Currently used criteria for active surveillance in men with low-risk prostate cancer
An analysis of pathologic features
Active surveillance (AS) represents a treatment option for select patients with low-risk, organ-confined prostate cancer (PCa). In this report, the authors addressed the rates of misclassification associated with the use of 5 different clinical criteria for AS. Misclassification was defined as the presence of either nonorgan-confined disease or high-grade PCa.
Between 1992 and 2007, 4885 patients underwent radical prostatectomy (RP) at 1 of 2 European academic centers, and the patients were identified who fulfilled the criteria for AS according to 5 different investigational groups (Hardie et al, Roemeling et al, Choo et al, Klotz, and D'Amico and Coleman). Statistics targeted the rates of misclassification for each of the 5 definitions.
Four thousand three hundred eight patients, 4047 patients, 3993 patients, 2455 patients, and 2345 patients fulfilled the AS criteria of Hardie et al, Roemeling et al, Choo et al, Klotz, and D'Amico and Coleman, respectively. Extracapsular extension was reported in 13.5% to 26% of patients, and seminal vesicle invasion was reported in 2.9% to 8.2% of patients. When PCa with Gleason scores from 8 to 10 at RP was considered high grade, the misclassification rates were 27%, 25%, 25%, 15%, and 14% for the 5 studies, respectively. Conversely, when PCa with Gleason scores from 7 to 10 was considered high grade, the misclassification rates increased to 56%, 55%, 45%, 42%, and 39%, respectively.
The currently available AS criteria are limited by a high rate of misclassification. The use of more selective AS criteria may reduce the rate of misclassification but also may reduce significantly the percentage of patients who may be considered for AS. Cancer 2008. © 2008 American Cancer Society.
Active surveillance (AS) has gained popularity over the past decade.1, 2 The premise used for AS rests on the finding that some patients who are diagnosed with clinically localized prostate cancer (PCa) harbor disease with limited biologic potential.3 Cancers that are amenable to AS usually are identified based on favorable prostate-specific antigen (PSA) level, clinical stage, and biopsy Gleason score.4 Currently, at least 5 different classification schemes can be used to define favorable clinical cancer characteristics.1, 5–8 The Klotz criteria commonly are cited when AS is considered. However, other clinically insignificant PCa criteria (eg, Choo et al) were suggested by other investigators.7 All borrow from the original definition of clinically insignificant PCa coined by Epstein et al.9, 10 For the current report, we investigated the ability of 5 AS criteria to correctly stage AS candidates. Correct staging was defined as the absence of unfavorable pathologic characteristics (Gleason score ≥8, extracapsular extension [ECE], seminal vesicle invasion [SVI], or lymph node invasion [LNI]). Our analysis was triggered by a recent report indicating a somewhat suboptimal performance of the original Epstein criteria.11 Specifically, our objective was to examine the rate of pathologic stage misclassification, defined as presence of nonorgan-confined PCa and/or high-grade Gleason, that may be associated with each of the AS criteria.
MATERIALS AND METHODS
Between 1992 and 2007, 4885 patients underwent an open retropubic radical prostatectomy (RP) at either the Hamburg University Medical Centre in Germany (3337 patients; 68.3%) or at the University Vita-Salute San Raffaele in Milan, Italy (1548 patients; 31.7%). Among these 4885 patients, we identified patients who fulfilled 1 of the 5 commonly used criteria for AS. The 5 AS definitions consisted of those described by Hardie et al, Roemeling et al, Choo et al, Klotz, and D'Amico and Coleman (Table 1).
