Adenoid cystic carcinoma: A retrospective clinical review

Authors


Abstract

Adenoid cystic carcinoma (ACC) are uncommon tumors, representing about 10% to 15% of head and neck tumors. We compare the survival and control rates at our institution with those reported in the literature, and examine putative predictors of outcome. All patients registered with the tumor registry as having had ACC were identified. Demographic and survival variables were retrieved from the database. Additionally, a chart review of all patients was done to obtain specific information. Minor gland tumors were staged using the American Joint Committee on Cancer's criteria for squamous cell carcinomas in identical sites. Histopathologic variables retrieved included grade of the tumor, margins, and perineural invasion. Treatment modalities, field sizes, and radiation doses were recorded in applicable cases. An effort to retrieve archival tumor specimens for immunohistochemical analysis was undertaken. A total of 69 patients were treated for ACC from 1955 to 1999. One patient, who presented with fatal brain metastasis, was excluded from further analysis. Of the remaining 68 patients, 30 were men and 38 were women. The average age at diagnosis was 52 years, and mean follow-up was 13.2 years. Mean survival was 7.7 years. Overall survival (OS) rates at 5, 10, and 15 years were 72%, 44%, and 34%, and cause-specific survival was 83%, 71%, and 55%, respectively. Recurrence-free survival rates were 65%, 52%, and 30% at 5, 10, and 15 years, with a total of 29 of 68 (43%) eventually suffering a recurrence. Overall survival was adversely affected by advancing T and AJCC stage. Higher tumor grades were also associated with decreased OS, although the numbers compared were small. Primaries of the nasosinal region fared poorly when compared with other locations. Total recurrence-free survival, local and distant recurrence rates were distinctly better in primaries of the oral cavity/oropharynx when compared with those in other locations. Reduced distant recurrence-free survival was significantly associated with increasing stage. No other variables were predictive for recurrence. Additionally, we found that nasosinal tumors were more likely to display higher stage at presentation, and were more often associated with perineural invasion. Also of interest was the association of perineural invasion with margin status, with 15 of 20 patients with positive margins displaying perineural invasion, while only 5 of 17 with negative margins showed nerve invasion (P = 0.02). On immunohistochemistry, 2 cases of the 29 (7%) tumor specimens found displayed HER-2/neu positivity. No correlation between clinical behavior and positive staining could be demonstrated. Our data concur with previous reports on ACC in terms of survival and recurrence statistics. Stage and site of primary were important determinants of outcome. Grade may still serve a role in decision making. We could not demonstrate any differences attributable to primary modality of therapy, perhaps due to the nonrandomization of patients into the various treatment tracks and the inclusion of palliative cases. Similarly, perineural invasion, radiation dose and field size, and HER-2/neu positivity did not prove to be important factors in our experience. © 2001 Wiley-Liss, Inc.

INTRODUCTION

Adenoid cystic carcinomas (ACC) are uncommon tumors, representing about 10% to 15% of the histology of head and neck tumors [1]. They are the most common malignant neoplasms of the minor salivary glands [2], and can also arise in a variety of sites in the head and neck including the major salivary glands, the tracheobronchial tree, the esophagus, the lacrimal gland, as well as sites outside the head and neck [3].

Previous literature has attempted to define factors predictive of outcome variables, such as rates of local control and survival. Pathologic factors examined have included histological grade, margins of resection, and perineural invasion. Tumor grade has generally been regarded as a poorer indicator than tumor stage [3, 4], although some reports have contended that grade is also significant. Positive margins and perineural invasion, especially of named nerves (cranial nerves or their named branches), have been shown to be adverse predictors of outcome in patients with ACC [5].

Among the clinical variables, tumor stage has proved to be a reliable and reproducible determinant of tumor control and survival [3, 6–8]. Although age and sex are independent of patient outcome, tumor site has been shown to be of significance, with tumors of the nasal cavities and sinuses associated with poorer outcomes [3, 4].

