Inability of shear-wave elastography to distinguish malignant from benign prostate tissue - a comparison of biopsy, whole-mount sectioning and shear-wave elastography.

AIM: This study was designed to assess the possible usefulness of shear-wave elastography in differentiating between benign and malignant tissue in prostate neoplasia. PATIENTS AND METHODS: A total of 120 prostate tissue samples were obtained from 10 patients treated by radical prostatectomy and investigated pre-operatively by ultrasound elastography followed by directed biopsy. After resection, whole-mount sectioning and histological examination was performed. The predictions based on shear-wave elastography were compared with biopsy and histological results.
RESULTS: The comparison between the results of shear-wave elastography and those of biopsy was performed by receiver operating characteristic analysis, which suggested an optimum cut-off tissue elasticity value of 50 kPa, in agreement with earlier studies aimed at distinguishing between benign and malignant tissue. However, the diagnostic selectivity (and thus the diagnostic power) was poor (area under the curve 0.527, which hardly differs from the value of 0.500 that would correspond to a complete lack of predictive power); furthermore, application of this cut-off value to the samples led to a sensitivity of only 74% and a specificity of only 43%. An analogous comparison between the results of shear-wave elastography and those of whole-mount histology, which itself is more reliable than biopsy, gave an even poorer diagnostic selectivity (sensitivity of 62%, specificity of 35%). Meaningful association with Gleason score was not found for D'Amico risk groups (p = 0.35).
CONCLUSIONS: The (negative) findings of this investigation add to the dissonance among results of studies investigating the possible value of shear-wave elastography as a diagnostic tool to identify malignant neoplasia. There is a clear need for further research to elucidate the diversity of study results and to identify the usefulness, if any, of the method in question.

Introduction

Prostate cancer is, among males in the Western world, the most frequent tumor and the third most frequent cause of cancer-related death (for a brief survey see Porsch et al.( and references therein). Although screening programs based upon digital rectal examination (DRE), transrectal ultrasound (TRUS) and serum prostate-specific antigen (PSA), along with modern diagnostic approaches, such as multiparametric magnetic resonance imaging, contrast-enhanced ultrasound (CEUS), computer-aided transrectal ultrasound and three-dimensional ultrasonographic histoscanning, have led to its more frequent and more accurate early diagnosis, the problem of distinguishing between benign and malignant tissue remains(.

A recent technique introduced for prostate cancer screening and diagnosis addition has been 3D ultrasonic elastography(, which however was critically dependent upon the skill of the operator in moving the ultrasound probe to provide the necessary compression of tissue. The more recent is shear-wave elastography (SWE)( in which an ultrasound probe sends a transverse wave into the tissue and detects the reflected wave, thus measuring the tissue’s distribution of elasticity in three dimensions and allowing this to be displayed by a computer-generated color map(, potentially showing any malignant growth, prostatitis or benign enlarged prostate nodules. SWE has shown promising preliminary results in the detection of prostate cancers and the assessment of their malignity in small-scale studies( and in larger cohorts of patients(.

Nevertheless, the mere detection of prostate cancer is of relatively little value in view of the high frequency of quiescent or non-aggressive neoplasms that develop extremely slowly, are not life-threatening and do not require treatment. At present this distinction is routinely made by multiple needle biopsy (the current “gold standard”), with the attendant risks and inconvenience, and also with frequent overdiagnosis leading to unnecessary treatment. There is therefore a clear need for a non-invasive method to recognize malignant prostate growth. We therefore conducted this study with the aim of finding out whether there is an elastographic cut-off value that allows the specific identification of malignant prostate tumor tissue.

Patients and methods

This study, registered under clinicaltrials.gov (NCT02425163), was conducted in the Department of Urology at Magdeburg University Hospital. It was approved by the institution’s local ethics committee. All patients were informed in adequate detail of the study’s aims and procedures, and gave their written informed consent to participate.

