Introduction

Biochemical failure (bF) following definitive therapy for localized prostate cancer is known to precede clinical evidence of tumor recurrence by months or years. ^1,2,3 Its significance for prediction of actual death has been challenged, perhaps partially in consequence of the long natural history of localized prostate cancer. Kupelian et al³ found no association between bF after definitive radiotherapy and increased mortality within the first 10 y after treatment. They suggested a trend toward worse outcome after 8 y in those with bF. Jhaveri et al² concluded patients with bF following radical prostatectomy have excellent overall survival, equivalent to those without a detectable PSA at 10 y. In this study, the 10-year actuarial overall survival curves for patients with and without bF were similar even among those at high risk. D'Amico et al,⁴ on the other hand, reported that patients at high risk for bF after radiation therapy are also at high risk for death from prostate cancer despite competing causes of morbidity.

The significance of biochemical failure may differ among different patient subgroups; more information is needed to further understand the impact of bF on long-term outcome after definitive therapy in prostate cancer. We therefore conducted a retrospective analysis covering patients who had undergone radical prostatectomy before 1998 at two Japanese institutions.

Patients and methods

Patient population

The study population comprised 223 consecutive men treated with radical retropubic prostatectomy and bilateral pelvic lymph node dissection at Kitasato University Hospital (n=155) and Kurashiki Central Hospital (n=68) between January 1988 and March 1998. Several different PSA assays were used at Kitasato during this period. Serum PSA was quantitated by Eiken polyclonal radioimmunoassay (Eiken, Tokyo, Japan) up to March 1993, with Dainapack IMx PSA assay (Dinabot, Tokyo, Japan) up to June 1997, and with AxSYM PSA assay (Dinabot) thereafter. For uniformity, Eiken PSA data were interconverted as follows: (IMx PSA)=1.39 (Eiken PSA)-1.02.⁵ Results of other assays were not interconverted because they are considered virtually identical. All values at Kurashiki were by Dainapack IMx PSA assay.

Pretreatment staging

A clinical stage was originally assigned by faculty urologists at each institution in accordance with the 1992 version of unified tumor node metastasis (TNM) system.⁶Hospital records were carefully reviewed in 20 patients who had undergone surgery prior to that date to assign clinical stages retrospectively. These are now translated into the updated 2002 version.⁷

Pathological examination

All hematoxylin and eosin (H&E) pathological slides of radical prostatectomy specimens were reviewed by a single pathologist (SK). Tumor grade was determined according to the Gleason grading system. The method used to prepare pathological slides was not necessarily consistent during this study period, especially at earlier dates. Thus, both whole mount step and random sectioning were seen. Since 1992, most slides have been prepared using the former technique. Tumor volumes were calculated using the method previously described.⁸ Tumor extent was determined according to the General Rule for Clinical and Pathological Studies on Prostatic Cancer.⁹ Briefly, the prostatic capsule on each slide was carefully inspected microscopically for areas of complete penetration of cancer through the capsule surface (extracapsular extension). The presence of seminal vesicle involvement, nodal involvement and the status of the surgical margins were also assessed.

Follow-up

After surgery, patients were evaluated at 3- to 6-month intervals with a PSA measurement and digital rectal examination. Bone scans were repeated annually as a regular test. Postoperative PSA values were considered elevated (bF) if values of 0.1 ng/ml or greater were obtained on two consecutive visits 1 month apart for IMx or AxSYM PSA assays and values of 0.6 ng/ml or greater for the Eiken PSA assay. If PSA never became undetectable postoperatively, then bF was considered to have occurred at time zero. Disease progression was defined clinically as finding biopsy-proven locally recurrent tumor (local recurrence), or appearance of new lesions on an annual bone scan or appearance of soft-tissue metastasis indicated by biopsy (systemic recurrence). Lymphadenopathy alone detected incidentally by computed tomography during the work-up of other medical conditions was not considered an indication of recurrence unless proven by biopsy. Causes of death were determined based on hospital records and death certificates.

