Transrectal sonoelastography in the detection of prostate cancers: a meta-analysis

Authors


  • J.T. and M.C. contributed equally to this article.

Danfeng Xu, Shanghai Changzheng Hospital, Second Military Medical University, Department of Urology, Fengyang 415, Shanghai 200003 China. e-mail: xudanfeng415@hotmail.com

Abstract

Study Type – Diagnostic (exploratory cohort)

Level of Evidence 3a

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

The accuracy of transrectal sonoelastography (TRSE) in the detection of prostate cancer is variable, with a sensitivity ranging from 51.1 to 91.7% and specificity ranging from 62.2 to 86.8%.

This is the first meta-analysis to assess the overall accuracy of TRSE in the detection of prostate cancer.

OBJECTIVE

  • • To assess the overall accuracy of transrectal sonoelastography (TRSE) targeted biopsy in the diagnosis of patients suspected of having prostate cancer (PCa).

METHODS

  • • A systematic search of electronic databases, including PubMed, Embase and The Cochrane Library, and manual bibliography searches were performed.
  • • All relevant studies assessing the diagnostic accuracy of TRSE in PCa detection were included in our meta-analysis.
  • • The data were pooled and sensitivity, specificity, area under the curve (AUC), positive likelihood ratio (LR) and negative LR were calculated.

RESULTS

  • • Pooled patient data analysis: the pooled (95% confidential intervals [95% CI]) sensitivity of TRSE targeted biopsy in patients suspected of having PCa was 62 (55–68) %; specificity was 79% (74–84%); AUC was 0.7696; positive LR was 2.92 (2.28–3.74); and negative LR was 0.49 (0.41–0.59).
  • • Pooled core data analysis: pooled (95% CI) sensitivity, specificity, positive LR and negative LR were 34% (30–38%), 93% (91–95%), 5.07 (3.91–6.57) and 0.71 (0.66–0.75), respectively.

CONCLUSION

  • • Transrectal sonoelastography is a promising technique in PCa detection and can be considered to be a valuable supplemental method to systemic biopsy.
Abbreviations
TRSE

transrectal sonoelastography

PCa

prostate cancer

AUC

area under the curve

LR

likelihood ratio

I2

inconsistency

SROC

summary receiver operating characteristic

PPV

positive predictive value

INTRODUCTION

Prostate cancer (PCa) is the second most common cancer and the sixth leading cause of cancer-related death in males worldwide [1]. It is generally believed that the incidence is much higher in most of the economically developed countries. With an estimated new cases of 217 730 in the USA in 2010, PCa is expected to overtake lung cancer as the most frequent cancer [2]. In other countries, such as Japan and several central and eastern European countries, the incidence is still on the rise [3,4].

Prostate cancer is potentially curable and has a better prognosis if early diagnosis is achieved. Unlike other malignant lesions, PCa is difficult to detect using conventional imaging techniques [5]. Currently, the recommended diagnostic strategy for PCa is TRUS-guided 10- to 12-core systemic biopsy [6]; however, up to 47% of clinically relevant cancers have been mis-diagnosed using this approach [7]. Meanwhile, increasing costs and biopsy-related complications also call for an alternative approach in PCa detection [8], so new strategies for effectively and safely detecting PCa are urgently required.

Sonoelastography is a promising method developed based on the fact that malignant tissues are generally stiffer than benign and normal tissues [9]. By comparing images obtained under compression and decompression, tissue elasticity is assessed. The accuracy of transrectal sonoelastography (TRSE) in PCa detection has been discussed in various studies. In patients suspected of having PCa, TRSE targeted biopsy yields a sensitivity ranging from 51.1 to 91.7% and a specificity ranging from 62.2 to 86.8% [5,8,10–17], while in patients with proven PCa, the sensitivity and specificity of TRSE are ∼80% based on step-section pathological analysis after radical prostatectomy [18,19].

The aim of the present study was to perform a meta-analysis of published papers to assess the overall accuracy of TRSE targeted biopsy in PCa detection.

METHODS

LITERATURE SEARCH

Relevant studies were identified by searching the electronic databases, including PubMed, Embase, Web of Science and The Cochrane Library using the search terms ‘elasticity imaging techniques’, ‘elastography’, ‘elasticity’, ‘strain imaging’, ‘prostate’ and ‘prostatic’ (Table 1). We also performed a full manual search from the reference list of each relevant article. No language restrictions were applied.

Table 1. Search strategies used and results
DatabaseDateSearch strategyResults
PubmedUp to Nov. 2011(‘Elasticity Imaging Techniques’[MH] OR elastography OR elasticity OR ‘strain imaging’) AND (prostate OR prostatic)222
EmbaseUp to Nov. 2011(elastography:ab,ti OR elasticity:ab,ti OR ‘strain imaging’:ab,ti) AND (prostate:ab,ti OR prostatic:ab,ti)170
Web of ScienceUp to Nov. 2011TS = (elastography OR elasticity OR ‘strain imaging’) AND TS = (prostate OR prostatic)243
Cochrane LibraryCochrane Library Issue 11, 2011(elastography OR elasticity OR ‘strain imaging’) AND (prostate OR prostatic) :ti,ab,kw2

STUDY SELECTION

Inclusion and exclusion criteria were defined before the literature search. Studies meeting the following criteria were included: study design: diagnostic clinical trials evaluating the accuracy of TRSE in the detection of PCa; and study population: patients with elevated serum PSA (>4 ng/mL), abnormal DRE, hypoechoic nodules on TRUS or low-intensity lesions on T2-weighted images on MRI. The exclusion criteria were: studies from which we were unable to construct 2×2 tables; studies in which the same samples were assessed; studies where patients were diagnosed with PCa or had previously undergone biopsy.

Two independent reviewers completed this procedure and resolved disagreements by consensus.

ASSESSMENT OF STUDY QUALITY

The methodological quality of the studies included in the meta-analysis was assessed using the QUADAS questionnaire. Recommended by The Cochrane Collaboration, the QUADAS is the only tool that includes evaluation criteria for quality assessment of diagnostic accuracy studies [20]. Items were rated as ‘yes’, ‘no’, or ‘unclear’.

DATA EXTRACTION

The following variables were recorded: authors, journal, year of publication, number of patients and number of TRSE positives, false-positives, false-negatives and TRSE negatives by patients and by cores. If necessary, the first author or the corresponding author was contacted for further information.

STATISTICAL ANALYSIS

A meta-analysis for evaluating the overall diagnostic accuracy of TRSE in the detection of PCa was performed. The effective sample size (Deeks' funnel plots) [21] and associated regression test of asymmetry were used to detect publication bias. A P value <0.10 indicated the presence of publication bias.

The pooled estimates of sensitivity, specificity, positive likelihood ratio (LR) and negative LR were calculated by applying the fixed-effect model, according to the Mantel–Haenszel method [22], when no significant heterogeneity presented. Otherwise, the random-effect model was applied as described by DerSimonian and Laird [23]. The Cochrane Q test was used to assess the heterogeneity among studies. A P value <0.10 indicated heterogeneity. Inconsistency (I2) was also calculated to evaluate heterogeneity. An I2 value <50% was considered acceptable. The summary receiver operating characteristic (SROC) curves were obtained using the Mantel–Haenszel method [24]. The area under the curve (AUC) was calculated. An AUC value of 0.5 indicated a poor diagnostic trial, while 1.0 indicated a perfect trial.

Deeks' funnel plots were obtained, and an associated regression asymmetry test was carried out using Stata 10.0 (Stata Corporation, College Station, TX, USA). The pooled sensitivity, specificity, positive LR, negative LR and SROC curve were obtained using Meta-DiSc 1.4 (Unit of Clinical Biostatistics, Ramony Cajal Hospital, Madrid, Spain).

RESULTS

The initial results of the literature search are shown in Table 1. After the exclusion of duplicated studies, 404 potentially relevant studies were identified and screened for retrieval. Six studies were finally included in the meta-analysis (Fig. 1). The baseline characteristics of patients in the included studies are shown in Table 2[10–15].

Figure 1.

Flow chart showing selection of studies for meta-analysis.

Table 2. Baseline characteristics of the included studies
StudyCompression methodReference standardThresholdGold standardNo. of patientsNo. of coresAge, yearsPSA, ng/mL
  1. NR, not reported.

Cochlin et al. 2002 [10]Free handColour patternMinimal or no colourBiopsy10062264NR
Kamoi et al. 2008 [13]Free handColour patternBlue areaBiopsy107NR68.411.4
Ferrari et al. 2009 [11]Free handColour patternBlue areaBiopsy8489461.3NR
Romagnoli et al. 2010 [14]Free handColour patternBlue areaBiopsy88NR66.86.5
Giurgiu et al. 2011 [12]Free handColour patternBlue areaBiopsy65NR68NR
Zhang et al. 2011 [15]Free handStrain index17.44Biopsy83NR68.61NR

The QUADAS questionnaire was used for quality assessment. With most items rating as ‘yes’, the included studies were of good quality (Table 3[10–15]). Because no study described whether the clinical data were available when test results were interpreted, the item clinical review bias was rated as unclear. Only two studies [13,15] clearly stated that the pathologists who analysed the prostate specimens were blinded to the results of TRSE.

Table 3. Quality assessment of included studies using QUADAS questionnaire
QUADAS itemStudy
Cochlin et al. (2002) [10]Kamoi et al. (2008) [13]Ferrari et al. (2009) [11]Romagnoli et al. (2010) [14]Giurgiu et al. (2011) [12]Zhang et al. (2011) [15]
  1. Y = yes; N = no; U = unclear.

 1. Spectrum compositionYYYYYY
 2. Selection criteriaYYYYYY
 3. Disease progression biasYYYYYY
 4. Partial verification biasYYYYYY
 5. Differential verification biasYYYYYY
 6. Incorporation biasYYYYYY
 7. Test review biasYYYYYY
 8. Reference standard review biasUYUUUY
 9. Clinical review biasUUUUUU
10. Uninterpretable resultsYYYYYY
11. WithdrawalsYYYYYY
‘Y’ rate, %81.890.981.881.881.890.9

The Deeks' funnel plots obtained are shown in Fig. 2. The effective sample size-weighted regression tests of funnel plot asymmetry gave a P value of 0.681.

Figure 2.

Deeks' funnel plots.

TRSE IN THE DETECTION OF PCa

Pooled data analysis by patient

The pooled (95% CI) sensitivity of TRSE biopsy in patients suspected of having PCa was 62 (55–68%), and no significant heterogeneity was found (chi squared = 8.75, df = 5, P= 0.1196, I2= 42.8%). The pooled (95% CI) specificity was 79 (74–84%), and no significant heterogeneity was found (chi squared = 8.69, df = 5, P= 0.1219, I2= 42.5% [Fig. 3]). The SROC curve was drawn and the calculated AUC was 0.7696 (Fig. 4).

Figure 3.

Forest plots of sensitivity (upper) and specificity (lower) for PCa detection.

Figure 4.

The SROC, with 95% CIs, for PCa detection. Q*, the point where the sensitivity and specificity are equal.

The pooled (95% CI) positive LR was 2.92 (2.28–3.74), and no significant heterogeneity was found (chi squared = 7.39, df = 5, P= 0.1935, I2= 32.3%). The pooled (95% CI) negative LR was 0.49 (0.41–0.59), and no significant heterogeneity was found (chi squared = 7.47, df = 5, P= 0.1879, I2= 33.1% [Fig. 5]).

Figure 5.

Forest plots of positive LR (upper) and negative LR (lower) for differentiation for PCa detection.

Pooled data analysis by core

We conducted analysis by core on the detection of PCa using TRSE. Disappointingly, only two studies [10,11] provided original data to construct the 2×2 tables, thus the pooled estimates were calculated based on these two studies. The pooled (95% CI) sensitivity, specificity, positive LR and negative LR were 34 (30–38)%, 93 (91–95)%, 5.07 (3.91–6.57) and 0.71 (0.66–0.75), respectively. No significant heterogeneity was observed (Table 4).

Table 4. Pooled estimates by cores
 No. of coresPooled estimates95% CI P I2
Sensitivity151634%30–38%0.215234.9%
Specificity151693%91–95%0.80890.0%
Positive LR15165.073.91–6.570.72790.0%
Negative LR15160.710.66–0.750.252523.6%

DISCUSSION

Early diagnosis of PCa still remains a challenge. To date, none of the commonly used imaging techniques, TRUS, CT or MRI, is able to provide optimum recognition of tumour masses within the prostate. This has fuelled interest in developing new imaging techniques that can offer high overall sensitivity and specificity in PCa diagnosis, and thus be of benefit when choosing the subsequent treatment.

Currently the ‘gold standard’ for PCa diagnosis is TRUS-guided systemic biopsy. Patients with elevated PSA levels, suspicious DRE or imaging findings are referred for this procedure, but because of the high sensitivity but low specificity of PSA, many patients with benign changes such as BPH and prostatitis are scheduled for biopsy [8,25–27]. In 2007, about 2 000 000 prostate biopsies were performed in the USA but only 220 000 cancers were detected [28]. Furthermore, the reported sensitivity of 12-core systemic biopsy is only 53% [7]. Increasing the number of biopsy cores, e.g. by extensive saturation biopsy, yields no significant improvement in cancer detection rates but increases morbidity [7,29].

Although TRUS is a convenient method, its diagnostic value in PCa has been reported to be limited to a sensitivity and specificity of 17–53% and 40–63%, respectively [30]. Colour Doppler is used to detect neovascularization and an enhanced blood flow, which are the main characteristics of various tumours but, in PCa, hypervascularity seems to be associated only with high Gleason score lesions [31–34] and it has been reported that colour Doppler targeted biopsy would miss up to 40% of patients with PCa [34]. MRI is a potential technique for detecting PCa, with an accuracy of 73% for local staging [35], but it requires an endorectal coil in conjunction with a pelvic phased array coil to obtain the image, and the cost of this procedure is very high.

Increased cell density of a tumour mass leads to a change of tissue elasticity and the stiffness of normal tissue is significantly different from neoplastic tissue [9,36]. On this basis, the technique of sonoelastography was developed to visualize tissue strains and has enabled cancer detection in many organs such as the thyroid gland, breast and prostate [10,37,38]. Nevertheless, the reported sensitivity and specificity of sonoelastography in PCa detection differ among studies. The present study was conducted to give a pooled estimates based on the existing data. To the best of our knowledge, it is the first meta-analysis evaluating the diagnostic value of TRSE in the detection of PCa.

After strict screening of the literature, six relevant studies were included. Deeks' funnel plots and regression tests of asymmetry showed no significant publication bias and according to the QUADAS questionnaire, the selected studies were of high quality. In addition, no significant heterogeneity was found among studies and these facts certified the quality and the reliability of our meta-analysis.

The sensitivity and specificity values found for the pooled TRSE data by patient were 62 and 79%, respectively, which are much greater than the sensitivity and specificity of the traditional TRUS for the detection of PCa.

The following factors can influence diagnostic accuracy in the studies included in the present meta analysis. Firstly, most PCa tends to grow along the prostate capsule, hence a round nodule-like appearance with a defined margin is not present [39,40] and this makes PCa detection by imaging techniques difficult. Secondly, most studies use a qualitative threshold such as ‘blue area predominant’ for diagnosis which might lead to vast inter- and intra-observer variability. Zhang et al. [15] used a quantitative value of strain index for diagnosis and yielded higher sensitivity (74%) and specificity (83%). Thirdly, the compression applied on the prostate is induced manually, which is poorly reproducible and compromises diagnostic quality. In a study by Masahiro et al. [41], using real-time balloon inflation elastography, the sensitivity and specificity were as high as 72.5% and 97.7%, respectively. We hypothesize, therefore, that a quantitative threshold for diagnosis and automatic compression would facilitate PCa detection.

With regard to analysis by core, only two studies provided the original data and the pooled estimates were calculated from those. Lower sensitivity (34%) and higher specificity (93%) were found. It is more important to diagnose PCa by patient rather than by core clinically, because all patients with PCa are scheduled for radical prostatectomy if no contraindication is found. Thus, the lower sensitivity might be compensated for by the multifocal feature of PCa.

A PSA level >10 ng/mL is considered to be an absolute indication for biopsy. Giurgiu et al. [12] reported higher sensitivity (76.47% vs 67.85%) and positive predictive value (PPV [84.25 vs 57.57%]) in a subgroup of patients with a PSA >10 ng/mL. Kamoi et al. [13] also reported a higher PPV (73 vs 69%) in this subgroup. This could be because tumours with higher Gleason score have higher cell density and might be detected better using TRSE. Unfortunately, we were unable to conduct a meta-analysis on this subgroup of patients because of incomplete data.

The present meta-analysis has several limitations. First, only six studies were included in the analysis by patient and only two studies were included in the analysis by core. Second, we did not carry out subgroup analysis in patients whose PSA was >10 ng/mL. Third, the standards of detecting a suspicious lesion in TRSE were different among the included studies. Fourth, different elastography systems, such as Hitachi EUB-8500, Hitachi EUB-6500, Hivision 900 and Toshiba tissue Doppler, were used in the included studies, and this might have led to bias. Large-scale trials are required to further explore the diagnostic value of TRSE.

In conclusion, TRSE is a promising technique in PCa detection. It could be a valuable supplemental method to systemic biopsy and we believe that, with the evolution of TRSE equipment, it might decrease the number of overall biopsy cores required for diagnosis in the future.

CONFLICT OF INTEREST

None declared.

Ancillary