Radiological imaging in primary parotid malignancy q

Transcription

1 The British Association of Plastic Surgeons (2003) 56, Radiological imaging in primary parotid malignancy q C. Raine a, *, K. Saliba b, A.J. Chippindale b, N.R. McLean a a Department of Plastic and Reconstructive Surgery, Royal Victoria Infirmary, Queen Victoria Road, Newcastle-upon-Tyne NE1 4LP, UK b Department of Diagnostic Radiology, Royal Victoria Infirmary, Queen Victoria Road, Newcastle-upon-Tyne NE1 4LP, UK Received 13 April 2003; accepted 8 July 2003 KEYWORDS Parotid cancer; Imaging; CT; MR Summary In a retrospective analysis of the preoperative imaging of patients presenting with primary malignant parotid disease, all relevant images were collected and reviewed by two experienced head and neck radiologists, blinded to the diagnosis. Forty-two patients (25 male, 17 female), median age 67.5 years (range 15 86), were included in the study of which 32 had undergone CT scanning, nine MR and three ultrasound. Forty tumours (93%) were correctly diagnosed as malignant when compared with histology. The two false negatives arose in a patient who had received an ultrasound scan only and in a patient with lymphoma, whose CT scan was reported as a pleomorphic adenoma. Neither had undergone prior radiotherapy. A poorly defined tumour boundary was the most consistent observation for both MR and CT images for the malignant tumours examined. Local infiltration was correctly correlated with pathological findings in eight of the nine MR scans, however, CT proved less reliable, correct in 14 of 24 cases ðp ¼ 0:01Þ: No correlation was identified between any of the imaging features examined and the final histological diagnosis. A poorly defined tumour boundary with evidence of local invasion was the best indicator of malignancy and was reported more frequently from MR scans than from CT. Imaging alone, however, proved unreliable in the prediction of final histological diagnosis or grade of tumour. Q 2003 The British Association of Plastic Surgeons. Published by Elsevier Ltd. All rights reserved. q This paper was presented at the Summer Meeting of the British Association of Plastic Surgeons, Royal Free Hospital, London, UK, on 3rd July 2002 and at the 34th Scientific Meeting of the British Association of Head and Neck Oncologists, Royal College of Physicians, London, on 26th April *Corresponding author. Tel.: þ ; fax: þ address: cr@raine.org.uk The clinical examination of a parotid mass is facilitated by the superficial location of the gland and when combined with an accurate history, can be expected to achieve a diagnostic accuracy in the region of 85%. 1 By performing fine-needle aspiration cytology (FNAC) of the lesion, an accuracy as high as 92% may be achieved. 2,3 In the case of suspected malignancy, these investigations are S /$ - see front matter Q 2003 The British Association of Plastic Surgeons. Published by Elsevier Ltd. All rights reserved. doi: /s (03)

2 638 C. Raine et al. frequently supplemented by imaging, serving both to map the lesion to its anatomical relations and stage the neck, thus providing accurate information upon which to base a treatment plan. 3,4 The most commonly used imaging modalities for these purposes are CT and MRI. To a lesser extent, ultrasound scanning may also be used. In addition to providing anatomical information, imaging data has also been used by some authors to offer an assessment of the histological nature and grade of parotid tumours. 5 7 There is, however, great controversy on this point and many have questioned the power of imaging alone to make such predictions. 8,9 The purpose of this study was to review our experience in the imaging of primary parotid cancer and to investigate the association between imaging features and final histological diagnosis based on the most commonly cited radiological features of malignancy. 