Statistical analysis

Mann–Whitney U and Kruskal–Wallis tests were used for comparison of variables, with P<0.05 as significant. A Cox proportional hazards model was used to assess the ability of parameters to predict outcome with other variables controlled. These parameters included age, PSA, specimen Gleason score, clinical stage, extracapsular extension, seminal vesicle involvement, nodal involvement, margin status, neoadjuvant and adjuvant endocrine therapy, salvage radiotherapy and bF. Another model including time to bF that substituted parameter bF was also investigated. The significance of risk-group classes⁸ based on PSA, specimen Gleason score and clinical stage were also tested in a separate model. In risk-group stratification, patients with clinical stage T1c or T2a, PSA10 ng/ml and specimen Gleason score 6 were defined as low-risk, and patients with T2c, a PSA >20 ng/ml or a specimen Gleason score 8 were defined as high-risk. The remaining patients were defined as the intermediate risk group. Analysis of clinical outcome after radical prostatectomy was also conducted using the Kaplan–Meier method and the results were tested using the log-rank test; the level of significance was again set at 5%. For comparison, expected survival curves of the Japanese general male population at an age of 65 y were also transposed.¹⁰ All statistical analyses were conducted using the statistics graphic data measurement software program (STATA Computing Resource, Los Angeles, CA, USA).

Results

Patient characteristics and pathological outcome

Median age of the 223 patients at radical prostatectomy was 65.0 y (range, 49–80 y; mean, 64.7 y). The median follow-up period after radical surgery was 93.0 months (range, 2.0–168.0 months; mean, 91.0 months).

The 223 patients were classified clinically as stage T1a (three patients), T1b (eight patients), T1c (96 patients), T2a (28 patients), T2b (47 patients), T2c (17 patients), T3a (seven patients), T3b (16 patients) and T4 (one patient). Specimen tumor grade was classified as Gleason score 2–4 in 27 patients (12.1%), 5 or 6 in 64 patients (28.7%), 7 in 89 patients (39.9%) and 8–10 in 38 patients (17.0%). Median preoperative serum PSA was 8.2 ng/ml (range, 0.5–136.0 ng/ml; mean, 14.0 ng/ml). The respective proportions of patients with PSA 4.0, 4.1–10.0, 10.1–20.0, and 20.1 ng/ml were 13.9 (31 patients), 45.3 (101 patients), 23.3 (52 patients) and 15.7% (35 patients). Four patients (1.8%) lacked preoperative PSA data.

Risk of progression⁸ was determined as low in 51 (22.9%), intermediate in 92 (41.3%) and high in 74 (33.2%) patients. Neoadjuvant endocrine therapy of any type was employed for longer than 3 months in 12 patients (5.4%). In all, 13 patients (5.8%) underwent lifelong adjuvant endocrine therapy with (one patient) or without (12 patients) adjuvant external beam local irradiation. After development of bF, some type of salvage therapy (eg, irradiation alone (63 patients) or endocrine therapy with or without chemotherapy) was utilized in 73 patients (32.7% of total cohort).

Five patients (2.2%) had pT0 disease. In total, 105 patients (47.1%) had locally advanced disease on final pathologic examination; extracapsular extension was seen in 99 patients (44.4%), seminal vesicle involvement in 52 (23.3%) and microscopic nodal involvement in 21 (9.4%). Positive surgical margins were found in 73 patients (32.7%, with rates of 9.7% in pT2 and 59.0% in pT3 disease). Median tumor volume was 2.6 cm³ (range, 0.01–59.9, data from 212 patients).

Disease progression and survival after radical prostatectomy

To date, 93 patients (41.7%) have experienced bF at a median of 16.0 months following surgery (range, 0.0–128.0). It occurred 60 months or longer after surgery in 18 patients (19.4%). Biopsy-proven local failure and systemic progression were diagnosed, respectively, in 12 patients (5.4%) at a median of 15.0 months (range, 5.0–54.0) and in 25 patients (11.2%) at a median of 38.0 months (range, 2.0–90.0) following surgery. To date, nine patients with local failure have developed metastasis, with local failure preceding systemic failure in all but one case. Only one patient with clinical recurrence did not also have bF; the lone exception had a positive biopsy at the anastomotic site despite undetectable PSA levels.

