Introduction

Prostate cancer (PCa) is the most common cancer in men and the second most common cause of death from tumors in the male population [1]. With the beginning of the PSA era, the early diagnosis of PCa has been made possible, widening the opportunities for local radical therapies such as RP or radiation therapy (RT). However, the scenario changes when PCa spreads to the lymph nodes or to the bone, becoming a systemic disease: In these cases, the opportunity window for a radical therapy is lost, although surgery and radiotherapy are progressively achieving a role in selected N + PCa patients [2–4]. A correct staging is therefore essential to properly focus the disease we are dealing with, and consequentially to decide the most appropriate treatment. Nowadays, computed tomography (CT) and bone scintigraphy (BS) are the imaging examinations commonly used in PCa for staging nodal and bone involvement, respectively. According to the current guidelines, these examinations are optional in intermediate-risk and recommended in high-risk PCa [5, 6]. However, their use has been questioned by several studies reporting a limited staging ability for both the procedures. In particular, BS is deeply influenced by PSA level and Gleason score [7], while the accuracy of CT scan has been shown to be really poor, even in those patients with a very high risk of nodal metastases [8]. Recently, whole-body magnetic resonance imaging (MRI) was shown to be more sensitive and specific than the combination of BS, targeted radiographs and abdomino-pelvic CT [9, 10], suggesting an “all-in-one” staging approach. Preoperative clinical models, such as Partin [11] or Briganti nomograms [12], have been adopted in the preoperative risk assessment of PCa for lymph node involvement (LNI), and showed possibly more appropriate and accurate than CT itself to decide for the need of pelvic lymph node dissection (PLND). All these things considered, we wonder whether it is time for BS and CT to be dismissed, or if they still have a role in the PCa staging. We tried to find an answer by evaluating their accuracy as staging examinations in a large cohort of RP patients.

Materials and methods

Patient population

Under institutional review board supervision, data from men who underwent RP between 1999 and 2012 with at least 2-year follow-up available (PSA records and clinical visits) at thirteen Italian Urology Departments located in Piedmont and Aosta Valley Italian regions were combined into the EUREKA-1 database. EUREKA-1 is an observational, multicentric, retrospectively derived dataset promoted by the CHIC project (sponsored by the European Union, 7th Framework Program, grant agreement no. 600841); it was approved by the Ethical Committee of FPO-IRCCS Cancer Center of Candiolo in July 2013 and amended in November 2014; data collected in EUREKA-1 included socio-demographic parameters, clinical tumor characteristics, surgery features, pathology, therapies, clinical and PSA follow-ups, and clinical outcomes.

To evaluate the ability of CT to predict LNI, we included 1091 patients treated with RP and PLND. All the patients were staged before surgery with abdomino-pelvic contrast-enhanced CT scan; CT was considered positive for LNI if at least one node had a short-axis diameter ≥10 mm [8]. CT technology greatly varied during the 14-year-long period under study. The CT devices used were: Philips Tomoscan AV (single detector row), Elscint Twin (2 slices), Philips Brilliance (6 slices), Toshiba Aquilion (16 slices), Philips Brilliance (40 slices), Philips Brilliance (64 slices), GE Discovery (64 slices), with slice thickness ranging in between 7 and 2 mm.

Following RP and PLND, the resected nodes were deemed positive for LNI if PCa metastases were found at the pathology examination (i.e., macro-metastases ≥2 mm or micro-metastases <2 mm [13]); finally, pathologic results were compared with CT findings.

Besides, patients were stratified according to the D’Amico clinical (preoperative) risk classification in low-, intermediate- and high-risk groups [14].

As for bone metastases, we included 1145 PCa patients scheduled for surgery, staged with planar total-body Tc-99m methylene diphosphonate BS. Most of the patients with positive or equivocal findings underwent targeted X-rays or MRI for confirmation. We looked only for false-positive and false-negative rates, considering that all patients successively underwent surgery.

Standard follow-up included PSA dosage and DRE every 3 months for 2 years, every 6 months until the fifth year and annually thereafter. Patients with clinical and/or serological suspect of bone metastasis at follow-up (i.e., fast severe arousal of post-prostatectomy PSA or clinical symptoms) underwent BS (and/or CT and/or MRI looking for osteoblastic metastases or secondary fractures) for radiologic confirmation.

Statistical analysis

Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were tested, and related P values were calculated with Chi-square test. Discrimination accuracy was assessed using the C-statistics of the ROC area under the curve (AUC). The overall accuracy of Partin and Briganti nomograms for LNI was assessed too. Univariate and multivariate logistic regression analyses tested the relationships between CT findings and LNI. Covariates consisted of preoperative PSA, clinical stage and bioptic Gleason score.

P values were considered statistically significant if <0.05. All statistical analyses were performed using SAS Foundation 9.3 (SAS Institute, Cary, NC, USA).

