Objective To evaluate p53 and bcl-2 immunohistochemistry in preoperative biopsies and radical prostatectomy specimens, as predictors of biochemical recurrence.
Patients and methods Preoperative biopsies from 73 men, and the radical prostatectomies from these men and from a further 47 men, were evaluated. The serum prostate specific antigen (PSA) level, Gleason score, pathological stage and margin involvement were recorded. The immunohistochemical expression of p53 and bcl-2 was studied on a representative area of tumour with the highest Gleason grade. The median follow-up was 53 months.
Results During the follow-up 47 of the 120 patients had a biochemical recurrence. Capsular penetration was present in 63 (53%) and the surgical margins were positive in 47 (39%). The Gleason score was < 7 in 81 (68%) patients; p53 was positive in 40 (66%) of 61 biopsies and 84 (71%) of 118 prostatectomy specimens. Bcl-2 was positive in eight (13%) of 63 biopsies and 20 (17%) of 118 prostatectomies. On multivariate analysis the biopsy p53, Gleason score and serum PSA were significant predictors of recurrence. On multivariate analysis, capsular penetration, PSA and margin status at prostatectomy were significant predictors of recurrence. There was also a significant interaction between PSA and margin status. Although univariately significant, neither p53 nor bcl-2 featured in the final multivariate model.
Conclusion Biopsy p53 status significantly predicts recurrence after radical prostatectomy, but its low specificity and technical issues suggest that it will not be useful in the clinical setting. However, a patient with negative p53 on biopsy is likely to have a good prognosis on prolonged follow-up.
Worldwide, more radical prostatectomies are being undertaken with the intention to cure clinically localized prostate cancer, and yet 16–42% of patients [1–4] have positive resection margins and up to 56% [5–7] show disease recurrence. The accurate preoperative prediction of disease relapse is urgently needed to select the right patients for curative treatment, enable appropriate preoperative counselling and select patients for adjuvant therapy. PSA, biopsy Gleason grade and clinical stage have often been identified as potential prognostic factors for biochemical recurrence. In addition, numerous attempts have been made to use tissue biomarkers to enhance the prediction of outcome after radical prostatectomy. Our group has previously investigated the role of p53, bcl-2, CD44 and E-cadherin immunohistochemical expression in biopsy and radical prostatectomy specimens on 76 patients with a median follow-up of 36 months . Multivariate analysis showed that Gleason score and p53 were independent preoperative predictors of PSA relapse. Significant postoperative predictors of biochemical recurrence were bcl-2, p53, Gleason score and margin status.
Several recent findings prompted us to re-examine our results; some studies cast doubt on the use of PSA as a surrogate for clinical relapse or overall survival. In a series of 132 consecutive patients, with up to 10 years of follow-up data available, Jhaveri et al. found no difference between the overall survival for patients with or with no PSA relapse after radical prostatectomy for clinically localized disease (P = 0.94). Biochemical recurrence-free survival estimates were highly dependent on the length of follow-up and may not reflect clinical recurrence rates after conformal radiotherapy .
In addition, in the largest series to date, with a median follow-up of 72 months, Stackhouse et al. found no relationship between either p53 or bcl-2 staining in biopsies and PSA recurrence. However, they found that both markers when measured in the prostatectomy specimen were associated with PSA relapse.
The present study represents an update of previously published results  with an extended follow-up and more radical prostatectomy specimens (120).
Patients and methods
Patients were selected on the basis of available pathological specimens and a full clinical follow-up. In total, 120 patients fulfilled these criteria and all had undergone radical prostatectomy at Southmead Hospital or nearby Weston General Hospital between 1987 and 1999. None of the patients had received neoadjuvant hormonal therapy. Preoperative biopsy material (either from TURP or biopsy cores) was available in 73 of these patients. Biochemical recurrence was defined as sustained increases in serum PSA of > 0.2 ng/mL on at least two occasions, with the second elevation measured at least 6 months after the first. All the radical prostatectomy specimens were reviewed by a histopathologist (J.D.O.), and the Gleason score and the pathological stage recorded. An area with the highest Gleason grade was selected for immunohistochemistry. The Gleason scores were grouped into three categories (2–4, 5–6 and 7).
