In Organ-Confined Prostate Cancer, Tumor Quantitation Not Found to Aid in Prediction of Biochemical Recurrence

Abstract

In 8^th edition AJCC staging, all organ-confined disease is assigned pathologic stage T2, without sub-classification. We investigated whether total tumor volume (TTV) and/or maximum tumor diameter of the index lesion (MTD) are useful in improving prediction of biochemical recurrence (BCR) in pT2 patients. We identified 1657 patients with digital tumor maps and quantification of TTV/MTD who had pT2 disease on radical prostatectomy (RP). Multivariable Cox regression models were used to assess whether TTV and/or MTD are independent predictors of BCR when adjusting for a base model incorporating age, pre-operative PSA, RP grade group, and surgical margin status. If either tumor quantification added significantly, we calculated and reported the c-index. 95 patients experienced BCR after RP; median follow-up for patients without BCR was 5.7 years. The c-index was 0.737 for the base model. Although there was some evidence of an association between TTV and BCR (p=0.088), this did not meet conventional levels of statistical significance and only provided a limited increase in discrimination (0.743; c-index improvement: 0.006). MTD was not associated with BCR (p>0.9). In analyses excluding patients with grade group 1 on biopsy who would be less likely to undergo RP in contemporary practice (622 patients; 59 with BCR), TTV/MTD was not a statistically significant predictor (p=0.4 and 0.8, respectively). Without evidence that tumor quantitation, in the form of either TTV or MTD of the index lesion, is useful for the prediction of BCR in pT2 prostate cancer, we cannot recommend its routine reporting.

INTRODUCTION

In the AJCC 8^th edition, the classification of pathologic stage T2 (pT2) prostate cancer was updated, with the removal of sub-categorization by laterality or extent of involvement (1). Formerly, organ-confined disease was divided in either two or three categories to parallel clinical T2 staging. In a 2009 conference on handling and staging of radical prostatectomy (RP) specimens, which foreshadowed the recent AJCC changes, the International Society of Urological Pathology (ISUP) advocated eliminating pathologic T2 sub-staging (2). This recommendation was based upon evidence from multiple studies showing: a) poor correlation between clinical and pathologic staging, b) that pathologic T2 sub-categorization did not convey prognostic information, c) that pathologists had poor reproducibility in determining less than vs. greater than ½ of one lobe (pT2a v. pT2b in 4^th, 6^th and 7^th editions), and to avoid the untenable scenario in which a unilateral large cancer would be assigned a lower T2 stage (pT2a or b) than two or more minute bilateral cancers (pT2c) (3–8). These recent changes have now been validated in at least one large study with long-term follow-up (9).

In manuscripts from the 2000s, multiple groups demonstrated that total tumor volume (TTV) was a predictor of biochemical recurrence (BCR) in univariate analysis yet did not add important predictive accuracy when controlling for known prognostic factors such as prostate specific antigen (PSA) levels, tumor grade, stage and surgical margin status (10–13). In the proceedings of the 2009 ISUP conference, a majority of experienced genitourinary pathologists were in favor of reporting tumor volume/quantitation in some form, yet there was no agreement on the ideal method (2). Subsequently, even though little new literature has come to light, recently revised College of American Pathologists (CAP) reporting protocols for RP recommend notating “tumor quantitation”, though it is left to the pathologist’s discretion whether one reports estimated TTV (%) or maximum tumor diameter in millimeters of the dominant/index tumor (MTD) (14).

Although there is currently no reproducible approach to quantitation, the elimination of pT2 subcategories in current AJCC staging presents a fresh opportunity to consider adding a measure of tumor quantitation to post-RP risk stratification models for organ-confined disease, as only tumor grade and margin status remain as pathology prognostic factors (15). As prior studies of TTV and MTD have not focused on their relevance in pT2 prostate cancer (10–13, 16–24), we examined a large cohort of entirely-submitted RP specimens with organ-confined disease to determine whether TTV and/or MTD are useful in improving prediction of biochemical recurrence (BCR) when controlling for standard clinicopathologic variables.

MATERIALS AND METHODS

After obtaining institutional review board approval, we identified 1657 patients who underwent primary RP between June 2000 and March 2012, had organ-confined (pT2) disease and had tumor maps available. All RP specimens were whole-mounted and entirely-submitted as previously described (25). Patients with microscopic bladder neck involvement (considered pT3a) and lymph node metastasis (clearly not organ-confined disease) were excluded, while patients with positive surgical margins (other than bladder neck) were included. TTV was calculated by planimetry (constructing digital tumor maps with delineation of tumor grade and applying imaging software to calculate TTV) using Photoshop CC (Adobe Systems, San Jose, CA). We compared this approach with measurement of two-dimensional MTD (assessed using a digital ruler on the same tumor maps), to see whether either parameter could improve our ability to predict BCR.

