Introduction

Aspiration cytology as a diagnostic method for head and neck tumors has been used since the 1920s [1]. Pioneering works by Eneroth [2] concerning morphologic studies on smears from salivary tumors led to the acceptance of fine-needle aspiration cytology as a safe and cost-effective diagnostic method. The technique is now commonly used as a first line of investigation in palpable head and neck masses.

Magnetic resonance imaging (MRI) provides anatomic information about the relationships of parotid tumors and can help to define the mass as benign or malignant. Many radiologists have attempted to find MRI features for predicting malignancy of the parotid, with success to some extent [3, 4, 5, 6].

To further assess the relative diagnostic value of fine-needle cytology (FNC) and MRI of parotid tumors, we retrospectively reviewed our experience with parotid tumor management during an 11-year period.

Materials and methods

From May 1991–July 2002, 549 patients with parotid tumor underwent a lateral or total parotidectomy for parotid tumor at our institution (CHU La Timone, Marseilles University Hospital Center, France). This retrospective study consisted of 181 patients with a parotid mass investigated with preoperative MRI and/or FNC. There were 83 men and 98 women. The mean age was 50 years old (range, 15–80) for women and 54 years old (range, 13–80) for men.

One hundred and forty-eight patients underwent preoperative fine-needle cytology, and 87 patients underwent a preoperative MRI. Of these patients, 54 had both fine-needle and MRI studies. Three groups of patients were defined depending on the preoperative study they underwent before parotidectomy (patients with fine-needle aspiration cytology, patients with MRI and patients with combined FNC and MRI studies). In patients with combined preoperative study, MRI was always performed prior to the fine-needle cytology in order to avoid any radiological modification. Distribution of tumors was as follows: 141 benign and 39 malignant parotid gland tumors.

Fine-needle aspiration

All patients had a palpable parotid mass, allowing a palpation-guided FNC. No image-guided FNC was performed. All specimens were procured by an experienced pathologist. Our technique of FNC didn’t use aspiration, contrary to the practice of many authors [7, 8]. The technique we routinely employed follows: The patient is positioned comfortably. The skin is disinfected with an alcohol wipe and the mass is grasped between the thumb and forefinger of the left hand for right-handers. A 25-gauge needle, not attached to a syringe, is inserted into the mass and a rotary up and down vertical needle move is then applied for 10 s (1 Hz frequency). Thus, cytological material is obtained by capillarity. A 10-ml syringe with 5 ml of air is then attached to the needle in order to expel the cytological material onto three glass slides with the bevelled edge of the needle touching the slides. The material is then smeared on the glass slides and air-dried before being placed in a box for transportation. Three punctures are performed as described for each tumor in order to optimize the smears and to reduce non-diagnostic smears because of insufficient material.

Four sorts of cytologic reports were available from the department of cytopathology: (1) benign FNC, with most probable histological diagnosis, (2) malignant FNC, with most probable histological diagnosis, (3) suspicious FNC, with most probable histological diagnosis and (4) non-diagnostic FNC (indeterminate). In all cases, cytological diagnoses were reviewed and compared with the final histologic diagnoses based on histologic sections from the parotidectomy specimens.

Magnetic Resonance Imaging study

Magnetic resonance images of the lesions were evaluated independently on a six-item weighted scale by an experienced head and neck radiologist who was unaware of the final diagnosis. MRI studies were performed using a 1.5-T MRI unit (Magnetom Symphony Maestro Class, Siemens Medical Solutions) with non-enhanced conventional spin-echo T1-weighted images (T1WI) and fast spin-echo T2-weighted images (T2WI). Images were obtained using a quadrature head coil with a matrix of 256×192, a field of view of 22 cm and a section thickness of 5 mm. The evaluation scale included the following radiological findings: (1) invasion of the surrounding structures: presence (5 points) or absence (0 points), (2) margin characteristics: irregular (5 points) or regular (0 points), (3) T2-weighted signal intensity of the lesions relative to the parotid gland: hypointense (2 points), hyperintense (0 points) or isointense (O points), (3) composition: heterogeneous (1 point) or homogeneous (0 points), number: multiple (1 point) or single (0 points) and (4) diameter (size): more than 2 cm (1 point) or less than or equal to 2 cm (0 points).

A total score was calculated for each patient by adding the sub-scores. Scores ranged from 0 to 15 points. After assessment of the magnetic resonance imaging, the radiologist had to decide for one of three available radiological reports: (1) benign MRI, with most probable histological diagnosis, (2) malignant MRI, with most probable histological diagnosis and (3) non-diagnostic MRI. In all cases, the MRI assessments were reviewed and compared with the final histological diagnoses based on histological sections from the parotidectomy specimens.

Statistical analysis

All results were analyzed in order to determine sensitivity, specificity, negative and positive predictive values and accuracy. To do this, each FNC and MRI conclusion has been labelled as “positive” or “negative.” We considered as positive results all cases with malignant diagnosis. All cases of benign diagnosis were considered negative. All cases of suspicious diagnosis were considered positive because they did not exclude the possibility of malignancy and treatment on that basis.