Standard immunohistochemical methods were used as described previously . When the preoperative biopsy was available this was placed on the same microslide as the radical prostatectomy specimen. The primary antibody for the study was either p53 (DO7, Dako UK Ltd, High Wycombe, Berks, UK) at a dilution of 1/60 for 30 min, or bcl-2 (124, Dako) at 1/40 for 120 min. The primary antibody was replaced with washing buffer in each case as negative controls. A high-grade prostatic carcinoma was used as a positive control for p53. Human tonsillar tissue was used as a positive control for bcl-2.
One pathologist, unaware of treatment outcome, assessed the immunohistochemical staining; 100 consecutive malignant cells in the area of strongest staining were counted. Cytoplasmic bcl-2 staining in malignant glands was considered positive (Fig. 1a). The same technique was used for scoring p53 but only nuclear staining was counted as positive (Fig. 1b). p53 and bcl-2 expression were considered positive regardless of the percentage of cells stained, as previously suggested ; the absence of staining was considered negative.
The outcome of interest was biochemical recurrence; patients not experiencing a recurrence were censored at death or the last follow-up. Cox proportional hazards regression was used to identify factors associated with biochemical recurrence. The models were built in stages using the modelling scheme suggested by Collett . Factors significant at P 0.1 were retained and interactions among variables examined. Survival curves for the factors found to be associated with biochemical recurrence were produced using the Kaplan-Meier method. The proportional hazards assumption was assessed using the method described by Grambsch and Therneau  and implemented in a statistical package (Stata Statistical Software version 6, Stata Corporation, College Station, Texas).
Clinical follow-up was available on all 120 patients in the series; the median (range) follow-up was 53 (< 1–150) months. Of the 120 patients, 47 had had a biochemical recurrence; 12 patients died, two of whom had no evidence of biochemical recurrence at the time of death. One patient died immediately after surgery.
Tumour involved the surgical margins in 47 (39%) patients and the seminal vesicles in 15 (13%). There was capsular penetration (pT3) in 63 (53%) of the patients. The pretreatment PSA level was < 11 ng/mL in 50 (42%) patients and the operative Gleason score was < 7 in 81 (68%). When the grouped Gleason scores (2–4, 5–6, 7) were compared between the 73 biopsy specimens and the radical prostatectomies, 46 (63%) showed concordance, whilst the biopsy score was lower in 22 (30%) and higher in 5 (7%).
P53 and bcl-2 biomarker expression
There was insufficient tumour in the core biopsies for immunohistochemistry of p53 in 12 of 73 (16%), leaving data available for 61 biopsies. p53 data at prostatectomy were available for 60 of these patients. For 42 patients the biopsy and prostatectomy results concurred (29 where both were positive, 13 where both were negative). There were 11 patients whose biopsy was positive and prostatectomy was negative, and seven patients for whom the reverse was true (biopsy negative, prostatectomy positive; sensitivity 0.81, specificity 0.54). Of the 120 patients, 118 had p53 results at prostatectomy, 84 (71%) of which were positive; the remaining two patients had insufficient tumour for immunostaining.
Data on bcl-2 immunohistochemistry were available for 63 biopsies; bcl-2 data at prostatectomy were available for 62 of these patients. The results were identical for 51 patients (five both positive, 46 both negative). There were three patients with a positive biopsy result and a negative prostatectomy result, and eight where the reverse was true. Of the 120 patients, 118 had bcl-2 results at prostatectomy, 20 (17%) of which were positive, and the remaining two patients had insufficient tumour for immunostaining.
