Materials and methods
Study cohort
This study was approved by our institutional review board, and informed patient consent was obtained before study participation according to institutional and native guidelines. Between August 2002 and December 2009, the 51 patients (29 men and 22 women aged 25–68 years; mean age, 44 ± 8.5 years) fulfilled the criteria for inclusion. 18 patients (13 men and 5 women aged 26–58 years; mean age, 41 ± 7.4 years) had pleomorphic adenomas, and 14 (6 men and 8 women aged 33–78 years; mean age, 55 ± 9.1 years) had Warthin’s tumours. 19 patients (11 men and 8 women aged 25–68 years; mean age, 46 ± 8.6 years) had malignancies, including adenoid cystic carcinoma (n = 6), mucoepidermoid carcinoma (n = 4), carcinoma ex pleomorphic adenoma (n = 2), ductal carcinoma (n = 2), acinic cell carcinoma (n = 1), sebaceous carcinoma (n = 1), lymphoma (n = 2) and squamous carcinoma (n = 1). Diagnoses for 51 patients with parotid gland neoplasms were established when histological proof was obtained at surgery.
Imaging protocol
All patients underwent multi-detector CT (LightSpeed ultra; GE Medical Systems, USA). All patients gave written informed consent to undergo CT. To reduce the radiation doses and preserve the diagnostic quality of the imaging, the protocol for the helical CT examinations performed in 51 patients with acquisition parameters: 5-mm slice, a 7.5 mm/s table speed, 120 kVp, 60 mA, a 0.8-s rotation time, a pitch of 0.625, and 5-mm reconstruction intervals before and after the intravenous bolus injection of contrast material (Ultravist 300, Schering, Germany; 300 mg of iodine per millilitre). Imaging began at the skull base and continued to the thyroid cartilage. To optimise the reproducibility of the starting measurements, each image was obtained while the patient held their breath and stopped deglutition.
First, a non-enhanced image was obtained through the parotid gland. An 18- or 20-gauge intravenous catheter was placed in an antecubital vein and manually tested by rapidly infusing 10 ml of saline. Subsequently, 80–120 ml of non-ionic contrast material (1.5 ml per kg) was administered at 2 ml/s by using a power injector (Medrad, Warrendale, PA, USA). In all patients, the second, third, and fourth images were obtained 30 s, 90 s, and 5 min after the start of the contrast material injection respectively. The imaging protocol was preprogrammed so that these images were obtained by using the same parameters that were used to obtain the non-enhanced images. The images were obtained with standard soft-tissue settings (window width, 300 HU; window level, 40 HU).
Image and data analyses
The CT were interpreted and values of diagnostic parameters were measured by using Advantage Windows 4.2 (GE Medical systems, USA). These tasks were performed by two radiologists who were experienced in performing CT of the parotid glands. Two radiologists had no knowledge of the clinical or histological findings, and they worked independently without consultation with one another.
To determine the size of the masses, a distance cursor was used to measure the diameter in the transverse plane. The section of the largest square area of the parotid gland mass was used for analysis. The measurements obtained by the two radiologists were averaged.
The attenuation values of all parotid gland masses detected at CT were measured by using circular region-of-interest (ROI) cursors placed over the area of disease (the biggest mass was selected if multiple lesions were present). The ROI circle was made as large as possible with sufficient margin to avoid partial volume effects. Cystic, necrotic and haemorrhagic components, as well as calcifications of the parotid gland masses, were excluded if they were present. Necrosis was defined as a region (within the mass) with an attenuation value similar to that of water (−20HU ~ 20HU) on the non-enhanced CT. Calcification was defined as a region with an attenuation value greater than 120 HU on non-enhanced CT. Attenuation values were recorded and averaged for final data analysis.
We calculated the following diagnostic parameters for all masses: the absolute percentage enhanced wash-out ratio (PEW) was calculated as \( {\hbox{PE}}{{\hbox{W}}_{{{3}0}}} = \left[ {{1} - \left( {{{\hbox{A}}_{{{3}00}}} - {{\hbox{A}}_{\rm{N}}}} \right)/\left( {{{\hbox{A}}_{{{3}0}}} - {{\hbox{A}}_{\rm{N}}}} \right)} \right] \cdot {1}00 \) or \( {\hbox{PE}}{{\hbox{W}}_{{{9}0}}} = \left[ {{1} - \left( {{{\hbox{A}}_{{{3}00}}} - {{\hbox{A}}_{\rm{N}}}} \right)/\left( {{{\hbox{A}}_{{{9}0}}} - {{\hbox{A}}_{\rm{N}}}} \right)} \right] \cdot {1}00 \), and the relative percentage enhanced wash-out ratio (RPEW30 or RPEW90) was calculated as follows: \( \left[ {\left( {{{\hbox{A}}_{{{3}0}}} - {{\hbox{A}}_{{{3}00}}}} \right)/{{\hbox{A}}_{{{3}0}}}} \right] \cdot {1}00 \) or \( \left[ {\left( {{{\hbox{A}}_{{{9}0}}} - {{\hbox{A}}_{{{3}00}}}} \right)/{{\hbox{A}}_{{{9}0}}}} \right] \cdot {1}00 \), where AN is the attenuation on non-enhanced CT, A
30
or A
90
is 30 or 90-s enhanced attenuation, and A
300 is 5-min enhanced attenuation.
