Digital Rectal Examination Remains a Key Prognostic Tool for Prostate Cancer: A National Cancer Database Review

Background

Adenocarcinoma of the prostate is the leading noncutaneous cancer diagnosis and the second leading cause of cancer mortality for men in the United States.^1,2 Patient outcomes are highly variable, and therefore risk stratification is critical in personalizing treatment to individual risk profiles. Risk stratification systems for prostate cancer have evolved over the past decades. In 1956, Dr. Willet Whitmore introduced the first prostate cancer classification system, which was subsequently revised in 1975 by Dr. Hugh Jewett.³ The Whitmore-Jewett system was succeeded by the TNM staging system for prostate cancer, first formulated in 1987 and implemented by the AJCC in the United States in 1992.^4,5

Pathologic T2 (pT2) subclassifications have been shown to lack prognostic value.^6,7 Accordingly, the newest 8th edition of the AJCC Cancer Staging Manual condenses the previous pT2 subclassifications into a single classification, “T2: organ-confined,” such that the pT2 category is no longer subclassified by extent of involvement or laterality.⁸ In contrast, the clinical tumor staging system has changed little over the decades, despite many advances in other aspects of clinical evaluation and treatment of prostate cancer (Table 1).

Table 1.

Comparison of Historical and Current Prostate Cancer Clinical T Classification

View Table

Since the initial implementation of the prostate cancer TNM staging system in 1992, revisions have been published in the 5th through 8th editions of the AJCC Cancer staging Manual.^8,9 An effort was made in the 1997 edition to improve the T2 classification by merging tumors occupying less than half of one lobe and those confined to one lobe into a single T2a category,⁹ thus labeling tumors confined to the prostate but involving both lobes as T2b. However, a series of 1,755 patients with clinical stage T2 (cT2) prostate cancer treated with radical prostatectomy subsequently demonstrated on multivariate analysis that the 1992 but not the 1997 classification was prognostic for recurrence-free survival.¹⁰ The revision was thus reversed in the 2002 edition and has remained unchanged.⁹

Minimal data are available in the modern era regarding prognostic value of the prostate cancer cT2 subclassifications, as determined by digital rectal examination (DRE). It is known that DRE has significant interexaminer variability,¹¹ and patients often find DRE to be invasive and undesirable. Multiple publications have examined trends in DRE use, finding that it is an underused physical examination component and may no longer be considered part of the core curriculum at certain medical schools.^12–14 Investigation is therefore critical to establish the presence of prognostic value for this examination in contemporary patients with prostate cancer.

We sought to determine the prognostic value of the cT2 subclassifications (cT2a–c) in a large cohort of men with high-risk clinical T2N0M0 prostate cancer treated with curative intent. The high-risk population was selected for this study because the end point of interest available was overall survival (OS), and this cohort included a large number of patients with cT2 disease and a high event rate for the end point of interest. Our results demonstrated the prognostic significance of the cT2 staging classifications in a large, contemporary cohort of men treated for high-risk prostate cancer, with important clinical and cost-conscious implications for the role of DRE in an era of oncologic practice in which expensive diagnostic tests are increasingly available and physical examination may be declining.

Methods

Patients

The patient population in the current study was obtained from the prostate cancer Participant Use Data File (PUF) from the National Cancer Database (NCDB), which is a nationwide oncology outcomes database that captures roughly 70% of all newly diagnosed cancer cases in the United States by obtaining data from >1,500 accredited sites. The database is supported by the American College of Surgeons and the American Cancer Society. Data are collected and transmitted using standardized coding, with variables including patient demographics, detailed staging, type of treatment administered, and OS outcomes.¹⁵

The Medical College of Wisconsin was granted user status for the prostate cancer PUF, which includes all incident cases of prostate cancer reported to the NCDB from 2004 through 2010. PUFs are entirely deidentified, HIPAA-compliant data files and are only available via an application process to investigators affiliated with accredited cancer programs.¹⁵ Exempt status was obtained from the Medical College of Wisconsin Institutional Review Board for the current study because it did not meet criteria for human subject research.

