What's known on the subject? and What does the study add?
Multifocality, age, PSA values, and biopsy protocols regarding the predictive value of high grade PIN have been discussed extensively in the literature.
Our study developed for the first time a predictive nomogram that could be helpful for patient counselling and to guide the urologist to perform rPBX after an initial diagnosis of isolated HGPIN.
• To evaluate factors that may predict prostate cancer (PCa) detection after the initial diagnosis of high-grade prostatic intra-epithelial neoplasia (HGPIN) on prostate biopsy (PBx) with six to 24 random cores.
PATIENTS AND METHODS
• We retrospectively evaluated 262 patients submitted from 1998 to 2007 to prostate re-biopsy (rPBx) after an initial HGPIN diagnosis in tertiary academic centres.
• HGPIN diagnosis was obtained on initial systematic PBx with six to 24 random cores.
• All patients were re-biopsied with a ‘saturation’ rPBx with 20–26 cores, with a median time to rPBx of 12 months.
• All slides were reviewed by expert uropathologists.
• Plurifocal HGPIN (pHGPIN) was found in 115 patients and monofocal HGPIN (mHGPIN) was found in 147 patients.
• In total, 108 and 154 patients, respectively, were submitted to >12-core initial PBx and ≤12-core initial PBx.
• Overall PCa detection at rPBx was 31.7%. PSA level (7.7 vs 6.6 ng/mL; P= 0.031) and age (68 vs 64 years; P= 0.001) were significantly higher in patients with PCa at rPBx.
• PCa detection was significantly higher in patients with a ≤12-core initial PBx than in those with a >12-core initial PBx (37.6% vs 23.1%; P= 0.01), as well as in patients with pHGPIN than in those with mHGPIN (40% vs 25.1%; P= 0.013).
• At multivariable analysis, PSA level (P= 0.041; hazards ratio, HR, 1.08), age (P < 0.001; HR, 1.09), pHGPIN (P= 0.031; HR, 1.97) and ≤12-core initial PBx (P= 0.012; HR, 1.95) were independent predictors of PCa detection.
• A nomogram including these four variables achieved 72% accuracy for predicting PCa detection after an initial HGPIN diagnosis.
• PCa detection on saturation rPBx after an initial diagnosis of HGPIN is significantly higher in patients with a ≤12-core initial PBx than those with a >12-core initial PBx and in patients with pHGPIN than in those with mHGPIN.
• We developed a simple prognostic tool for the prediction of PCa detection in patients with initial HGPIN diagnosis who were undergoing saturation rPBx.
High-grade prostatic intra-epithelial neoplasia (HGPIN) has been traditionally considered as a precursor of prostate cancer (PCa) [1–3]. TRUS-guided needle biopsies, performed after increased PSA levels or an abnormal DRE, detected HGPIN in 4–25% of patients [4–6]. Previous studies have shown a 22–100% cancer detection rate on prostate re-biopsy (rPBx) in patients with an initial HGPIN diagnosis [7,8].
Several clinical variables, such as abnormal DRE, anomalies on TRUS, patient age, PSA level and HGPIN focality, have been investigated as markers for predicting the presence of PCa on rPBx, although no consensus has yet been reached [9–12].
Moreover, the prognostic value of HGPIN in prostate biopsy (PBx) cores has been recently questioned because several studies have shown a lower cancer yield on rPBx, especially when the first sampling is performed using an extended biopsy technique [12–15].
The present study aimed to evaluate the impact of the number of cores taken at the initial PBx on the subsequent PCa diagnosis in patients with initial diagnosis of isolated HGPIN. Furthermore, we investigated which factors may predict the risk of PCa detection at rPBx in this subgroup of patients.
PATIENTS AND METHODS
We retrospectively evaluated 270 patients who underwent a prostate rPBx after an initial HGPIN diagnosis in three tertiary academic centres between 1998 and 2007. HGPIN diagnosis was obtained on initial systematic TRUS-guided PBx with six to 24 random cores (median, 12 cores). All patients were re-biopsied with a saturation scheme consisting of a mean (median, range) of 22 (20, 20–26) cores, with a median (range) time to rPBx of 12 (3–30) months, according to the urologist's preference. All patients undergoing a third or fourth biopsy received a 20–26-core biopsy.
All slides were reviewed by expert uropathologists. A total of eight patients were excluded; six as a result of a concomitant atypical small acinar proliferation (ASAP) and two as a result of a concomitant PCa microfocus.