Statistical analyses
Statistical analyses were performed with software (SPSS 13.0; Chicago, IL, USA). Primary statistical analysis of the pooled data (mean±standard deviations) was performed with the analysis of variance (ANOVA) to determine mean differences in objective measurements of parotid gland tumour groups by helical CT, followed by Student-Newman-Keuls –q (SNK-q) test for multiple comparisons. P < 0.01 was considered to indicate statistical significance. Receiver operating characteristic (ROC) analysis was employed to investigate the differentiation capability of the diagnostic parameters value for distinguishing malignant from benign parotid tumours when significant differences existed simultaneously in the malignant tumours versus pleomorphic adenoma groups and malignant tumour versus Warthin’s tumour groups. The optimal threshold value, determined by the maximal Youden index defined as sensitivity plus specificity minus 1, was selected and the area under the ROC (AUC) was calculated. Then, corresponding sensitivity and specificity were calculated for individual parameters. In addition, we examined whether the diagnostic accuracy could be improved by combining these individual diagnostic parameters.
Discussion
Parotid gland masses are common: about 80% of parotid tumours are benign, and most of them, about 60%, are pleomorphic adenoma, followed by Warthin’s tumour (about 10%). Malignant tumours comprise about 15 to 30% of all kinds of parotid tumours. The most common parotid gland malignancy is mucoepidermoid carcinoma [
11]. The cystic adenoid carcinoma is the second most frequent malignancy; other parotid tumours include acinar cell carcinomas, adenocarcinoma not otherwise specified (NOS), lymphoma, and the carcinoma ex pleomorphic adenoma [
12].
The main symptom in patients with parotid neoplasms is a lump in the parotid gland area. Other symptoms such as pain, facial palsy and skin ulcers may emerge in malignancies. According to the extent of the tumour, partial or total parotidectomy is the mainstream treatment measure for benign and malignant parotid tumours. Radiotherapy may aid in malignant cases as adjuvant therapy; chemotherapy is rarely applied. Parotid gland neoplasm recurrence rates of local, regional and distant are 40%, 15% and 11%, respectively. Patients with recurrence have a poor prognosis [
13].
The use of CT and MR imaging has resulted in the detection of parotid gland masses. More recent literature [
3‐
10,
14,
15] has focused on different approaches to characterising parotid masses, including assessing lesion size, describing morphological criteria, measuring attenuation or signal intensity, calculating the contrast material change of single or dual enhancement. Kress et al. [
14] have demonstrated that CT sialography is dispensable in the diagnosis of suspected malignant tumours. Casselman and Mancuso [
15] reported that CT and MR imaging were the first choices for providing the same diagnostic information in all cases of parotid gland neoplasms. Choi et al. [
4] demonstrated that the ratio of CT numbers in enhancement patterns and the attenuation change was significantly different between Warthin’s tumours and pleomorphic adenomas, and between Warthin’s tumours and malignant tumours. Yabuuchi et al. [
7] showed that parotid neoplasms could be differentiated by the 120 s of MR time of peak enhancement and a 30% wash-out ratio (WR), and benign and malignant tumours were recognised by a combined classification of time signal intensity curve and 30% WR. However, to our knowledge, there are no reports of comparisons for benign and malignant parotid gland tumours by the attenuation and percentage enhancement wash-out ratios.
The results of our study indicated that the mean attenuation values of pleomorphic adenomas were similar to those of malignant tumours, but lower than those of Warthin’s tumours on non-enhanced CT. However, there was considerable overlap in attenuation among pleomorphic adenomas, Warthin’s tumours and malignant tumours, and, hence, there was no difference between benign and malignant parotid gland tumours (P = 0.047). In our limited data, multiple masses were found in patients with Warthin’s tumours and malignant tumours. Necrotic portions were viewed in three types of parotid tumours. Calcifications were detected in some pleomorphic adenomas, we presumed that the finding was related to ossification of cartilage structure in pleomorphic adenomas.