Table 1. Distribution of Clinical and Pathologic Variables in Patients Qualified for Active Surveillance According to Specific Selection Criteria From the Total of 4885 Men
|PSA, ng/mL||—||≤20||≤15||≤15||If age >70 y, ≤15; if age <70, ≤10||≤10||≤4|
|Biopsy Gleason score||—||≤7||≤7||≤7||If age >70 y, ≤7 (3+4); if age <70, ≤6||≤6||≤6|
|No. of patients included who fulfilled criteria (%)||4885 (100)||4308 (88.2)||4047 (82.8)||3993 (81.7)||2455 (50.3)||2345 (48)||359 (6.9)|
|Clinical features|| || || || || || || |
|Age, y|| || || || || || || |
| Mean [median]||63 ||62.9 ||62.1 ||62.9 ||63.3 ||62.8 ||61.8 |
|PSA, ng/mL|| || || || || || || |
| Mean [median]||7.4 [6.5]||7.3 [6.4]||6.7 [6.2]||6.7 [6.2]||5.8 [5.7]||5.6 [5.6]||2.7 |
|No. with clinical stage (%)|| || || || || || || |
| T1c||2911 (66)||2911 (67.6)||2758 (68.1)||2758 (69.1)||1876 (76.4)||1812 (77.3)||359 (100)|
| T2a||1025 (23.3)||1025 (23.8)||958 (23.7)||958 (24||579 (23.6)||533 (22.7)||0 (0)|
| T2b||311 (7.1)||311 (7.2)||277 (6.8)||277 (6.9)||0 (0)||0 (0)||0 (0)|
| T2c||61 (1.4)||61 (1.4)||54 (1.3)||0 (0)||0 (0)||0 (0)||0 (0)|
| T3||100 (2.3)||0 (0)||0 (0)||0 (0)||0 (0)||0 (0)||0 (0)|
|No. with biopsy Gleason score (%)|| || || || || || || |
| 2-6||3022 (57.6)||2974 (69)||2831 (70)||2805 (70.2)||2354 (95.9)||2345 (100)||359 (100)|
| 7 (3+4)||1035 (23.5)||999 (23.2)||915 (22.6)||893 (22.4)||101 (4.1)||0 (0)||0 (0)|
| 7 (4+3)||351 (8)||335 (7.8)||301(7.4)||295 (7.4)||0 (0)||0 (0)||0 (0)|
| 8-10||284 (5.8)||0 (0)||0 (0)||0 (0)||0 (0)||0 (0)||0 (0)|
|Pathologic features: No. (%)|
|ECE||1561 (32)||1120 (26)||1003 (24.8)||969 (24.3)||368 (15)||316 (13.5)||25 (6.9)|
|SVI||603 (12.3)||353 (8.2)||301 (7.4)||288 (7.2)||78 (3.2)||69 (2.9)||7 (1.9)|
|LNI||258 (5.7)||127 (2.9)||109 (2.7)||107 (2.7)||25 (1)||20 (0.9)||3 (0.8)|
|Pathologic Gleason score|| || || || || || || |
| 2-6||2074 (42.5)||1453 (12.6)||1398 (34.5)||1892 (47.4)||1520 (61.9)||1475 (62.9)||257 (71.6)|
| 7 (3+4)||2022 (41.4)||1780 (41.3)||1660 (41)||1635 (40.1)||814 (33.2)||725 (30.9)||91 (25.3)|
| 7 (4+3)||553 (11.3)||378 (8.8)||341 (8.4)||325 (8.1)||104 (4.2)||87 (3.7)||11 (3)|
| 8-10||236 (4.8)||75 (1.8)||66 (1.6)||66 (1.7)||17 (0.7)||13 (0.7)||0 (0)|
|Overall unfavorable*|| || || || || || || |
| No. (%)||3020 (61.8)||2422 (56.2)||2241 (55.4)||1795 (45)||1040 (42.4)||925 (39.4)||108 (30.1)|
| 95% CI||60.4-63.2||54.7-57.7||53.8-56.9||43.4-46.5||40.4-44.3||37.5-41.5||28.7-31.6|
|Overall unfavorable†|| || || || || || || |
| No. (%)||1607 (32.9)||1149 (26.7)||1029 (25.4)||995 (24.9)||378 (15.4)||325 (13.9)||26 (7.2)|
| 95% CI||31.6-34.2||25.4-28||24.1-26.8||23.6-26.3||14-16.9||12.5-15.3||4.8-10.4|
Clinical and Pathologic Evaluation
Pretreatment PSA was measured before digital rectal examination and transrectal ultrasound with the use of the Abbott AxSym assay and the Elecsys Roche Assay in Hamburg and in Milan, respectively. The clinical stage was assigned prospectively before surgery by the attending urologist according to the 2002 TNM staging system. For patients who were diagnosed before 2002, the 1997 and the 1992 stages were recoded according to the 2002 version. All biopsy and RP specimens were staged and graded by dedicated genitourinary pathologists at each institution. All prostatectomy specimens were processed according to the Stanford protocol.12 Tumor volume was not measured routinely. The misclassification rate associated with each of the AS definitions consisted of patients with high-grade PCa and/or nonorgan-confined disease at RP. Two definitions of high-grade PCa were used. One definition included patients with Gleason score 8 to 10 PCa at RP. The other definition encompassed patients with Gleason 7 to 10 PCa. On the basis of 2 different definitions for high-grade PCa, the rate of misclassification was calculated for patients with Gleason 8 to 10 and/or nonorgan-confined disease and then was recalculated for patients with Gleason 7 to 10 and/or nonorgan-confined disease. Nonorgan-confined disease included ECE, and/or SVI, and/or LNI.13, 14
Statistical methods consisted of descriptive and categorical analyses targeting the rates of unfavorable outcomes according to the 5 different AS criteria. Ninety-five percent confidence intervals were calculated using the exact binomial method.