Combined databases from the three centres were analyzed to retrospectively retrieve PSA and PSA density values at biopsy time, patient age, DRE or TRUS results, number of cores with HGPIN, and the time interval between initial and rPBx.
The HGPIN was classified as plurifocal when neoplastic foci were present in ≥2 cores.
When patients with initial HGPIN diagnosis underwent rPBx, four diagnoses were made: benign prostate tissue, HGPIN, ASAP or PCa. We combined the first three findings in a ‘no-cancer’ group, aiming to perform univariable and multivariable analyses comparing the cancer and no-cancer groups. Clinical data were analyzed using chi-squared and Student's t-tests. A multivariable logistic regression analysis was performed to address the impact of the above-mentioned variables on PCa detection in patients who were submitted to saturation rPBx. Multivariable logistic regression coefficients were used to generate a predictive nomogram. Each variable was assigned a scale of points according to a prognostic effect in the range 0–100. The point values determined for each individual case were added to give a total sum. The total sum calculated was correlated to the probability of PCa detection. The accuracy of this nomogram was quantified with receiver area under the curve. Internal validation was performed using 200 bootstrap resamples.
Statistical analyses were performed with S-Plus Professional software (MathSoft Inc., Seattle, WA, USA). P < 0.05 was considered statistically significant.
Patient characteristics are reported in Table 1. Plurifocal HGPIN (pHGPIN) was found in 115 (43.9%) patients and monofocal HGPIN (mHGPIN) in 147 (56.1%) patients. In total, 154 patients and 108 patients, respectively, underwent ≤12-core and >12-core initial PBx.
Table 1. Patient characteristics and descriptive statistics
Mean (SD, range) age was 65.3 (0.7, 47–83) years; mean (SD, range) PSA level was 7.1 (4.5, 1.5–26.4) ng/mL; and mean (SD, range) prostate volume was 55.5 (27.3, 17–160) mL.
The overall PCa detection at rPBx was 31.7% (83 patients). There were eight patients who had ASAP diagnosis at rPBx, whereas 11 presented HGPIN. A Student's t-test showed significant differences in initial PSA level (6.6 vs 7.7 ng/mL; P= 0.031), age at PBx (64 vs 68 years; P= 0.001) between the ‘no-cancer’ group and PCa patients at rPBx. No differences were found with respect to prostate volume and the time to rPBx.
Univariable logistic regression analysis showed that the higher number of cores taken at the initial PBx (continuous variable) was associated with a lower PCa detection at saturation rPBx (P= 0.03). A threshold of ≤12-core vs >12-core was chosen according to the most informative threshold, using anova for every possible threshold and choosing the lowest P-value.
Chi-squared analysis showed that PCa detection was significantly higher in patients who executed a ≤12-core initial PBx than in those with a >12-core initial PBx (37.6% vs 23.1%; P= 0.01) and in patients with pHGPIN than in those with mHGPIN (40% vs 25.1%; P= 0.013).
When analyzing PSA level, age, prostate volume and DRE findings, no significant difference was found between patients with a ≤12-core initial PBx and those with a >12-core initial PBx between patients with pHGPIN or mHGPIN.
At multivariable analysis, PSA level (P= 0.041; hazards ratio, HR, 1.08), age (P < 0.001; HR, 1.09), pHGPIN (P= 0.031; HR, 1.97) and ≤12-core initial PBx (P= 0.012; HR, 1.95) were independent predictors of PCa detection. By contrast, DRE findings, prostate volume and the time to rPBx did not achieve the independent predictor status of PCa detection (Table 2).
Table 2. Univariable and multivariable logistic regression analysis in the overall population (N= 262)
Variables not included in the model: DRE findings, prostate volume, time to re-biopsy. HGPIN, high-grade prostatic intra-epithelial neoplasia; OR, odds ratio.
PSA (ng/mL) (continuous variable)
Number of cores with HGPIN, plurifocality vs monofocality
Number of cores at initial biopsy, ≤12 vs >12
Age (years) (continuous variable)
A nomogram including the four predictive variables achieved 72% accuracy in predicting PCa detection after an initial HGPIN diagnosis (Fig. 1).
The bootstrap corrected accuracy of the nomogram including these four variables was 72% (Fig. 1). According to this nomogram, patients with monofocal HGPIN, aged 65 years old, with a PSA level of 5 ng/mL and >12-core initial PBx received a total of 60 points, corresponding to a less than 20% risk of PCa detection at saturation rPBx. A patients with plurifocal HGPIN, aged 70 years old, with PSA level of 10 ng/mL and ≤12-core initial PBx received a total of 110 points, with a 60% risk of PCa detection at saturation rPBx.