Over the study period, 30 patients (13.5%) died from all causes combined. Of the 93 patients with bF, 24 (25.8%) died, while only six (4.6%) of the 130 without bF died (P<0.0001). One patient without bF committed suicide 2 months after surgery and this represented the only mortality within a 6-month postoperative period. The observation periods were not significantly different between those with and without bF. Prostate cancer was the cause of death in 20 patients (66.7%). One patient died with both advanced prostate cancer and colorectal carcinoma. Since there was no autopsy to identify the specific cause of death, both conditions were treated as the probable cause of his death for purposes of statistical analysis. Median times to prostate cancer death and all cause death were, respectively, 60.5 months (range, 33.0–109.0) and 60.0 months (range, 2.0–168.0) after surgery.

Table 1 shows the 5- and 10-y rates for bF-free survival, local progression-free survival, systemic progression-free survival, cancer-specific survival and overall survival. The 5- and 10-y overall survival rates for patients with and without bF were 87.1 (95% CI: 78.3–91.4) and 69.1% (95% CI: 55.8–79.1), and 96.9 (95% CI: 91.9–98.8) and 94.6% (95% CI: 85.3–97.5), respectively (P<0.05).

Table 1 - Outcomes of 5 and 10 y.

Full table

Results of Cox proportional hazards multivariate analysis

Pretreatment PSA was found to be a significant predictor for bF (P<0.05), but not for overall survival. Risk group stratification had the same predictive value (data not shown). Specimen Gleason score and clinical stage were not predictive of any clinical outcomes. Advanced pathological stage, as well as neoadjuvant and adjuvant endocrine therapy, was predictive of bF (P<0.05). Seminal vesicle involvement was found to be a significant predictor of systemic progression, cancer death and overall survival (P<0.05). Positive surgical margin and bF were also independent predictors of overall survival (P<0.05). Age, neoadjuvant and adjuvant endocrine therapy, and salvage radiotherapy were not correlated with overall survival. Biochemical failure was excluded from pretreatment variables in other outcome analyses, since it almost invariably accompanied these outcomes. Salvage radiotherapy was excluded for the same reason in the analysis for bF. Time to bF 24 months following surgery was also an independent predictor of systemic progression, cancer death and overall survival (P<0.05) in a different model.

Figure 1 displays the overall survival curves by bF status. After radical prostatectomy, patients with bF had significantly worse survival than those without bF (P<0.0001). Survival of patients without bF was even superior to the expected survival of the 65-y-old Japanese general male population (P=0.002).¹⁰

Figure 1.

Kaplan–Meier projection of overall survival of patients after radical prostatectomy according to biochemical failure. bF, patients with biochemical failure, non-bF, patients without biochemical failure, age 65 y, expected survival curves of the Japanese general male population at age 65 y.

Full figure and legend (25K)

Table 2 summarizes the clinical outcome of patients with and without bF according to risk-group stratification. Patients with bF and higher risk had significantly worse outcome with respect to systemic failure, cancer-specific survival and overall survival (P=0.03). Table 3 shows the results according to the pathological stage and status of seminal vesicle involvement. Patients with bF had significantly worse outcome with respect to systemic failure (P=0.03), cancer-specific survival (P=0.04) and overall survival (P=0.04) when their diseases were pathologically advanced (pT3). Those with seminal vesicle involvement had much worse outcomes with respect to systemic failure (P=0.0006), cancer-specific survival (P=0.008) and overall survival (P=0.01). Positive margin status did not allow any further stratification (data not shown). Table 4 shows the results according to the time to bF following surgery. Those with shorter time to bF (24 months) had significantly worse outcome with respect to systemic failure (P=0.003) and cancer death (P=0.011). The difference in overall survival approached statistical significance (P=0.08).

Table 2 - Biochemical failure and clinical outcome according to risk-group stratification after radical prostatectomy.

Full table

Table 3 - Biochemical failure and clinical outcome according to pathological findings after radical prostatectomy.

Full table

Table 4 - Clinical outcome according to time to biochemical failure following radical prostatectomy.

Full table

Discussion

Prostate-specific antigen has been used effectively as a surrogate to measure individual success or failure of surgical treatment. ^1,11,12,13 However, the natural history of the patient with bF after radical prostatectomy remains poorly defined, and no criteria for implementing treatment of such patients are uniformly accepted.

