Aim: Two-thirds of patients with a gray-zone prostate-specific antigen (PSA) level undergo unnecessary biopsy. Sensitivity is not yet sufficient to permit the use of modified PSA parameters or magnetic resonance (MR) imaging alone for prostate cancer screening. Thus, we evaluated the combination of MR imaging and PSA density (PSAD) for specificity and sensitivity.
Methods: During the period April 2004 through March 2006, 185 patients with a PSA level of 4.0–10.0 ng/mL underwent MR imaging and transrectal ultrasonography-guided 8-core biopsy (systemic sextant biopsy of the peripheral zone plus two cores of transition zone). All MR images were interpreted prospectively by two radiologists. An image was considered positive for prostate cancer if any feature indicated a cancerous lesion. Receiver operating characteristic (ROC) curves were used to compare the usefulness of the PSA level, PSAD and PSA transitional zone density (PSATZ) for the detection of prostate cancer.
Results: Of the 185 patients, 62 had prostate cancer. Sensitivity and specificity of the axial T2-weighted MR imaging findings for cancer detection were 79.0% and 59.4%, respectively. The area under the ROC curve was 0.590 for the PSA level, 0.718 for PSAD and 0.695 for PSATZ. MR imaging findings and PSAD were shown by multivariate analysis to be statistically significant independent predictors of prostate cancer (P < 0.001). With a PSAD cut-off value of 0.111, sensitivity was 96.8%, but specificity was 19.5%. Combining MR imaging findings with PSAD increased the specificity to 40% and retained 95% sensitivity.
Conclusion: MR imaging findings combined with PSAD provide high sensitivity and improve the specificity for the early detection of prostate cancer.
The discovery of prostate-specific antigen (PSA) in 1979 resulted in significant changes in the detection of organ-confined prostate cancer. PSA is the most useful tumor marker for prostate cancer (PCa). However, two-thirds of patients with a gray-zone PSA level (4.0–10.0 ng/mL) undergo unnecessary biopsy because of the marker's lack of disease specificity. Prostate biopsy is very uncomfortable for patients. If possible, most patients would choose non-invasive modality for the detection of PCa over prostate biopsy. Although the complication rate for prostate biopsy is very low, severe complications, such as sepsis, occasionally occur. Thus, a highly sensitive and specific diagnostic method that avoids unnecessary biopsy is needed.
Various methods have been proposed to enhance the usefulness of the serum PSA level for discrimination between PCa and benign disease, including the determination of PSA density (PSAD), the age-modified PSA level, PSA velocity and the free-to-total PSA ratio.1,2 Sensitivity and specificity of these PSA parameters are better than those of the PSA level alone, but they do not have enough sensitivity to avoid missing cancers.3,4
Advances in magnetic resonance (MR) imaging have provided for improved visualization of the prostate anatomy. Originally, MR imaging was performed simply to help determine the stage of PCa. More recently, MR imaging has also been used for early detection of PCa. However, the sensitivity of MR imaging is not sufficient to permit use of this modality alone for PCa screening.5
In cancer screening, high sensitivity is necessary to reduce the risk of missing cancers. Sensitivity of at least 90% is needed.6 Morphological evaluation by means of MR imaging for the detection of PCa does not include identification of biochemical factors that alert screeners to the possibility of PCa. Thus, we hypothesized that combined analysis of the morphological (MR imaging findings) and biochemical (PSAD) factors would enhance both sensitivity and specificity for the early detection of PCa.
During the period April 2004 through March 2006, 370 consecutive patients were referred to either Gifu University Hospital or Kizawa Memorial Hospital, Gifu, Japan, because of an elevated serum PSA level or lower urinary tract symptoms and underwent prostate biopsy. Patients with urinary tract infection, acute urinary retention, prior prostatic surgery, re-biopsy or known PCa were excluded. Consequently, 185 patients with a PSA level greater than 4.0 ng/mL and less than or equal to 10.0 ng/mL were evaluated. The risks and benefits of the biopsy procedure were explained to each patient, and written informed consent was obtained. These 185 patients underwent MR imaging and then transrectal ultrasonography (TRUS)-guided biopsy.
Blood samples were obtained before digital rectal examination (DRE). The PSA level was measured by means of the Abbott ARCHITECT Total PSA assay (Abbott Laboratories, Abbott Park, IL, USA). DRE was performed by one of several urologists. The results were classified as either normal or suspicious if the prostate was diffusely hard or showed irregular contours or prominent lobe asymmetry.