There were 1,294,126 incident cases in the prostate PUF for 2004 through 2010. Of these, the subset of men with high-risk clinical T2N0M0 prostate cancer treated with curative-intent external-beam radiotherapy (EBRT) and androgen deprivation therapy (ADT) ± surgery was extracted (Figure 1). Disease risk was defined based on the risk groups defined in the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines), wherein high-risk disease is defined by the presence of any of the following clinical factors at diagnosis: Gleason score >7, prostate-specific antigen (PSA) level >20 ng/mL, or clinical stage T3a–T4.² Curative-intent therapy was filtered for based on radiation dose thresholds. Patients outside the age range of 18 to 90 years were excluded. The high-risk population was selected for this study because the end point of interest available in the PUF was OS, and this cohort included a high number of patients with cT2 disease and a high event rate. The final study population included 5,291 patients.

Patient selection.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 7; 10.6004/jnccn.2018.7278

View Full Size

Patient selection.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 7; 10.6004/jnccn.2018.7278

Patient selection.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 7; 10.6004/jnccn.2018.7278

Results

A total of 5,291 men with high-risk clinical T2N0M0 prostate cancer treated with EBRT and ADT ± surgery were included. Baseline patient demographics across the cT2 subclassifications are displayed in Table 2. A few demographic and clinical differences were noted among the different cT2 subclassification groups, supporting our controlling for these effects by including them subsequently in the multivariate model. Median age at diagnosis was 70 years (range, 38–90 years) for the entire cohort, with significantly younger age for men with higher cT2 subclassification (P<.001). Most patients were white (79%) and a minority were black (17%), with race not found to be associated with cT2 subclassification (P=.126). Patient medical insurance was primarily Medicare (62%) or private (29%), with private and Medicaid insurance being more common in men with a higher cT2 subclassification (P<.001). This was a generally healthy cohort, with a CDCI score of 0 in 88% of men, indicating no comorbid conditions recorded (P=.415).

Table 2.

Baseline Patient Characteristics

View Table

Baseline tumor characteristics across cT2 subclassifications (Table 2) were typical for a high-risk cohort. Gleason score was ≤7 in 19.5% and >7 in 80.6% of patients for whom Gleason score data were available (P=.064). PSA level at diagnosis was <10 ng/mL in 15.7% of patients, 10 to 20 ng/mL in 7.0%, and >20 ng/mL in 76.8% (P<.001), with the P value indicating higher PSA level at diagnosis for men with a higher cT2 subclassification. There were 1,368 patients with the cT2a subclassification, 1,380 with cT2b, and 2,543 with cT2c. Treatment consisted of EBRT + ADT in 95.9% and surgery + EBRT + ADT in 4.1% (P=.138).

At a median follow-up of 5.4 years, there were 308 (22.5%), 355 (25.7%), and 757 deaths (29.8%) in the cT2a, cT2b, and cT2c groups, respectively. In the bivariate and multivariate analyses, the cT2a, cT2b, and cT2c subclassifications were found to have prognostic significance, with increasing HRs for the outcome of death (Tables 3 and 4). Multivariate values were 1.00 (reference), 1.25 (95% CI, 1.07–1.45; P=.0046), and 1.43 (95% CI, 1.25–1.63; P<.0001), respectively. This finding was independent of other known prognostic variables included in the multivariate model. Kaplan-Meier analysis also demonstrated statistical significance (P<.0001) of cT2 subclassification for OS (Figure 2).

Table 3.

Bivariate Factors for Overall Survival

View Table

Table 4.

Multivariate Factors for Overall Survival

View Table

OS by cT2 classification.

Abbreviations: cT2, clinical T2; OS, overall survival.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 7; 10.6004/jnccn.2018.7278

View Full Size

OS by cT2 classification.

Abbreviations: cT2, clinical T2; OS, overall survival.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 7; 10.6004/jnccn.2018.7278

OS by cT2 classification.

Abbreviations: cT2, clinical T2; OS, overall survival.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 7; 10.6004/jnccn.2018.7278

Additional factors of prognostic significance for OS identified in the multivariate model included age group (P<.0001), insurance type (P=.0037), CDCI score (P<.0001), Gleason score (P=.0182), and PSA level at initial diagnosis (P=.0040). HRs and confidence intervals for these factors are listed in Table 4. For Gleason score subgroups, multivariate HRs were 1.00 for Gleason score 7 (reference), 0.72 for Gleason score ≤6 (95% CI, 0.31–1.68; P=.4432), 1.61 for Gleason score 8 to 10 (95% CI, 1.12–2.32; P=.0107), and 1.19 for Gleason score missing (95% CI, 0.88–1.61; P=.2605). On post hoc analysis of interaction between Gleason missingness and cT2 subclassification, no significant interaction was present (P=.3796; df=2). Factors assessed that did not meet the statistical significance threshold for OS on multivariate analysis were race (P=.0907) and treatment (P=.3062).