We created a base multivariable Cox proportional hazards model adjusted for age, pre-operative PSA, RP grade group (grade group 1 vs 2 vs 3 vs 4/5), and surgical margin status (positive vs. negative). A linear predictor was calculated from this model to use as a single covariate in subsequent models to overcome the challenge of a limited number of BCR events. Two subsequent Cox proportional hazards models were then created, including the linear predictor to represent current predictors and each of the two tumor quantification measurements, TTV and MTD. If either parameter added significantly to the base model, we calculated and reported the c-index. As a sensitivity analysis, we repeated all calculations excluding patients with grade group 1 (Gleason score 3+3=6) on biopsy, as these patients would be less likely to undergo definitive surgical intervention in contemporary practice. All analyses were performed using R version 3.3.2.

RESULTS

In our cohort of 1657 patients, 95 patients experienced BCR after RP. Median follow-up time for patients who did not have BCR was 5.7 years (IQR 3.9, 8.3). Patient and disease characteristics are summarized in Table 1. Associations between BCR and grade and margin status were as expected (see Supplemental Table 1). Of note, just over 60% of these organ-confined (pT2) patients had RP grade group ≥ 2, while 11% had a positive surgical margin.

Table 1.

Patient and disease characteristics, N=1657. Data are presented as frequency (%) or median (IQR).

Age at surgery	59 (54, 64)
Pre-RP PSA (ng/ml)	4.9 (3.6, 6.7)
RP grade group
1	639 (39%)
2	853 (51%)
3	136 (8.2%)
4	20 (1.2%)
5	9 (0.5%)
Positive surgical margins	175 (11%)
Tumor volume (cc)	0.79 (0.31, 1.66)
Maximum tumor diameter (cm)	1.19 (0.73, 1.67)

RP: radical prostatectomy; PSA: prostate specific antigen

The multivariable analyses are summarized in Table 2. To assess whether adding TTV or MTD improved our ability to predict BCR, we first calculated a c-index of 0.737 for the base model that included age, pre-operative PSA, RP grade group, and surgical margin status. When adding TTV to this base model, there was some evidence of an association between TTV and BCR and the c-index increased minimally to 0.743; however, this association did not meet conventional levels of statistical significance (p=0.088). When adding MTD to the base model, we found no association with time to BCR when controlling for other prognostic factors (p>0.9).

Table 2.

Multivariable analysis for all patients and the subset of patients with RP grade group ≥ 2: models adjusted for age, PSA, RP grade group, and surgical margin status.

	All patients (N=1657)			RP grade group ≥ 2 (N= 622)
Model	HR	95% CI	p value	HR	95% CI	p value
Base model + Tumor volume (cc)	1.12	0.98, 1.28	0.088	1.08	0.91, 1.28	0.4
Base model + Maximum tumor diameter (cm)	0.99	0.71, 1.38	>0.9	1.07	0.69, 1.67	0.8

RP: radical prostatectomy; HR: hazard ratio; CI: confidence interval; cc: cubic centimeter; cm: centimeter

On repeat analyses, we excluded patients with grade group 1 disease on biopsy. Of the remaining 622 patients, 59 had BCR, and the median follow-up for patients without BCR was 5.5 years (IQR 3.6, 8.0). The c-index of the base model at 5 years was 0.698 for this sub-cohort. Neither TTV nor MTD added significantly to the base model (p=0.4 and p=0.8, respectively).

DISCUSSION

Prediction of outcome after RP in patients with localized prostate cancer is important for tailoring follow-up and future management, such as whether to consider adjuvant therapy in patients at high risk for recurrence. In the post-operative setting, the most common RP pathology characteristics used to predict progression are Gleason score/grade group, pathologic tumor and nodal stage, and surgical margin status (26–31).

Among other pathology parameters that may be gleaned from RP specimens, it is well-known that increased tumor volume is associated with several prostate cancer grading and staging parameters which portend adverse clinical outcomes such as BCR (10, 13, 17). However, the degree to which TTV aids prediction of BCR in the post-RP setting is more controversial, with variable published results over the past two decades. While some studies have demonstrated statistically significant independent predictive value for BCR prediction (16–17), others have argued that TTV provides no additional prognostic information when controlling for pathologic tumor grade and stage (10–11). Our recent findings that tumor volume-related parameters, including measures of high grade tumor volume, do not provide benefit in decision-making post-RP (32) are concordant with the latter findings. Moreover, there is currently no accepted standard for the determination of tumor quantification in RP specimens (2). The prior literature on RP tumor quantification yields a wide variety of methods, including streamlined three-dimensional volume estimation (33), assessment of maximum tumor area using single or multiple slides (34), counting the number of involved blocks (35), assessment of the positive to negative block ratio (36), use of a measurement grid (37) or point counting (38), and visual estimation of percent microscopic involvement, with or without prior circling of identifiable tumor (39).