4,6,7,10,11 Materials and methods Patients presenting with suspected primary malignant disease of the parotid salivary gland in the period were identified from computer databases held at the Northern Centre for Cancer Treatment, and the departments of histopathology and radiology at the Royal Victoria Infirmary, Newcastle. Patients were excluded if appropriate CT/MR or ultrasound images were not available and if the clinical or radiological features were those of metastatic disease. The images were pooled from a number of referring hospitals and, therefore, were produced by differing modalities, machines and protocols. All CT scans were, however, obtained with 5 mm contiguous slice thickness. Two experienced head and neck radiologists, guided only by the information contained in the original imaging request form, systematically reviewed each image and the data was recorded. Clinical and pathological data were retrieved from the case notes and recorded separately. In particular, any previous radiotherapy to the parotid area prior to imaging was noted and whether there was clinical evidence of facial nerve and/or skin involvement at the time of presentation was similarly recorded. In three patients, a pathological diagnosis was made on FNAC, in all other cases the results of formal histological analysis of the tumours were available from operative specimens. The following imaging features were systematically analysed for each patient, tumour size (in 2 dimensions, using 5 mm digit preference this being the smallest gradation on the image scale); position within the gland (superficial, deep or involving both lobes); shape (round, lobulated or complex); boundary definition (well defined or ill defined); presence or absence of contrast enhancement (when contrast had been administered); the pattern of contrast enhancement (marked, mild or absent, uniform, marginal or complex); infiltration of adjacent structures and the presence of lymph node enlargement (using a 1 cm short axis diameter as the cut off point). In the case of CT images, the attenuation of the lesion and the presence of calcification were also recorded. For the MR images signal intensity on the T1 and T2/STIR sequences as well as the presence of an intra-tumour fluid signal were also noted. The size of the lesion was measured using the computer generated MR/CT scale and the position of the lesion within the parotid gland was classified as involving the superficial lobe or the deep lobe or both; attempts were also made to establish where the epicentre of the lesion was when both lobes were involved. The radiological appearances were then correlated with the histological findings and the results submitted to statistical inferencing using a chisquared test of association. A p value,0.05 was considered to represent statistical significance. As previously stated, parotid cancer is a rare tumour and, therefore, by necessity this study is observational in design. Results Forty-two patients (25 male, 17 female) with a median age 67.5 years (range 15 86), were included in the study of whom 32 had undergone CT scanning, nine MR and three ultrasound. Six patients (14%) had previously received radiotherapy to the parotid region. Twenty-eight of the 32 CT scans were contrast enhanced, whereas, none of the MR scans were. One third of patients were noted clinically to have facial nerve dysfunction and four patients (10%) had skin involvement at the time of presentation. The final histological diagnoses are given in Table 1. The commonest malignancies seen were carcinoma ex-pleomorphic salivary adenoma and primary adenocarcinoma. In 12 patients (29%) the clinical impression of primary malignant parotid disease was found to be incorrect when compared to final histology, of these, eight patients (19%) were found to have metastatic carcinoma and four patients (10%) had lymphoma. The TNM staging at presentation of the remaining 30 patients is shown in Table 2.

3 Radiological imaging in primary parotid malignancy 639 Table 1 Histological diagnoses of parotid malignancies Histological diagnosis Patients ðn ¼ 42Þ Carcinoma ex-pleomorphic adenoma 10 Adenocarcinoma 8 Primary parotid squamous cell carcinoma 7 Metastatic carcinoma 7 Lymphoma 4 Adenoid cystic carcinoma 3 Small cell undifferentiated 1 Mucoepidermoid carcinoma 1 Metastatic sarcoma 1 Three patients in the CT group and none of the patients in the MR group, required upstaging of the neck following scanning. Radiological and pathological agreement Forty patients (93%) were correctly diagnosed as having malignant parotid disease when compared to the final histological diagnosis. The two false negatives arose firstly, in a patient early in the series who had received an USS only which was interpreted as showing a pleomorphic adenoma by the radiologists, however, following excision, the tumour was found to contain significant areas of invasive squamous cell carcinoma. In the second patient the histological diagnosis