Joel B. Nelson, Frederic N. Schwentker Professor & Chairman, Department of Urology, University of Pittsburgh Medical Center, Shadyside Medical Building, Suite 209, 5200 Centre Avenue, Pittsburgh, PA 15232, USA. e-mail: email@example.com
To determine whether the predictive value of isolated high-grade prostatic intraepithelial neoplasia (HGPIN) for an unsampled prostate cancer on an extended biopsy is lower due to more thorough prostate sampling, and whether the proportion of cores with HGPIN is associated with prostate cancer, as isolated HGPIN on sextant prostate biopsy is associated with a 27–57% risk of prostate cancer on repeat biopsy.
PATIENTS AND METHODS
All extended prostate biopsies taken by one urologist over 6 years were reviewed for patients with isolated HGPIN on initial biopsy. Biopsies were evaluated for histological features and the proportion of cores with HGPIN. The clinical characteristics and pathological findings from subsequent biopsies were determined.
Of 577 men having extended biopsies, 48 had isolated HGPIN, followed by one to four site-directed repeat biopsies. Although only 10 (21%) had cancer on the first repeat biopsy, overall 15 (31%) had cancer. Those with cancer on repeat biopsy had a significantly higher proportion of cores with HGPIN, i.e. 29% vs 15%, cancer vs no cancer, respectively (P = 0.04).
Isolated HGPIN on extended biopsy conferred a 31% risk of unsampled prostate cancer. The proportion of cores with HGPIN on initial biopsy was significantly associated with the risk of cancer. The same was not true for age, race, prostate-specific antigen level, or the findings on digital rectal examination. The significant association between the proportion of cores with HGPIN and the risk of cancer suggests that patients with unifocal HGPIN on extended biopsy be managed expectantly, whereas those with multifocal HGPIN be re-biopsied.
High-grade prostatic intraepithelial neoplasia (HGPIN) is characterized by atypical proliferation of ductal and acinar epithelial cells with preservation of the glandular basal cell layer. The association between HGPIN and prostate cancer is generally accepted, with studies showing it to be a stronger predictor of cancer than any other variable . Studies of autopsy and radical prostatectomy (RP) specimens have shown HGPIN to be present in 63–94% of prostates with adenocarcinoma, and 25–43% of benign prostates [2,3]. When the diagnosis of isolated HGPIN is made after sextant TRUS-guided prostate sampling, the standard practice is to re-biopsy, as various studies reported a 27–57% incidence of an initially unsampled cancer on the second biopsy [1,4,5].
Increasingly many practices have advocated the use of more extensive biopsy sampling, as obtaining more cores has been shown to increase the rate of cancer detection by 35%. Given the more thorough sampling of the prostate, it would be intuitive to assume the diagnosis of isolated HGPIN would have a higher negative predictive value for cancer after an extended-pattern needle biopsy than it would were it obtained in the standard sextant fashion. We report our experience with site-directed repeat biopsies after the diagnosis of isolated HGPIN on extended-needle biopsies. We also determined the predictive significance of clinical factors, including race, PSA level, age, DRE result, extent of HGPIN, histological subtype, and presence of atypical small acinar proliferation (ASAP).
PATIENTS AND METHODS
Institutional Review Board approval was obtained to review the medical records of all patients who had TRUS-guided prostate needle biopsies by one urologist (J.B.N.) and isolated HGPIN with or without ASAP, on initial biopsy (IRB♯ 0507148). All patients with previous biopsies taken by another urologist were excluded.
Between August 1999 and June 2005, 577 patients had 721 prostate needle biopsies; all initial biopsies were taken in an extended peripheral scheme. Seventy-eight patients (14%) had HGPIN with no malignancy on initial biopsy; 57 of these (73%) had no previous biopsies and met the inclusion criteria. All patients were offered immediate re-biopsy; nine either refused the procedure or were lost to follow-up, but the remaining 48 had at least one re-biopsy. All repeat biopsies were site-directed, where three to four cores were sampled from each area of the prostate that was positive for HGPIN. Thirty-seven patients had one repeat biopsy, nine had two, one had three, and one had four. The decision to have further biopsies depended on the PSA value, PSA velocity, abnormal DRE finding, and/or the finding of HGPIN on subsequent biopsies. Of the 48 patients, a median (mean, range) of 12 (12, 8–17) cores were sampled during the initial biopsy.
One pathologist (S.I.B.) reviewed all 48 cases for confirmation of initial diagnosis and determination of histological subtype. The extent of HGPIN was calculated as the ratio of the number of cores with HGPIN to the total number of cores taken. Age, race, DRE and serum PSA level were also determined retrospectively. Univariate and multivariate logistic regression analysis was used to determine the correlation between these variables and the presence of cancer on repeat biopsies.