Results

The characteristics of the 1091 CT-staged patients are listed in Table 1. Overall, LNI was detected in 102 patients (9.3 %). CT scan was suggestive of LNI in 29 patients (2.7 %); of them, only 9 (31 %) had histologically proven LNI at PLND. False positives were found in 20 patients and false negatives in 93. Sensitivity (Se), specificity (Sp), PPV and NPV in predicting LNI for the whole cohort were 8.8, 98.0, 31.0 and 91.2 %, respectively.

Table 1 Descriptive characteristics of our series of 1091 patients staged with CT split according to D’Amico risk categories
Full size table

Subgroup analysis disclosed a significant association only for the high-risk subgroup of 334 patients (P 0.009) with a Se 11.8 %, Sp 96.2 %, PPV 44.4 % and NPV 81.0 % (Table 2).

Table 2 Cross-tabulation and measure of association (Chi-square test) between CT finding and LNI at PLND split according to risk categories
Full size table

Of consequence, we further compared the CT predictive ability with clinical risk factors in the high-risk group only. Logistic multivariate regression analysis including CT finding and PSA value, bioptic Gleason score and clinical staging left out any significant additional predictive ability of CT (P value of 0.40; see Table 3). These data were further confirmed by AUC analysis of the ROC curves in the high-risk group, showing a low overall accuracy of 54 % for CT, compared with 69 % for Partin tables and 80 % for Briganti nomogram (Fig. 1).

Table 3 Univariate and multivariate logistic regression in the high-risk group (N = 334) including CT finding, PSA, bioptic Gleason score and clinical stage (A), Partin tables (B) and Briganti nomogram (C)
Full size table
Fig. 1
figure 1

Area under the curve of the ROC graph in high-risk patients: CT finding, 54 % (a); Partin tables, 69 % (b) and Briganti nomogram, 80 % (c)

Full size image

Regarding BS, patients’ features are given in Table 4. BS was performed in 1145 patients (192 low-risk, 567 intermediate-risk and 386 high-risk), deemed fit for surgery after clinical and/or further radiologic examinations (targeted X-rays or MRI). Among them, BS showed some positivity in 74 cases; only four of them developed bone metastases successively, after an average of 27.5 months from surgery (PPV 5.4 %). Finally, 49 patients with negative scintigraphy developed bone metastases during their follow-up (NPV 95.4 %), six of them within 6 months after surgery.

Table 4 Descriptive characteristics of our series of 1145 patients staged with BS split according to D’Amico risk categories
Full size table

Discussion

Historically, abdomino-pelvic CT scan and BS played an important role as staging procedures in PCa patients, to rule out any extra-prostatic presence of disease. As the risk of nodal or bone metastases has been associated with bad clinical features such as elevated PSA, high clinical stage and poor tumor differentiation, CT and BS use has been confined to the cases at higher risk, namely some of the intermediate-risk and all of the high-risk PCa [5, 7, 11–15]. In the last years, however, their role has been strongly questioned, as a poor staging ability has been demonstrated for both procedures [7–10, 16, 17]. Nevertheless, despite the criticism of the scientific community, the last EAU guidelines still recommend both CT and scintigraphy performance for selected groups of PCa patients [5]. Following the lead of previous studies [7, 8, 16], we therefore decided to assess the usefulness of these examinations as predictors of LNI/bone metastases in our retrospective series of patients collected in the Eureka-1 database.

In 2008, a comprehensive meta-analysis on 1024 patients warned about the poor performance of presurgical CT in the detection of lymph node metastases: Pooled sensitivity and specificity were 0.42 and 0.82, respectively. The authors discredited the use of CT, concluding that it misrepresented the true nodal metastases involvement, thus misdirecting the therapeutic strategies offered to the patient [16]. Another study focusing on patients at high-risk of locally advanced disease had already shown a low accuracy of CT even in this high-risk group, with consequences on the correct treatment choice [17]. More recently, Briganti et al. [8] evaluated a single-center, homogeneous series of 1541 patients undergoing RP and extended PLND, all preoperatively staged with CT scan. Including only patients who received a standardized anatomically defined PLND, their study overcame the false-negative results due to a limited PLND. According to their results, sensitivity, specificity and accuracy of CT scan were 13, 96 and 54.6 %, respectively. Again, their conclusions deposed against the use of CT scan to decide on the need for, or extent of PLND in patients with PCa. In our analysis, the performance of CT was confirmed to be very poor, with low sensitivity and accuracy of 11.8 and 54 % even in the high-risk group.

The limited sensitivity of CT is due to its inability in detecting micro-metastases: A threshold of 10 mm in the short-axis nodal diameter for oval nodes and 8 mm for round nodes are used as criteria for the diagnosis of lymph node metastases [18], but such accuracy is obviously far inadequate to detect smaller metastases. On the contrary, the risk of false positives due to reactive lymph nodes needs to be taken into account, considering the 20 false-positive cases in our series.