was of non- Hodgkins lymphoma, however, the radiologists reported the CT scan as showing features consistent with a pleomorphic adenoma (Fig. 1). Neither patient had undergone radiotherapy prior to scanning. Tumour size recorded radiologically correlated well with the pathology specimen with an 81% level of agreement. The radiological assessment of tumour position within the parotid gland agreed with the pathologist s report in 76% of cases. Disagreement occurred most frequently when the radiologists had documented both lobes as being involved, however, analysis of the operative specimen found only single lobe involvement. In one case, only the superficial lobe was thought to be involved radiologically whilst the pathologist documented diffuse involvement of the whole gland. This case was the only one Fig. 1 Axial CT showing a mass arising in the left parotid gland. Reported radiologically as showing features consistent with pleomorphic adenoma this lesion proved to be non-hodgkins lymphoma. in our series that had been assessed by ultrasound alone. Histological assessment of local tumour invasion was available for 33 tumours and the imaging findings were in agreement with the pathologist s report in 67% of cases across all modalities (Table 3). When analysed separately, however, the radiologists accurately reported the presence of local invasion in eight of the nine tumours imaged with MR compared to 14 of the 24 CT scans when compared with histology. The difference between these two groups was statistically significant ðp ¼ 0:01Þ: Imaging characteristics The different imaging characteristics obtained for each modality were classified and assessed for their sensitivities in the detection of malignant parotid disease. The radiological shape of the lesion for all modalities did not correlate with malignancy. MR features Nine patients from our series had undergone preoperative MR imaging. Local infiltration at the Table 2 Disease stage at presentation Stage Tumour Nodes Metastasis T1 T2 T3 T4 N0 N1 M0 M1 Patients ðn ¼ 30Þ 0 11 (37%) 3 (10%) 16 (53%) 21 (70%) 9 (30%) 28 (93%) 2 (7%)

4 640 C. Raine et al. Table 3 Correlation of imaging with pathological findings Image feature Correlation with pathology Sensitivity (%) Tumour size ðn ¼ 26Þ 21/26 81 Tumour position ðn ¼ 29Þ 22/29 76 Local infiltration ðn ¼ 33Þ 22/33 67 tumour margin was the most sensitive feature on MR and correlated with pathological findings in eight of the nine patients (89%). Boundary definition was the next most sensitive feature with seven of the parotid lesions having ill-defined boundaries (78% sensitivity) (Fig. 2). By comparison the other MR features examined, T1-weighted signal (six intermediate and three low signal); T2-weighted signal (five high and four mixed signal); intra-tumour fluid signal (three contained fluid and three did not); and lymph node enlargement (five had enlarged nodes and four did not) were unreliable (Table 4). Ill-defined tumour boundary with evidence of local invasion was the most sensitive feature associated with malignancy observed on MR (Fig. 3). CT features Thirty-two patients had undergone preoperative imaging using CT scanning. Once again the various characteristics were categorised and evaluated. Twenty of the parotid lesions (63%) had illdefined boundaries on CT whilst in the other 12 cases the boundaries were well defined. The presence of contrast enhancement (18 enhanced, 10 were nonenhancing and four were post-contrast only scans) was also an important feature, however, the degree and pattern of contrast enhancement was not consistent (Fig. 4). The attenuation of the parotid lesion was similarly found to be largely inconsistent (Table 5). Seven of 32 patients (22%) were noted to have enlarged loco-regional lymph nodes. The presence of calcification (nine patients showed calcification) was also a poorly sensitive marker for malignancy. In four patients, post contrast films only were available and, therefore, the presence of calcification could not be assessed. Contrast enhancement and ill-defined tumour boundaries were the most significant findings associated with malignancy identified on CT (Fig. 5). Local tumour invasion present pathologically, however, frequently went unrecognised on CT scans. Ultrasound features Only three patients underwent ultrasound as part of their preoperative assessment, one patient having ultrasound as the sole imaging investigation