Transrectal ultrasonography-guided biopsy does not reliably identify dominant cancer location in men with low-risk prostate cancer

Authors


Peter R. Carroll, Department of Urology, University of California, San Francisco, Box 1695, San Francisco, CA 94143-1695, USA. e-mail: pcarroll@urology.ucsf.edu

Abstract

Study Type – Diagnostic (exploratory cohort)

Level of Evidence 2b

What's known on the subject? and What does the study add?

The widespread use of serum PSA testing followed by TRUS-guided biopsy have resulted in profound prostate cancer stage migration with many patients presenting with focal rather than multifocal disease. There is increasing interest in the use of focal rather than whole-gland treatment. However, current biopsy schemes may still miss cancer or, even when cancer is identified, its extent or grade might not be accurately characterized. In order for focal therapy to be effective, the area of highest tumour volume and/or grade needs to localized accurately. The aim of this study was to assess how well biopsy, as currently performed, locates the focus of highest prostate cancer volume and/or grade.

OBJECTIVE

  • • To evaluate the ability of transrectal ultrasonography (TRUS)-guided extended core biopsy to identify the dominant tumour accurately in men with early stage prostate cancer.

PATIENTS AND METHODS

  • • Patients with early stage, low-risk prostate cancer who subsequently underwent radical prostatectomy (RP) and had complete surgical specimens were identified.
  • • Re-review was performed by a single uropathologist using ImageJ software to identify tumour location, dominant grade (DG) and dominant volume (DV).
  • • Pathology findings were then compared with biopsy results.

RESULTS

  • • A total of 51 men with early stage, low-risk prostate cancer, who had undergone RP, had complete specimens for review and a median of 15 biopsy cores taken for diagnosis and grading.
  • • Sixteen men had a single diagnostic biopsy, 21 had one repeat biopsy, and 14 had two or more repeat biopsies.
  • • Compared with surgical findings, biopsy correctly identified the sextant with the largest tumour volume in 55% (95% CI 0.5–0.6) of specimens and the highest grade in 37% (95 CI 0.3–0.5).
  • • No demographic or clinical factors were significantly associated with identification of DG. Interval between last biopsy and RP, total tissue length taken and total length of tumour identified were significantly associated with correct identification of DV.

CONCLUSIONS

  • • Our findings show that TRUS-guided biopsy detects and localizes DV better than it does DG.
  • • Even with an extended scheme, TRUS-guided biopsy does not reliably identify dominant cancer location in this low-risk cohort of men with early stage prostate cancer.
  • • TRUS-guided biopsy may perform better in similar men with low stage, but higher volume disease.
Abbreviations
RP

radical prostatectomy

DG

dominant grade

DV

dominant volume

UCSF

University of California, San Francisco.

INTRODUCTION

In the majority of patients with prostate cancer, the cancer is multifocal with the tumour focus of highest grade and volume considered to be the dominant or index tumour [1–4]. With effective identification and treatment of the dominant tumour, smaller secondary tumours may have little effect on overall disease progression [4–7]. By localizing treatment to the dominant tumour, focal therapy could eliminate the need for radical or whole gland treatment , thus reducing unintended damage to healthy tissue [8]. There is increasing interest in the use of focal therapy, especially for the management of those with early stage prostate cancer, perhaps in lieu of active surveillance. TRUS-guided prostate biopsy is used to characterize tumour grade, location and volume, and predict clinical stage, disease extent and prognosis [2,8–10]. Correct identification of cancer by TRUS biopsy is critical to the accurate characterization and localization of the dominant tumour for effective focal therapy [11].

Several studies have compared pathological findings from diagnostic biopsies with radical prostatectomy (RP) specimens, with varying reports of concordance [6,12]. Two recent reports have suggested that TRUS-guided prostate biopsy may either underestimate [13,14] or overestimate [15,16] tumour grade in as many as 35% of cases. These findings were limited by the lack of standardization of biopsy technique [13], and the authors analysed two sequential biopsies, rather than a single biopsy and the subsequent RP specimen [15]. An additional study found that extended pattern biopsies increase the chance of cancer detection [17], yet 25% of anterior tumours were still missed by peripherally directed biopsy strategies [13]. It therefore remains unclear whether TRUS biopsy, as currently performed, can accurately identify the dominant tumour for targeting during focal therapy.

