140 (55.1) 64 (25.2) 26 (10.2) 24 (9.4)
64 30 4 2
Abbreviations as in Table 1. Data in parentheses are percentages. * Seventeen patients with clinical Stage T1a-T1b or T3a excluded during validation of 2001 Partin tables because new tables no longer provide predictive values for them.
The AUC of the 1997 Partin tables was 0.723 (95% confidence interval [CI] 0.664-0.782), 0.674 (95% CI 0.603-0.744), 0.749 (95% CI 0.667-0.830), and 0.781 (95% CI 0.688-0.874) for OCD, ECP, SVI, and LNI, respectively. The corresponding AUCs of the 2001 Partin tables were consistently low at 0.604 (95% CI 0.5350.673), 0.539 (95% CI 0.463-0.616), 0.654 (95% CI 0.553-0.755), and 0.631 (95% CI 0.522-0.740). These statistical data have confirmed that the 1997 Partin tables had reasonably predictive accuracy for OCD, SVI, and LNI (AUC ⬎0.7) and that the 2001 Partin tables had lower predictive accuracy in the present cohort. Figure 2 shows the Loess smoothing plots of the prediction of the observed pathologic stage using the 1997 Partin tables. The nomogram tended to underestimate the probability of OCD and overestimate the probability of SVI and LNI. For the prediction of ECP, an important overestimation was noted for 0%-60%, but for ⬎60%, an underestimation was noted.
COMMENT Of the patients diagnosed with clinically localized PCa, ⬎40% are found to have pathologic extraprostatic extension or pelvic LNI. For these patients, RP alone cannot cure the cancer, and about 50% will develop recurrence within 5 years after surgery.8,9 Thus, the prediction of the pathologic PCa stage is of great importance in 1097
Figure 1. Receiver operating characteristic curves of (Left) 1997 Partin tables and (Right) 2001 Partin tables for predicting different pathologic features: (A) Organ-confined disease, (B) established capsular penetration, (C) seminal vesicle involvement, and (D) lymph node involvement.
Figure 2. Nonparametric, local regression Loess smoothing plots of pathologic stage prediction using 1997 Partin tables. Table-predicted probabilities shown on x-axes; y-axes represent observed pathologic stage. Perfect predictions correspond to slope of 1. (A) Organ-confined disease (OCD), (B) established capsular penetration (ECP), (C) seminal vesicle involvement (SVI), and (D) lymph node involvement (LNI).
1098
UROLOGY 72 (5), 2008
treatment decision making. Because imaging studies are not accurate for staging PCa, preoperative clinical and pathologic parameters are often used to predict the pathologic stage and thus identify the patients most likely to benefit from the RP. Many predictive models have been published to predict the pathologic outcomes of patients with clinically localized PCa.10-14 These predictive tools can help counsel individual patients in choosing the most appropriate management option. Among these predictive tools, the most widely used is the 1997 Partin tables. This nomogram was constructed in the 1990s and had been validated externally and interracially several times, reflecting its good performance in different scenarios.3,4,6,7 As the only predictive tool of the pathologic stage that has been extensively validated, the 1997 Partin tables have been recommended for their clinical application in the seventh and eighth editions of Campbell’s Urology.15,16 This nomogram has also gained acceptance as a useful guide in clinical practice in the United States and other countries. However, in the PSA era, a tremendous change has occurred in the stage of PCa at presentation, especially in the United States and European countries, with more PCa detected in men with a lower serum PSA level and biopsy Gleason score. In accordance with this change, the 1997 Partin tables were updated in 20012 to provide a more accurate prediction of the pathologic stage under this “migration” phenomenon. The 2001 Partin tables have not been as thoroughly evaluated for their predictive accuracy5-7; however, they have been substituted for the older Partin tables in the ninth edition of CampbellWalsh Urology.17 Unlike in Western countries, PCa is not a common malignancy in China. However, several reports have shown a trend in China toward an increasing incidence of PCa. Sim and Cheng18 reviewed the registered cases of PCa in some Asian countries, including China, from 1978 to 1997 and found that the incidence had increased by 5%-118%. In addition, the wide spread of the PSA test, which promotes the early diagnosis of PCa, has resulted in RP being performed more widely in China than before.18-20 Therefore, Chinese urologists urgently need to adopt a predictive pathologic outcome tool in their clinical practice, such as the Partin tables. Thus, 2 key questions caused our concern. First, the 1997 and 2001 Partin tables were derived in the United States, and it remains uncertain whether these nomograms would be accurate for Chinese patients, because the biologic features and disease detection could vary among these patients. Second, the 2001 Partin tables were constructed from a more contemporary cohort of U.S. patients to reflect the tremendous change in the stage of PCa at presentation in the United States. The intent of this new version is to provide more accurate information about the pathologic stage for patients durUROLOGY 72 (5), 2008
ing this stage migration. However, whether this newer version is more applicable than the older version when applied in Chinese clinical practice remains unknown. The present study was conducted to explore these questions. Although the possible limitations of this study were the relatively small sample size and the single-center study, we believe that our results are valid. Considering the low incidence of PCa in China, and the consecutive enrollment of the present cohort, 271 patients could serve as a representative sample for validation. Another limitation of our study was that it was a single-center study. If the present study had been conducted as a multicenter study, although the number of the cases could have been greater, the validation results might have been less convincing. The use of a single center meant that all surgery was performed exclusively by 2 experienced urologists (X.G. and Y.H.S.), nearly all patients (only 9 patients [3%] were excluded from the 302 cases because of missing parameters) had the clinical data required for the validation, and central pathologic review could be easily achieved. Our results have demonstrated that the 1997 Partin tables could be applied to Chinese patients, although this predictive tool was derived in the United States. The AUC for OCD (0.723), SVI (0.749), and LNI (0.781) showed the accurate predictive ability of this nomograms when applied to our cohort. Similar results could be seen from several previous validation studies performed in and outside the United States, confirming that the 1997 Partin tables are able to adjust for patient selection differences across different clinical scenarios.3,4,6,7 The Loess plots in Figure 2 showed that the nomogram tended to underestimate OCD and generally overestimate ECP, SVI, and LNI. This should be noted during clinical application because patients with predicted late-stage disease might select alternative, potentially less curable, treatment modalities. However, we have concluded that the 2001 Partin tables are not accurate enough to apply to urologic practice in China. The AUC of the 2001 Partin tables was 0.604, 0.539, 0.654, and 0.631 for OCD, ECP, SVI, and LNI, respectively, consistently inferior to those of the 1997 Partin tables. Because an AUC of ⬍0.700 could be considered a poor result in ROC analysis, we believe that the 2001 Partin tables do not have a reasonable predictive value for the final pathologic features in Chinese patients with PCa. The decreased accuracy of the 2001 Partin tables was likely because of the following. First, similarly lower pretreatment variables was seen in the 2001 Partin cohort compared with both to our cohort and the 1997 Partin cohort (Tables 1 and 2). For example, 62.3% of our patients had Stage T2a PCa or less and a comparable ratio of 66% was seen in the Partin 1997 cohort. However, a significantly greater ratio of 95% was seen in the development cohort of the 2001 Partin tables. Also, the percentage of 1099
patients with a Gleason score of 2-4 and ⱖ7 was 11.1% and 25.8% in our study, respectively. The corresponding ratios had decreased to 5% and 22% in the 1997 Partin cohort and had decreased further to 0.6% and 17.4% in the 2001 Partin cohort. Again, the corresponding ratio in the 1997 Partin cohort was between that of our study and the 2001 Partin cohort. According to Tables 1 and 2, the other ratios in the 1997 Partin cohort could also be seen between that of our patients and that of the 2001 Partin cohort, such as the percentage of patients with a PSA of 4-10 ng/mL or ⱖ10 ng/mL and with SVI and LNI, except for the percentage of patients with OCD, which was lowest in the 1997 Partin cohort (48%) and the greatest in the 2001 Partin cohort (64%). Differing population characteristics between the validation cohort and the development cohort are quite common and also have been well described in other validation studies.3-7 To a certain extent, these differences could be compensated for by the transportability of a well-performed predictive tool, such as the 1997 Partin tables in the present study. However, as the differences show a trend toward more significance, the accuracy of a predictive tool will be compromised, even to the point of no longer qualifying for clinical practice. In relation to our validation study, this is one of the main reasons accounting for the lower accuracy of the 2001 Partin tables. Second, the basic structural changes of the 2001 Partin tables have reduced their predictive accuracy for the present cohort. During the past decades, a tremendous PCa stage migration has occurred in western countries. In accordance with this change, Partin et al.1,2 revised their former nomogram with a more contemporary cohort in 2001. In this revised nomogram, some basic structural changes were made. For example, the PSA groups were stratified into narrower intervals when ⬍10 ng/mL. However, the new tables put those patients with a PSA of ⬎10 ng/mL into the same group. This might be more suitable to a validation setting with a low PSA level. Nonetheless, when applied to the present validation cohort, in which nearly 65% of the patients had a PSA level of ⬎10 ng/mL, the theoretical advantage related to narrower divisions in the pretreatment PSA level has not yielded an appreciable gain in predictive accuracy. Finally, the 2001 Partin tables provided low coverage for the present validation cohort. In our cohort, 17 patients (6.3%) in our validation cohort had clinical Stage T1a-T1b or T3a. These patients were excluded from the validation of the 2001 tables because the new tables no longer provided predictive values for individuals with these stages. Furthermore, the probabilities were not available for another 29 and 31 patients (10.7% and 11.4%) for the predicted SVI and LNI using the newer tables according to their clinical variables. These data were considered as missing during the final ROC analysis. Even though the missing data do not necessarily have 1100
negative effect on the accuracy of the predictive tool, a more comprehension one (ie, the 1997 Partin tables), would be more suitable for clinical use from intuitive and practical considerations.
CONCLUSIONS Our results have indicated that the 1997 Partin tables adequately predict the pathologic outcomes in Chinese men who undergo RP for PCa. They are particularly useful for the prediction of OCD, SVI, and LNI. However, the 2001 version of the Partin tables, which were updated in 2001 to “better represent the contemporary presentation of PCa with stage shift phenomena,” might not be applicable to general urologic practice in China because their predictive values for each pathologic feature were consistently lower than those of the 1997 version. References 1. Partin AW, Kattan MW, Subong EN, 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;277:1445-1451. 2. Partin AW, Mangold LA, Lamm DM, et al. Contemporary update of prostate cancer staging nomograms (Partin tables) for the new millennium. Urology. 2001;58:843-848. 3. Blute ML, Bergstralh EJ, Partin AW, et al. Validation of Partin tables for predicting pathological stage of clinically localized prostate cancer. J Urol. 2000;164:1591-1595. 4. Graefen M, Augustin H, Karakiewicz PI, et al. 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. Eur Urol. 2003;43:6-11. 5. Eskicorapci SY, Karabulut E, Turkeri L, et al. Validation of 2001 Partin tables in Turkey: A multicenter study. Eur Urol. 2005;47: 185-189. 6. Ayyathurai R, Ananthakrishnan K, Rajasundaram R, et al. Predictive ability of Partin tables 2001 in a Welsh population. Urol Int. 2006;76:217-222. 7. Augustin H, Eggert T, Wenske S, et al. Comparison of accuracy between the Partin tables of 1997 and 2001 to predict final pathological stage in clinically localized prostate cancer. J Urol. 2004; 171:177-181. 8. Hull GW, Rabbani F, Abbas F, et al. Cancer control with radical prostatectomy alone in 1,000 consecutive patients. J Urol. 2002; 167:528-534. 9. Eastham JA, Scardino PT. Radical prostatectomy. In: Walsh PC, Retik AB, Vaughan ED, et al. eds. Campbell’s Urology, Vol. 4. Philadelphia: WB Saunders; 2002:3080-3106. 10. Peller PA, Young DC, Marmaduke DP, et al. Sextant prostate biopsies: A histopathologic correlation with radical prostatectomy specimens. Cancer. 1995;75:530-538. 11. Bostwick DG, Qian J, Bergstralh E, et al. Prediction of capsular perforation and seminal vesicle invasion in prostate cancer. J Urol. 1996;155:1361-1367. 12. Gilliland FD, Hoffman RM, Hamilton A, et al. Predicting extracapsular extension of prostate cancer in men treated with radical prostatectomy: Results from the population based prostate cancer outcomes study. J Urol. 1999;162:1341-1345. 13. Rabbani F, Bastar A, Fair WR. Site specific predictors of positive margins at radical prostatectomy: An argument for risk based modification of technique. J Urol. 1998;160:1727-1733. 14. Badalament RA, Miller MC, Peller PA, et al. An algorithm for predicting organ confined prostate cancer using the results obtained
UROLOGY 72 (5), 2008
from sextant core biopsies with prostate specific antigen level. J Urol. 1996;156:1375-1380. 15. Carter HB, Partin AW. Diagnosis and staging of prostate cancer. In: Walsh PC, Retik AB, Vaughan ED, et al., eds. Campbell’s Urology. Philadelphia: WB Saunders; 1998:2519-2537. 16. Carter HB, Partin AW. Diagnosis and staging of prostate cancer. In: Walsh PC, Retik AB, Vaughan ED, et al., eds. Campbell’s Urology. Philadelphia: WB Saunders; 2002:3055-3079. 17. Carter HB, Allaf ME, Partin AW. Diagnosis and staging of prostate cancer. In: Wein AJ, Kavoussi LR, Novick AC, et al.,
UROLOGY 72 (5), 2008
eds. Campbell-Walsh Urology. Philadelphia: Saunders Elsevier; 2007:2912-2931. 18. Sim HG, Cheng CW. Changing demography of prostate cancer in Asia. Eur J Cancer. 2005;41:834-845. 19. Peyromaure M, Debre B, Mao K, et al. Management of prostate cancer in China: A multicenter report of 6 institutions. J Urol. 2005;174:1794-1797. 20. Tsuji X, Li I, Kuwahara M, et al. Mass screening of prostate cancer in Changchun city of China. Int Urol Nephrol. 2004;36: 541-548.
1101