Between March 2012 and March 2013, a total of 73 patients received an SWE investigation using the SuperSonic Imagine Ultrasound System AIXPLORER and an endocavity sonic head SE12-3 (SuperSonic Imagine, Aix-en-Provence, France). Of these patients, 38 were diagnosed as having prostate cancer, and of these 38 cases a total of 10 were treated by radical prostatovesiculectomy (RPVE) between July 2012 and December 2013 on the basis of malignancy revealed by needle biopsy. From the resected tissue, 120 points/regions were investigated; each of these yielded both an elastographic result and a histopathological result. In this study we compared these results region by region: both the malignant and the benign regions were included in the analysis. At the same time, we examined the benign regions found in needle biopsy and having high elasticity values for possible benign prostate nodules due to benign prostate hyperplasia and inflammation.

The prostate glands were divided notionally into 12 regions of approximately equal volume (details are given elsewhere(; for a summary see Fig. 1). Each region was considered separately, and the point with the highest elastography value (stiffness in kPa) was biopsied. In this way, each anatomically defined region yielded a kPa value and a corresponding histological result. For each patient, 12 TRUS-guided random biopsies were then taken by using a device fitted to the sonic head (Magnum™; C.R. Bard GmbH, Karlsruhe, Germany) with twelve 18gauge/25 cm needles (Angiotech, Medical Device Technologies Inc., Gainesville, Florida, USA). The device produces a two-dimensional map of the whole prostate gland and its surroundings, on which the tissues’ elasticity (strictly: their stiffness, measured in kilopascals) is shown. This information is color-coded and overlaid on the B-mode image of the prostate in real time. An elastometric analysis of all the regions of the prostate glands has been described in an earlier publication(; here, we focus on the ten patients in whom the radical prostatectomy allowed a detailed, region-by-region comparison between histological and elastometric results. It was previously reported that elasticity is a weak criterion for distinguishing between benign and malignant prostate tissue, as well as between different D’Amico risk groups or Gleason scores, respectively(.

Fig. 1

Comparison of whole-mount section histology and TRUS and SWE scans. Upper left: The sonograms obtained before operation by TRUS and SWE. Centre: Schematic view of the prostate defining points/regions of biopsy. Right: The transversal section (indicated as a plane in the TRUS figure and the central sketch) after HE staining; in this example an apical section is shown. Below: Comparison of the results for each region and from each assessment method. The correlation between elasticity and the findings of biopsy or sectioning is poor (see text)

All histology samples were assessed, after hematoxylin–eosin (HE) staining, by the same two experienced physicians and discussed, if necessary, so that consensus was attained. Gleason grading and scoring were performed for each of the 12 regions separately.

Finally, the results from TRUS, SWE, biopsies and histological sections were compared by matched mapping.

A statistical analysis was performed with the software IBM SPSS Statistics, version 22. Initial results were presented by descriptive statistics (absolute and relative frequencies, mean, standard deviation, median). The receiver operating characteristic (ROC) analysis was then performed. The “optimum” cut-off elasticity value is determined mathematically by the maximum of the Youden Index (defined as “sensitivity + specificity – 1” and corresponding to the maximum distance between the plotted line and the diagonal); this was found to be 48 kPa, rounded to 50 kPa(. Quantitative variables were initially investigated for normal distribution by the Shapiro–Wilk test; this showed that a normal distribution could not be assumed, especially in the case of elasticity. Therefore, comparison of elasticity was performed by the Kruskal–Wallis test with a subsequent Dunn–Bonferroni test as a pairwise comparison (post hoc test). For comparison of Gleason scores in biopsy and histopathology, we used the symmetry test (Bowker) and, in addition, the Wilcoxon matched-pairs signed-rank test, as the symmetry test requires the same numbers of rows and columns in the contingency table and thus the exclusion of some cases. In all statistical tests, the significance level was defined by an error probability of α = 0.05.

Results

Patients and histological samples

An illustrative typical histological section and the corresponding TRUS and SWE scan results are shown in Fig. 1.

Findings for the 10 study patients are summarized in Tab. 1. In spite of the small size of the cohort, the distributions of age and disease characteristics are representative of such patients in general clinical routine, both in our experience and according to the literature(.