Pound et al¹ reported that in radical-surgery patients the median time to discovery of distant metastases is 8 y from the time of isolated bF and the median actuarial time to death is 5 y thereafter. Thus, those with bF after radical surgery may have a reasonably long life expectancy. An elderly man may die from an unrelated cause before he develops symptoms that require treatment. This is supported by other studies indicating an excellent overall survival, equivalent to those without detectable or increasing PSA after definitive therapy.^2,3 This lack of bF impact on overall survival in the surgical series may partially reflect more favorable tumor characteristics and lead time,^2,3 but the 10-y actuarial overall survival curves for patients with and without bF were found to be similar even in high-risk patients (T2b or greater, PSA greater than 10 ng/ml or specimen Gleason score 7 or greater).² A trend toward worse outcome in those with bF was found after 8 y in the radiotherapy series.³ This suggests that longer follow-up may show bF becoming increasingly significant in predicting overall survival.

Contrary to previous reports from Western countries, the current data on the natural history of bF indicate a significant association with overall survival. Biochemical failure occurred at a median of 16.0 months after surgery. Median time to prostate cancer death and all cause death were 60.5 and 60.0 months. The 5- and 10-y overall survival rates for patients with bF were significantly worse than for those without bF (87.1 vs 96.9, and 69.1 vs 94.6%, respectively (P<0.05). The reason for better overall survival in patients without bF than in the general male population aged 65 y is likely due to selection of healthier patients at surgery. The 10-y overall survival rates in our study appear comparable to Western surgical series (82.5 vs 75–95%) despite lower bF-free survival.^{1,2,14,15,16,17}

As expected, tumor characteristics most significantly affected overall survival. In multivariate analysis, patients with seminal vesicle involvement had significantly worse outcomes regarding bF, systemic progression, cancer death and overall survival (P<0.05). Both bF and positive surgical margin were associated with worse overall survival (P<0.05). The use of neoadjuvant or adjuvant endocrine therapy, or salvage irradiation, is an independent predictor for local progression, probably reflecting more advanced disease. Age, neoadjuvant and adjuvant endocrine therapy, and salvage radiotherapy, however, did not affect overall survival. Although preoperative PSA, risk-group stratification, extracapsular extension and adjuvant and neoadjuvant endocrine therapy were significant predictors of bF, multivariate analysis failed to show a correlation with death. Factors associated with bF were thus not necessarily associated with worse overall survival.

To predict systemic failure, cancer-specific survival and overall survival, patients with bF could further be stratified according to clinicopathological characteristics such as risk-group classes, pathological stage or seminal vesicle involvement (Tables 2 and 3, P<0.05). Adverse findings always accompanied a worse outcome.

It is recognized that some patients with bF have a worse prognosis than others. Pound et al¹ proposed an algorithm to predict a man's likelihood of developing metastatic disease following bF that was based on the specimen Gleason score (7 vs 8), the time of initial bF (2 vs >2 y) and PSA doubling time (PSADT, 10 vs >10 months). Specimen Gleason score was not an independent predictor for systemic progression in our study population. Advanced pathological stages had stronger impact in the multivariate analysis. Time to bF 24 months following surgery was confirmed an independent predictor of systemic progression, cancer death and overall survival (P<0.05). One of the problems with using PSADT is its limited applicability. In only 53 of 93 cases with bF was PSADT successfully calculated based on calculation over a 6-month postoperative period in this study (data not shown).¹⁸ A more practical model is thus needed.

The existence of a significant correlation between bF and overall survival in our findings may partly result from the inclusion of larger and more advanced tumors. This, in turn, reflects different practice patterns (eg, prevalence of PSA-based screening) in different countries.^11,18 The median tumor volume in our surgical series was more than twice that seen in the series' Western counterparts (2.6 vs 1.2 cm³).⁸ This difference in volume is presumably attributable to years-earlier detection of prostate cancer in Western countries, since the median PSA doubling time of clinically resectable, nonmetastatic prostate cancer has been calculated as approximately 36 months.¹⁸ About half (47.1%) of our patients had pathologically advanced tumors (pT3). Higher incidence of seminal vesicle and nodal involvement compared with Western series^1,2 is also noted (23.3 vs 5–15 and 9.4 vs 4–6%, respectively). All these findings suggest prostate cancer was detected at least several years later in our series.¹⁸

Our study is limited owing to the small number of cases and its retrospective nature. The use of neoadjuvant or adjuvant endocrine therapy and salvage therapy at the time of bF, which is carefully excluded in the study of Pound et al,¹ may have obscured the impact of this event on overall survival. No uniform and common policy between institutions existed regarding the use of such treatment in this study. Nevertheless, the use of such additional therapies for adverse pathological findings or bF, with the intent to improve survival, is common. The analysis of such population is thus relevant to patients in general practice. More efforts should be directed toward the study of the natural history of these patients with bF. Specifically, accurate prediction of the risk of clinical progression and death after bF is of utmost importance. Patients with worse prognosis after bF are most suitable for clinical trials rather than mere observation.