Discussion

Since its introduction in 1987, the TNM staging system for prostate cancer has changed little. pT2 subclassifications were recently revised in the 8th edition AJCC Cancer Staging Manual, because these had been shown to lack prognostic value.^6–8 Literature on the prognostic value of the cT2 subclassifications in a contemporary patient cohort is lacking. Risk stratification of prostate cancer is critical in guiding treatment decisions because prognoses can span a broad spectrum. A key component of the current NCCN risk stratification system is the clinical T stage as assessed by DRE,² and therefore it is important to establish the value of this parameter as a prognostic factor. Our study sought to evaluate the prognostic value of the prostate cancer cT2 subclassifications.

In the modern era, many advances have been made in the evaluation and treatment of prostate cancer. New biomarkers detected in urine are being evaluated in an effort to improve prostate cancer screening, including prostate cancer antigen 3 (PCA3) and TMPRSS2-ERG gene fusion.¹⁶ The 4Kscore has gained popularity as a prognostic tool, incorporating testing of kallikrein blood biomarkers along with clinical factors, including findings on DRE.¹⁷ Multiparametric MRI and MRI-guided prostate biopsy have become integral components of patient evaluation at many institutions.^18,19 Multiparametric MRI can be used to confirm low-risk disease before enrollment in active surveillance or to provide additional pretreatment information for patients with intermediate- to high-risk disease, and can help identify extracapsular extension and indicate aggressiveness of identified lesions.

In light of these new technically complex and expensive tests, it is particularly worthwhile to reevaluate traditional low-cost methods of disease assessment, such as DRE. Moreover, it is also prudent to investigate the utility and necessity of DRE, because patients often find it to be invasive and undesirable. Our series demonstrates that cT2 subclassification has statistically significant prognostic value for the outcome of OS. This finding was incrementally true for cT2a versus cT2b versus cT2c, and remained significant on multivariate analysis independent of other known prognostic variables, including PSA level and patient age, comorbidities, and race.

Our findings can be placed in the context of existing data that have supported current AJCC staging. In the 1997 edition of the AJCC Cancer Staging Manual, tumors occupying less than half of one prostate lobe and those confined to one lobe were merged into a single T2a category. However, Cagiannos et al¹⁰ reported on a series of 1,755 patients with cT2 prostate cancer treated with radical prostatectomy, demonstrating on multivariate analysis that the previous 1992 subclassification was prognostic for recurrence-free survival but the 1997 subclassification was not. Thus the cT2 subclassification was reversed in the 2002 edition and has remained unchanged.⁹

Demonstration of the prognostic significance of the cT2 subclassifications has important clinical and cost-effectiveness implications. In an era in which physicians have come to rely on new and expensive diagnostic tests, such as multiparametric MRI and genomic prognostics assessments, it is critical to highlight that traditional DRE remains a simple and valuable clinical tool for predicting prostate cancer outcome. Literature on how MRI compares with DRE for T2 subclassification is sparse, but MRI has been shown to outperform DRE in localization of cancer within the gland.²⁰ Data show that for clinical detection of T3 disease, subsequently confirmed on pathology, MRI has higher sensitivity than DRE; however, DRE has higher specificity.²¹ Given the inexpensive, efficient, and widely available nature of DRE, we feel this physical examination component remains critical for oncologists to consider in evaluation of patients with prostate cancer.