Tab. 1

Patient and tumor characteristics

Patients (N = 10)
Age [years]	mean ± SD	61.5 ± 6.3
	median (range)	62.5 (49–71)
PSA level [ng/ml]	mean ± SD	7.2 ± 2.3
	median (range)	6.0 (4.9–12.0)
Prostate volume [ml]	mean ± SD	32.6 ± 13.4
	median (range)	31.0 (18 – 65)
Suspicious finding in DRE [no – yes]	n	4 – 5*
Risk assessment† after biopsy [low – intermediate – high]	n	4 – 6 – 0
Risk assessment† after histology [low – intermediate – high]	n	1 – 7 – 2
All samples (N = 120)
Biopsy specimens benign – malignant	n (%)	67 (56%) – 53 (44%)
RPVE specimens benign – malignant	n (%)	62 (52%) – 58 (48%)
Biopsy specimens (N = 53)
Left – right	n (%)	25 (47%) – 28 (53%)
Gleason score 3 + 3 = 6	n (%)	41 (77%)
Gleason score 3 + 4 = 7a	n (%)	6 (11%)
Gleason score 4 + 3 = 7b	n (%)	6 (11%)
RPVE specimens (N = 58)
Gleason score 3 + 3 = 6	n (%)	19 (33%)
Gleason score 3 + 4 = 7a	n (%)	16 (28%)
Gleason score 4 + 3 = 7b	n (%)	7 (12%)
Gleason score 4 + 5 = 9	n (%)	4 (7%)
Gleason score 5 + 4 = 9	n (%)	12 (21%)

1 missing.

According to D’Amico(. SD, standard deviation

The Gleason scores found in biopsy and in histopathology differed, as shown in Tab. 2. This difference was significant, with p = 0.001 (two-sided McNemar–Bowker test). The Wilcoxon test (see Methods) to include all samples gave a similar result.

Tab. 2

Lack of correspondence between Gleason scores as determined by needle biopsy and by histopathology

		Result according to RPVE specimen
		Benign	Gleason 6	Gleason 7	Gleason 9	Total
Result according to needle biopsy	BenignGleason 6Gleason 7Gleason 9	40175–	613––	1571–	–––16	6137616
	Total	62	19	23	16	120

Elasticity cut-off as a possible predictor of malignancy

The elasticity values found at the points where the 120 samples were taken lay between 11.5 and 215.1 kPa. Values between these were used as notional cut-off values to generate standard ROC plots in the same manner as described previously(. The results are shown in Fig. 2A. The upper curve shows the ROC analysis for elasticity as a criterion for malignancy compared with actual malignancy as assessed by needle biopsy. The closeness of the ROC curve to the diagonal gives a prima facie impression that elasticity is not a useful predictive criterion for this. This observation was pursued quantitatively: the area under the curve was 0.527 (95% confidence limits 0.423 and 0.631); this hardly differs from the value of 0.500 that would correspond to a complete lack of any predictive power.

Fig. 2

ROC analyses: A. ROC analysis with elasticity used as a criterion for malignancy compared with actual malignancy as assessed by needle biopsy. B. ROC analysis with elasticity used as a criterion for malignancy compared with actual malignancy as assessed by histology. For details, see text

The ROC curve in Fig. 2A was found to lie furthest from the diagonal (Youden index, the criterion usually applied for optimizing the cut-off value; see Methods) at an elasticity of 50 kPa, and the result of classifying the needle biopsy samples for malignancy according to this criterion are shown in Tab. 3A. Inspection of Tab. 3A makes it clear that the criterion “elasticity >50 kPa” does not distinguish well between benign and malignant tumors as identified by needle biopsy. Moreover, the sensitivity thus obtained (detected true positives / total true positives) was 38/53 = 72% and the specificity (detected true negatives / total true negatives) was 29/67 = 43%. (The positive and negative predictive values were calculated to be 50% and 66%, respectively.) Thus only 72% of malignant neoplasms and only 43% of benign neoplasms were recognized correctly.

