New aspects in shear-wave elastography of prostate cancer

Abstract

Aim

This study was designed to evaluate the performance of shear-wave elastography as a diagnostic tool for prostate cancer in a larger cohort of patients than previously reported.

Patients and methods

Seventy-three patients with suspected prostate carcinoma were investigated by ultrasound elastography followed by directed biopsy. The elastographic and histological results for all biopsies were compared.

Results

After exclusion of invalid and non-assessable results, 794 samples were obtained for which both a histological assessment and an elastometric result (tissue stiffness in kPa) were available: according to the histology 589 were benign and 205 were malignant. Tissue elasticity was found to be weakly correlated with patient's age, PSA level and gland volume. ROC analysis showed that, for the set of results acquired, elastometry did not fulfil literature claims that it could identify malignant neoplasia with high sensitivity and specificity. However, it did show promise in distinguishing between Gleason scores ≤6 and >6 when malignancy had already been identified. Unexpected observations were the finding of a smaller proportion of tumours in the lateral regions of the prostate than generally expected, and also the observation that the elasticity of benign prostate tissue is region-sensitive, the tissue being stiffest in the basal region and more elastic at the apex.

Conclusions

Shear-wave elastography was found to be a poor predictor of malignancy, but for malignant lesions an elasticity cut-off of 80 kPa allowed a fairly reliable distinction between lesions with Gleason ≤6 and those with Gleason >6. We demonstrate an increase in elasticity of benign prostate tissue from the basal to the apical region.

Introduction

Prostate cancer affects many men world-wide; for example, Germany (population 80 million) has an annual incidence of 70,000 new cases and mortality of 13,000⁽¹⁾. Among males it is thus the most frequent tumour and the third most frequent cause of cancer-related death, after lung and intestine cancers⁽¹⁾. The introduction of screening measures in the mid-1980s, such as the prostate-specific antigen (PSA) test, led to a substantial increase in the number of positive diagnoses; however, despite many early therapeutic options, the mortality figures remain largely unchanged^{(2, 3)}. One reason for this is the difficulty of distinguishing between benign and malignant tissue, which is expressed as a high rate of false positive results in the PSA and other tests. Moreover, the result of even a multiple-needle biopsy can, on account of the nature of the sampling, easily be a false negative. Thus the consequences of diagnostic uncertainty are on the one hand the many possible side effects of active therapy such as radical prostatectomy or radiotherapy, and on the other the risk of delaying necessary surgical removal of neoplasms of unrecognised malignancy.

Modern diagnostic approaches have included multiparametric magnetic resonance imaging⁽⁴⁾, contrast-enhanced ultrasound – CEUS⁽⁵⁾, computer-aided transrectal ultrasound (C-TRUS)⁽⁶⁾ and three-dimensional (3D) ultrasonographic histoscanning⁽⁷⁾. According to current guidelines^{(8, 9)} diagnosis should include PSA level measurement, digital rectal examination (DRE) and transrectal ultrasound (TRUS). By combining PSA and DRE a positive predictive value (PPV) of 60.6% can be achieved; DRE alone attains 31.4% and PSA alone 42.1%⁽¹⁰⁾.

A recent addition to this armoury has been 3D ultrasonic elastography. Here a 3D image is generated that reveals regions of low tissue elasticity, corresponding to high cell density and potentially neoplasm – ultimately a refinement of the classical digital rectal examination. This was introduced two decades ago⁽¹¹⁾, but it depended upon the skill of the operator in moving the ultrasound probe to provide the necessary compression of tissue.

A recent refinement is shear-wave elastography (SWE)⁽¹²⁾ in which the elasticity of tissue is detected by its response to an (operator-independent) compression wave sent into the tissue by the ultrasound probe⁽¹³⁾; the reflected wave is used to obtain a computer-generated 3D colour map of the prostate tissue's elasticity and thus to reveal potentially malignant growth. SWE has shown promising preliminary results in the detection and characterisation of prostate cancers^(14–16) in a limited number of patients. The present study was aimed at assessing SWE performance in a larger cohort of patients. In the meantime a report of a study of 1040 biopsies has been published⁽¹⁶⁾; see Discussion.

Patients and methods

This study, registered under ClinicalTrials.gov, was conducted in the Department of Urology at Magdeburg University Hospital. It was approved by the institution's local ethics committee. All patients gave written informed consent.

Patients were included if there was suspicion of prostate carcinoma, based on DRE, on a high PSA level (>4 ng/ml) or a high PSA velocity (increasing by >0.75 ng/ml year), including Patients with earlier (negative) biopsies.

Between March 2012 and March 2013 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). Biopsies were then taken by a device fitted to the sonic head (Magnum™; C.R. Bard GmbH, Karlsruhe, Germany) with twelve 18-gauge/25 cm needles (Angiotech, Medical Device Technologies Inc., Gainesville, Florida, USA). The device yields a twodimensional map of the prostate gland and its surroundings, on which the tissues’ elasticity (strictly: their stiffness, measured in kilopascals, kPa) is shown. This information is colour-coded and overlaid on the B-mode image of the prostate in real time.

The prostate glands were divided notionally into 12 regions of approximately equal volume (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. All histology samples were assessed by the same experienced physician.

Fig. 1.

Regions of the prostate gland as defined for this study. Schematic representation of the prostate in coronal section, showing the regions defined for this study. Nos. 1–3 and 7–9 are lateral; nos. 4–6 and 10–12 are medial. The 12 regions were of approximately equal volume. No distinction between front and back is made

Statistical analysis was performed with the software IBM SPSS Statistics, version 21.

Results

Patients and histological samples

Of the 73 men for whom data were acquired, 4 were withdrawn from analysis because of EBRT (external beam radiation therapy). The study population thus comprised 69 men aged 65 ± 8 years (mean ± standard deviation, SD), with a range from 43 to 79 years. They had PSA values from 0.83 to 323 ng/ml (median 7.7 mg/ml, mean ± SD 18 ± 42 ng/ml). Gland size was 15.6 to 127 ml (median 44 ml, mean ± SD 50 ± 24 ng/ml). According to DRE, 31 patients (45%) showed findings giving rise to suspicion of prostate cancer, while the other 38 (55%) did not.

Biopsy yielded 827 valid samples: 589 benign, 205 malignant and 33 with prostatic intra-epithelial neoplasia. To allow clear-cut testing of the method at this early stage, the latter were excluded, giving 794 samples used for the analysis of which histologically 589 (74%) were benign and 205 (26%) malignant.

Elasticity and location of benign or malignant tissue

Figure 2 shows an example of SWE output. The two areas shown in dark red have a high stiffness: the smaller, oval one is inside the gland and is a candidate for histological investigation, while the larger, peninsular one is outside the gland and therefore irrelevant (the mere fact of a high kPa value does not in itself imply neoplasia).

Fig. 2.

Elastographic measurement. The lower part of the figure shows the conventional ultrasonographic image of a prostate gland. In the upper part of the figure, the elastographic image is superposed onto the lower figure. The small, oval region in the centre and the large peninsular region on the left, shown in red, are stiffer (low elasticity; compare the scale at upper right). See text for details

We first examined the elasticity of the benign tissue. Unexpectedly, values were associated with the depth of the section where they were measured, as shown in the left-hand block in Table 1: basal tissue (i.e. closest to the bladder) was stiffest, mid-gland less so, and apical tissue the most elastic. A Kruskal–Wallis test gave p < 0.001; the low value is not unexpected in view of the very large number of samples. Pairwise, α-adjusted comparison of the three regions gave p < 0.001 for apex/base, p = 0.001 for mid-gland/base and p = 0.069 for apex/mid-gland.

Tab. 1.

Elasticity (stiffness in kPa) of benign and malignant tissue

	Benign (n = 588)	Malignant (n = 204)
Basal	81 ± 37	104 ± 58
Mid-gland	68 ± 30	82 ± 37
Apical	61 ± 34	65 ± 38

The elasticity values of the malignant neoplasias were compared in a similar way (Table 1, right-hand block). Again, a statistically significant difference was found (Kruskal–Wallis, p = 0.004). Pairwise, α-adjusted comparison gave p = 0.003 for apex/base, p = 0.112 for mid-gland/base and p = 0.706 for apex/mid-gland.

In collecting our data set we noticed that the frequency of malignant lesions appeared to be fairly equal between the medial and lateral regions (see Fig. 1). Details are shown in Table 2.

Tab. 2.

Benign and malignant lesions according to region

	Lateral regions			Medial regions
	Benign	Malignant	All	Benign	Malignant	All
Base, right	49	18	67	49	18	67
Base, left	50	15	65	48	17	65
Base, total	99	33	132	97	35	132
Mid-gland, right	45	21	66	47	20	67
Mid-gland, left	53	13	66	51	15	66
Mid-gland, total	98	34	132	98	35	133
Apex, right	48	18	66	49	16	65
Apex, left	51	16	67	49	18	67
Apex, total	99	34	133	98	34	132

Elasticity and patient/tumour characterisitcs

For the entire set of points biopsied (irrespective of malignancy status), the correlation of stiffness with other characteristics was investigated: patient's age, PSA level and gland size. Respective Spearman's coefficients (ρ) were calculated. These were: for age, 0.216; for PSA level, 0.241; and for gland size, 0.369. Thus all three correlations were weak (it is generally assumed that to conclude even moderate correlation a ρ value of at least 0.5 is required). This results was checked by using Kendall's τ, which gave similar results (0.151, 0.165, 0.248).

Prostate tissue elasticity as a predictor

We pooled the benign and malignant tumour samples and ranked them by elasticity. Various cut-off values of elasticity were then applied to all samples in order to “predict”, on the basis of their elasticity alone, which samples were benign and which malignant. Results were then assessed by using a two-by-two truth table (true/ false positive, true/false negative) to calculate according to the standard definitions the sensitivity and specificity of the test. Cut-off values were raised in small steps from a very low to a very high value and the corresponding pairs of sensitivity and specificity values were compared by using a standard rate of change (ROC) plot (Fig. 3A). It is immediately seen that the elasticity is in this case not a good criterion for distinguishing between benign and malignant tissue, as the line obtained lies only slightly above the diagonal, with an area under the curve (AUC) of 0.604 ± 0.023 (close to the area of 0.500 for a completely valueless criterion). Owing to the large number of samples this is statistically significant (p < 0.001; 95% confidence limits 0.559 and 0.649), but it is clearly not clinically so. 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 distance between the line plotted and the diagonal); this was found to be 48 kPa, but – as the plot immediately shows – this has little practical meaning as closely similar values can be seen elsewhere in the plot; moreover, the low value of the Youden index implies that no meaningful distinction can be made.

Fig. 3.

ROC analyses. A. A wide range of elasticity cut-off values were tested as criteria for distinguishing benign from malignant neoplasms. B. A wide range of elasticity cut-off values were tested on known malignant tumours as criteria for distinguishing “Gleason ≤6” from “Gleason >6” neoplasms. For each plot the point corresponding to the highest Youden index is indicated. For details, see text

A corresponding analysis of the known malignant tissues gave a qualitatively different result. The ROC plot is shown in Fig. 3B. Here, elasticity was used in the same way as described above as a criterion for “predicting” the Gleason score of these tissues – specifically, whether the score was ≤6 and >6. An ability to discriminate between these Gleason scores, with an AUC of 0.828 ± 0.030 (p < 0.001; 95% confidence limits 0.769 and 0.886) was found. The optimum cut-off for elasticity was found to be 80 kPa; this corresponded to a sensitivity of 80% and a specificity of 76%, with a positive predictive value of 73% and a negative predictive value of 83%.

Discussion

When this study was commenced, large-scale investigations into the correlation of prostate elasticity with malignity were lacking. In the meantime a report of a study of 1040 biopsies has been published⁽¹⁶⁾. The present study was smaller overall, but (as described below) more biopsies per patient were taken (5.65, presumably with a planned 3 per side, versus 10.87 in our study, with a planned 6 per side). We therefore consider the resolution of our “elasticity maps” to be better; it has been reported that the standard sextant procedure misses 10% to 30% of cancers^{(17, 18)}. Barr et al.^{(14, 19)} likewise used sextant biopsies(18) in contrast to the standard 12-core biopsy in the present study.

Our multivariate analysis showed the expected correlation between PSA level, gland size and the presence of malignancy. Interestingly, we report – as far as we are aware, for the first time – that the measured elasticity of nonmalignant prostate-gland tissue differs between the various regions of the gland. We believe that this should be taken account of when considering benign and malignant tissue. However, it is not yet clear whether different cut-off values should be used for the three different regions (base, mid-gland, apex), or to what extent the scatter of values is relevant, and these points will require further investigation. A similar effect was noted for malignant tissue as well, but it was smaller and quantitatively unimportant.

Zheng et al. ⁽²⁰⁾ also describe in their investigation of 120 healthy subjects that they found no difference in elasticity between the inner (medial) and outer (lateral) gland. They did not report any investigation of possible differences between base, mid-gland and apex. In agreement with our result, they found that stiffness increased with age.

The very high sensitivity and specificity reported by other authors for SWE did not turn out to be reproduced in our investigation. For example, Correras et al.^{(16, 18)} claim almost ideal values (sensitivity >95%, specificity >80%) for prostate cancer sized 2–12 mm at an elasticity cut-off of 35 kPa. There appears to be no obvious explanation for this apparent discrepancy, and further work will be required to resolve this and to identify the factors that gave rise to it.

Nevertheless, we found a strong correlation between elasticity and Gleason score of tumours known to be malignant. This could prove to be of value in the regular postdiagnostic screening of patients with known malignancies whose disease is under active surveillance and who frequently return to their physician or hospital for assessment; it may in such cases prove possible to dispense with a biopsy.

Finally, our survey of nearly 800 biopsies in 69 patients revealed – in contrast to the generally accepted belief – that malignant lesions were distributed roughly equally between the medial and lateral regions. This was not expected on the basis of an earlier survey⁽²¹⁾, which suggested that lesions occur more frequently close to the edge of the gland.

Conclusions

We demonstrate an increase in elasticity of benign prostate tissue from the basal to the apical region. The incidence of malignant lesions was approximately the same in the medial and lateral regions of the gland. In contrast to other published data, shear-wave elastography was found to be a poor predictor of malignancy (irrespective of the cut-off value chosen), but for malignant lesions an elasticity cutoff of 80 kPa allowed a fairly reliable distinction between lesions with Gleason ≤6 and those with Gleason >6.

Conflict of interest

The authors have no financial or personal connections with other persons or organizations that might negatively affect the contents of this publication and/or their claim to authorship rights to this publication.