Unfortunately, studies suggest that DRE is an underused component of the physical examination, and may not be emphasized as a critical skill in the training of future physicians.^12–14 We believe that DRE should continue to be performed and documented as a key component of the physical examination for all patients with prostate cancer and should be recognized as a low-cost prognostic tool for men with adenocarcinoma of the prostate. DRE also continues to have an important role in screening for prostate cancer, because abnormal findings in conjunction with abnormal PSA levels have been shown to increase detection of clinically significant prostate cancer.²²

Our study has several limitations. Unfortunately, the NCDB prostate cancer data file does not contain data on biochemical failure, which would be another helpful prognostic end point to evaluate in addition to OS. However, OS may be the most clinically meaningful end point to correlate with the AJCC T2 classification. In addition, our study excluded low- and intermediate-risk cohorts. Given the relatively low incidence of T2 disease in these cohorts, it remains unknown whether our findings would hold statistical significance for all risk groups of patients with prostate cancer. As previously outlined, a further requirement for inclusion in this series was that radiotherapy be a component of treatment delivered, either upfront or postoperatively. Therefore, our results may not be as applicable to low- and intermediate-risk patients or those treated with surgery alone.

An additional limitation of this study was the absence of Gleason score data in a large number of NCDB subjects. We used several statistical strategies to evaluate any effect this might have had on the conclusions. Patients with a missing Gleason score were included in both the bivariate and multivariate models as an independent variable. This variable was not found to be a significant predictor of OS in either analysis. If a collection of patients with high Gleason score was somehow hidden in the “Gleason missing” cohort, we would expect that group to independently impact OS; such an influence on OS was clearly seen in the patients with a known Gleason score of 8 to 10.

An interaction test between Gleason missingness and cT2 subclassification was also performed and results were nonsignificant. If the presence of a missing Gleason score influenced the significance of clinical T stage on OS, we would have expected a statistically significant finding on the interaction test. Moreover, on multivariate analysis, the prognostic value of cT2 subclassification was highly significant and independent of the patients with Gleason score missing. Given our focus on an OS end point rather than a biochemical failure end point, any influence of missing Gleason scores should have been seen in this analysis. Despite missing Gleason scores, our data are sufficient to reasonably conclude that clinical T stage should be assessed routinely by oncologists for its potential prognostic significance.

Finally, it is possible that some patients may have been excluded from this analysis if they had a PSA level <20 ng/mL and Gleason score of 8 to 10 that was missing. In addition, the NCCN Guidelines define a “very-high risk category,” characterized by primary Gleason pattern 5, >4 cores with Gleason score of 8 to 10, or T3b–T4.² We likely included some very-high risk patients in our analysis because we did not make an effort to exclude or uniquely identify these patients.

Our study has many noteworthy strengths. Most importantly, it involves a contemporary cohort of men, with cases reported from 2004 to 2010. These men are likely to have undergone a modern staging workup, including pelvic CT and bone scan, making the findings applicable to today’s patients. Use of the NCDB data file allowed for evaluation of a large cohort of 5,291 high-risk men, making this the largest series to our knowledge to examine this question.

Next, the NCDB data represent a spectrum of clinical oncologic practice as opposed to a single major academic center or a group of major academic centers, and thus the conclusions are more applicable across a variety of clinical practices. Finally, the end point of OS is a strength; it is often the main end point of interest in clinical practice and is less subject to misinterpretation than surrogate end points often used in prostate cancer, such as biochemical progression-free survival.

Our 5-year OS values were lower than those reported in a recent institutional series from a major academic center.²³ There are a number of reasons for this. First, we used a landmark time of 10 months after diagnosis from which to evaluate OS. Second, all men in the other series had undergone radical prostatectomy, whereas <5% in our series underwent treatment including radical prostatectomy. Thus, our cohort was likely a less healthy and higher-risk population in general. Third, single-institution data from a prestigious and accomplished academic center are unlikely to be reflective of outcomes seen in the NCDB, which encompasses a broad range of clinical practices and levels of oncologic expertise.

Another noteworthy finding is verification of previously established prostate cancer prognostic factors. Age, CDCI score, Gleason score, and initial PSA level have all been established as prognostic factors for OS in prostate cancer.^24,25 Our analysis contributes to this supporting body of data. Current literature is unclear as to the prognostic value of race and insurance type in prostate cancer.²⁶ We found insurance type but not race to be a statistically significant prognostic factor on multivariate analysis. However, this was not the focus of our study and warrants further study.

Conclusions

Results of our series demonstrate prognostic significance of the cT2 subclassifications as determined by DRE in a large, contemporary cohort of men treated for high-risk prostate cancer. In an era of resource-conscious oncologic practice, DRE remains an effective and low-cost prognostic tool in the management of men with adenocarcinoma of the prostate, and thus should be performed routinely by oncologists treating this malignancy.