Of 4885 assessable patients, 4308 patients fulfilled the criteria of Hardie et al, 4047 patients fulfilled the criteria of Roemeling et al, 3993 patients fulfilled the criteria of Choo et al, 2455 patients fulfilled the criteria of Klotz, and 2345 patients fulfilled the criteria of D'Amico and Coleman. The latitude of permitted PSA, clinical stage, and biopsy Gleason score criteria affected the proportions of included patients for each of the definitions. For example, when the Hardie et al criteria were used, 88.2% of our initial cohort could be included. Conversely, when the D'Amico and Coleman criteria or the Klotz criteria were used, the proportion dropped to 48% and 50.3%, respectively.
The characteristics of the patients selected according to each of the AS criteria reflected the criteria strictness. For example, the D'Amico and Coleman criteria and the Klotz criteria were the strictest. Consequently, when these criteria were used, the characteristics of the selected cohort were more favorable than if the Hardie et al criteria or the Roemeling et al criteria were used (Table 1).
Table 1 also summarizes the pathologic findings at RP and the rates of unfavorable pathologic outcomes according to the 5 AS criteria. ECE was were reported in 13.5% to 26% of patients, and SVI was reported 2.9% to 8.2% of patients within each of the 5 cohorts. High-grade PCa (RP Gleason score, 8-10) was reported in 0.7% to 1.8% of patients who would have qualified for AS according to 1 of the criteria.
Like the clinical characteristics, the pathologic characteristics of selected patients depended heavily on the strictness of the selection criteria. The fewest adverse pathologic characteristic were observed in patients who were selected according to the strictest criteria (Klotz, D'Amico and Coleman). Conversely, the least strict and more inclusive AS criteria resulted in higher rates of various pathologically adverse findings. Specifically, the rate of misclassification at RP, defined as the presence of Gleason 8 to 10 or nonorgan-confined disease despite qualifying for AS, ranged from 13.9% to 26.7% from the strictest to the least strict of the definitions. This implies that 26.7% of patients who were classified as fit for AS according to by the Hardie et al criteria AS actually had 1 of the characteristics (ECE, SVI, LNI, or Gleason 8-10 at RP) that clearly represented a contraindication for AS. It is noteworthy that, when a new definition for AS was generated using the strictest elements from the 5 different definitions, only 7.2% of patients who would have qualified for AS using these criteria demonstrated adverse pathologic characteristics at RP. Despite a lower rate of misclassification, this new definition for AS also resulted in the inclusion of only 359 patients (6.9%) versus up to 4308 patients (88.2%) when the least strict definition (Hardie et al) was used.
The objective of our study was to examine the ability of various AS inclusion criteria to correctly predict the pathologic stage of AS candidates. Incorrect predictions (misclassifications) were defined first as the presence of Gleason 8 to 10 PCa and of ECE, SVI or LNI. Pathologic stage was predicted incorrectly in 13.9% to 26.7% of men who were identified as candidates for AS by 1 of the 5 examined schemes.1, 5–8 When Gleason 7 to 10 PCa and ECE, SVI, or LNI were used as inclusion criteria, pathologic stage was predicted incorrectly in 39.4% to 56.2% of patients. The combination of a serum PSA level <4 ng/mL, clinical stage T1c disease, and a biopsy Gleason score ≤6 represented the most stringent elements of each criteria and were used to develop a new definition for AS inclusion. The use of this newly developed definition resulted in incorrect staging of 7.2% of patients who fulfilled the AS characteristics.
