it is important to recognize that these patients operated on from1983 to 1992 reflect a different group of patients with prostate cancerthan is currently seen in practice. These tumors were huge, with amean tumor volume of 4.7 cc., and 18% of patients had seminalvesicle invasion. Only 25% of these patients had nonpalpable tumors,with men having a mean serum PSA of 10.6 ng./ml. The authorsfound that Gleason pattern 4/5 was the best predictor of prognosisafter radical prostatectomy. When this pattern was less than 10%,patients had an excellent prognosis in contrast to the poor prognosiswith greater than 70% of the pattern. However, for the majority ofthe remaining patients with Gleason patterns 4/5, the prognosisamong those with various percentages of the pattern was not suffi-ciently different to be clinically useful for the individual patient.Furthermore, with less than 5% Gleason pattern 4/5 on the radicalprostatectomy, the prognosis was equivalent to those with a Gleasonscore of 3 1 3 5 6. This finding suggests that their threshold fordiagnosing Gleason pattern 4/5 is too low. It has also not beendemonstrated that classifying tumors based on the percentage of thispattern is more predictive than stratifying patients into Gleasonscore 3 1 4 versus 4 1 3. Assessing the percent of Gleason pattern 4is often difficult and not likely to be performed routinely. Often,patterns 4 and 3 are intimately admixed, as the relative percentagesare not readily calculable. Additional difficulty with asking pathol-ogists to derive a specific percentage of pattern 4/5 stems fromstudies demonstrating interobserver variability in grading tumorswith Gleason scores 5 to 7. Therefore, although accurate meas-urements of the percent of Gleason pattern 4 may not be practical,distinguishing score 3 1 4 versus 4 1 3 is simpler, more reproducibleand more likely to be performed at routine pathological examina-tions. We have demonstrated that Gleason score 7 tumors are het-erogeneous, with significant differences in prognosis between thosewith score 4 1 3, as compared to those with score 3 1 4.5

I agree with Noguchi et al that “systematic biopsies are not usefulfor representation of cancer in the prostate.” We must continue todevelop better algorithms for the prediction of “potentially insignif-icant” tumor. Hopefully, biopsy findings obtained from expandedsampling techniques in conjunction with either existing or newlyderived serum markers can improve our prognostication.

Jonathan I. EpsteinDepartments of Pathology, and Urology and OncologyThe Johns Hopkins Medical InstitutionsBaltimore, Maryland

1. Carter, H. B., Sauvageot, J., Walsh, P. C. et al: Prospectiveevaluation of men with stage T1C adenocarcinoma of the pros-tate. J Urol, 157: 2206, 1997

2. DiGiuseppe, J. A., Sauvageot, J. A., and Epstein, J. I.: Increasingincidence of minimal residual cancer radical prostatectomyspecimens. Am J Surg Pathol, 21: 174, 1997

3. Chan, T. Y., Chan, D. Y., Lecksell, K. et al: Are more insignifi-cant prostate tumors detected with increased biopsy sam-pling? Mod Pathol, suppl., 14: 103A, 2001

4. Rubin, M. A., Mucci, N. R., Manley, S. et al: Predictors of Gleasonpattern 4/5 prostate cancer on prostatectomy specimens: canhigh grade tumor be predicted preoperatively? J Urol, 165:114, 2001

5. Chan, T. Y., Partin, A. W., Walsh, P. C. et al: Prognostic signif-icance of Gleason score 3 1 4 versus Gleason score 4 1 3 tumorat radical prostatectomy. Urology, 56: 823, 2000

REPLY BY AUTHORS

We agree with Smith that too many men are treated unnecessarilyfor prostate cancer. The most recent SEER mortality data show thatthe annual death rate from prostate cancer for men 65 years old orolder is only 226/100,000 (0.23%),1 which is a small fraction of the 3%death rate most of us quoted for years.

Exactly half of Epstein’s comment is devoted to criticizing ourevaluation of Gleason grade 4/5 cancer and specifically our 1999report (reference 7 in article). In that study we showed that for every10% increase in Gleason grade 4/5 cancer, there was approximatelyan equivalent 10% PSA failure rate in 379 men with peripheral zonecancers followed for a mean and median of 5 years. This single figuremay represent the most quantitative insight into the extraordinaryusefulness of the Gleason grading system ever published. Epsteinsays that our 1999 report included “huge tumors” but he ignores thepenultimate paragraph of this current biopsy article in which wesummarized all changes in cancer volume and percent Gleason grade

4/5 cancer between 1988 and 2000 in our radical prostatectomies.Cancer volume decreased from 4.8 to 2.5 cc but mean percent Glea-son grade 4/5 cancer remained constant throughout the 12 years at33% (median values for percent grade 4/5 cancer have also remainedconstant at 20%).