The calibration plot showed that the performance characteristics of the developed nomogram were close to achieving perfect prediction in almost all patients with a nomogram predicted probability ≤50% (Fig. 2). Conversely, in patients with a PCa predicted probability >50%, the nomogram tend to underestimate the real risk of tumour.
Furthermore, among the 179 patients with no cancer on rPBx, 85 were followed-up and received a third biopsy over a mean (range) period of 14.5 (3–29) months. Eleven patients had PCa diagnosis at third PBx (12.7%). Among those patients, eight have had an initial diagnosis of HGPIN with a ≤12-core initial PBx.
Eleven patients received a fourth biopsy. In one patient, who received ≤12-core initial PBx, PCa was found (9%).
The clinical importance of HGPIN is related to its predictive value as a marker for PCa. The HGPIN incidence and its correlation with PCa have been very variable in the literature, in the range 0.7–24% [16–18]. Many trials have reported a PCa yield on rPBx (as a result of a previous HGPIN diagnosis) in the range 22–100% [7,8,18–20]. Thus, the PCa detection rate after an initial HGPIN diagnosis has decreased from 40% to 50% in the early 1990s to 10% to 30% in recent studies [9–12]. The change in prostate sampling from sextant to extended or double-sextant protocol is considered to be largely responsible for this decrease .
The results of the present study are similar to those reported in contemporary series, with a mean PCa detection rate of 31% after an initial HGPIN diagnosis. The relatively higher percentage, compared to values previously reported by Gallo et al.  and De Nunzio et al. , could be explained by the higher number of cores taken at rPBx (18–26 cores), as well as the longer time to rPBx (range 3–30 months; median 12 months). Moreover, the results of the present study appear to be in accordance with those proposed by Price et al.  in a prospective study.
Furthermore, the data obtained in the present study show the impact of the number of cores taken at the initial PBx with respect to the subsequent risk of PCa detection in patients with an initial diagnosis of HGPIN.
A higher number of cores taken at the initial PBx (considered as a continuous variable) was associated with lower PCa detection at saturation rPBx (P= 0.03). Furthermore, PCa detection at rPBx was significantly higher in patients who executed a ≤12-core initial PBx than in those with a >12-core initial PBx. A possible objection is that many (if not most) current biopsy schemes consist of 12 cores, which is considered to represent an adequate initial sampling of the prostate. Therefore, it could have been more useful to look at the risk of cancer on rPBx for <12 cores vs >12 cores. Unfortunately, when dichotomizing the population according to the <12-core biopsy vs ≥12-core biopsy, statistical significance was not achieved, probably as a result of the relatively low number of patients in the <12-core biopsy group.
Moreover, a number of biopsies with a ≤12 cores taken at the initial PBx achieved the independent predictor status of PCa detection. Patients with isolated HGPIN diagnosis after an initial PBx set with ≤12 cores taken had a twofold higher risk of PCa detection at saturation rPBx compared to those patients undergoing a >12-core initial PBx. This result is in agreement with the guidelines of the National Comprehensive Cancer Network, which recommends extended rPBx, including the transition zone, if HGPIN is found in TRUS-guided PBx with <10 cores .
It may be concluded that a >12-core initial PBx appears to represent an extensive sampling, thus providing a high negative predictive value. In the case of an isolated HGPIN diagnosis, we can reasonably assume that the risk of missing concurrent PCa at rPBx is low and requires no aggressive rPBx protocol. Moore et al.  proposed that patients could be monitored with yearly PSA and DRE. Nonetheless, DRE was not predictive of PCa detection in the present study. However, there are a lack of good long-term data on the risk of subsequent cancer. Until such studies are completed, it may be reasonable to perform a rPBx within 3 years after the initial diagnosis of HGPIN  and, at present, an initial HGPIN diagnosis remains a significant risk factor for rPBx within 3 years of the initial PBx .
The need for a close follow-up and rPBx protocol in patients with an initial isolated HGPIN diagnosis obtained with a lower number of cores is further supported by the finding that most PCa detected at the third or fourth rPBx have been diagnosed with a ≤12-core initial PBx.
Moreover, several studies have analyzed the prognostic value of PSA level, PSA density, patient age, prostate volume, abnormal DRE and/or TRUS findings, as well as the time to rPBx in patients with an initial HGPIN diagnosis [13–21,26].
In the present study, both older age and higher PSA levels were independently associated with a higher risk of PCa detection at saturation rPBx. By contrast, DRE findings, prostate volume and the time to rPBx did not achieve the independent predictor status of PCa detection.
The re-biopsy follow-up interval is one of the main concerns in the case of an initial, isolated HGPIN diagnosis. The most aggressive re-biopsy protocol reported in the literature is the performance of follow-up biopsies at 3- to 6-monthly intervals for 2 years, followed by 12-monthly intervals for life . Lefkowitz et al.  confirmed that HGPIN is a risk factor in the development of PCa and recommended a 3-year follow-up interval biopsy. In a previous mono-institutional experience , a 12-month follow-up rebiopsy appeared to provide a higher detection rate for pathologically organ-confined cancer in the case of a ≤12-core initial PBx, avoiding unnecessary negative biopsies (as a result of biopsying too soon) and reducing the risk of missing curable PCa (as a result of biopsying too late).
In the present study, no impact on PCa detection was found according to the interval between the initial PBx and rPBx. The increased number of cores taken at the initial PBx probably provides a high negative predictive value, and a longer follow-up time to rPBx is needed to increase PCa detection in case the initial PBx had missed a very low-volume cancer.
The role of HGPIN plurifocality is still controversial. Merrimen et al. [10,27] reported a higher likelihood of PCa when multiple prostatic sites (≥2 cores) were involved. De Nunzio et al.  suggested that a 6-month PBx is recommended in patients with HGPIN when four or more cores with HGPIN are detected in the initial PBx sample, regardless of the PSA level. By contrast, HGPIN focality did not appear to influence the subsequent diagnosis of PCa according to the findings of Gallo et al. . However, the first two studies [10,27] analyzed more than 500 patients, whereas that of Gallo et al.  analyzed only 65 patients.
We previously reported a statistically significant difference in the PCa detection rate in patients with mono- or plurifocal HGPIN in a population of patients with an initial 10–12-core PBx. The data obtained in the present study support the predictive role of HGPIN plurifocality. Patients with ≥2 cores with HGPIN had an almost twofold higher risk of PCa detection at rPBx compared to patients with monofocal HGPIN. According to these results, we developed a simple tool for the prediction of PCa detection (with an accuracy of 72%) in patients undergoing saturation rPBx after an initial diagnosis of isolated HGPIN.
This prognostic tool may help to better select those patients who would need rPBx after an initial diagnosis of isolated HGPIN and also could be useful in patient counselling. It should be noted that the performance characteristics of the developed nomogram were close to achieving perfect prediction in almost all patients with a nomogram predicted probability ≤50%. Conversely, in patients with a PCa predicted probability >50%, the nomogram tends to underestimate the real risk of a tumour. However, this observation does not limit the use of the nomogram. In clinical practice, all patients with a PCa predicted risk of >50% will receive a biopsy, regardless of their exact predicted value. Conversely, the nomogram predictions were almost perfect in patients with a predicted value ≤50%. Consequently, a biopsy can be safely omitted in patients with a low PCa predicted probability, without significantly increasing the risk of missing positive diagnosis.
In addition to accuracy, simplicity and the use of readily available clinicopathological criteria, the use of a multi-institutional cohort allows the current predictive tool to be applied to the general patient population.
The present study is not devoid of limitations. The conclusions drawn may be somewhat limited by the relative small study population and the retrospective nature of the study. The number of cores taken at the initial PBx is determined by the treating urologist and may be affected by several parameters such as PSA level, prostate volume or DRE findings. Furthermore, patients in the present study underwent a second set of PBx based on the opinion of their urologist. Even if these biases may affect the results of the present study, no significant differences in abnormal DRE findings, prostate volume and PSA levels were detected between the group of patients with a 12-core initial PBx or >12-core initial PBx and between patients re-biopsied more (or less) than 12 months after the first set of biopsies.
In conclusion, PCa detection on saturation rPBx after an initial diagnosis of HGPIN is significantly higher in patients with a ≤12-core initial PBx than those with a >12-core initial PBx. Moreover, patients with ≥2 cores with HGPIN had a more than threefold higher risk of PCa detection at rPBx compared to patients with monofocal HGPIN. Older age and higher PSA levels are also associated with a higher risk of PCa detection on subsequent rPBx.
We have developed a simple and accurate nomogram based on the four available clinicopathological variables of age, PSA level, mono- or plurifocality of HGPIN and the number of cores taken at the initial PBx. This simple prediction model can be used for patient counselling, as well as for guiding the urologist to perform rPBx after an initial diagnosis of isolated HGPIN.