Statistical diagnostic values were calculated for each of the three groups of patients: patients with fine-needle aspiration, patients with MRI and patients with combined cytology and MRI. In this last preoperative combined study group, patients were considered as positive for malignant tumor when at least one of the two studies was positive, whatever the other study result was. They were considered as negative when both studies were negative.

Results

Patients with fine-needle cytology

In this group, 148 smears were analyzed and included. Fifteen specimens were non-diagnostic (10%) because of insufficient cellularity for diagnosis. One hundred and thirty-three specimens provided adequate material for diagnosis, corresponding to 102 benign smears and 31 malignant smears (cytological diagnoses). Results of FNC for benign parotid tumors are reported in Table 1, while results of FNC for malignant parotid tumors are reported in Table 2. The distribution of patients was as follows: 25 true positive, 97 true negative, 5 false positive and 6 false negative.

Table 1 Results of FNC for benign parotid tumors. Numbers in parentheses indicate number of cases with correct histological correlation
Full size table
Table 2 Results of FNC for malignant parotid tumors. Numbers in parentheses indicate number of cases with correct histological correlation
Full size table

In this series of patients with preoperative fine-needle cytology, sensitivity for malignant tumor was 83%, specificity was 95%, positive predictive value was 83% and negative predictive value was 94%. Global accuracy of differentiation of benign from malignant was 92%. No complications after FNC have been reported in our series.

Among the benign FNC, diagnoses were confirmed by histological study in 85 of the 102 conclusive benign smears (83%). The most characteristic diagnosis was pleomorphic adenoma with 55 correct cytological diagnoses (histologically confirmed) on 60 conclusive smears (92%) originating from pleomorphic adenomas. Warthin’s tumor is less characteristic with 15 correct cytological diagnoses on 20 conclusive smears originating from Warthin’s tumors (75%).

Diagnostic correlation between malignant FNC and histological diagnosis was reported in 11 of the 31 conclusive smears of malignant tumors (35%).

Patients with preoperative MRI

In this group, 87 MRI studies were analyzed and included. Only one patient (1%) was considered as non-diagnostic as the parotid tumor was very anterior and the radiologist didn’t conclude that this tumor originated from the parotid gland. The results of MRI for benign parotid tumors are reported in Table 3, while the results of FNC for malignant parotid tumors are reported in Table 4. The distribution of patients was as follows: 13 true positive, 67 true negative, 4 false positive and 2 false negative.

Table 3 Results of MRI for benign parotid tumors. Numbers in parentheses indicate number of cases with correct histological correlation
Full size table
Table 4 Results of MRI for malignant parotid tumors. Numbers in parentheses indicate number of cases with correct histological correlation
Full size table

In this series of patients with preoperative MRI, sensitivity for malignant tumor was 87%, specificity was 94%, positive predictive value was 76% and negative predictive value was 97%. Global accuracy of differentiation of benign from malignant was 93%. No complication after FNC has been reported in our series. Among the benign MRI reports, diagnoses were confirmed by histological study in 48 of the 71 conclusive benign MRI (68%). The most characteristic MRI diagnosis was pleomorphic adenoma with 39 correct “imaging” diagnoses on 47 cases (83%). In malignant MRI reports, the radiologist never made any precise malignant diagnosis.

Patients with fine-needle cytology and MRI

In this group, 54 patients were analyzed and included. No patient was considered as “non-diagnostic” (indeterminate) as all patients had at least one of the two studies with complete conclusion. Among them, 43 had a benign tumor while, 11 had a malignant one. The results of FNC combined with MRI as a preoperative diagnostic study for benign parotid tumors were reported as follows: concerning the benign tumors, 35 patients had both benign cytology and MRI results (negative cases), 5 had either malignant MRI or cytology results (positive cases), and 3 had indeterminate cytology with valid benign MRI results (negative cases). Concerning the malignant tumors, six patients had both malignant cytology and MRI (positive cases), two had either malignant MRI or cytology (positive cases), two had an indeterminate cytology with valid malignant MRI (positive cases), and one had an indeterminate MRI with valid malignant cytology (positive case). The distribution of patients was as follows: 11 true positive, 38 true negative, 5 false positive and no false negative patients.

In the series of patients with preoperative combined fine-needle cytology and MRI, sensitivity for malignant tumor was 100%, specificity was 88%, positive predictive value was 69% and negative predictive value was 100%. Global accuracy of differentiation of benign from malignant was 91%.

Discussion

Indeterminate patients

Non-diagnostic (indeterminate) smears reached 10% in our series. In all these cases, smears were either hemorrhagic or performed on cystic lesions, so that the cellularity was insufficient for diagnosis despite the number of smears (10 to 12 smears per tumor were obtained in those cases, as the material originating from cystic or hemorrhagic lesions is abundant). This rate is compatible with those reported in the literature, ranging from 2 to 10% [9, 10], whatever the aspiration technique was. In order to limit non-diagnostic FNC, smears must be sufficient in number. Jandu et al. reported that the experience and practice of the practitioner performing the FNC improved sensitivity and specificity [11]. All statistic calculations of diagnostic values were performed without taking into account these non-diagnostic smears in order to reflect a more “practical” value.