In a recent unpublished analysis of 1,182 consecutive radical pros-tatectomies at Stanford the largest cancer was 0.5 cc or less in 73men, of which 70 were located in the peripheral zone. Gleason grade4/5 cancer was a constituent part of 44% of these 73 cancers, com-prising from 1% to 90% of the index cancer volume. It is clear thatGleason grade 4/5 cancer is an early change in a significant propor-tion of prostate cancers. However, Epstein claims that assessingpercentage of Gleason grade 4 is difficult and unlikely to be per-formed. He recommends the traditional Gleason 314 versus 413,ignoring the fact that 314 5 5% to 49% grade 4, and 413 5 50% to95% grade 4 in the traditional system as first described by Gleason.Thus, the basic percent grade 4 decision has already been made andhas even become routine. We suggest only 2 further minor modifica-tions: 1) stop worrying about diagnosing grades 1, 2 and 3 (whichshould make things “simpler”) and 2) establish 2 more cut points athalfway between 5% grade 4/5 and 50% grade 4/5 (30% grade 4/5),and at halfway between 50% and 95% (70% grade 4/5). Epstein hasconfirmed our findings (and those of Gleason) that prognosis is worsefor grade 413 over 314 (that is greater than 50% grade 4 versus lessthan 50% grade 4). Further subdivisions of 5%, 30%, 50%, 70% and95% grade 4 can be estimated with practical accuracy in most cases.

It is also a major error to attribute “inter-observer variability” tomistakes in estimating percent of a histological pattern. In fact, suchvariability is due to disagreement on the criteria for different grades.Different pathologists often use different criteria to guide them inassigning a grade or score level to a biopsy. Often such criteria arelargely intuitive since a good written description of the Gleasonsystem has not been published. Lack of current scientific validationand consensus about Gleason grading criteria is by far the mostserious defect and the most under recognized problem with thissystem. We believe that any institution with a prostate cancer pop-ulation in which less than 70% to 80% of clinically diagnosed cancersat prostatectomy have some element of grade 4 are not recognizingall of the grade 4 because their criteria are not valid.

To be fair, as a pathologist Epstein has been handicapped by TheJohns Hopkins radical prostatectomy series, in which young mendominate their series because of the nerve sparing radical prostatec-tomy. Prostate cancer size and percent Gleason grade 4/5 cancer aredirectly proportional to age when examined in 5-year increments,starting at a median age of only 51 years (reference 3 in article),2which is precisely why at least 30% of Epstein’s radical cases atHopkins have an index (largest) cancer of less than 0.5 cc comparedto our rate of only 10%. We now learn that in 6% of these radicalcases the cancer was so small that they could not find it in thestep-sections. Although we are responsible for the original 0.5 ccvolume cutoff suggested for clinically insignificant cancers, that es-timate was reported more than 8 years ago (reference 3 in article). Infact, Epstein et al reduced our original 0.5 cc estimate to 0.2 cc(reference 2 in article), which conveniently reduced their 30% rate ofunnecessary radical prostatectomies at a cutoff of 0.5 cc. However, asso few men with histological evidence of invasive cancer (at least 40%overall) actually die annually of prostate cancer (0.23%), it is likelythat our original 0.5 cc estimate of cancer volume significance ismuch too small and perhaps substantially so. It is quite possible thatcancer volumes 2 cc or larger with small amounts of Gleason grade4/5 cancer may be clinically insignificant as an eventual cause ofdeath. We suggested in our 1999 report that the PSA cure rate of94.4% was so high in the 19% of 379 men who had no Gleason grade4/5 cancer that perhaps they needed no treatment.

The only purpose of our current article is to show with convincingstatistics how poorly multiple systematic biopsies of the prostaterepresent cancer volume and grade of the index (largest) tumor inthe prostate. Our analysis adds 1 more uncertainty (in addition tothe vagaries of serum PSA) to our ability to advise a man correctlywith newly diagnosed prostate cancer.

1. National Cancer Institute: SEER Cancer Statistics Review 1973–1997. http://seer.cancer.gov/publications/csr1973_1997/prostate.pdf. Accessed Dec. 2000

2. Stamey, T. A., Raimondo, M., Yemoto, C. M. et al: Effect ofageing on morphologic and clinical predictors of prostate can-cer progression. The Prostate J, 2: 157, 2000

BIOPSY FINDINGS AND HISTOPATHOLOGICAL FEATURES OF RADICAL PROSTATECTOMY SPECIMEN110