whilst the other two had the ultrasound together with a CT and MRI, respectively. The small numbers precluded any useful analysis of this group. Tissue characterisation The different histological parotid tumour types were also categorised and assessed for any differences in imaging characteristics. None were identified. Discussion It is generally accepted that pre-operative imaging is an important tool in the investigation of patients presenting with suspected malignant disease of the parotid gland. The planning of treatment and counselling of patients is greatly enhanced by the additional information provided. The nodal status of the neck may also be examined, providing useful staging, and thus, prognostic information. More contentious is the use of imaging data by some authors to provide a histological diagnosis and an assessment of the grade of a particular tumour. 5 7 We were able to identify only three previous Table 4 Imaging features seen on MR Image feature Sensitivity (%) Local infiltration Confirmed histologically 8/9 Not confirmed histologically 1/9 89 Boundary definition Ill defined 7/9 Well defined 2/9 78 Signal intensity T1 High 0/9 Intermediate 6/9 Low 3/9 Signal intensity T2 High 5/9 Intermediate 4/9 Low 0/9 Intra-tumour fluid signal Present 3/9 Absent 6/9 34

5 Radiological imaging in primary parotid malignancy 641 Fig. 2 Axial (a) T1 weighted and (b) STIR MR sequences showing a complex mass involving both lobes of the right parotid gland. The posterolateral border is indistinct suggestive of local invasion. The histological diagnosis was carcinoma arising in a pleomorphic salivary adenoma. studies on radiological imaging of the parotid gland where statistical data has been reported.2,4,8 Byrne et al. reviewing 24 cases, reported a poorly defined tumour boundary and contrast enhancement as features of malignant parotid disease on CT scanning, although tumour enhancement was noted to be nondiscriminatory when compared to benign disease. Takashima et al. found that a poorly defined tumour boundary and evidence of local Fig. 3 (a) T1 weighted and (b) coronal STIR MR images showing a large mass arising in the left parotid gland involving both superficial and deep lobes and encroaching on the carotid sheath. The edges of the lesion are ill defined and there is evidence of local infiltration. The histological diagnosis was primary parotid squamous cell carcinoma.

6 642 C. Raine et al. Fig. 4 Post contrast CT scan showing a uniformly enhancing right parotid mass. The histological diagnosis in this case was of primary parotid adenocarcinoma. invasion observed on MR were both features of malignant parotid disease with the former showing the strongest correlation in the 14 cases of parotid malignancy reviewed. Freling et al. also found that local invasion showed a significant correlation with malignancy on MR imaging, however, their data did not support the association with poor boundary definition in the 30 cases studied. In our series a poorly defined tumour boundary showed the strongest correlation with malignancy on MR but was also a significant finding for CT although the association was weaker. Local tumour invasion, however, was reliably identified only for the MR scans reviewed. In addition, we found no correlation between signal intensity or tumour homogeneity with malignancy as observed on histology. We were also unable to identify any imaging features that could reliably predict the final histological diagnosis from the scans reviewed. Fig. 5 Contrast enhanced CT scan showing a complex right parotid mass with marginal tumour enhancement, a poorly defined medial border and evidence of local infiltration. The histological diagnosis was carcinoma arising in pleomorphic adenoma. The sensitivity of imaging in the diagnosis of parotid malignancy in our study was 100% for MR although our numbers in this group were small, and 91% for CT with one false negative. Our overall sensitivity across all imaging modalities was 93% which is in line with other published series and compares favourably with fine needle aspiration cytology as a diagnostic tool. 1,12 It is our practice, however, to perform FNAC on all suspected parotid malignancies in conjunction with imaging on the grounds that the additional cytological information obtained may prove helpful during the planning stage of treatment. The radiologists were able to correctly identify the location of the tumour relative to the facial Table 5 Imaging features seen on CT Image feature Sensitivity (%) Contrast enhancement Enhancing 18/28 Nonenhancing 10/28 64 Boundary definition Ill defined 20/32 Well defined 12/32 63 Local infiltration a Present 14/24 Absent 10/24 58 Calcification b Present 9/28 Absent 19/28 32 Signal attenuation High 3/32 Intermediate 13/32 Low 7/32 Mixed 4/32 a Pathological assessment of tumour invasion was not available in eight cases ðn ¼ 24Þ: b In four cases post contrast films only were available for review ðn ¼ 28Þ:

7 Radiological imaging in primary parotid malignancy 643 nerve in 76% of the scans reviewed when compared to operative findings. The intraglandular segment of the facial nerve cannot be reliably imaged directly and the radiologists are, therefore, required make a prediction of its course through the gland based on fixed anatomical landmarks. This has been the subject of previous work in our department and serves to emphasise the importance of understanding the limitations of imaging in this regard when planning surgery. 13 The strengths of MRI as an imaging tool in the investigation of parotid disease have been well documented. 11,14,15 Of particular relevance to the imaging of malignant parotid disease is the superior soft tissue contrast resolution, providing detailed information in the critical region of the tumour boundary. Our results support the use of MR as the imaging modality of choice for the investigation of suspected malignant parotid disease. CT remains a valuable imaging modality in the preoperative assessment of primary parotid malignancies, however, the significant correlations established from our results are weaker than for MR. It may have a role to play if bone invasion is a consideration. Nonetheless, CT was used to investigate the majority of patients in this study (76%), reflecting the differing availabilities of the two modalities during the period under review. Our present practice, however, is to undertake MR scanning for the investigation of all patients presenting with suspected parotid malignancy prior to treatment. MRI emerges as the strongest and most reliable imaging modality for the assessment of primary parotid malignancy. Recent and future developments in MR sequences and contrast agents will surely contribute to enhance its role still further. References 1. McGuirt WF, Keyes Jr. JW, Greven KM, et al. Preoperative identification of benign versus malignant parotid masses: a comparative study including positron emission tomography. Laryngoscope 1995;105: Takashima S, Takayama F, Wang Q, et al. Parotid gland lesions: diagnosis of malignancy with MRI and flow cytometric DNA analysis and cytology in fine-needle aspiration biopsy. Head Neck 1999;21: Malata CM, Camilleri IG, McLean NR, et al. Malignant tumours of the parotid gland: a 12-year review. Br J Plast Surg 1997;50: Byrne MN, Spector JG, Garvin CF, et al. Preoperative assessment of parotid masses: a comparative evaluation of radiologic techniques to histopathologic diagnosis. Laryngoscope 1989;99: Joe VQ, Westesson PL. Tumors of the parotid gland: MR imaging characteristics of various histologic types. Am J Roentgenol 1994;163: Som PM, Biller HF. High-grade malignancies of the parotid gland: identification with MR imaging. Radiology 1989;173: Soler R, Bargiela A, Requejo I, et al. Pictorial review: MR imaging of parotid tumours. Clin Radiol 1997;52: Freling NJ, Molenaar WM, Vermey A, et al. Malignant parotid tumors: clinical use of MR imaging and histologic correlation. Radiology 1992;185: Yousem DM. Dashed hopes for MR imaging of the head and neck: the power of the needle. Radiology 1992;184: Wittich GR, Scheible WF, Hajek PC. Ultrasonography of the salivary glands. Radiol Clin North Am 1985;23: Mandelblatt SM, Braun IF, Davis PC, et al. Parotid masses: MR imaging. Radiology 1987;163: Bartels S, Talbot JM, DiTomasso J, et al. The relative value of fine-needle aspiration and imaging in the preoperative evaluation of parotid masses. Head Neck 2000;22: Ragbir M, Dunaway D, Chippindale AJ, et al. Prediction of the position of the intraparotid portion of the facial nerve on MRI and CT. Br J Plast Surg 2002;55: Schaefer SD, Maravilla KR, Close LG, et al. Evaluation of NMR versus CT for parotid masses: a preliminary report. Laryngoscope 1985;95: Casselman JW, Mancuso AA. Major salivary gland masses: comparison of MR imaging and CT. Radiology 1987;165: