The Prostate 35:50–55 (1998) Diagnoses Rendered on Prostate Needle Biopsy in Community Hospitals Michael H. Weinstein,1* David L. Greenspan,2 and Jonathan I. Epstein2 1 Department of Pathology and Laboratory Medicine, University of Kentucky College of Medicine, Lexington, Kentucky 2 Department of Pathology, Johns Hopkins Medical Institutions, Baltimore, Maryland BACKGROUND. Reported incidences of various diagnoses made on needle biopsy of the prostate vary significantly in the literature, most of which has originated from large, academic medical centers. METHODS. We recorded all the prostate needle biopsy results from three community hospitals for 1990–1993 to determine the rates of, and trends in, various diagnoses in these practices. RESULTS. Hospital H1 (1,192 cases) halved the rate of atypical, nondefinitive diagnoses from 11.8% in 1990 to 5.7% in 1993 (P < 0.001). The rate at H2 (2,792 cases) remained essentially unchanged at 5.95 ± 0.55%, and H3 (1,306 cases) went from 2.3% to 6.0% (0.1 < P < 0.2). In the setting of an atypical, nondefinitive diagnosis, H1 and H2 recommended repeat biopsy less than 7% of the time. H3 made this recommendation in an average of 22.1% of atypical cases. Annual rates of high-grade prostatic intraepithelial neoplasia (PIN) showed no trend over time, and averaged 2.0% (1.2–3.25%) at H1 and 1.2% (0.3–2.0%) at H2. The diagnosis was never made at H3. The fraction of cancers diagnosed as low-grade (Gleason sum ø4) showed a statistically significant decreasing trend over time at all three hospitals (P < 0.05). These data are compared with those from the Johns Hopkins Hospital (JHH), a large academic center in geographic proximity to hospitals H1–H3. CONCLUSIONS. At these three community hospitals, we discerned (1) convergence to a rate of approximately 6.0% of atypical, nondefinitive diagnoses; and (2) a progressively more appropriate fraction of carcinomas diagnosed as low-grade on needle biopsy. The rates of diagnosis of high-grade PIN and recommendation of repeat biopsy varied. These rates of PIN, atypical, nondefinitive diagnoses, and low-grade cancer represent an assessment of diagnostic habits. Prostate 35:50–55, 1998. © 1998 Wiley-Liss, Inc. KEY WORDS: prostate cancer; prostatic intraepithelial neoplasia; atypical; Gleason grade INTRODUCTION The last 5–10 years have seen an increase in the number of needle biopsies of the prostate performed in the United States each year. This has been due to widespread screening for prostate cancer by measurement of serum prostate-specific antigen (PSA) concentration, as well as a burgeoning awareness of this disease. While the amount and grade of adenocarcinoma of the prostate seen on needle biopsy are of some use in predicting the extent of disease found in subsequent prostatectomy specimens, even limited cancer on needle biopsy may be associated with moderate to large amounts of unsampled cancer in the whole © 1998 Wiley-Liss, Inc. gland. Therefore, the goal of pathologic examination of biopsy material continues to be the identification of every patient with adenocarcinoma of the prostate without erroneously labeling anybody who is free of cancer of the prostate. Surgical pathologists have also begun to identify markers seen on needle biopsy, such as high-grade *Correspondence to: Michael H. Weinstein, Department of Pathology and Laboratory Medicine, University of Kentucky College of Medicine, MS 117, Chandler Medical Center, Lexington, KY 405360084. E-mail: email@example.com Received 19 May 1997; Accepted 7 August 1997 Community Hospital Diagnoses on Prostate Needle Biopsy prostatic intraepithelial neoplasis (PIN), that indicate a great likelihood of carcinoma within the gland. It has become common for a nondefinitively malignant diagnosis to be rendered that suggests an increased likelihood of unsampled or inadequately sampled carcinoma. This usually leads to repeat biopsy. Physicians in community hospitals administer a great fraction of the primary health care, and there is a paucity of published information regarding their habits of diagnosis for needle biopsy of the prostate. We concentrated in this study on three main aspects: rates of atypical, nondefinitive diagnoses; rates of diagnosis of highgrade PIN; and habits of grading of adenocarcinomas of the prostate. MATERIALS AND METHODS The record of every set of needle biopsies of the prostate examined from 1/1/90 to 12/31/93 by the departments of pathology at three Baltimore–Washington area community suburban hospitals was obtained by computer search and entered into a database (Paradox, Borland International, Scotts Valley, CA). The demographics (e.g., age, racial/ethnic background, financial resources) of the populations served by these three suburban hospitals are subjectively similar. There were a total of 1,192 cases, each comprised of a set of needle biopsies, from the first community hospital (H1), 2,792 cases from the second (H2), and 1,306 from the third (H3). The database was used to tally statistics on several diagnostic categories. We tallied diagnoses of atypia suspicious for carcinoma, atypia of unspecified character, and atypia probably benign. It was not our purpose to assess the diagnostic accuracy of the interpretation of these needle biopsies, but rather to delineate the diagnostic habits and extent of variation at these community hospitals. Hence, we did not perform a histologic review of any of their material. Even had we wanted to review slides from these hospitals, none would have permitted this exercise because of the potential medicolegal risk. Cases were counted as atypical, indefinite for carcinoma when the report was neither benign, nor definitively malignant, nor PIN. That is, there was a concern expressed that some of the glands in the biopsy might be malignant, but no specific diagnosis was given. In cases in which an outside consultation had been requested, we recorded the original diagnosis. Thus, cases initially diagnosed as atypical, indefinite for carcinoma were recorded that way, even if subsequent consultation lead to a definitive diagnosis. Recommendations for repeat biopsy were counted, as were diagnoses of adenocarcinoma of the prostate of each Gleason score. Also tallied were diagnoses of 51 PIN, which fell into six categories: grades 1–3, highand low-grade, and unspecified grade. We grouped diagnoses of PIN 1 with those called low-grade to give our category of low-grade PIN. We also grouped diagnoses of PIN 2, PIN 3, and high-grade PIN to give our category of high-grade PIN. Diagnoses of atypia were counted only in the absence of any carcinoma in the same set of biopsies, and PIN was counted only if there was no concurrent diagnosis of atypia or carcinoma. The Gleason grade assigned to each case with carcinoma was the highest of the grades given to any of the needle biopsies in the set. Statistical analyses were performed using the chisquare statistic for trend analysis. RESULTS We obtained records of all the needle biopsies of the prostate performed from 1/1/90 to 12/31/93 at the three community hospitals studied. There were 1,192 sets of needle biopsies at the first community hospital (H1), 2,792 at the second (H2), and 1,306 at the third (H3). The average number of biopsies per case (set) was difficult to assess because of incomplete information in some of the reports, but was approximately 2 at each of the three hospitals. Tabulation of the diagnoses is given in Tables I–III. Table I lists by year and category the diagnoses of atypia in the absence of a concurrent diagnosis of carcinoma. Note that the fraction of biopsy sets that were diagnosed as atypical, indefinite for carcinoma varied in 1990 from 2.3% at H3 to 11.8% at H1. The range shrank in each of the succeeding three years, so that by 1993 only 5.7–6.3% of all biopsies carried the diagnosis of atypical, indefinite for carcinoma. Table II lists by year and grade the diagnoses of PIN in the absence of concurrent carcinoma or atypia of other type. At H1 and H2, the relative frequencies of the diagnoses of high-grade and low-grade PIN varied over the years, without a discernible pattern. They made the diagnosis of high-grade PIN in 1–2% of needle biopsies in most years. Hospital H3 did make the diagnosis of low-grade PIN in 0–1.6% of needle biopsies, but never diagnosed high-grade PIN. The least common diagnosis at all three hospitals in every year was PIN of unspecified grade. The greatest change in the fraction of carcinomas labeled as low-grade occurred at H1, as seen in Table III. In 1990, 28.0% of all the sets of needle biopsies with carcinoma at H1 were graded as Gleason score 2–4. Hospitals H2 and H3 made the diagnosis of low-grade carcinoma in 9.0% and 5.7% of the carcinomatous sets of needle biopsies seen in that year. By 1993 the frac- 52 Weinstein et al. TABLE I. Percentage of Total Cases Diagnosed as Atypical, Indefinite for Carcinoma and Rates of Recommendation of Repeat Biopsy in These Cases by Year and Hospital* Hospital H1 Atypia, favor benign Atypia, unspecified character Atypia, favor malignant Total atypical diagnoses by year Recommendations for repeat Bx Total cases Hospital H2 Hospital H3 90 91 92 93 90 91 92 93 90 91 92 93 2.0 5.3 4.5 11.8 0 247 2.7 4.5 3.0 10.2 2.9 334 1.0 1.5 2.7 5.2 4.8 404 1.4 1.9 2.4 5.7 0 207 1.2 2.1 2.4 5.7 0 328 0.9 3.2 1.3 5.4 0 685 1.3 2.7 2.5 6.5 3.6 1268 1.4 3.1 1.8 6.3 6.2 511 0.5 0.9 0.9 2.3 40.0 222 1.5 3.0 1.0 5.5 27.3 404 0.8 1.9 1.7 4.4 6.2 361 0.3 4.4 1.3 6.0 21.1 319 *There was a statistically significant trend in the fraction of total cases diagnosed as atypical at H1 (P < 0.001). TABLE II. Percentage of Total Cases Diagnosed as PIN by Year and Hospital Hospital H1 LG PIN HG PIN Unspecified grade Total no. of cases Hospital H2 Hospital H3 90 91 92 93 90 91 92 93 90 91 92 93 2.0 1.2 0 247 2.8 2.1 0.3 334 1.5 3.2 0.2 404 0.5 1.5 0 207 0.6 0.3 0.3 328 0.7 2.0 0.6 685 0.9 1.5 0.4 1268 3.6 1.0 0.4 511 0 0 0 222 0.2 0 0 404 0 0 0 361 1.6 0 0 319 HG PIN, high-grade PIN; LG PIN, low-grade PIN. TABLE III. Percentage of Total Cases of Adenocarcinoma Assigned Each Gleason Score and Percentage of Total Cases Diagnosed as Cancer by Year and Hospital* Hospital H1 Gleason scores 2–4 5–6 7 8–10 Carcinoma of any grade Total no. of cases Hospital H2 Hospital H3 90 91 92 93 90 91 92 93 90 91 92 93 28.0 39.0 14.6 18.4 33.2 247 27.8 44.3 13.4 14.5 29.0 334 6.8 50.0 18.2 25.0 32.7 404 1.3 40.8 34.2 23.7 36.7 207 9.0 51.7 23.6 15.7 27.1 328 7.5 58.8 26.9 6.8 23.4 685 1.5 64.5 27.5 6.5 25.6 1268 2.3 65.9 26.4 5.4 25.2 511 5.7 37.7 41.5 15.1 23.9 222 2.9 53.9 32.4 10.8 25.2 404 3.6 50.0 28.2 18.2 30.5 361 1.9 47.6 32.0 18.5 32.3 319 *The trend in the fraction of carcinomas diagnosed as Gleason score 2–4 over 1990–1993 was statistically significant at H1 (P < 0.001) and H2 (P < 0.001), but not at H3 (P > 0.3). tion of malignant needle biopsies with Gleason score 2–4 only varied from 1.3% at H1 to 2.3% at H2. DISCUSSION The ability to interpret thin needle biopsies of the prostate has increased considerably during the past several years. The advent of immunohistochemical staining for high-molecular-weight cytokeratin has allowed pathologists to reliably identify very small collections of atypical glands as carcinoma, and expertise has increased to the point where this tool is often not necessary even in some very small foci. However, there do remain a significant number of biopsies with atypical glands that cannot be characterized definitively as either benign or malignant. There are multiple reasons for this. Not only are the biopsies themselves small, but the foci in question are very small and are not likely to be present on deeper levels obtained for the purpose of immunohistochemical staining. Few laboratories are willing to routinely obtain unstained intervening sections on every needle biopsy in case special stains are required. Second, many laboratories have limited capacity to perform immunohistochemical stains. Even under the best possible cir- Community Hospital Diagnoses on Prostate Needle Biopsy cumstances some lesions cannot be characterized with certainty, since failure of staining for high-molecularweight cytokeratin to demonstrate basal cells in a small focus of atypical glands is not itself diagnostic of carcinoma. The goal in the present diagnostic milieu of examination of needle biopsies of the prostate is to identify every patient with adenocarcinoma of the prostate. We are willing to have some patients who do not have carcinoma be rebiopsied because of a focus of atypical glands that are not identifiably benign with certainty. However, unnecessary biopsy is clearly undesirable, so it is important to keep the number of nondefinitive diagnoses low. These conflicting demands require that each pathologist set a threshold for the diagnosis of ‘‘atypical, indefinite for carcinoma.’’ In 1990, the diagnosis of ‘‘atypical, indefinite for carcinoma’’ was made in 11.8% of the sets of needle biopsies seen at hospital H1. There was a statistically significant trend, with the rate going from 11.8% in 1990 to 5.7% in 1993 (P < 0.001) (Table I). The rate was essentially constant at H2 from 1990 to 1993 (5.95 ± 0.55%), and there was a fluctuating but overall slightly increasing trend at H3 from 2.3% in 1990 to 6.0% in 1993 (0.10 < P < 0.20). Personal communication suggests these changes may be due, at least in part, to self-education of the pathologists rendering these diagnoses. Means of education may have included attendance at conferences, interdepartmental consultation, verification by immunohistochemical stains for high-molecular-weight cytokeratin, and the accumulation of experience in interpreting needle biopsies of the prostate. Although there may be some variation between the patient populations seen at these three hospitals that are located in the same geographic area, they appear grossly similar. The large number of needle biopsies included in this study (5,290) implies that a large number of patients were screened. Consequently, these data can be expected to be fairly representative of a large segment of the pertinent population in terms of variables such as serum PSA concentration and patient age. By way of comparison, recent data from the Johns Hopkins Hospital (JHH), a large academic center in geographic proximity to the three community hospitals in this study, indicated a rate of ‘‘atypical, indefinite for carcinoma’’ diagnoses of 4.6% in 439 consecutive sextant needle biopsies seen there from 1/1/92 to 7/15/94 . This slightly lower rate may be due to one or more of several conflicting factors. One reason for the lower rate may be that the number of biopsies in each set at JHH was six, instead of the two that were usual in the sets of biopsies performed during the time period of this study at the three community hospitals. More biopsies increase the likelihood 53 that a carcinoma in the gland will be sampled in a fashion that permits a definitive diagnosis. On the other hand, more biopsies make it more likely that a small, atypical focus that is difficult to categorize will show up in one of the needle biopsies of a set. At JHH, H+E-stained sections are routinely obtained on three levels for every needle biopsy of the prostate performed there. Unstained slides of sections from intervening levels are also obtained, making it more likely that a section of a difficult lesion will be available to stain for high-molecular-weight cytokeratin. This was not routine at the three community hospitals. It is of course possible that there is a referral bias, and that the patient populations differ more than is apparent. Finally, one of the authors of this study (J.I.E.) has extensive experience in prostate pathology, including an active consultation practice, and he reviews most of the problem cases at JHH. It is clear from Table I that there was no standardization on recommendations for repeat biopsy when a diagnosis of atypical, indefinite is made. The likelihood of finding carcinoma on repeat biopsy depends heavily on the threshold for making the diagnosis of ‘‘atypical, indefinite for carcinoma.’’ It is the experience of the authors of this work in reviewing consult material that the frequency of such diagnoses varies considerably from one institution to another, implying a variation in threshold. We consider it appropriate to recommend repeat biopsy in all cases that are suspicious but not diagnostic for prostate cancer. We reason that the same clinical indications that warranted the initial set of needle biopsies still pertain. In fact, there is more cause for concern, since the pathologist has recognized something in the initial biopsies that may likely represent carcinoma. The morbidity of needle biopsy is quite low , and generally is not a significant factor in this decision. High-grade PIN is known to be a marker for the presence of carcinoma, and should be reported if found on needle biopsy in the absence of clear-cut malignancy. There are several reports detailing the risk of unsampled adenocarcinoma of the prostate when a needle biopsy of the gland contains highgrade PIN [3–7]. It is generally agreed that this risk is approximately 33–50%. However, it is clear that if the accepted diagnostic criteria for high-grade PIN are not adhered to and more cases are included, then the diagnosis becomes less meaningful. Published data reporting the incidence of high-grade PIN are scarce, and the reported incidences vary. In one study, 25,000 men were screened by measuring serum PSA concentration, and 0.15% of them were found on needle biopsy to have high-grade PIN without carcinoma on any of the biopsies in the initial set . Another study included biopsies from an academic center and a pri- 54 Weinstein et al. vate practice laboratory and reported rates of 16.5% and 9.5%, respectively, of high-grade PIN on needle biopsy in the absence of atypia or identifiable carcinoma . A third study reported a rate of 5% in consult material , and a fourth found an incidence of 11% in men with hypoechoic lesions of the prostate . Recent data from JHH yielded an incidence of 5.5% . Notably, this study from JHH included review of all 439 cases, which disclosed an additional 2.8% of the cases that met criteria for high-grade PIN over and above the reported rate of 2.7%. Table II shows that the rate of diagnosis of highgrade PIN hovered around 1–2% at hospitals H1 and H2 for most of the study period, and the averages over the four years were 1.7% and 1.4%, respectively. By contrast, hospital H3 never made the diagnosis of high-grade PIN, and this probably represents underreporting. High-grade PIN has been shown to be a reproducible diagnosis in two recent studies [11,12]. The first of these studies demonstrated a good degree of interobserver agreement in differentiating between low-grade PIN and benign on the one hand and highgrade PIN on the other . Low-grade PIN was diagnosed in approximately 1% of the cases seen at H1– H3, and there were a few cases of PIN of unspecified grade. The diagnosis of low-grade PIN is not reproducible and has not been shown to be of clinical significance [3,6]. A diagnosis of PIN of unspecified grade is problematic and probably should not be made. Left untreated, many lesions histologically identifiable as adenocarcinoma of the prostate will never lead to clinical disease. We are limited in our ability to discern which lesions are biologically benign but that have a histologically malignant appearance. One feature well known to be important in predicting behavior is the Gleason sum of the tumor. Some patients with tumors of Gleason sum <5 on needle biopsy and low PSA densities are now, with the advice of their physicians, electing to be cared for with ‘‘watchful waiting’’ and no definitive therapy. Therefore, correct identification of those patients with low-grade tumors is important. As with high-grade PIN, reported rates vary. One group found that none of 67 consecutive patients who underwent radical prostatectomy with needle biopsy available for review had tumors of Gleason sum ø4 either on needle biopsy or in the corresponding radical prostatectomy specimens . Another found that 14% of their patients had lowgrade tumors on needle biopsy with a corresponding rate of 3% in the prostatectomy specimens , and a third group  found that 12% of their radical prostatectomy specimens had carcinomas of Gleason sum ø4. It is possible that these differences are due to differences in patient populations, but we believe it is likely that they are due, at least in part, to differences in judgment on the parts of the pathologists. The experience of the authors (based on the consult material of J.I.E.) is that some pathologists factor the quantity of tumor present in an needle biopsy into the grading process. However, a very small focus of small, crowded, well-formed glands clearly infiltrating between larger, benign glands should be categorized as Gleason pattern 3, regardless of the quantity. We have seen that it is not rare for a lower grade to be assigned in such a situation. Probably more common is a consistent difference in the grade given to each pattern. In order to apply the findings contained in the published literature, interobserver reliability in Gleason grading is necessary. Though data are scarce, it is probable that the threshold for the diagnosis of low-grade adenocarcinoma of the prostate (Gleason sum ø4) on needle biopsy should be set so that, in most patient populations, no more than approximately 5% of patients with cancer are classified as having low-grade lesions [13– 16]. This number accurately reflects the fraction of low-grade carcinomas seen in radical prostatectomy specimens, in which more tumor is visible for accurate grading. The fraction of low-grade cancers seen in TURP specimens is higher, and, indeed, it is known that transitional zone carcinomas tend to be of lower grade. Table III shows a wide spectrum among the hospitals in the fraction of carcinomas seen on needle biopsy that were labeled as low-grade (Gleason sum ø4) in 1990. The fraction ranged from 28% at H1 to 5.7% at H2. This difference is unlikely to be completely attributable to differences in patient population. The proportion of low-grade tumors steadily decreased at H1 to 1.3% in 1993. The trend from 28% in 1990 at H1 to 1.3% in 1993 was statistically significant (P < 0.001) Similar, though less dramatic decreases were seen at the other two hospitals, with the trends being statistically significant at H2 (P < 0.001) but not at H3 (0.3 < P < 0.5). Thus, in 1993, the range among the three hospitals was 1.3–2.3%. The series of consecutive needle biopsies at JHH contained 3.9% that were categorized as low-grade adenocarcinoma . In the cumulative experience with radical prostatectomy specimens from 721 patients at JHH, the fraction that had tumors of Gleason grade 2–4 was 6.8% . This is a greater fraction than was reported on needle biopsy at any of the three hospitals, H1–H3, in 1993, and the difference is difficult to interpret. There may be some limitations in reproducibility for the Gleason grading system. CONCLUSIONS We found a temporal convergence among three community hospitals in the rates of ‘‘atypical, indefi- Community Hospital Diagnoses on Prostate Needle Biopsy nite for carcinoma’’ diagnoses to approximately 6% of all sets of biopsies. Similarly, the fraction of carcinomas labeled as low-grade also converged over the 4 years of the study period. The rates in 1993 were all approximately 1–2%. High-grade PIN, the diagnosis of which may still be in a state of evolution, was found in approximately 1–2% of the sets of biopsies seen at two of the community hospitals and was never diagnosed at the third during the four years that were studied. This is a lower number than would be expected based on data from several academic institutions. These data represent an early step in assessing practice patterns, and in most respects approximate that obtained from a series of sextant needle biopsies seen at JHH, a large academic center in geographic proximity to the three hospitals studied here. Similar studies involving larger numbers of institutions from varied locations may be valuable in defining the extent of variation in the diagnostic habits of pathologists . The authors recognize that the identification and limitation of diagnostic variation (reproducibility) is beneficial. We also recognize that diagnostic accuracy is a related but different attribute of testing that requires validation, ideally by reference to patient outcomes . Additional efforts at standardization of diagnostic criteria as well as education appear to be warranted. 55 5. Aboseif S, Shinohara K, Weidner N, Narayan P, Carroll PR: The significance of prostatic intra-epithelial neoplasia. Br J Urol 1995;76:355–359. 6. Keetch DW, Humphrey P, Stahl D, Smith DS, Catalona WJ: Morphometric analysis and clinical follow up of isolated prostatic intraepithelial neoplasia in needle biopsies of the prostate. J Urol 1995;154:347–351. 7. Davidson D, Bostwick DG, Qian J, Wollan PC, Oesterling JE, Rudders RA, Siroky M, Stilmant M: Prostatic intraepithelial neoplasia is a risk factor for adenocarcinoma: Predictive accuracy in needle biopsies. J Urol 1995;154:1295–1299. 8. Bostwick DG, Qian J, Frankel K: The incidence of high grade prostatic intraepithelial neoplasia in needle biopsies. J Urol 1995;154:1791–1794. 9. Epstein JI: Diagnostic criteria of limited adenocarcinoma of the prostate on needle biopsy. Hum Pathol 1995;26:223–229. 10. Lee F, Torp-Pedersen ST, Carroll JT, Siders DB, Christensen-Day C, Mitchell AE: Use of transrectal ultrasound and prostatespecific antigen in diagnosis of prostatic intraepithelial neoplasia. Urology 1989;34(suppl):4–8. 11. Epstein JI, Grignon DJ, Humphrey PA, McNeal JE, Sesterhenn IA, Troncoso P, Wheeler TM: Interobserver reproducibility in the diagnosis of prostatic intraepithelial neoplasia (PIN). Am J Surg Pathol 1995;19:873–886. 12. Allam CK, Bostwick DG, Hayes JA, Upton MP, Wade GG, Domanowski GF, Klein MA, Boling EA, Stilmant MM: Interobserver variability in the diagnosis of high-grade prostatic intraepithelial neoplasia and adenocarcinoma. Mod Pathol 1996;9: 742–751. 13. Spires SE, Cibull ML, Wood DP, Miller S, Spires SM, Banks ER: Gleason histologic grading in prostatic carcinoma. Arch Pathol Lab Med 1994;118:705–708. REFERENCES 14. Bostwick DG: Gleason grading of prostatic needle biopsies. Am J Surg Pathol 1994;18:796–803. 1. Wills ML, Hamper UM, Epstein JI: Incidence of high-grade prostatic intraepithelial neoplasia in sextant needle biopsy specimens. Urology 1997;49:367–373. 2. Renfer LG, Kiesling VJ, Kelley J, Vaccaro JA, Belville WD: Digitally-directed transrectal biopsy using Biopty gun versus transrectal needle aspiration: Comparison of diagnostic yield and comfort. Urology 1991;138:108–112. 3. Brawer MK, Bigler SA, Sohlberg OE, Nagle RB, Lange PH: Significance of prostatic intraepithelial neoplasia on prostate needle biopsy. Urology 1991;38:103–107. 4. Weinstein MH, Epstein JI: Significance of high-grade prostatic intraepithelial neoplasia on needle biopsy. Hum Pathol 1993;24: 624–629. 15. Garnett JE, Oyasu R, Grayhack JT: The accuracy of diagnostic biopsy specimens in predicting tumor grades by Gleason’s classification of radical prostatectomy specimens. J Urol 1984;131: 690–693. 16. Epstein JI, Partin AW, Sauvageot J, Walsh PC: Prediction of progression following radical prostatectomy: A multivariate analysis of 721 men with long-term follow-up. Am J Surg Pathol 1996;20:286–292. 17. Wennberg J, Gittelsohn A: Variations in medical care among small areas. Sci Am 1982;246:120–134. 18. Greenfield S: The state of outcome research: Are we on target? N Engl J Med 1989;320:1142–1143.