Also, 728 randomly selected retropubic RP specimens, resected over the same period by the same surgeon (J.B.N.) for cancer, were evaluated. The maximum diameter of the dominant tumour, Gleason score, margin status, seminal vesicle involvement, and clinical stage were assessed and compared with those of the resected prostate cancer specimens from the HGPIN group.
The median (mean, range) age of all 577 patients was 62.1 (62.4, 40.6–88.4) years and the PSA level before biopsy was 5.4 (11.8, 3.3–1329) ng/mL. The clinical variables of the 48 patients in the study are shown in Table 1, which also gives the results of first repeat biopsies. Overall, 10 (21%) patients who had isolated HGPIN on initial biopsy had cancer on their first repeat biopsy. The time to the second biopsy was 20 (41, 7–462) days, with all but one patient having a repeat biopsy by 49 days. Univariate analysis showed no significant differences in any of the measured clinical variables between patients who had cancer, benign, or HGPIN on first repeat biopsies. Reported P values are determined from the analysis among all three categories.
Table 1. The characteristics of the 48 patients and the outcome results on univariate and multivariate analyses
The results of all repeat site-directed biopsies for each patient were combined into the categories ‘cancer’ and ‘no cancer’, the latter consisting of patients with only benign or HGPIN diagnoses (the last four columns of Table 1). At the final repeat biopsy, 15 (31%) of patients had adenocarcinoma. Final biopsies were obtained at 28 (170, 7–1471) days from the initial biopsy date. After nonparametric univariate analysis, all covariates significant at P < 0.07 were evaluated in a multiple logistic regression model. While age, race, PSA level, presence of ASAP, HGPIN subtype and DRE result were no different between the outcomes, the proportion of cores with HGPIN on initial biopsy was significantly higher in those with cancer on final biopsy (P = 0.04).
A simple logistic regression analysis model based on the 48 patients suggests that the risk of finding cancer on repeat biopsy was significantly related to the proportion of cores with HGPIN (%HGPIN) found on initial biopsy (P = 0.02), expressed as (e−2.05+0.06 ×%HGPIN)/(1 + e−2.05+0.06 × %HGPIN). Figure 1 shows this predicted risk of having cancer, based on the proportion of cores with HGPIN on initial biopsy.
Of the 15 patients with cancer, 14 had a Gleason score of ≤7; 11 had a retropubic RP, and the remainder had (one each) either external beam radiation with or without brachytherapy, hormonal therapy, or were lost to follow-up. The 11 RP specimens were compared with those of 728 randomly selected patients undergoing RP during the same interval by the same urologist (Table 2). In the sample the maximum tumour diameter was 0.6 (0.7, 0.2–1.8) cm, significantly smaller than the maximum tumour diameter for the overall RP group (P < 0.01). No other evaluated tumour characteristic was significantly different between the groups.
HGPIN occurs as an isolated diagnosis in 0.7–24% of prostate biopsies [7,8]. When a sextant biopsy yields isolated HGPIN the standard practice is to re-biopsy, as reports have cited a 27–57% incidence of cancer on follow-up [1,4,5]. However, after extended-biopsy sampling, studies report an incidence of 2.3–11% cancer on the follow-up of patients with HGPIN [2,9–11], thus questioning the utility of re-biopsy. By contrast, we found that 15 of 48 patients with isolated HGPIN (31%) had cancer on subsequent site-directed biopsies. Ten of the 15 cancers (21% overall) were discovered on the first repeat biopsy. Our results are similar to those of Roscigno et al., who reported a 44.6% incidence of cancer on repeat extended biopsies, concluding that repeat biopsies are warranted in patients with isolated HGPIN.
By contrast with the present study, repeat biopsies in these previous reports were taken in the same extended pattern as the initial ones [2,9–12]. While the present patients had initial biopsies taken by an extended pattern, repeat biopsies were site-directed, where three to four cores were taken only from areas with HGPIN on initial sampling. There are several reasons why we used this practice. In our cohort, all biopsies were taken by one urologist who was confident of the location of the HGPIN foci found initially. Also, by limiting the number of cores sampled, the urologist was able to minimize biopsy-associated morbidity. The appropriate re-biopsy strategy is controversial. Chan and Epstein  found a significantly greater incidence of malignancy in areas localized to sites of atypical biopsy findings. Similarly, Park et al. found significantly greater rates of cancer after site-directed repeat biopsies, but also noted a 10–12% incidence of cancer at a location away from the location(s) of the initial atypical biopsy. Roscigno et al. reported a 24% incidence of remote cancer, including a 14% rate of cancer on the contralateral side. While these values support an extended re-biopsy technique, 75–85% of malignancies occurred in regions localized to the original abnormal biopsy site. The increased rate of cancer in regions of HGPIN might explain the higher incidence of cancer in our study. Also, our study was the first to use one urologist and one pathologist, thereby reducing inter-investigator variability in sampling and analysis, respectively.
