Fax: (507) 284-0079
The relevance of prostatectomy findings for brachytherapy selection in patients with localized prostate carcinoma
Version of Record online: 23 JUL 2002
Copyright © 2002 American Cancer Society
Volume 95, Issue 3, pages 513–519, 1 August 2002
How to Cite
Pisansky, T. M., Blute, M. L., Hillman, D. W., Davis, B. J., Haddock, M. G., Suman, V. J., Wilson, T. M. and Zincke, H. (2002), The relevance of prostatectomy findings for brachytherapy selection in patients with localized prostate carcinoma. Cancer, 95: 513–519. doi: 10.1002/cncr.10698
- Issue online: 23 JUL 2002
- Version of Record online: 23 JUL 2002
- Manuscript Accepted: 11 MAR 2002
- Manuscript Revised: 17 JAN 2002
- Manuscript Received: 29 OCT 2001
- lymph nodes;
- biologic models;
- serum prostate specific antigen;
- prostatic neoplasms;
- seminal vesicles
The efficacy of brachytherapy for patients with localized prostate carcinoma depends on adequate radiotherapeutic coverage of the primary tumor and its subclinical extraprostatic extensions. Predictive models based on pretherapy factors may be useful to estimate the likelihood for clinically relevant extraprostatic disease and may be incorporated into selection criteria for this procedure.
Multivariate logistic regression model building was performed using pretherapy factors in 2905 surgically staged patients with localized prostate carcinoma to estimate the probability of seminal vesicle and/or lymph node involvement. Bootstrap methods were employed to assess the stability of the final model parameters and to determine the sensitivity and specificity of the final model.
Clinical tumor classification, biopsy Gleason score groupings, and serum prostate specific antigen (PSA) levels were associated with seminal vesicle and/or pelvic lymph node involvement. These factors were incorporated into a multivariate model that predicted for these adverse histopathologic features. Allowing for up to a 10% likelihood for seminal vesicle and/or pelvic lymph node involvement, patients with tumors classified as T1c–T2a, Gleason scores of 2–6, and PSA ≤ 16 ng/mL; or with tumors classified as T1c–T2a, Gleason scores of 7–10, and PSA ≤ 4 ng/mL; or with tumors classified as T2b–T2c, Gleason scores of 2–6, and PSA ≤ 6 ng/mL would be potential candidates for brachytherapy alone.
The predictive model presented may provide criteria whereby an adequately performed prostate brachytherapy procedure is expected to encompass the intraprostatic and adjacent extraprostatic disease. Prostate brachytherapy alone may be considered in these circumstances, whereas the addition of external beam radiotherapy may be reserved for patients with disease that is apt to extend beyond the brachytherapy target volume. Cancer 2002;95:513–9. © 2002 American Cancer Society.
The ideal management of patients with clinically localized prostate carcinoma remains a subject of controversy. Although external beam radiotherapy and radical prostatectomy have been the conventional treatment options, prostate brachytherapy (also referred to as seed implantation) has gained increased acceptance in recent years. Although factors like prostate volume and urinary symptomatology may affect the patient selection process, the American Brachytherapy Society considers that patients with a high likelihood of having disease that is well encompassed by implants are candidates for seed implantation alone.1 The solitary presence of extracapsular tumor extension does not preclude prostate brachytherapy as sole treatment,1–4 although seminal vesicle or lymph node involvement does. Therefore, patients with tumor (T) classifications of T1–T2a,5 with prostate specific antigen (PSA) levels < 10 ng/mL, and with biopsy Gleason scores ≤ 6 have been deemed appropriate for brachytherapy alone.1, 6 In contrast, the addition of external beam radiotherapy to brachytherapy has been advocated for patients with T2c–T3 tumors,5 with biopsy Gleason scores of 8–10, or with PSA > 20 ng/mL.1 The preferred radiotherapeutic approach for patients with disease-related characteristics that lie between these two groupings is uncertain. However, modification of the American Brachytherapy Society criteria may be necessary as new information is made known.1
The determination of tumor classification, histologic grade of the biopsy, and serum PSA level is an established component of the evaluation of patients with a newly established diagnosis of prostate carcinoma.7 These factors are associated with tumor involvement of extraprostatic sites that may not be included within an otherwise satisfactory prostate brachytherapy procedure.8–10 Although tumor involvement of the fibroadipose tissues immediately surrounding the prostate is common,10 this component of extraprostatic disease often is confined to the surgical specimen2 and may be encompassed by a tumoricidal brachytherapy dose.3 However, tumor cells present in the seminal vesicles or regional lymph nodes are not treated effectively with prostate brachytherapy alone. Therefore, the pretherapy prediction of tumor involvement of these structures is important in defining the criteria used to select patients for prostate brachytherapy or for further pretherapy evaluation.
MATERIALS AND METHODS
Between January, 1994 and December, 1997, 3195 consecutive patients with clinically organ-confined (T1–T2,N0–NX,M0)5 prostatic adenocarcinoma underwent radical prostatectomy with pelvic lymph node dissection at the Mayo Clinic in Rochester, Minnesota (Mayo Clinic-Rochester). However, 290 patients (9%) with incomplete diagnostic biopsy information, with unknown postprostatectomy seminal vesicle or pelvic lymph node status, or who received preoperative hormonal therapy or radiotherapy were excluded. Thus, the study group consisted of 2905 patients.
The diagnosis was established by needle biopsy (2862 patients) or through transurethral resection (43 patients). The primary tumor classification was assigned solely by digital rectal examination findings and was performed according to a minor modification of the 1992 American Joint Committee on Cancer Staging System5 as follows: T1a was defined as clinically inapparent primary Gleason Grade 1, 2, or 3 tumor in ≤ 5% of the resection specimen and/or < 1 cm3 tumor volume; T1b was defined as incidental, nonpalpable carcinoma in > 5% of the resection specimen, and/or ≥ 1 cm3 tumor volume, and/or primary Gleason Grade 4 or 5; T1c was defined as nonpalpable tumor (with or without findings on transrectal ultrasound) identified on needle biopsy performed due to an elevated serum PSA level; T2a was defined as palpable tumor confined to one-half or less of a prostatic lobe; T2b was defined as palpable tumor confined to more than one-half of one prostatic lobe; and T2c was defined as palpable tumor confined to the prostate with involvement of both lobes.
A low midline abdominal extraperitoneal incision was made, and a modified bilateral pelvic lymph node dissection was performed.11 The margins of resection were as follows: lateral, junction between external iliac artery and vein; inferior, femoral canal; posterior, obturator nerve and vasculature; and superior, bifurcation of the common iliac vessels. Thereafter, the endopelvic fascia was incised, and radical prostatectomy was accomplished in standard fashion.
All diagnostic prostatic biopsy specimens were reviewed at the Mayo Clinic-Rochester, and histologic grade was based on glandular differentiation in association with the pattern of stromal tumor growth, as described by Gleason.12 The Gleason score was obtained by multiplying the primary Gleason grade by 2 when a secondary grade was not identified (92 patients).
The radical prostatectomy specimen was promptly submitted and measured, the external surface was inked, and the specimen was placed in neutral buffered formalin for overnight fixation. The apical and bladder base portions initially were removed, and the remainder of the prostate was sectioned at 3–5 mm intervals perpendicular to the longitudinal (apical-basal) axis. The seminal vesicles were sectioned in a similar manner parallel to their junction with the prostate. Histologic examination was performed to determine tumor size and location, with special attention to capsular invasion, capsular perforation, surgical margin status, and seminal vesicle invasion. Pelvic lymphadenectomy tissues were submitted separately were and handled as frozen sections. The fibroadipose tissue was dissected, and all identified lymph nodes were removed, sectioned, and histologically examined for the presence and extent of tumor involvement. The numbers of lymph nodes involved and the numbers identified were reported with respect to laterality.
Method of Analysis
The objective of this study was to develop a model based on pretherapy factors that predicted the likelihood of tumor involvement of the seminal vesicles and/or pelvic lymph nodes. To avoid an overly optimistic view of the model's performance, the study population from each year was randomly divided in half, and the model was developed on 50% of the patients (developmental data set). The remaining patients were used to assess the performance of the developed model (examination data set).
Univariate logistic regression analysis was used to assess whether the likelihood of seminal vesicle or lymph node involvement differed with respect to patient age, clinical tumor classification, Gleason primary grade, Gleason secondary grade, Gleason score, prior transurethral prostatic resection, number of examined lymph nodes, or pretherapy PSA level in the developmental data set. These factors were examined as polychotomous variables represented by design variables in the model. Age and PSA level also were examined as continuous variables. The loge scale for PSA values was used in all logistic regression analyses, because they exhibited a symmetric, near normal distribution. Multivariate logistic regression13 model building was performed with several different sets of pretherapy factors in the developmental data. Each contained the pretherapy PSA level, Gleason score, clinical tumor classification, patient age, and history of transurethral prostatic resection. The pretherapy PSA level appeared as a continuous variable on the loge scale or as four levels (< 4.0 ng/mL, 4.0–9.9 ng/mL, 10.0–19.9 ng/mL, and ≥ 20 ng/mL). Clinical tumor classification (T1 vs. T2, T1a–T2a vs. T2b vs. T2c, and T1a–T2a vs. T2b–T2c) and Gleason score (2–4 vs. 5–6 vs. 7 vs. 8–10, 2–6 vs. 7 vs. 8–10, and 2–6 vs. 7–10) appeared as one of three different representations. Primary Gleason grade alone (Grade 1–2 vs. 3 vs. Grade 4–5) also was considered, and age was a dichotomous variable split at 60 years. Forward selection, backward elimination, and the score method model selection procedures were used, as were classification and regression tree analysis. The residual plots from each potential model were examined. The examination data set was then used to assess the performance of each potential model found with the developmental data set. Receiver operating characteristic (ROC) curves14 were constructed, and the most parsimonious model was chosen. Thus, the data were combined, the data from the entire cohort were fit to the final model, and parameter estimates were obtained. The estimated probability that a patient would have seminal vesicle and/or lymph node involvement using these parameter estimates will be referred to as the likelihood.
The stability of the final model parameters was examined using bootstrap methods.15 One thousand bootstrap samples of 2905 patients from the entire cohort were generated, and the final model was fit to each sample data set. The distribution of each parameter estimate was examined and was stable. Bootstrap methods also were used to assess the ability of the final model to classify patients properly. One thousand bootstrap samples of 2905 patients were generated from the entire cohort, and the likelihood was calculated for each data point. For each bootstrap sample and prespecified cut-off point, the sensitivity and specificity associated with this final model were determined. For each cut-off point, the median and the 2.5th and 97.5th percentile values for sensitivity and specificity were determined, as were the percent of false negatives.
In the entire study population, 197 patients (7%) had tumor involvement of the seminal vesicles only, and 105 patients (4%) had lymph node involvement with (61 patients; 2%) or without (44 patients; 2%) seminal vesicle invasion. Clinical tumor classification, biopsy Gleason score, serum PSA level, and patient age were associated univariately with seminal vesicle and/or lymph node involvement, as shown in Table 1. Model building results from the developmental data set, assessment of performance with the examination data set, and the stability exhibited in model parameters using bootstrap methods led to a model that included tumor classification (T1a–T2a vs. T2b–T2c), Gleason score (2–6 vs. 7–10), and PSA value (Table 2). The likelihood for tumor involvement of seminal vesicles and/or pelvic lymph nodes is expressed by the following equation:
where T1–T2a tumor classification risk = 0, T2b–2Tc tumor classification risk = 1, Gleason score 2–6 risk = 0, and Gleason score 7–10 risk = 1.
|Factor||No. of patients (%)||Seminal vesicle invasion No. (%)||Lymph node involvement No. (%)||Seminal vesicle or lymph nodel involvement No. (%)||P valuea|
|All patientsb||2905 (100)||258 (9)||105 (4)||302 (10)|
|T1a||21 (1)||0 (0)||0 (0)||0 (0)|
|T1b||22 (1)||1 (5)||0 (0)||1 (5)|
|T1c||1117 (38)||47 (4)||19 (2)||58 (5)||< 0.001|
|T2a||587 (20)||29 (5)||11 (2)||36 (6)|
|T2b||767 (26)||106 (14)||46 (6)||122 (16)|
|T2c||391 (13)||75 (19)||29 (7)||85 (22)|
|2–4||318 (11)||10 (3)||5 (2)||11 (3)|
|5–6||1849 (64)||98 (5)||45 (2)||119 (6)|
|7||592 (20)||113 (19)||46 (8)||131 (22)||< 0.001|
|8–10||125 (4)||37 (30)||9 (7)||41 (33)|
|Unknown||21 (1)||0 (0)||0 (0)||0 (0)|
|< 4.0||516 (18)||15 (3)||6 (1)||18 (3)|
|4.0–9.9||1697 (58)||118 (7)||34 (2)||136 (8)|
|10.0–19.9||513 (18)||66 (13)||33 (6)||79 (15)||< 0.001|
|20.0–49.9||155 (5)||48 (31)||27 (17)||57 (37)|
|≥ 50||17 (1)||11 (65)||5 (29)||12 (71)|
|Unknown||7 (0)||0 (0)||0 (0)||0 (0)|
|< 50||77 (3)||4 (5)||1 (1)||4 (5)|
|50–59||699 (24)||43 (6)||23 (3)||55 (8)|
|60–69||1565 (54)||146 (9)||64 (4)||171 (11)||0.004|
|70–79||564 (19)||65 (12)||17 (3)||72 (13)|
|Factor||Odds ratioa||95% CI||P value|
|Gleason score 7–10||3.37||2.59–4.39||< 0.001|
|PSA (loge)||2.65||2.19–3.19||< 0.001|
Graphic displays of the results from fitting the model to the entire study cohort are presented in Figures 1 and 2. For example, a patient with clinically classified T1c disease, a diagnostic prostatic tumor biopsy Gleason score of 6, and a pretherapy serum PSA value of 10 ng/mL would have a 5% risk (95% confidence interval, 4–6%) of seminal vesicle and/or lymph node involvement if surgical staging was considered.
The ability of the model to predict seminal vesicle and lymph node involvement was evaluated by determining the sensitivity and specificity for various prespecified likelihood values in 1000 bootstrap samples, as shown in Table 3. Those patients who were misclassified as free of seminal vesicle and/or lymph node involvement (false negatives) are also presented. For example, among the patients who actually had seminal vesicle and/or lymph node involvement, 11% had a predicted likelihood of < 5%. The percent of false negative results with only seminal vesicle invasion also was examined. For each predicted likelihood examined, the percent of false positive results with only seminal vesicle invasion was at least 67%, which represented the proportion of patients in the study cohort with seminal vesicle invasion only. Table 4 provides combinations of clinical tumor classification, Gleason score, and PSA values that resulted in the estimated likelihood, with the upper limit of the 95% confidence interval residing below a given threshold value.
|Predicted likelihood (%)||Sensitivity, %||Specificity, %||False negatives, %||Seminal vesicle only false negatives, %|
|≥ 5||89 (85–92)||50 (42–56)||11 (8–15)||8 (6–12)|
|≥ 10||72 (66–78)||73 (68–77)||28 (22–34)||19 (14–25)|
|≥ 15||58 (51–64)||83 (81–86)||42 (36–49)||29 (23–35)|
|≥ 20||49 (41–56)||89 (87–91)||51 (44–59)||35 (29–41)|
|Likelihood (%)||No. of patients (%)||T1–T2a||T2b–T2c|
|Gleason score 2–6||Gleason score 7–10||Gleason score 2–6||Gleason score 7–10|
|≤ 5||1293 (45)||10 (8)||3 (2)||4 (2)||1 (0)|
|≤ 10||1967 (68)||21 (16)||6 (4)||8 (6)||2 (1)|
|≤ 15||2275 (78)||34 (25)||10 (7)||12 (10)||3 (2)|
|≤ 20||2437 (84)||49 (34)||14 (11)||17 (14)||5 (4)|
Currently, it is possible to estimate the likelihood for certain adverse histologic findings in patients with newly diagnosed prostate carcinoma.8–10 This information may be used to select patients for radical prostatectomy7, 10 or to assist in the design of external beam radiotherapy treatment fields.8, 9 Although studies like these were applied to prostate brachytherapy selection criteria,1 they were not conducted to specifically address issues unique to this procedure. Selecting patients who are likely to have organ-confined disease10 may be unduly restrictive when considering prostate brachytherapy alone as a management option.1 For example, we previously observed that tumor extension into the immediately surrounding periprostatic tissues may be readily included readily within an effective radiotherapeutic dose level using conventional prostate brachytherapy treatment-planning parameters.2, 3 These observations were confirmed by other authors,4 and they indicate that tumor rarely extends more that a few millimeters away from the edge of the prostate into the periprostatic tissues. Consequently, a satisfactorily performed brachytherapy procedure is unlikely to “miss” this component of extraprostatic disease. Therefore, an accurate assessment of disease spread to the seminal vesicles or regional lymph nodes is the primary factor in determining whether a patient is an otherwise suitable candidate for prostate brachytherapy. That is, patients with a “high risk” for disease spread to these structures should not be considered for brachytherapy as sole treatment. With these considerations in mind, a properly validated predictive model was developed to identify these “high-risk” patients and (conversely) to identify patients who were apt to have the primary tumor site adequately encompassed by brachytherapy alone.
Multivariate analysis demonstrated that T classification determined by digital rectal examination, Gleason score of the diagnostic specimen, and serum PSA level was associated with tumor involvement of the seminal vesicles or of the pelvic lymph nodes in this study group, as noted previously.8–10 These pretherapy factors are particularly relevant elements for inclusion in the predictive model developed in this investigation, because they are the basic elements of the initial evaluation of any patient with prostate carcinoma.7 At the beginning of our model-development process, several representations of T classification, Gleason score, and PSA level were examined. However, the model was simplified greatly by identifying those factors that could be grouped together without sacrificing predictive accuracy. For example, multivariate models with Gleason score classified as 2–6 versus 7 versus 8–10 and 2–6 versus 7–10 were compared, and essentially no additional predictive accuracy was identified in the more complex Gleason score categorization (the area under the ROC curve was 0.793 in both models). This resulted in four combinations of T classification (T1–T2a vs. T2b–T2c)5 and Gleason score (2–6 vs. 7–10) while retaining the value of considering PSA level as a continuous variable predictor.16 Thus, the relevant information was distilled into two likelihood estimate plots (Figs. 1, 2) that can be applied reliably to the new patient in whom prostate brachytherapy is a consideration.
Different criteria were used during the past decade to select patients for prostate brachytherapy as a sole treatment method for clinically organ-confined disease.17 Some investigators promoted the view that supplemental external beam radiotherapy was of benefit for nearly all patients undergoing prostate brachytherapy,18 whereas others significantly limited its use.19 In the context of such varying perspectives, the American Brachytherapy Society suggested that patients with a high probability of organ-confined disease, as determined by studies available at that time,10 were appropriate candidates for brachytherapy alone.1 Thus, the combination of clinical classifications T1c–T2a, diagnostic specimen Gleason scores of 2–6, and pretherapy serum PSA levels < 10 ng/mL were proposed as sound selection criteria for this procedure, despite the recognition that up to 50% of patients who fulfill these criteria have some degree of extraprostatic tumor extension.10 Despite this high risk for extraprostatic tumor extension, prostate brachytherapy as sole treatment may be highly effective.20–22 This suggests that extraprostatic tumor extension, per se, may not be a contraindication to prostate brachytherapy but that the specific site (periprostatic fibroadipose tissue vs. seminal vesicles vs. regional lymph nodes) of extraprostatic disease carries greater importance for brachytherapy.
In concert with our prior studies,2, 3 the current report contributes to an understanding of extraprostatic tumor spread in patients with clinical Stage II prostate carcinoma and to its implications for the radiotherapeutic management of patients with this condition. When taken together, these observations provide a solid foundation for prostate brachytherapy selection criteria. The information provided herein directly relates to the identification of patients in whom the intrapelvic component of the neoplastic process is likely to be included fully within the brachytherapy target volume, thus reducing the chance for a “geographic miss.” Once it is determined that a particular patient's disease extent is likely to be encompassed by the intended brachytherapy dose, then measures to improve overall outcome can be addressed through appropriate avenues of investigation.
An examination of the likelihood estimate plots (Figs. 1, 2) and of Table 4 raises the prospects for modifying the American Brachytherapy Society selection criteria,1 which were based largely on the work of Partin et al.10 Although the definition of an “acceptable” risk level is open to interpretation, a likelihood ≤ 10% for seminal vesicle and/or pelvic lymph node involvement is well within the guidelines used for external beam radiotherapy field design.8, 9 A conservative revision of the American Brachytherapy Society guidelines that takes into account the confidence limits of the predictive model may provide selection criteria for prostate brachytherapy alone in patients with the following combinations of tumor-related characteristics: 1) tumor classification T1c–T2a, Gleason scores 2–6, and PSA ≤ 16 ng/mL; 2) tumor classification T1c–T2a, Gleason scores 7–10, and PSA ≤ 4 ng/mL; and 3) tumor classifications T2b–T2c, Gleason scores 2–6, and PSA ≤ 6 ng/mL. Implementation of these guidelines would affect substantially the proportion of patients in whom brachytherapy may be considered as the sole treatment method. Because approximately two-thirds of patients with false negative results (for seminal vesicle and/or lymph node involvement) in the current study cohort would have seminal vesicle invasion without lymph node involvement, it would be possible to evaluate this risk with further imaging studies,23 or by performing image-guided biopsy of the seminal vesicles24 in selective situations. It appears that forms of therapy other than brachytherapy alone are warranted for patients with T2b–T2c disease when Gleason scores of 7–10 are identified in the biopsy specimen. However, the information provided herein should be used with caution in patients with Gleason scores of 8–10 disease due to the relatively small number of such patients available for analysis. We also encourage continued assessment of these criteria as new information becomes available.
The selection criteria put forth in this report also may serve as a foundation on which future prostate carcinoma cancer research efforts may be based. In a first step, investigators with access to study populations treated solely with prostate brachytherapy may investigate the clinical outcome of patients with the selection criteria put forth in this report. Furthermore, appropriate study populations may be identified by application of the risk estimates provided by the predictive model put forth herein, and clinical studies may be fashioned to address issues germane to certain patient subsets.25 For example, patients with a substantial risk for extraprostatic disease limited to the immediate periprostatic area may be suitable candidates for the investigation of external beam radiotherapy and/or androgen suppression combined with prostate brachytherapy. In contrast, patients with larger intraprostatic tumors may be considered for androgen suppression as a cytoreductive measure in conjunction with prostate brachytherapy26 or for dose-escalation approaches. In this manner, large-scale, comparative clinical trials can focus on selected patient groups to explore relevant therapeutic strategies that hold the potential to identify improvements in patient care.
The authors express their gratitude to Ms. Sandra K. Martin for assistance with data collection and to Mr. Jeffrey M. Slezak for statistical support.
- 5American Joint Committee on Cancer. Manual for staging of cancer. Philadelphia: JB Lippincott, 1992.
- 7NCCN practice guidelines for prostate cancer. Oncology. 2000; 14: ( 11A Suppl) 111–119., , , et al.
- 11Pelvic lymphadenectomy. In: CrawfordED, DasS, editors. Current genitourinary cancer surgery. Philadelphia: Lea and Febiger, 1990: 162–169., .
- 12Histologic grading of prostatic carcinoma. In: BostwickDG, editor. Pathology of the prostate. New York: Churchill Livingstone, 1990: 83–93..
- 15An introduction to the bootstrap. New York: Chapman & Hall, 1993., .