Ki-67 Staining Is an Independent Correlate of Biochemical Failure in Prostate Cancer Treated with Radiotherapy¹

The clinical progression of prostate cancer is slow, and this is reflected in various proliferation markers. Pretreatment serum PSA⁵ doubling is usually >12 months (1, 2), as is tumor doubling time based on cell kinetic parameters (3). Likewise, the in situ nucleoside analogue labeling index is usually <6% (4, 5, 6), and potential doubling time is most often >24 days (6). Flow cytometric analysis, as well as image analysis, has shown that most tumors are DNA diploid (7) and are composed of <5% S-phase cells (8, 9). The Ki67-LI is another proliferation marker that is determined via a rapid, simple immunohistochemical method. The Ki67-LI, measured using MIB-1 antibody, provides an accurate estimate of growth fraction (10, 11, 12, 13) and in many studies has been found to be a predictor of outcome for patients treated with radical prostatectomy (14, 15, 16, 17, 18, 19, 20, 21, 22). Preliminary data indicate that this static approximation of growth fraction is more strongly associated with patient outcome than DNA ploidy (23). However, few studies have investigated whether Ki67-LI is an independent predictor of prostate cancer patient outcome after treatment with radiotherapy in the PSA era (21). In this report, the Ki67-LI from pretreatment biopsy specimens in 106 prostate cancer patients treated with radiotherapy in the PSA era was established to be a strong independent correlate of biochemical failure.

Archival paraffin-embedded pretreatment prostate cancer tissue was available in 106 patients treated with external beam radiotherapy from 1987 to 1993. Every patient had a pretreatment PSA. No patient received androgen ablation neoadjuvantly or adjuvantly, underwent lymph node dissection for staging, or underwent radical prostatectomy. The diagnosis of prostate cancer was made by TURP in 42 patients (40%) and by ultrasound-guided transrectal prostate needle biopsies in 64 patients (60%).

Median and mean pretreatment PSA levels were 7.6 and 11.8 ng/ml for all patients, 9.8 and 13.7 ng/ml for the needle biopsy group, and 4.1 and 8.8 ng/ml for the TURP group. Median follow-up was 62 months for the entire group, 62 months for the needle biopsy group, and 61 months for the TURP group. There were 74 patients with stage T₁/T₂ disease and 32 patients with T₃/T₄ (only one had T₄) disease.

The monoclonal antibody, MIB-1 (Immunotech, SA MAC, Inc., Germany), was used to determine the proportion of tumor cells staining positive for Ki-67, Ki67-LI. As described previously (23), slide-mounted, paraffin-embedded prostatic tissue sections were deparaffinized in xylene, rehydrated sequentially in ethanol (100, 90, and 70%) and placed into a 1% phosphate-buffered solution (PBS, pH 7.4). The sections were then heated in a conventional 600 W microwave oven at maximum power for 3 × 5 min. The sections were left at room temperature for 40 min, and 2% normal horse serum was added to block nonspecific protein binding. The sections were incubated with MIB-1 antibody (1:50 dilution) overnight at 4°C in a humidified chamber. Detection of the bound MIB-1 antibody involved applying the Vectastain Elite ABC reagents (Vector Laboratories, Inc., Burlingame, CA) using avidin DH:biotinylated horseradish peroxidase H complex with 3,3′-diaminobenzidine (Polysciences, Inc., Warington, PA) and Mayer’s hematoxylin (Fisher Scientific, Fair Lawn, NJ). Appropriate positive controls (HeLa cells) were included in each immunohistochemical run to verify the specificity of MIB-1 and negative controls were produced by substituting the primary antibody with PBS in duplicate sections.

Needle biopsy and TURP sections were reviewed by the study pathologist (P. T.) and graded according to the Gleason system. Sections representative of the tumor with the highest grade were selected for immunohistochemical analysis. When possible, 2000 tumor cells were counted for the determination of Ki67-LI. Any nuclear staining, regardless of intensity, was considered positive for MIB-1. The Ki67-LI was expressed as a percentage of immunoreactive tumor cells to the total counted tumor cells. Two of the investigators (V. K. and D. C.) independently scored the slides without any prior knowledge of the patient data or treatment-related outcomes. The mean (±SE) Ki67-LIs for the two counts were 2.5 ± 0.3% (±SE) and 3.8 ± 0.3% and were statistically different (Student’s t test and Wilcoxon signed ranks test, for paired samples). The averages of these independent counts were used for the analyses.

The χ² test was used to assess the significance of differences between proportions (24). Nonparametric comparisons between independent groups were performed using the Mann-Whitney test. Kaplan-Meier curves were calculated from the completion of radiotherapy, with tests of statistical significance based on the log-rank statistic (25). Biochemical failure was defined as three PSA rises on follow-up (26). The onset of a rising PSA was defined as the average time between the date of the PSA obtained prior to the first risen value and the date of the first risen value.

The mean Ki67-LI was 3.2%, and there was no statistically significant difference between the mean Ki67-LIs from the TURP and needle biopsy specimens (Table 1). For the entire group, as well as for those diagnosed by TURP, mean Ki67-LI was significantly higher in the presence of T₃/T₄ and Gleason score 7–10 disease and when there was evidence of biochemical failure. Also, no correlation was seen between pretreatment PSA and Ki67-LI for the entire cohort; however, a significant relationship was observed for those diagnosed from TURP. For patients diagnosed by needle biopsy, no associations between Ki67-LI and the other pretreatment prognostic factors were discerned. The reason for the differences in the relationships of Ki67-LI in the patients diagnosed by needle biopsy and TURP is uncertain but could be based on inherent biological divergence or technical factors, such as the amount of tissue available for analysis. One correlation that was evident for patients diagnosed by needle biopsy, TURP, or for the entire group was between Ki67-LI and treatment failure. The mean Ki67-LI was significantly higher for those manifesting a rising PSA after radiotherapy. Radiotherapy dose has been shown to be a determinant of outcome (27) and therefore was investigated here. There was no relationship between dose and mean Ki67-LI.

Previously, we found (23) that, for patients diagnosed with prostate cancer by TURP, a Ki67-LI >3.5% was associated with a poor prognosis. Table 2 shows that the correlations between dichotomized Ki67-LI and stage, Gleason score, pretreatment PSA, radiotherapy dose, and treatment failure were the same as for mean Ki67-LI in Table 1.

Fig. 1 shows the Kaplan-Meier freedom from biochemical failure (bNED) survival analysis for the entire cohort, the patients diagnosed by TURP, and those diagnosed by needle biopsy. There was no difference statistically between the two diagnostic groups. The univariate 5-year bNED survival results for the factors associated with patient outcome are shown in Table 3. Stage T₃/T₄, Gleason score 7–10, pretreatment PSA >10 ng/ml, and Ki67-LI >3.5% predicted strongly for reduced bNED rates. Radiotherapy dose did not correlate with bNED in this cohort. Fig. 2 displays the bNED survival curves for dichotomized Ki67-LI, subdivided by diagnostic group. A high Ki67-LI consistently was associated with a lower bNED rate, independent of whether diagnosis was based on TURP or needle biopsy. The results of Cox proportional hazards regression for bNED are shown in Table 4. Ki67-LI was a highly significant correlate of bNED, along with pretreatment PSA, stage, and Gleason score.

The relationship of Ki67-LI to freedom from distant metastasis was also examined to determine whether the reduction in bNED associated with a high Ki67-LI was attributable to metastasis. The absolute percentage of patients with distant metastasis and nodal metastasis identified in the follow-up period was 4% (n = 4) and 3% (n = 3), respectively. Fig. 3 shows that 98 and 94% were free of distant metastasis by Kaplan-Meier analysis when the Ki67-LI was low and high, respectively. This difference was not significant. Of the 37 patients that had a rising PSA, 20 were investigated by prostate biopsy and imaging, and 13 were found to have local disease persistence, 1 with concurrent nodal and distant spread. Therefore, the initial rise in PSA appears to be attributable to incomplete eradication of local disease in most cases.

The Ki67-LI by immunohistochemical staining provides a noninvasive, relatively rapid determination of growth fraction, which has prognostic value. In the vast majority of reports (14, 16, 18, 19, 20, 21), Ki67-LI has been observed to be a correlate of biochemical and/or disease outcome for patients with prostate cancer. Table 5 summarizes a number of contemporary series (14, 16, 18, 21, 28) showing that in the majority Ki-67 immunostaining was also independent of other prognostic factors in multivariate analysis.

The predictive usefulness of immunohistochemical Ki-67 staining of diagnostic tissue for patients treated with radiotherapy has only been reported by one other group (21). Scalzo et al. (21) classified Ki-67 staining into low and high groups based on the number positive cells per high powered microscopic field. Although the classification of Ki-67 staining by the number per high powered field is less exacting than the quantification of labeling index, they found a correlation of Ki-67 staining with biochemical failure in univariate and multivariate analyses. Our data in 106 cases also demonstrated that Ki-67 is a significant predictor of biochemical relapse in patients with clinically localized prostate cancer treated with radiotherapy. The Ki67-LI cutpoint of 3.5% that we used was taken from a prior analysis of the patients diagnosed by TURP (23). In the analysis described here, this Ki67-LI cutpoint was also strongly associated with biochemical failure in patients diagnosed from needle biopsy tissue. The pooled TURP and needle biopsy Ki67-LI analysis was of sufficient power to document that Ki67-LI is independent of pretreatment PSA, Gleason score, and stage as a correlate of a rising PSA after radiotherapy.

The series displayed in Table 5 included patients followed for progression after observation (deferred treatment) or androgen ablation (18, 19), radical prostatectomy (14, 15, 16, 17, 20, 22, 28), or radiotherapy (21). The predictive merit of Ki-67 immunohistochemical staining appears to be unaffected by the treatment used. Factors of concern in the application of Ki67-LI clinically are interobserver variability in the estimation of Ki67-LI and the way the data are categorized. The two investigators that quantified the staining in our study had slightly, but statistically, different estimates of Ki-67-LI on a case-by-case basis. Disparity in data categorization is also evident; Scalzo et al. (21) and Kallakury et al. (22) used the number of positive cells per high powered field, whereas the quantification of Ki67-LI is the more typical and reproducible method. Moreover, there has been considerable variation in the Ki67-LI cutpoints used to assess failure risk. Ki67-LI cutpoints have ranged from 1% (17) to 25% (15). These cutpoint differences are most likely reflective of median Ki67-LI differences. Because inconsistency in Ki67-LI cutpoints is apparent among patients treated by radical prostatectomy, the median differences do not appear to be solely a consequence of inherent attributes of the patient populations examined. A number of other technical variables might contribute to this diversity in absolute Ki67-LI levels, such as loss of Ki-67 antigen staining with storage (29), monoclonal antibody, and classification of positive staining (interobserver variation). Stricter standardization of the method needs to be established before widespread clinical application is feasible.

In conclusion, Ki67-LI is strongly associated with biochemical relapse after radiotherapy. A high Ki67-LI, which was defined here as >3.5%, resulted in a bNED rate of only 33% versus 76% for those with a low Ki67-LI. Our data suggest that local disease persistence most commonly accounts for a rising PSA after radiotherapy. Because prostate cancer proliferation rates are exceptionally low when compared with those of other tumor sites, even in the high Ki67-LI group, it is unlikely that accelerated tumor repopulation is responsible for the presumed resistance to radiotherapy. The more probable mechanism is that a high Ki67-LI is associated with intrinsic tumor aggressiveness and radioresistance. Abnormalities in the expression of p53 (30), MDM2 (31), and other cell cycle-related proteins are associated with high Ki67-LI in diagnostic biopsy material, and Ki-67-LI is elevated in locally recurrent prostate cancer after primary radiotherapy (32, 33). Routine clinical use of Ki67-LI remains hampered by the lack of standardization in the immunohistochemical assays, quantification methods, and cutpoints used for risk classification.

We thank Kuriakose Abraham, Department of Experimental Radiation Oncology, for preparation of the histological material.