The interval to repeat biopsy is important in the follow-up of patients with HGPIN. Extending the interval allows for a synchronous cancer to enlarge to a size that would be more likely to be sampled on repeat biopsy, but it also confounds the results, as developing tumours might be metachronous and not present when the HGPIN is initially detected. Previous studies concluded that there is a significant increase in the number of cancers detected after a 6-month delay between initial HGPIN diagnosis and repeat biopsy . Lefkowitz et al. found that patients with HGPIN on initial extended biopsy had a 2.3% incidence of cancer at 1 year on repeat extended biopsy, and a rate of 25.8% at 3 years. In the present study, the urologist recommended that patients be re-biopsied immediately. While there were a few ‘outliers’ who had final biopsies taken at >1 year after initial biopsy, the median time to final biopsy was 28 days; thus, the cancers detected were most likely unsampled synchronous, rather than metachronous tumours. This suggests that waiting for small tumours to enlarge might be helpful, but unnecessary.
We report that the proportion of cores with HGPIN on initial biopsy correlated with the chance of finding cancer on subsequent biopsies. Prange et al. reported that the number of cores positive for HGPIN on sextant biopsy corresponded with the total volume of HGPIN in cystoprostatectomy specimens. Furthermore, they showed that the rate of cancer was also associated with the number of cores positive for HGPIN on initial sextant biopsy. Our results are again similar to those of Roscigno et al., who concluded that, of the patients who had cancer on repeat biopsy, 90% had multifocal HGPIN on initial extended sampling. Our model illustrates that an individual with one of 12 cores positive would have a 17.2% risk of having cancer detected on subsequent biopsy, a rate similar to that reported from the follow-up of initially normal biopsies [1,17]. Consequently, we recommend that those men with unifocal HGPIN on extended biopsy be followed conservatively with serial PSA levels and DRE. Conversely, patients with multifocal HGPIN should be re-biopsied immediately. However, our study has an inherent bias; patients with more HGPIN foci are more extensively re-biopsied. This might artificially inflate the correlation between the proportion of cores with HGPIN and risk of cancer; however, the previous studies of Prange et al. and Roscigno et al. did not have the same site-directed repeat biopsy bias, but still reported a significant association between the variables [12,16]. Importantly, the proposed risk model might not be applicable to patients whose clinical features vary from those of the 48 in the present study (Table 1, first column).
Serum PSA levels were significantly different between the groups with cancer and without in the univariate, but not in the multiple logistic regression, analysis. PSA levels might be predictive of synchronous cancer in patients with HGPIN; our small sample of 48 men limits our ability to dismiss any clinical variable from having no predictive value. This might also be the case with the presence of ASAP, as only five patients had this on initial biopsy. Previous studies showed that the finding of ASAP with HGPIN is associated with a significantly higher rate of cancer on re-biopsy than HGPIN alone . The histological subtype of HGPIN did not correlate with final outcome either, a finding that might also be a function of the sample size. Only two patients had isolated flat HGPIN subtype, and none had cribriform or micropapillary patterns. Previous studies assessed the predictive significance of the histological subtype of HGPIN for prostate cancer on subsequent biopsies. Most report no significant differences in cancer rates between the different morphological patterns; however, Kronz et al. reported cribriform and micropapillary HGPIN subtypes to have a higher positive predictive value after univariate analysis of outcomes.
Of the cancers found on subsequent biopsy and removed, the median maximum tumour diameter was only 0.6 cm. This is significantly smaller than that of the 728 RP specimens resected by the same urologist. The smaller tumours might explain why these malignancies evaded detection on initial biopsy. Previous studies showed that prostate tumour volumes correlate with the number of positive cores in men undergoing extended biopsy . It is likely that men with isolated HGPIN and initially undetectable prostate tumours would have smaller cancers than men whose tumours were identified on initial biopsies. Extended sextant biopsy appears to increase the rate of detection of patients with HGPIN and cancer in a proportionate fashion, which might explain why our positive predictive value for cancer was comparable to data obtained from standard sextant biopsy samples. It is likely that extended sextant biopsy detects neoplasia early in its natural history. However, the overall Gleason scores and clinical stages were not significantly different between the groups. This might signify that these initially undetected cancers, while smaller than those found on initial biopsies, are histologically no less dangerous, further supporting the notion that patients with isolated HGPIN must be followed aggressively.
In conclusion, we report that 31% of patients with isolated HGPIN on initial biopsy had prostate cancer on subsequent biopsy. The incidence of cancer was directly associated with the number of cores with HGPIN involvement. Therefore, we recommend that patients with multifocal HGPIN on initial biopsy be re-biopsied immediately, whereas those with unifocal HGPIN be managed conservatively with serial estimates of PSA level and DREs.