After DRE, patients underwent MR imaging of the prostate. Either a 1.5-T Philips Intera Achieva Pulser scanner (Philips Medical Systems, Eindhoven, the Netherlands) or a 1.5-T superconducting system (Signa, General Electric, Milwaukee, WI, USA) was used. T2-weighted images were obtained with a phased array coil.
With the Intera Achieva 1.5-T Pulsar scanner, axial, T2-weighted fast spin-echo MR images were obtained with the following parameters: TR 4125 ms, TE 100 ms, field-of-view 220 × 220 mm, matrix 204 × 256 pixels, slice thickness 4 mm, interslice gap 1 mm and a stack of 20 contiguous slices. With the 1.5-T superconducting system, axial, T2-weighted fast spin-echo MR images were obtained with the following parameters: TR 2400 ms, TE 102 ms, field-of-view 240 × 240 mm, matrix 224 × 256 pixels, slice thickness 5 mm, interslice gap 1 mm and a stack of 11 contiguous slices.
All MR images were interpreted at the time they were obtained by two radiologists (observer 1 and observer 2) who were informed of patients' clinical data, including age and PSA level, but blinded to DRE findings. They had 10 and 8 years of experience, respectively, as radiologists. Observer 1 prospectively interpreted 112 MR images, and observer 2 interpreted 73. Low signal intensity areas within the relatively high signal intensity peripheral zone, homogenous low signal intensity areas invading the peripheral zone and a low signal intensity surgical capsule on the T2-weighted images were interpreted as cancerous areas. An image was considered positive for PCa if any features indicated a cancerous lesion (Fig. 1).
After MR imaging, patients underwent prostate biopsy. All biopsy specimens were obtained under TRUS guidance with an automatic biopsy gun and an 18-gauge needle. In addition to systematic sextant biopsy specimens, two core transition zone specimens were obtained.
We considered each of the 185 prostates to be divided into halves (right and left), with each half including the transition zone and peripheral zone (at the base, middle and apex). We compared DRE findings and biopsy results, and MR imaging findings and biopsy results in 370 (185 × 2) prostate regions.
Total prostate volume and transition zone volume were determined by planimetry, through tracing of the central gland margin and whole prostatic margin, respectively, on axial, T2-weighted MR images. PSAD and PSA transition zone density (PSATZ) were determined by dividing the serum PSA level by the volume of the entire prostate and by the transition zone volume, respectively.
We retrospectively compared the accuracy of the PSA level, PSAD and PSATZ for distinguishing PCa via receiver operating characteristic (ROC) curves. Values are shown as mean ± SD. Differences in variables between the biopsy-negative and biopsy-positive patients were analyzed using the Mann–Whitney U-test. Areas under the curves (AUCs) were calculated and compared as described by Hanley and McNeil.7 The probability of detection of PCa by prostate biopsy was calculated by using a multivariate logistic regression model. Variables included patient age, DRE, MR imaging findings and PSAD. Statflex 5.0 for Windows (Artec, Osaka, Japan) was used for all statistical analyses and significance was set at P < 0.05.
The overall study group comprised patients aged 34–84 years (mean, 68.7 ± 7.7 years). On the basis of histological features of biopsy specimens, PCa was diagnosed in 62 (33.5%) of the 185 patients.
Mean age, the serum PSA level, total prostate volume and transition zone volume of patients with cancer and those without prostate cancer are shown in Table 1. Patient age, serum PSA level, total prostate volume, transition zone volume, PSAD and PSATZ differed significantly between patients with and without prostate cancer.
Table 1. Characteristics of patients with benign and malignant disease
Total patients n = 185
Cancer-negative n = 123
Cancer-positive n = 62
P-value (Mann–Whitney U-test) Cancer negative vs Cancer-positive
mean ± SD; PSA, prostatic-specific antigen; PSAD, PSA density; PSATZ, PSA transition zone density.
On the basis of T2-weighted images, 99 cases were judged to be cancer positive and 86 cases were judged to be cancer negative. Of the 99 cancer-positive cases, 49 were diagnosed pathologically as cases of PCa. Of the 86 cancer-negative cases, 73 were determined pathologically to be cancer negative.
DRE was performed in 184 of the 185 patients. It was not performed in one case due to severe anal pain. Forty-eight of the 184 patients were judged to be DRE positive. Of these 48 DRE-positive cases, 20 were diagnosed pathologically as cases of PCa. Of the 136 DRE-negative cases, 95 were determined pathologically to be cancer negative. Sensitivity, specificity and accuracy of DRE findings and axial T2-weighted MR imaging for detecting PCa are shown in Table 2.
Table 2. Predictive values of digital rectal examination, magnetic resonance imaging, and prostate-specific antigen (PSA) density
DRE (n = 184)
MR Imaging (n = 185)
DRE, digital rectal examination; MR, magnetic resonance; PSAD, prostate-specific antigen density.
When we considered the prostate to be divided into halves, histopathological results were positive in 76 of the 370 halves. On the basis of T2-weighted images, 113 regions were judged to be cancer positive. Of these 113 regions, 49 were diagnosed pathologically as PCa. On the basis of DRE, 48 regions were judged to be cancer positive. Of these 48 regions, 20 were diagnosed pathologically as PCa. Sensitivity, specificity and accuracy of DRE and MR imaging in the 370 prostate regions for detecting PCa are shown in Table 3. The odds ratio of MR imaging findings in the two halves remains better than that in the whole prostate.
Table 3. Digital rectal examination vs magnetic resonance imaging in divided prostates regions
DRE (n = 368)
MR Imaging (n = 370)
CI, confidence interval; DRE, digital rectal examination; MR, magnetic resonance.
The AUC was 0.590 for the PSA level, 0.718 for PSAD and 0.695 for PSATZ. PSAD and PSATZ were significantly better than the PSA level alone in differentiating between benign and malignant prostate disease (P < 0.05). The AUC did not differ significantly between PSAD and PSATZ (P = 0.271) (Fig. 2). The best cut-off for PSAD, taken as the point on the ROC curve with the highest odds ratio, was 0.184. When a PSAD cut-off value of 0.184 was used, sensitivity was 77.4% and specificity was 66.7%. The results of multivariate analysis relating to patient age, DRE findings, MR imaging findings and PSAD are shown in Table 4. MR imaging findings and PSAD were shown to be statistically significant predictors of prostate cancer.
Table 4. Multivariate analysis of prebiopsy variables associated with prostate cancer
CI, confidence interval; DRE, digital rectal examination; MR, magnetic resonance; PSAD, PSA density.
Age (every 10 years)
MR imaging findings
PSAD (every 0.1 ng/mL/cm3)
If we had used a combination of DRE findings and PSAD and avoided prostate biopsy in patients with negative DRE findings and a PSAD less than 0.184, sensitivity and specificity for the detection of PCa would have been 82.0% and 49.6%, respectively. If we had used a combination of MR imaging findings and PSAD and avoided prostate biopsy in patients with negative MR imaging findings and PSAD less than 0.184, sensitivity and specificity for detection of PCa would have been 95.2% and 40.7%, respectively (Table 5).
Table 5. Specificity of PSA density alone vs magnetic resonance imaging findings combined with PSA density
DRE (n = 184) +PSAD (0.184cut-off)
MR Imaging (n = 185) +PSAD (0.184cut-off)
CI, confidence interval; DRE, digital rectal examination; MR, magnetic resonance; PSA, prostatic-specific antigen; PSAD, PSA density; PSATZ, PSA transition zone density.
MR imaging is a useful modality for staging PCa, owing to its excellent depiction of zonal anatomy.8 PCa is usually of low signal intensity on T2-weighted images of the prostate gland. Post-biopsy hematoma may interfere with the interpretation of MR images for as much as one month after the procedure. Thus, MR images should be obtained before biopsy.9
Wafer et al. compared the accuracy of endorectal MR imaging with that of sextant biopsy for cancer localization in 47 patients who subsequently underwent prostatectomy. In their study, sensitivity and specificity of MR imaging were 67% and 69%, respectively.10 Ikonen et al. reported sensitivity and specificity of 56% and 70%, respectively.5 Both reports included the overall range of PSA levels among patients studied. There are few reported studies pertaining to the accuracy of MR imaging for the detection of PCa, and even fewer in which MR images were obtained before biopsy.
Hoshii et al. estimated prostatic volume via MR imaging and via TRUS.11 They reported that there was no statistical difference between MRI-based PSAD and TRUS-based PSAD, however, AUCs of MRI-based PSAD were higher than that of TRUS-based PSAD. Thus, we not only interpreted the MR images but also performed the accurate measurement of total prostate volume and transition zone volume on the images.
Our study showed that PSAD more accurately distinguished benign and malignant disease than the PSA level alone. However, PSAD and PSATZ do not differ significantly in this respect. Hoshii et al. also reported no significant difference between MRI-based PSAD and MRI-based PSATZ.11 Therefore, we used PSAD, not PSATZ, in combination with MR imaging findings. Use of PSAD can reduce the number of unnecessary biopsies at the cost of increasing the number of missed cancers. In our study, the use of a PSAD cut-off of more than 0.184 resulted in 14 cancers being missed (sensitivity, 77.4%). For cancer screening, sensitivity of 90% or higher is needed. In previous studies of PSAD in patients with a gray-zone PSA level, use of a PSAD cut-off of more than 0.15 did not yield more than 80% sensitivity, resulting in many cancers being missed. Because of low sensitivity, PSAD and PSATZ are not specific enough for clinical use as diagnostic factors.3,4,12 For example, a small PCa producing a small amount of PSA and accompanied by significant benign prostatic hyperplasia could be missed due to a low PSAD.
We hypothesized that the excellent visualization provided by MR imaging would allow detection of PCa in patients with a low PSAD and could make up for the low sensitivity of PSAD. Ikonen et al. reported that endorectal MR imaging detects poorly differentiated PCa lesions more accurately than it detects clinically insignificant tumors.13 MR imaging findings and PSAD can each enhance the power of the other for early detection of PCa. Our multivariate analysis indicated that both MR imaging findings and PSAD are strongly associated with PCa. To our knowledge, no reports have been published on the combination of volume-modified PSA and MR imaging findings.
Retaining more than 90% sensitivity in previous series of patients evaluated on the basis of PSAD resulted in decreasing specificity by approximately 20%. Brawer et al., Zlotta et al. and Gohiji et al. found that with the use of a PSAD cut-off value of 0.10, sensitivity and specificity were 93% and 21%, 90% and 20% and 97% and 23%, respectively.3,14,15 In our study, when we achieved greater than 90% sensitivity, specificity fell below 20%. However, if we had used a combination of MR imaging findings and PSAD and avoided prostate biopsy in patients with negative MR imaging findings and PSAD < 0.184 (Fig. 3), we could have maintained sensitivity above 95% and improved specificity to 40%. That is to say, we could have avoided 50 negative biopsies and reduced the number of undetected cancers by three. The three missed cancers had a Gleason score of 6, and the patients were considered to be at low risk.
A limitation of our study is that there might have been too few biopsy cores for histopathological evaluation. The standard biopsy method is nowadays an extended one with around 12 biopsy cores. Matsumoto et al. reported the efficacy of transrectal ultrasound-guided 12-core biopsy of the prostate in Japanese men. They found a 7.7% increase in cancer detection using an additional far lateral sextant biopsy in the peripheral zone.16 Takenaka et al. also reported a 20.4% increase in cancer detection using the systematic 12-core transperineal method rather than the sextant method.17 The sextant biopsy technique with two transition zones is subject to sampling error and contains false-negative cases. Particularly in the cephalad-caudal dimension, sampling error might occur in the wrong direction. Furthermore, a small low signal intensity area that is not cancerous will result in a false positive diagnosis. Thus, we considered the prostate to be divided into halves and evaluated the consistency between MR imaging findings and biopsy results. Accurate evaluation of cancer sites should be performed on the basis of histopathological features of the prostatectomy specimen. However, prostatectomy is not performed in all patients. Extensive biopsy for histological correlation might improve sampling errors.18 Another limitation is that we did not evaluate interobserver variability in the MR imaging findings. Two independent observers interpreted the MR images prospectively. We should have investigated interobserver variability between several readers.19
Although DRE is an easily applied tool for detection of PCa from the standpoint of cost, sensitivity is low and many cancers are missed. Even the combination of PSAD and DRE findings missed 11 cancers, including high-grade PCa with a Gleason score of 8. When cost is not a consideration, MR imaging is clearly a better modality than DRE for the localization of PCa. MR imaging findings combined with PSAD could have reduced the number of undetected cancers and avoided unnecessary biopsies in our patients. However, prospective studies are needed to prove the scientific utility of our algorithm.
Use of MR imaging for cancer screening is expensive, however, MR imaging is a non-invasive examination that enables screeners to localize the PCa and perform accurate volume measurements which provide useful information for the treatment of PCa.20 Performing an MR imaging study of all patients who are at risk for PCa is impossible in terms of cost. However, we suggest that avoiding unnecessary biopsies by combining MR imaging findings and PSAD may be valuable in men of advanced age for whom prostate biopsy is a high-risk procedure or who have significant comorbidity, such as those receiving anticoagulant therapy. In Japan, MR imaging of the prostate costs approximately $20 US. Outpatient prostate biopsy costs approximately $30 US. Complications such as urinary retention or prostatitis can increase the cost. If unnecessary biopsies are avoided and the complication rate reduced, the cost of performing MR imaging might be acceptable.