The stringency of selection criteria represents a trade-off for their ability to correctly predict the absence of Gleason score 8 to 10 PCa, ECE, SVI, or LNI. Very highly selective and strict inclusion criteria, like those defined by Klotz,1 will allow the inclusion of a relatively small proportion of patients from the pool of men with clinical localized PCa. The application of the Klotz criteria resulted in the inclusion of 50% of the RP cohort. These criteria were associated with a 15.4% misclassification rate (15.4% of patients had Gleason score 8-10 PCa, or ECE, or SVI, or LNI, or more than 1 unfavorable PCa characteristic). Conversely, the Hardie et al criteria6 resulted in the inclusion of 88% of men in the RP cohort. However, the misclassification rate also increased to 26.7%. The use of the most selective and exclusive criteria, defined as a PSA level <4 ng/mL, clinical stage T1c disease, and a biopsy Gleason score ≤6, reduced the rate of misclassification significantly (7.2%). However, at the same time, the cohort of AS candidates would have been reduced to only 6.9% of RP patients. These AS criteria are meant only to illustrate the proportion of men who will be considered for AS and the proportion that will harbor unfavorable pathology at RP. Taken together, our findings indicate that the numbers of potential candidates for AS need to be reduced severely if misclassification error, defined as adverse pathologic staging, is to be avoided.
It could be argued that pathologic staging does not represent the only valid endpoint in AS. Recurrence and progression may represent better endpoints and were described by Klotz and other investigators,1, 5–8 who addressed the cancer control outcomes of AS. Finally, mortality represents the ultimate endpoint, and avoidance of excess mortality in patients who are treated with AS, relative to patients who are treated with definite therapy, should represent one of the ultimate therapeutic goals. Unfortunately, the surgical nature of our series makes the assessment of either biochemical recurrence (BCR) rates, or distant progression, or mortality impossible, because therapy did not consist of AS but of RP. It is noteworthy that, in the Klotz series, of the patients who demonstrated evidence of biochemical progression (PSA doubling time, <2 years), 58% had pathologically nonorgan-confined disease at RP, and 8% had lymph node metastases.1 Therefore, evidence of BCR appears to herald a very advanced stage of PCa that mostly is not curable by most forms of salvage therapy.15–17
The current study has limitations, and lack of consideration of cancer recurrence represents one of them. Patients with either advanced pathologic stage or high-grade PCa may not demonstrate clinically meaningful disease progression over a long time. Advanced pathologic stage or high PCa grade do not always correlate with prognosis. Therefore, some patients who were misclassified according to AS criteria and who demonstrated unfavorable PCa characteristics may not have demonstrated disease recurrence after even a long period of AS. Lack of data regarding detailed cancer information in biopsy cores represents another limitation. According to the Klotz algorithm, the number of involved cores and the percentage of involvement of each core by PCa represent additional AS criteria, especially for patients with life expectancy in excess of 15 years.1 Our study also spanned a long period, and the grading of biopsy and prostatectomy specimens did change over that time.18, 19
Finally, our series consists of men who were selected for RP, which may be viewed as a limitation by some. Others may consider the target population as a strength of the study, because men who undergo RP also represent candidates for AS. Our study demonstrated that, depending on the definition of AS criteria and on the definition of what constitutes a misclassified AS patient, between 14% and 56% of patients could have been selected for AS despite harboring a pathologic variant of PCa that clearly is not amenable to AS. Finally, the observed rates of misclassification apply to European patients. It is possible that these rates may be lower in men from the United States or elsewhere.20, 21
Despite these and other potential limitations, the current study represents a valuable contribution to the urologic literature. Our findings are useful for patients and clinicians who deal with complex treatment decisions, and they demonstrate the limitations of AS criteria. From a clinical perspective, the informed consent for AS should include a discussion about the potential of misclassification, in which AS criteria are fulfilled but the pathologic stage may not be compatible at all with AS. On an individual basis, it is impossible to quantify the probability of misclassification. Nonetheless, patients should be informed about misclassification rates that can be expected if 1 of the 5 classifications is used. Ideally, the most stringent of the definitions should be used (the D'Amico and Coleman criteria). Such practice still will be associated with misclassification rates of 14% (if the definition for high-grade PCa includes only Gleason 8-10 disease at RP) and 39% (if the definition is expanded to Gleason 7-10 at RP). Whether a rate of 14% or 39% is too high will depend on the clinician and the patient's perception, and only individuals who are involved in the decision-making can judge whether such rates are acceptable or are too high. Finally, prospective cohorts will be required to address the appropriateness of each of the acceptable AS criteria over time.