MRI studies reported only one patient considered as indeterminate and corresponded to 1% of the series. This patient had a very anterior small parotid gland tumor developed along the parotid duct that appeared on the MR images as an extra-parotid tumor. The diagnostic values presented further should be read taking into account that MRI is much more reliable in terms of efficiency.

Prediction of malignancy

In our series, diagnostic values of FNC for differentiating malignant to benign parotid tumors were consistent with previously published series. Concerning the diagnostic value of FNC, sensitivity was 81%, specificity 95% and global accuracy 92%. The highly reliable accuracy of differentiation, sensitivity and specificity for each of the three groups of patients are reported in Table 5. Other series reported sensitivity rates ranging from 79 to 100% and specificity rates ranging from 86 to 100% [12, 13, 14, 15, 16] for fine-needle cytology. The validity of FNC is now accepted as a well-known fact despite the disparity of opinion about the use of FNC for parotid tumors [17]. Questions about such issues as tumor seeding have disappeared because of the lack of evidence and reports with modern techniques [18, 19].

Table 5 Diagnostic value of MRI, FNC and combination for parotid tumors. Without indeterminates
Full size table

The diagnostic value of MRI in this study appeared to be similar to the FNC value in terms of prediction of malignancy. The most suggestive findings of malignancy were the tumor invasion of surrounding structures, margin characteristics and T2-weighted signal intensity. The distribution of MRI score percentages is reported in Fig. 1 for patients with benign tumors and patients with malignant ones. The mean score of the 64 patients with benign parotid tumors was 3.5, while the mean score of the 15 patients with malignant tumors was 10.5. Many authors described that imaging characteristics optimize the sensitivity and specificity of MRI [5, 13, 20, 21, 22, 23, 24, 25]. The results of this study suggest that MRI is the test of choice because it is as effective as FNC at confirming the suspicion of malignancy and has a much lower rate of indeterminate results. In addition, MRI provides an evaluation of the full extent of lesions and site of the tumor, determines a surgical approach and predicts possible complications after surgery. MRI studies appear more reliable in terms of efficiency than FNC.

Fig. 1
figure 1

Distribution of MRI scores of benign and malignant tumors

Full size image

Combining both FNC and MRI preoperative studies allows better efficiency and reduces false-negative cases as patients are considered as negative when both FNC and MRI are negative. In our series, no false negative patient was reported in the third group (preoperative combined study).

Preoperative histological diagnosis

The distinction between benign and malignant parotid tumors was highly reliable in our series. However, discrimination between various histological types of benign and malignant tumors is often more difficult. In the benign tumor group, diagnostic correlation between FNC and histology was reported in 85 of 102 conclusive smears (83%). The most characteristic diagnosis was pleomorphic adenoma with 55 correct cytological diagnoses among the 60 smears (92%). Warthin’s tumor is less characteristic with 15 correct cytological diagnoses on 20 smears (75%). The global value for correct histological diagnosis of FNC appears quite reliable for benign tumors, even if limited.

Conversely, for malignant tumors, correct histologic diagnosis with FNC wasn’t reliable. Only 11 of 31 conclusive smears of malignant tumors led to a correct diagnosis (35%). Concerning MRI, a diagnostic correlation between MRI and histology was reported in 48 MRI reports (72%). The global value for correct histological diagnosis appears quite reliable for benign tumors, while histological diagnosis for malignant tumors has never been made with MRI. The most characteristic diagnosis was pleomorphic adenoma with 39 correct MRI diagnoses in 47 cases (83%). Other diagnoses such as lipoma, branchial cyst or dermoid cyst were made with MRI in most cases, allowing a reduction of surgical indications. FNC appears more reliable than MRI as a method of histological diagnosis, but remains clearly limited, mainly for malignant diagnoses.

Preoperative management of parotid tumors has to be adapted to the medical environment. The reliability and associated anatomic information of MRI in parotid gland tumor diagnosis make MRI the test of choice in our practice. Cytology is of main importance, despite having a higher indeterminate rate than MRI. Conversely, FNC leads to more specific histological diagnoses. The choice of modality has to be made by the surgeon depending on the availability of local medical resources and the cost-effectiveness of management.

Conclusion

FNC and MRI are simple and well-tolerated studies. Both preoperative studies have an impact on the management and treatment of salivary gland masses, enabling more confident preoperative patient counselling, better surgical planning and better selection of parotidectomy candidates. In our series, MRI appears more efficient because of the very low rate of indeterminate patients, while FNC procures better histological information. Combining both studies lowers the number of false negative patients to the detriment of positive predictive value and cost efficiency. The use of FNC and MRI must be adapted to the medical environment in order to optimize the preoperative diagnosis. Both techniques provide excellent information concerning the differentiation between benign and malignant tumors, but poor information for the histological diagnosis, particularly for malignant neoplasms.