The present study aims to determine the accuracy with which TRUS-guided serial prostate biopsy identifies the location of the dominant tumour grade (DG) and volume (DV) when compared with the RP specimen in men with low volume, low-risk prostate cancers.

PATIENTS AND METHODS

Study participants were selected from the Urologic Oncology Database, a clinical and patient-consented research resource at the University of California, San Francisco (UCSF). The study was approved by the institutional review board to gather diagnostic, surgical, pathological and outcomes data on men treated for prostate cancer with surgery or active surveillance. Patients diagnosed in 1994–2009 with low or intermediate grade, low volume disease who were on active surveillance for at least 6 months and later underwent RP were included in this analysis [2]. Men who had surgery may or may not have had disease progression; detailed data about decision-making was unavailable. In addition, most men who underwent surgery within 6 months and had matching clinical risk characteristics, year of diagnosis and age were included. Those with a history of external beam radiation therapy or androgen ablation therapy before surgery were excluded. Of those remaining, patients whose surgical specimens were available for review formed the study cohort.

All prostate specimens had been received previously from the operating room at the time of surgery, weighed, fixed, inked and sectioned at 3–4-mm intervals perpendicular to the urethral axis according to standard institutional protocol. Slides of RP specimens were reviewed by a single genitourinary pathologist (M.B.) who identified areas of prostate cancer in each specimen. Areas of tumour were then colour-coded by grade: Gleason 3: green; Gleason 4: blue; and Gleason 5: red. Slides were labelled by sextant and scanned to produce digital pictures. The area of each coloured region was then quantified using ImageJ, a picture analysis program based upon the NIH Image program, using methods previously described [18,19]. Calculated areas for each coloured region were multiplied by 0.4 mm, the standard thickness of the pathology specimen, to obtain the calculated volume of each colour region. This calculated volume was the equivalent of the volume of tumour burden. DG and DV were then identified within each RP specimen. DV was the sextant containing the largest volume of tumour in each RP specimen. DG was the sextant containing the highest tumour grade in each RP specimen. If multiple sextants had matching grade, the largest volume was used to determine DG.

Clinical data such as diagnostic age, PSA and TRUS prostate volume were acquired from medical records. Biopsy specimens obtained from outside institutions were submitted for re-review within the Department of Pathology at UCSF as part of standard clinical care. During re-evaluation, biopsy cores were examined and grouped by sextant. Primary and secondary Gleason scores were reported by sextant as percentages of tumour involvement per sextant grouping. Follow-up biopsies were performed at UCSF. Histological findings used in the present study were based upon those provided in the histology reports of re-reviewed biopsy specimens. Pathology findings per sextant for biopsies containing at least 10 cores were compared with the surgical pathology findings for DG and DV. DG was correctly identified when biopsy findings matched sextant location and Gleason grade as those of DG in the RP specimen. DV was correctly identified when biopsy and RP findings were positive for tumour burden in the same sextant, irrespective of tumour grade.

Rates of DG and DV identification were calculated per area of the prostate (apex, midgland and base), as well as by lobe. Overall rates of identification were calculated as composite rates from all biopsy–RP specimen comparisons, with 95% CIs for percentages. Frequency tables and chi-squared tests were used to describe categorical variables. Means and t-tests were used to describe continuous variables. Statistical analysis was performed using Stata SE 10 and sas 9.2 for Windows.

RESULTS

Of 3486 prostate cancer patients who gave consent for research at UCSF, 83 had early stage, low volume disease, initially managed with active surveillance and subsequently treated with RP. An additional 29 patients with matching clinical characteristics, year of diagnosis and age underwent RP as primary treatment. The final study cohort consisted of 51 patients who met the study criteria and who had available tissue. This final cohort did not differ with respect to clinical and surgical pathology characteristics from the 61 excluded men without tissue available for review, though a trend towards significance was noted in proportion of Caucasian patients (94% final vs 84% excluded cohort, P= 0.08).