Six variables recorded at biopsy were considered as potential prognostic factors for biochemical recurrence; patient age, PSA group (0–10, 11–20, > 20 ng/mL), clinical stage (T1, T2), Gleason group (2–4, 5–6, 7–10), p53 > 0 (yes, no) and bcl-2 > 0 (yes, no). To allow for patients with insufficient tumour in the biopsy to be included in the analysis, it was necessary to include indicators for missing data for the variables p53 and bcl-2 in the regression models fitted (with 12 and 10 missing, respectively).
The PSA group, Gleason group and p53 were univariately significant at P 0.1 (Table 1); the final multivariate model contained the variables PSA group, Gleason group and p53. The relative risks for recurrence are also shown in Table 1. The likelihood of recurrence was similar for patients in Gleason groups 2–4 and 5–6, but those with a score of 7–10 were nearly five times more likely to experience a recurrence at any given time. The likelihood of recurrence was similar for patients with a PSA of 0–10 or 11–20 ng/mL, but patients with a PSA of > 20 ng/mL were twice as likely to experience a recurrence. Patients with a positive p53 were twice as likely to experience a recurrence as those with a negative p53. The Kaplan-Meier survival estimates by p53 status are shown in Fig. 2a.
Table 1. Relative risks for preoperative (73 men) and postoperative variables (120 men), and for significant pre- and postoperative factors (adjusting for other factors in the model)
Seven variables recorded at prostatectomy were considered as potential prognostic factors for biochemical recurrence; PSA group (0–10, 11–20, > 20 ng/mL), Gleason group (2–6, 7–10), capsular penetration (yes/no), seminal vesicle involvement (yes/no), margin involvement (yes/no), p53 >0 (yes/no) and bcl-2 >0 (yes/no). Because there were too few patients with a Gleason score of < 5 it was necessary to combine groups 2–4 and 5–6 for this analysis. Patients with missing data for p53 and bcl-2 (two each) at prostatectomy were assigned to the most common categories for these variables (p53 >0 and bcl2 = 0, respectively).
Each variable was univariately significant at P < 0.1; the final multivariate model contained the variables PSA, capsular penetration, margin involvement and an interaction between PSA and margins (Table 1). Patients with no capsular penetration had better survival than those with penetration (Fig. 2b) and patients with negative margins had better survival than those with positive margins, provided their PSA was 20 ng/mL. For those with a PSA of > 20 ng/mL, there was no difference in the survival of patients with positive and negative margins. The overall percentage (95% CI) of biochemical recurrence-free survival at 1, 3 and 5 years after prostatectomy were 81 (72–88), 62 (52–72) and 53 (42–64), respectively.
The ability to predict which patients will have recurrent disease after radical prostatectomy has been the goal of many studies over the past two decades. The advent of immunohistochemistry has led to many studies using tumour biomarkers, but no marker has moved from the research setting into the routine assessment of prostatic carcinoma, unlike oestrogen receptor status in breast carcinomas. Several groups have studied p53 and bcl-2 (reviewed in ) and the results have been conflicting.
In the earlier analysis of our data, p53 was the only significant preoperative tumour biomarker on multivariate analysis , although bcl-2 was significant univariately. With a further follow-up p53 remains significant in multivariate analysis. Previously p53 was also a significant postoperative marker but this was not found in the present analysis when more prostates were assessed. Stackhouse et al. examined p53 in 129 and bcl-2 in 103 preoperative biopsies and corresponding postoperative radical prostatectomies. Bauer et al. had previously studied these markers in the same radical prostatectomy specimens. Stackhouse et al. found bcl-2 positivity in 17% and p53 positivity in half the biopsies, whilst in the present study bcl-2 was positive in 13% and p53 in 67%. Stackhouse et al. did not find either to be significant univariately in predicting biochemical relapse (defined as a PSA of 0.5 ng/mL, or two serial measurements of 0.2 ng/mL). That paper also updated the series of 199 patients reported by Bauer et al., and confirmed that both p53 and bcl-2 were significant postoperative predictors on multivariate analysis. They concluded that the differences were probably caused by sampling error, as the tumours were heterogeneous and multifocal. This theory was supported in that they found p53 positivity in half the biopsies but 68% of the radical prostatectomies, and bcl-2 positivity in 17% of the biopsies and 33% of the corresponding radical prostatectomies . The present data show small differences between the specimen types, with p53 positivity in 67% of the biopsies compared with 60% of the corresponding radical prostatectomies, and bcl-2 positivity in 13% and 21%, respectively. In several patients there was no concordance between the biopsy and the corresponding radical prostatectomy sample for the markers. These differences may have been caused by sampling or technical issues, such as differences in fixation between the specimen types .