As opposed to the gross tumor dimensions and depth of invasion that are more easily determined and routinely reported for cancers involving hollow organs, the complex, infiltrative growth characteristics of prostate cancer make three-dimensional measurement of tumor extent difficult. While the proceedings of the 2009 ISUP conference on RP reporting favored inclusion of tumor volume/quantitation in some form, the variability of approaches among centers, lack of one well-accepted mode, and the crucial issue of interobserver reproducibility were acknowledged (2). Although no consensus was reached, it was suggested that measurement of two-dimensional MTD of the index lesion may be a method that could more easily be incorporated into routine clinical practice (2).

Indeed, initial cohort reports from the group at Brigham and Women’s Hospital in the late 1990s (n=57 and n=434 patients, respectively) (18–19) concluded that MTD was a simple, inexpensive, and independent predictor of BCR in RP specimens. However, when the same group subsequently evaluated a larger group of 781 patients with similar follow-up to that in our study, MTD lost statistical significance when surgical margin status was added to the multivariable analysis (21). Eichelberger et al, in a cohort of 364 patients, also found that MTD was an independent predictor of BCR when controlling for typical clinicopathologic variables (median follow-up: 12 months) (20). As in the literature regarding TTV however, subsequent reports by Fukuhara et al (n=364 patients; median follow-up: 37.4 months) and van Oort et al (n=542 patients; median follow-up: 39.5 months) have found that when MTD is treated as a continuous variable, it is not an independent prognostic factor and/or adds no discriminative value in BCR prediction beyond standard clinical and pathologic factors (23–24). It is now clear that the majority of early studies (18–22) examining MTD were characterized by limited follow-up (20), multivariable analyses that did not include one or more readily available clinicopathologic parameters (18–19, 22) and/or cohorts with incomplete submission of RP tissue (18–19, 21), such that grading, staging and margin parameters may not have been robustly represented in prediction models.

Interestingly, no prior study of tumor quantitation methods in prostate cancer has focused exclusively on organ-confined disease. Considering the recent condensation of all organ-confined cases into pathologic stage T2 in the AJCC 8^th edition (1), we took the opportunity to examine the two most commonly utilized tumor quantitation methods (TTV and MTD) in this population. Measurements in this study, accomplished by planimetry using commercial software, are a complex and time-consuming task requiring construction of digital tumor maps and application of imaging software to calculate tumor volume. Tumor quantitation in this fashion, could only be deemed appropriate for routine practice if it importantly increased prediction of clinical outcomes. Nonetheless, it may be considered more standardized, if not more objectively accurate, than the visual estimation utilized in many studies. After analyzing a large cohort of organ-confined RP specimens in this manner, including MTD determination by digital ruler (which may be more scalable to routine pathology workflow), we found that neither TTV nor MTD significantly improved prediction of BCR. While current CAP guidelines, despite conflicting data from prior literature, recommend some estimation of tumor quantitation (14), based on our results we are unable to recommend routine evaluation and reporting of TTV and/or MTD in patients with organ-confined disease at RP.

Any study with a retrospective single institution design may be influenced by the selection of a cohort of patients. In this study, limited to those patients with organ-confined prostate cancer who had available RP tumor maps and tumor quantitation data, this raises the possibility of whether the patients included are representative of all RP patients. Nonetheless, the current cohort of 1657 cases are by far the largest pT2 cohort analyzed for the association of tumor quantitation with BCR and our results for TTV mirror those in our recent work on tumor volume-related parameters in localized prostate cancers of all grade groups and stages (32). As in that study, a main strength of this report is that all RP specimens were examined by dedicated urologic pathologists, minimizing variation in interpretation. Furthermore, nearly two-thirds of cases studied had grade group ≥ 2 and our median follow-up (66 months in the sensitivity analysis examining contemporary-type patients) is comparable to or longer than other single institution studies. These latter factors are especially important considering recent work showing a very low risk of early (2-year) BCR even for patients with Gleason score 8–10/grade groups 4–5 (40).

In sum, in a large cohort of RP patients with organ-confined disease controlled for standard clinicopathologic variables, neither total tumor volume nor maximum tumor diameter of the index lesion added predictive value for BCR. We cannot recommend routine reporting of tumor quantitation in this clinical scenario.