A total of 107 biopsy reports, the majority of which (73%) were performed at UCSF, were reviewed. Most patients (41%) had one repeat biopsy, while 16 (31%) men had a single diagnostic biopsy and 14 (28%) had two or more repeat biopsies. The median number of cores per biopsy increased with successive biopsies: 14 cores for the first biopsy, 16 for the second and 17 for the third. Few patients had Gleason grade 7 or higher on diagnostic biopsy (10%), and all of these were very low volume. Nineteen men were upgraded on repeat biopsy. The median interval between diagnostic biopsy and RP was 19 months. Total volume of tumour found in the RP specimen was also low (median 0.6 mL; Table 1). Twelve men with <10 biopsy cores taken and 26 men with no prostate tissue available were excluded from analysis.

Table 1.  Clinical, pathological and demographic characteristics of patients undergoing RP
CharacteristicValue
  1. IQR, interquartile range.

Mean (sd)age, years61.3 (7)
PSA, ng/mL, n (%) 
 <412 (24)
 4.1–618 (35)
 6.1–1011 (22)
 >1010 (20)
Clinical stage, n (%) 
 T128 (55)
 T223 (45)
 Median (IQR)prostate volume, mL33 (24–57)
Biopsy resultsValue
 Biopsy Gleason grade, n (%) 
  2–646 (90)
  7 (3 + 4)4 (8)
  7 (4 + 3)1 (2)
 Median (IQR) no. of cores taken15 (13–17)
 Median (IQR)biopsy tissue length, mm69 (40–120)
 Median (IQR)biopsy tissue length positive (mm)10 (5–17)
 Median (IQR)% biopsy length positive14 (9–21)
Final pathology resultsValue
 Pathological stage, n (%) 
  T239 (76)
  T3a10 (20)
  T3b2 (4)
 Positive margin, n (%)6 (12)
 Median (IQR)tumour volume, mL2.8 (1.3–3.5)
 Median (IQR) volume DG, mL0.3 (0.1–0.8)
 Median (IQR) volume DV, mL1.1 (0.4–1.6)

The majority of biopsies had positive findings in the same lobe as the DG (81%, 95% CI 0.7–0.9); however, DG was correctly identified in 37% of cases (95% CI 0.3–0.5). Identification of DG varied greatly by location (Fig. 1) yet did not differ significantly (P= 0.4). In 22% (15/67, 95% CI 0.1–0.3, P < 0.01) of cases, the location of DG was correctly identified but biopsy Gleason grade was incorrect. Most biopsies had positive findings on the lobe ipsilateral to DV (82%, 95% CI 0.7–0.9). DV was correctly identified in 55% (95% CI 0.5–0.6). Accuracy of DV identification did not vary significantly by location (Fig. 1) (P= 0.4). In cases where DV was not identified, 54% (26/48, 95% CI 0.4–0.7, P < 0.01) had positive cores in adjacent sextants. The last biopsy before RP identified both DG and DV more often than earlier biopsies (Fig. 2). When accuracy was assessed by individual patient, DG was found in 58% (95% CI 0.4–0.7) of subjects and DV in 73% (95% CI 0.6–0.9 [Fig. 3]). For all biopsies, the agreement between biopsy and RP for DG and DV did not vary significantly by lobe (P > 0.05).

Figure 1.

Identification of DG and DV by location of tumour for any biopsy (n= 107). There was no significant difference in identification of DG or DV by location. L, left; R, right.

Figure 2.

Identification of DG and DV by location of tumour for last biopsy (n= 51). There was no significant difference in identification of DG or DV by location. L, left; R, right.

Figure 3.

Identification of DG and DV by total number of biopsies taken. There was no significant difference in identification of DG or DV by location.