Variation in the results among authors may also be caused by differences in the thresholds for positive immunostaining, the number of preoperative biopsies and patient selection. Most of the patients in the series of Stackhouse et al. only had 2–4 biopsies and the PSA follow-up was not available for the early years, which may have had an effect on predictive ability. Some authors use a positive immunostaining threshold of > 0% , whilst others have used a threshold of 10% . Quinn et al. assessed p53 and used a threshold of > 2%, but they also included a group with < 2% staining showing ‘cluster positive’ (defined as a focus of 12 p53-positive cells within a × 200 field). They found that p53 was an indicator of poor prognosis in 263 patients, regardless of whether the patients had received neoadjuvant hormonal therapy. The present scoring system would have included the ‘cluster positive’ as true-positives, as the strongest staining areas were scored and the threshold set at 0%. p53 is a prognostic indicator in various patient groups, including neoadjuvant therapy combined with radical prostatectomy , locally advanced disease , and salvage prostatectomies after radiotherapy . These studies are not directly comparable with the present but suggest that p53 may be useful in particular patient groups.
Capsular penetration and the combination of preoperative PSA with the margin status were the only significant clinicopathological factors identified on multivariate analysis of the 120 radical prostatectomies. Epstein et al. evaluated 721 men who had undergone radical prostatectomy and found the Gleason score, capsular penetration and surgical margins to be significant predictors of biochemical recurrence. Banerjee et al. studied a group of 485 men and found that race, preoperative PSA, pathological stage and postoperative Gleason score were all significant multivariately. Both these large studies compare favourably with the present results. In a recent analysis of our entire series of 350 patients (including the 120 from this study) with a follow-up of 60 months, the preoperative PSA, Gleason score and surgical margin status were significant predictors of PSA recurrence on multivariate analysis (data not shown).
The present study and those of others include a group of patients who no longer reflect those currently undergoing biopsy and radical prostatectomy. There was a significant stage migration in the entire series of 350 men, with clinical stage T2 tumours reduced from 67% to 47% and positive surgical margins reduced from 38% to 24%, comparing patients undergoing prostatectomy before and after 1995. With the advent of PSA testing, improved selection criteria and increased use of systematic extended field biopsy, more tumours are identified at an earlier stage (stage T1c). The use of systematic extended field biopsy (10–12 cores) rather than lesion-directed or sextant biopsy allows for more representative tissue to be taken. This increases the likelihood that all the tumour foci could be sampled. This may allow greater accuracy for immunostaining in the future. These populations of patients still have limited follow-up and, given the controversy about PSA recurrence as a surrogate for clinical recurrence [9,10], there may be a long wait for future data, with perhaps different results.
At present, the noted limitations of current studies and the low specificities of p53 make p53 immunostaining in preoperative biopsies an unrealistic marker of recurrence in routine practice. However, from the present and previous studies, a patient with negative preoperative p53 staining is likely to have a good prognosis on prolonged follow-up.
The authors thank Hanna Lord for her technical assistance. The Southmead Hospital Research Foundation provided funding for this study. The Research and Development Support Unit at North Bristol NHS Trust is supported by a grant from the South-west NHS R&D Directorate.