To overcome the limitations of the available imaging techniques, several nomograms or predictive models have been proposed to assess the risk of LNI, such as Briganti or Partin tools [11, 12]. Both these nomograms have shown a very high accuracy for the LNI prediction (up to 88 %), and they are currently an important input for the decision-making process driving the choice to perform PLND. It has already been suggested that such preoperative clinical models should completely replace the imaging procedures in the assessment of the risk of LNI [8]. Interestingly, our study confirmed in the high-risk cohort a good discrimination accuracy for Briganti nomogram (80 %), but not so high for Partin model (69 %).

Maybe, the imaging predictive capability will improve with the progressive replacement of CT by multiparametric MRI, which is the gold standard for prostate imaging, and of planar BS with the more specific SPECT–CT BS [19]. However, CT and MRI performances are currently similarly poor compared with histologic diagnosis [20], even if soon iron oxide particles should enhance the detection of enlarged lymph nodes [10, 21]. Up to date, choline PET and PET–CT have also provided low sensitivity (49.2 %) in the detection of lymph node metastases prior to surgery in PCa patients [22].

As for bone metastases detection, the sensitivity of BS completed with targeted X-rays has been shown to be 86 %, with a 98 % specificity [10]. According to a recently developed prognostic tool, staging bone scans might be considered only for patients with a biopsy Gleason score >7, or with a PSA > 10 ng/ml and palpable disease (cT2/T3) prior to treatment. Such an approach would avoid the use of staging bone scans in about 81 % of the patients with a NPV of 99.6 % [23]. On the basis of these results, the recent guidelines recommend bone scan in asymptomatic patients only if the PSA level >10 ng/ml or Gleason score ≥8 or clinical stage ≥T3, namely the intermediate- and high-risk situation [5, 6]. However, even if the indications are restricted accordingly, a considerable number of performed bone scans are unnecessary. In our analysis, we could not evaluate the sensitivity and specificity of bone scan, as our cohort was composed of patients deemed fit to undergo RP, and so excluding all the patients positive to bone scan and treated primarily with medical therapy, i.e., androgen deprivation therapy. Nevertheless, we thought interesting to report the low accuracy of BS in our series, where 74 cases initially found positive were successively cleared for surgery, whereas six cases already harboring bone metastases were not detected.

Our study does have several limitations, mainly due to the selection criteria of our database, the heterogeneousness of the cohort and the retrospective nature of our data.

The main issue is inherent to our database, consisting in a population of men having been deemed fit for surgery; this criterion introduces substantial selection bias (i.e., men found to harbor metastatic disease and not eligible for RP, or who underwent primary RT instead of RP); this trouble is more relevant for BS than for CT, and so data collected on RP patients are inadequate to perform a proper evaluation of the accuracy of BS. A second limitation regarding bone scan is the lack of histologic confirmation from bone biopsy specimens; nevertheless, the long follow-up of our series allowed us to assess which patients harbored bone metastases not discovered during the preoperatory staging. Furthermore, the staging imaging procedures were performed by different physicians and with different equipments without any central review. One final limitation is related to the PLND template, which was not anatomically defined as recently recommended [8]: after all, the recommendations about the extension of PLND have changed several times in the last 15 years. PLND extension and template varied significantly between the participating urology divisions, while no statistical trend was identified in the mean number of nodes resected according to preoperatory D’Amico risk class (14 vs. 12.5 vs. 13.5 nodes in low-, intermediate- and high-risk classes, respectively). On the other hand, our retrospective data provide a snapshot about the usefulness of PCa staging examinations in a vast region of Italy over a 15-year period.

According to our findings, these examinations have only a minor effect on the correct diagnostic and therapeutic strategy. Of consequence, we advise to stage patients with CT just if they belong to high-risk category (i.e., PSA > 20 ng/ml or GS ≥ 8 or clinical stage ≥cT2c) and no functional MRI or CT–PET scan is available; this result is in agreement with AUA recommendations recently reported by Risko et al. [24].

Besides, bone scan too may be considered useful in the high-risk group only, so to decrease in lower-risk classes the incidence of misleading false-positive examinations, which can hamper or delay the choice of a radical therapy. In addition, a negative impact of a widespread use of these radiologic examinations on healthcare costs should be taken into account.

Conclusions

Our study confirms the poor accuracy of CT scan and BS as staging examinations for PCa. Their role should be restricted to selected high-risk cases, whereas their avoidance as routine examinations would simplify the diagnostic work-flow and save resources and money for the National Health Systems. As previously suggested in the literature, clinical predictive nomograms about lymph node involvement are more reliable to plan the best surgical therapeutic strategy.