Clinical and pathological features were evaluated by t-test for association with DG and DV. No factors were significantly associated with the identification of DG (Table 2). Total tissue length taken at last biopsy (P= 0.01), total length of tumour identified at last biopsy (P < 0.01), and interval between last biopsy and RP (P= 0.02) were significantly associated with correct identification of DV (Table 3). Lower prostate volumes and higher tumour volumes were observed in patients in whom DV was correctly identified, though the associations did not reach significance.

Table 2.  Clinical and pathological features of patients by identification of DG for any biopsy
CharacteristicDG not identifiedDG identifiedP
 N N
  • *

    Values taken from last biopsy before RP.

Mean (sd) age, years59 (8)2162 (6)300.1
Median (IQR) PSA, ng/mL5 (4.3–6.5)215.6 (4.1–0.5)300.1
Median (IQR) prostate volume, mL36 (29–59)2027 (22–57)280.4
Median (IQR) PSA density, ng/mL20.1 (0.1–0.2)200.2 (0.1–0.2)280.2
Median (IQR) no. of biopsies per person2 (1–2)212 (2–3)300.1
Median (IQR) no. of cores taken*15 (12–17)2116 (15–17)300.2
Median (IQR) length tissue taken*, mm61 (46–92)18105 (59–142)300.1
Median (IQR) length tumour*, mm7 (3–15)1817 (9–26)300.1
Median (IQR) % cores positive*14 (5–23)1817 (10–31)280.1
Median (IQR) prostate weight, g61 (48–84)2146 (40–78)300.4
Median (IQR) tumour volume, mL0.9 (0.4–1.7)201.3 (0.7–2.8)290.9
Median (IQR) interval between biopsy* and RP, months6 (2–11)215 (3–10)300.7
Table 3.  Clinical and pathological features of patients by identification of DV for any biopsy
CharacteristicDV not identifiedDV identifiedP
 NMedian (IQR)N
  • *

    Values taken from last biopsy before RP.

Mean (sd) age, years59 (9)1362 (6)380.2
Median (IQR) PSA, ng/mL5.6 (3.6–7.1)135.4 (4.3–9.6)380.4
Median (IQR) prostate volume, mL36 (33–72)1327 (22–54)350.1
Median (IQR) PSA density, ng/mL20.1 (0.1–0.2)130.2 (0.1–0.2)350.2
Median (IQR) no. of biopsies per person1 (1–2)132 (2–3)380.1
Median (IQR) no. of cores taken*14 (12–17)1316 (15–17)380.09
Median (IQR) length tissue taken*, mm56 (31–65)12105 (60–141)360.01
Median (IQR) length tumour*, mm5 (2–10)1216 (10–26)36<0.01
Median (IQR) % cores positive*12 (5–20)1217 (10–29)340.1
Median (IQR) prostate weight, g62 (52–100)1347 (42–78)380.2
Median (IQR) tumour volume, mL0.8 (0.2–1.1)121.4 (0.7–2.8)370.2
Median (IQR) interval between biopsy* and RP, months9 (6–16)134 (2–8)380.02

DISCUSSION

The present study assessed the ability of TRUS-guided biopsy to accurately identify tumour location within the prostate. Biopsy showed relatively poor localization of either DG and DV by sextant or lobe. Serial biopsy slightly improved the ability to accurately identify the dominant tumour. No diagnostic, clinical or pathological factors were found to be significantly associated with better DG identification. By contrast, total length of tissue taken, total length of tumour identified, and time from biopsy to RP were significantly associated with more accurate identification of the DV. Based upon these results, one can speculate that extended core biopsy, as currently performed, does not reliably identify dominant tumour location with sufficient accuracy to direct focal therapy in those with early stage prostate cancer.

Our results are supported by previous work that showed that positive biopsies occur in the same lobe as the dominant tumour in only 53% of cases in patients with bilateral disease [20]. Furthermore, the ability to localize to the correct sextant in the present study varied widely. This has also been observed previously [3,8,21]. In general, although a biopsy may accurately identify tumour within a sextant, other locations of higher grade or volume are often missed. This highlights the ability of extended biopsy to detect cancer, but also shows its limited ability to localize the dominant tumour [11,15].

The higher rate of agreement for DV compared with DG shows that larger tumour size improves the rate of localization. In the present study, smaller lesions of higher grade were often missed at biopsy and only detected in the RP specimen. More frequent identification of DV compared with DG could account for the changes in Gleason score observed when comparing biopsy and RP specimens. Smaller lesions are less likely to be detected and would also have lower likelihood of clinical significance [22–24]. This has been reported previously, though it was noted to be dependent on the biopsy scheme [15,16,25,26]. While it is conceivable that treating the DV only by focal therapy is sufficient for good outcomes, it is equally likely the small DG lesions, which are more frequently missed, are inevitably responsible for prostate cancer morbidity and mortality.

Interestingly, we were unable to identify any diagnostic, clinical or pathological factors that were significantly associated with better DG identification. Various biopsy schemes have been evaluated, all with consistently low levels of accuracy that did not exceed those found in our study [3,8,11,12,20,21,25,26]. A previous study reported that increasing the number of cores taken per biopsy increases the detection rate of prostate cancer [25]. By contrast, total length of tissue taken, total length of tumour identified and time from biopsy to RP were significantly associated with more accurate identification of the DV. Although neither the volume of tumour identified nor the time from biopsy to RP can be modified in order to improve localization of DV, the total length of tissue taken, which depends entirely on the number of cores taken, can be improved. Unfortunately, the present study is not able to identify the ideal number of cores to take for DV localization.

One limitation of the present analysis is the small size of the cohort, which restricts the analysis and generalizability of these findings. The present study is the second largest of those directly addressing the issue of tumour localization by TRUS-guided biopsy [11,21]. A potential source of error is that review of entire RP specimens was not possible; specimens were sectioned into slide mounts and whole mounts. Only slide mounts were used for the present study, preventing complete visualization and mapping of tumour involvement within the prostate. To combat this, the volume of tumour burden per prostate sextant was characterized using ImageJ software rather than attempting to ascertain the three-dimensional shape of tumour within the prostate. A significant limitation is that ≈25% of biopsies were taken at outside institutions, generating variation in the exact biopsy scheme used between patients and biopsies, as well as a potential source of error in tumour identification. Additionally, some patient data were not supplied with these outside reports, decreasing the number of patients with data available for incorporation into Tables 2 and 3. Variation in diagnosis was minimized owing to re-review of all histology slides by in-house pathologists upon referral to UCSF as part of standard clinical practice. Follow-up biopsies were performed at UCSF. To compensate further, we analysed only those biopsies that resembled the extended biopsy scheme by excluding biopsies with <10 cores. Similarly, RP specimens were re-evaluated by a single genitourinary pathologist blinded to biopsy results, eliminating inter-observer variation in the determination of DG and DV.

A limitation discussed in previous similar studies is the lack of standardization in the assignment of prostate region [11,21]. Some systematic differences probably exist between the sextant designated by a urologist during biopsy and by a pathologist after RP. Also of note, the present study was done in a low-risk population with low volume disease. Though this population may have an increased the number of biopsies performed relative to a higher risk population, overall rates of identification might be underestimated as a result of pooling findings from multiple biopsies. The last biopsy before prostatectomy is assumed to have the highest rates while earlier biopsies would have a higher likelihood of missing DG and DV, resulting in lower overall rates of correct identification when all data is pooled together. To compound this, the low volume of disease makes detection more difficult, thereby lowering the rate of correct identification. Thus, our rates of correct identification are likely to underestimate those of populations of equal or higher risk and larger tumour volumes.

In conclusion, the present findings show that TRUS-guided biopsy detects and localizes DV better than it does DG. Yet even with an extended scheme, TRUS-guided biopsy does not reliably identify dominant tumour location in men with early stage prostate cancer. TRUS-guided biopsy may perform better in similar men with low stage, but higher volume disease.

CONFLICT OF INTEREST

Peter R. Carroll is an Investigator for the National Cancer Institute and TAP Pharmaceuticals.

Ancillary