Widespread use of PSA screening and early detection in Japan may lead in the near future to a shift toward detection at stages more typical of those seen in Western practice, as has been observed in the African-American population.¹⁹ In such patients, prostate cancer may simply need more time to affect overall survival. Further work is warranted.

References

1. Pound CR et al. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999; 281: 1591–1597. | Article | PubMed | ISI | ChemPort |
2. Jhaveri FM, Zippe CD, Klein EA, Kupelian PA. Biochemical failure does not predict overall survival after radical prostatectomy for localized prostate cancer: 10-year results. Urology 1999; 54: 884–890. | Article | PubMed | ISI | ChemPort |
3. Kupelian PA et al. Impact of biochemical failure on overall survival after radiation therapy for localized prostate cancer in the PSA era. Int J Radial Oncol Biol Phys 2002; 52: 704–711. | Article |
4. D'Amico AV et al. Pretreatment predictors of time to cancer specific death after prostate specific antigen failure. J Urol 2003; 169: 1320–1324. | Article | PubMed |
5. Machida T et al. Clinical evaluation of a new kit (IMx PA Dainapack) for detection of serum prostate specific antigen. Acta Urol Japon 1993; 39: 977–984.
6. International Union against Cancer. TNM Atlas, 3rd edn., 2nd revision. Springer: New York 1992; pp 241–250.
7. UICC International Union. In: Sobin LH, Wittekind CH (eds). UICC International Union Against Cancer, TNM classification of malignant tumors. 6th edn. John Wiley & Sons, Inc: New York 2002; pp 184–187.
8. Egawa S et al. Infrequent involvement of the anterior base in low-risk patients with clinically localized prostate cancer and its possible significance in definitive radiation therapy. Jpn J Clin Oncol 2000; 30: 126–130. | Article | PubMed | ChemPort |
9. Japanese Urological Association. General Rule for Clinical and Pathological Studies on Prostatic Cancer, 3rd edn. Japan: The Japanese Pathological Society 2001.
10. National Institute of Population and Social Security Research: Population Projections for Japan: 2001–2050 (with long-range population projects: 2051–2100).
11. Egawa S. Detection of prostate cancer by prostate-specific antigen. Biomed Pharmacother 2001; 55: 130–134. | Article | PubMed |
12. Egawa S et al. Impact of life expectancy and tumor doubling time on the clinical significance of prostate cancer in Japan. Jpn J Clin Oncol 1997; 27: 394–400. | Article | PubMed |
13. Partin AW, Oesterling JE. The clinical usefulness of prostate specific antigen: update 1994. J Urol 1994; 15: 1358–1368.
14. Zincke H et al. Long-term (15 years) results after radical prostatectomy for clinically localized (stage T2c or lower) prostate cancer. J Urol 1994; 152: 1850–1857. | PubMed | ISI | ChemPort |
15. Catalona WJ, Smith DS. Cancer recurrence and survival rates after anatomic radical retropubic prostatectomy for prostate cancer: intermediate-term results. J Urol 1998; 160: 2428–2434. | Article | PubMed | ISI | ChemPort |
16. Gerber GS et al. Results of radical prostatectomy in men with clinically localized prostate cancer. JAMA 1996; 276: 615–619. | Article | PubMed | ISI | ChemPort |
17. Hull GW et al. Cancer control with radical prostatectomy alone in 1000 consecutive patients. J Urol 2002; 167: 528–534. | Article | PubMed | ISI |
18. Egawa S et al. Observations of prostate specific antigen doubling time in Japanese patients with nonmetastatic prostate carcinoma. Cancer 1999; 86: 463–469. | Article | PubMed | ChemPort |
19. Bianco Jr FJ et al. Radical prostatectomy nomograms in black American men: accuracy and applicability. J Urol 2003; 170: 73–77. | Article | PubMed |

Acknowledgements

We thank WA Thomasson, PhD, for expert editorial assistance.