Tab. 3

Classification of lesions according to elasticity = 50 kPa using needle biopsy or histopathology as reference

A	Actual status according to needle biopsy		Total
	Benign	Malignant
Elasticity < 50 kPa inferred: benign Elasticity ≥ 50 kPa inferred: malignant	29 38	1538	4476
Total	67	53	120
B	Actual status according to histopathology		Total
	Benign	Malignant
Elasticity < 50 kPa inferred: benign Elasticity ≥ 50 kPa inferred: malignant	2240	2236	4476
Total	62	58	120

A similar analysis of the samples was performed, but using the histopathological results (whole-mount sectioning) as reference. Results are shown in Fig. 2B and Tab. 3B. In this case, the failure of the ROC curve to lie above the diagonal is even clearer (area under the curve 0.457 with 95% confidence limits 0.363 and 0.571), and it is impossible to find an optimum cut-off value for elasticity. Using the criterion of 50 kPa, as before, led to the assignments shown in Tab. 3B, which in turn implied a sensitivity of 36/58 = 62% and a selectivity of 22/62 = 35%, both of which may be regarded as valueless. (The positive and negative predictive values were calculated to be 47% and 50%, respectively.)

Association of tissue elasticity with Gleason grading

The distributions of elasticity values by Gleason grading/ score – the latter determined both by biopsy and by histopathology – are shown in Tab. 4. For the biopsy-derived scores, a clear trend is seen, p = 0.001 (Kruskal–Wallace test; for details see Methods); for the pairwise comparison between Gleason 3+3=6 and 3+4=7, p = 0.003 was obtained. For the histopathology-derived scores, the difference is less clear, but significance threshold was nonetheless reached (p = 0.022 and 0.017, respectively). However, in the latter assessment no difference was found between the elasticity values in a comparison between Gleason scores of 6 and 9, which were clearly very similar, with identical means and similar ranges (Tab. 4).

Tab. 4

Elasticity by Gleason scores as determined by needle biopsy and by histopathology

		n	Mean ± SD	Median	Range
Result according to needle biopsy	BenignGleason 3 + 3 = 6Gleason 3 + 4 = 7Gleason 4 + 3 = 7	674166	67 ± 3759 ± 2495 ± 10105 ± 59	63549783	12 – 21518 – 13983 – 10853 – 206
	All	120	67 ± 35	58	12 – 215
Result according to histopathology	BenignGleason 3 + 3 = 6Gleason 3 + 4 = 7Gleason 4 + 3 = 7Gleason 4 + 5 = 9Gleason 5 + 4 = 9	6219167412	69 ± 3777 ± 2848 ± 1953 ± 3268 ± 1877 ± 51	598045376663	12 – 21531 – 14813 – 9014 – 10150 – 9130 – 207
	All	120	67 ± 35	58	12 – 215

Tissue elasticity as a possible predictor of risk distribution in needle biopsy and histological section

As was found for the Gleason scores (see above), the distribution of risk assessment differed according to whether it was based on needle biopsy or histopathological examination (p < 0.001; details not shown). A breakdown of elasticity values by risk assessment is shown in Tab. 5.

Tab. 5

Elasticity by risk assessment as determined by needle biopsy and by histopathology

		n	Mean ± SD	Median	Range
Risk according to needle biopsy	LowIntermediate	3672	55 ± 3268 ± 34	5069	13 – 21512 – 207
	All	108	64 ± 34	56	12 – 215
Risk according to histopathology	LowIntermediateHigh	187416	60 ± 4163 ± 2975 ± 45	515764	31 – 21512 – 14830 – 207
	All	108	64 ± 34	56	12 – 215

A significant difference was found between the elasticity values of the low-risk and intermediate-risk groups as assessed by biopsy (p = 0.019, Mann–Whitney test). However, elasticity did not differ significantly between high-risk and other carcinomas as assessed by histopathology (p = 0.35, Kruskal–Wallace test). This implies that the elasticity of high-risk carcinoma is not significantly higher than that of intermediate-risk carcinoma. This again shows that the ultrasound procedure does not allow risk predictions in this case.