From our limited data, the mean attenuation of Warthin’s tumour attained the peak of enhancement during 30-s imaging, and most malignant tumours attain peak attenuation in 90-s imaging, but significant differences did not exist simultaneously between the malignant tumour vs the pleomorphic adenoma groups and the malignant tumour vs the Warthin’s tumour groups for a sustainable overlap. On the 5-min delay imaging, there was no difference among the parotid gland tumour groups. The discrepancy of mean attenuations between ours and a previous study [
4] was attributed to CT parameters, such the total volume of contrast material and the injection rate of contrast material. Our findings suggested that reliable cut-off values were not helpful in differentiating benign from malignant tumours of the parotid gland on non-enhanced, 30-s, 90-s and 5-min imaging series.
The PEW and RPEW are the percentage enhanced wash-out ratio of enhanced attenuation according to the timing of the early and delayed imaging. Our results showed that there were significant differences in PEW and RPEW of parotid gland tumour groups, and significant differences simultaneously in the malignant tumour versus pleomorphic adenoma groups and the malignant tumour versus Warthin’s tumour groups.
From 30-s to 5-min helical CT of parotid gland tumours, applying the threshold value of −70% of PEW30 and −30% of RPEW30 between pleomorphic adenomas and malignant tumours, 89% and 89% of sensitivity, 56% and 61% of specificity were procured for the diagnosis of malignant tumours, relatively, 36% of PEW30 and 19% of RPEW30 between Warthin’s tumours and malignant tumours, 93% and 86% of specificity company with 100% sensitivity was diagnosed for malignancies. Likewise, from 90-s to 5-min helical CT of parotid gland tumours, using an optimal cut-off value of 12% of PEW90 and 4% of RPEW90 between pleomorphic adenomas and malignant tumours, 95% and 95% of sensitivity, 89% and 94% of specificity were obtained for diagnosis of malignant tumours, and 31% of PEW90 between Warthin’s tumours and malignant tumours, 63% of sensitivity and 71% of specificity were diagnosed for malignancies; conversely, with RPEW90 there was much more of an overlap between Warthin’s tumours and malignant tumours.
Although, the enhanced wash-out ratio of tumours was altered for CT parameters [
16], for instance, volume, injection rate and time of contrast material, the foundation of different enhanced wash-out ratios was determined by many inherent traits of parotid gland neoplasms, such as angiogenesis [
17,
18]. With pleomorphic adenoma it is widely accepted that both epithelial and mesenchymal (myxoid, hyaline, chondroid, osseous) elements often arise from the same cell clone, which may be a myoepithelial or ductal reserve cell [
19,
20]. Warthin’s tumours, also known as papillary cystadenoma lymphomatosum, consist of an epithelial, monomorphic, oncocytic component and a lymphoid stroma [
21,
22]. Parotid gland malignancies include diverse tumour types, with the mucoepidermoid carcinoma being the most common tumour. The incidence of adenocarcinoma NOS, the acinar cell carcinoma, the cystic adenoid carcinoma, the carcinoma ex pleomorphic adenoma, and the undifferentiated carcinoma is obviously disparate in different research [
23‐
27]. The enhancement wash-out ratio may mirror different histological components of tumour cells and different architecture of tumoral angiogenesis in parotid gland tumours. Hereby, the percentage-enhanced change could provide the pathological information and differentiate benign from malignant tumours in the parotid gland.
Although we set the optimal threshold criteria to obtain diagnostic values of individual percentage enhanced wash-out ratios, they had much more distinction with regard to sensitivity and specificity. We believe that identification of patients with malignant parotid tumours is more important than missing benign parotid tumours. For this reason, we reanalysed cut-off values of the individual percentage enhanced washout ratios to yield 100% specificity, and examined whether a combination of some of these parameters improved the sensitivity. We tried to employ one or two of three percentages enhanced changes to get 100% specificity, but failed. The final research demonstrated that a parotid gland tumour was defined as a malignant mass when the following parameters were satisfied: −70% > PEW30 < 36%, −30% > PEW30 < 19%, PEW90 > 12%, and 74% sensitivity.
Our study had some limitations. First, parotid gland malignancies include various types of tumours, such as mucoepidermoid carcinoma, adenoid cystic carcinoma, carcinoma ex pleomorphic adenoma, squamous cell carcinoma, acinic cell carcinoma, duct carcinoma and lymphoma. Second, some selected masses were biased in Warthin’s tumours and malignant tumours, only the largest masses of the parotid gland were evaluated. Third, because of the small sample size and overlap in data among the three study subgroups, validation of the proposed cut-off values of CT diagnostic parameters needs to be performed in a larger scale trial. Lastly, if multiple-phasic CT were performed along with a diagnostic study, the patients would be exposed to an additional radiation dose.
In summary, the percentage wash-out ratios in contrast material-enhanced CT may reflect various characters of parotid gland neoplasms, and assist in differentiating the benign from malignant tumours. Combining the percentage enhanced wash-out ratios of the CT protocol can yield diagnostic results in malignant parotid gland tumours.