Comparison of high stiffness with the corresponding areas that histologically appeared benign

Seventy-six regions (cf. Tab. 3) investigated revealed SWE values above 50 kPa and were to be regarded as potentially malignant. However, microscopy showed that they all consisted entirely of benign tissue. We investigated more closely the regions with SWE values between 70 kPa and 120 kPa. In all of these cases, the tissue was typical of benign prostatic hyperplasia, with a high proliferation rate and dense cell architecture. The looser glandular structure was replaced by functionless hyperplasic nodes, mainly with stroma and myofibroblast proliferation. The tissue also contained many corpora amylacea (prostatic concretions), glandular ducts with inspissated prostate secrete that can lead to prostate stones, all of which are signs of inflammation. The sections examined revealed clear signs of inflammation in the regions of high elasticity, with a high occurrence of lymphocyte aggregates and of granulocytes of leucocyte alkaline phosphatase activity. However, none of them were found to contain malignant tissue. A detailed investigation of the region with the highest elasticity value (215 kPa) likewise revealed diverse inflammatory and post-inflammatory processes, while both biopsy and histological sectioning characterized it as benign.

Discussion

Although we found that the clearest distinction between benign and malignant tissue was obtained at an elasticity value of 50 kPa, in agreement with others(, this was nevertheless found to be inadequate to differentiate usefully between the two. Moreover, the identification of 50 kPa was based on the apparent relationship observed between SWE and needle biopsy only; when the whole-mount histopathological analysis was used instead of needle biopsy as a reference for the accuracy of the SWE result, then no such relationship was found. The previous analysis showed unexpectedly that the elasticity of benign prostate tissue depends on the anatomical prostate region, the tissue being the stiffest in the basal region and more elastic at the apex(.

The problematical nature of SWE is illustrated by the fact that the region with the highest elasticity value (215 kPa, well above the empirical cut-off value of 50 kPa) was in fact found by both biopsy and sectioning to be benign. Instead of malignancy, it revealed inflammatory and post-inflammatory processes. This poses a new problem in the interpretation of SWE: far from distinguishing solely between benign and malignant regions, it is also influenced by (benign) inflammatory and fibrotic tissue, which obviously also can cause peak elasticity values in this highly inhomogeneous tissue.

In a study by the Correas group(, a maximum Youden index of 81% (95%CI: 78%, 84%) was found, and the use of a cut-off value of 35 kPa for SWE and AUC of 95% (95% CI: 93%, 97%) was proposed for the distinction between malignant and benign tissue, in contrast to the value put forward by Boehm et al. (. An earlier study by our group suggested a cut-off of 50 kPa for distinguishing between prostate carcinoma with Gleason scores of 6 and 7(. These results did indeed suggest that SWE may have a discriminatory potential.

Similar studies have been conducted in other organs. These have included, inter alia, the liver (reviewed by Conti et al. () and the thyroid gland(. In both cases potentially promising results have been obtained. However, Conti et al. conclude that there is a need for further and higher-quality prospective studies to assess the potential value of SWE in differentiating between benign and malignant lesions and among different lesions. Dobruch-Sobczak et al. showed medullary thyroid carcinoma to reveal lesions that were stiffer than the surrounding tissues but were unable to draw clear conclusions about the usefulness of SWE. In both cases the authors recommend caution and call for larger-scale studies at a level of evidence higher than that attained until now: for liver, 4 on the Oxford scale(; for thyroid(, and for the present study the level is similar, as all of these studies are based on case series. In contrast to this, the guidelines of the European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) posit that elastometry can be useful in distinguishing between mild liver fibrosis (or none) on the one hand and significant liver fibrosis on the other(. Similarly, in the diagnosis of mammary tumors, the World Federation for Ultrasound in Medicine and Biology (WFUMB) regards SWE as a potentially important diagnostic technique(.

Conclusion

Despite early results from various groups suggesting a potential usefulness of SWE in differentiating between benign and malignant prostate tissue – or between grades of potential malignancy – this work supports the assertion that the usefulness of SWE for the prostate gland may initially have been overestimated.