In the search for more reliable prognostic indicators, several investigators have explored the possible relevance of molecular markers in both salivary gland carcinomas in general [9] and ACC in particular. Some markers evaluated using immunohistochemistry include the epidermal growth factor receptor (EGF-R), HER-2, HER-3, p53, and proliferating cell nuclear antigen (PCNA) [10–16]. The resultant data for ACC have not been concordant.

Optimum treatment of ACC has been subject to considerable debate, probably because of the paradoxical nature of the disease, being both neurotropic and infiltrative, yet having an indolent, protracted course. Radiosensitivity of these tumors has now been firmly established, despite early reports to the contrary [6, 17]. While radical surgery was the mainstay of adequate therapy, combined therapy with radiation has led to superior results in many studies [1, 5, 6, 18, 19]. In light of the difficulties in obtaining clean margins, some centers are performing routine postoperative radiation therapy [5]. However, others have insisted that surgery alone is sufficient in early tumors where wide margins can be obtained without major functional loss, reserving radiation therapy (RT) for cases in which pathologic indications mandate its use [6]. In addition, adequate field-size and dose of radiation have been reported as determinants of better local control and relapse-free survival [18, 19].

The purpose of this retrospective analysis was twofold. First, we report clinical data from our institution on this rare tumor type, thus elucidating its biological behavior and comparing our survival and control rates with those reported in the literature. Second, we examine putative predictors of outcome, including the possible role of Her-2/neu overexpression.

PATIENTS AND METHODS

All patients registered with the tumor registry as having had ACC were identified. A total of 69 patients were treated for ACC of the head and neck, with both curative and palliative intent, from 1955 to 1999. Demographic and other variables including primary tumor site, dates of diagnoses, death, and local and distant recurrence were retrieved from the database provided by the tumor registry. Additionally, a comprehensive chart review of all patients was done to obtain specific information. One patient, who presented with extensive brain metastasis, was excluded from further analysis due to a brief 1-week hospital course that ended in death, and inadequate documentation of previous disease progression. Of the remaining 68 patients, 30 were men and 38 were women. The average age at diagnosis was 52 years (range 11 to 82 years), and mean follow-up was 13.2 years. Only two patients had less than 24 months of follow-up.

Major gland tumors represented 38% (26 of 68) whereas minor gland tumors comprised 62% of the cohort. The distribution of tumors according to site of origin is shown in Table 1. When adequate information was available, major salivary gland tumors were staged retrospectively according to the American Joint Committee for Cancer's (AJCC) staging protocols [20]. Minor gland tumors were staged using the AJCC criteria for squamous cell carcinomas in identical sites, which has been shown to be prognostically useful in ACC [7].

Table 1. Distribution by Tumor Site
SiteNumber
Major salivary glands26
 Parotid15
 Submandibular11
Minor salivary glands42
 Oral cavity7
  Upper lip3
  Anterior tongue1
  Other3
 Oropharynx16
  Base of tongue5
  Palate8
  Other3
 Nasal Cavity/sinus13
  Nasal cavity2
  Maxillary sinus10
  Sinus, multiple1
 Larynx/trachea6

A total of 49 patients had adequate information in their charts for stage assignment. The extent of disease according to T stage is shown in Table 2. Histopathologic variables retrieved from patient charts when available included grade of the tumor, margin status, and perineural invasion. Margin status was recorded as either positive, negative or close (< 5 mm), as has been done by previous authors [5]. Tumor grade was reviewed when pathology reports were available and assigned as well differentiated, moderately differentiated, poorly differentiated or undifferentiated (grades 1–4). The terms tubular, cribriform, and solid were roughly assigned grades 1, 2, and 3–4, respectively. For higher grade lesions, “areas of solid histology” or unspecified percent of solid component was listed as grade 3, whereas “predominantly solid” and similar descriptions, grade 4. In tumors described as “high grade,” grade 4 was assigned, whereas in those reported as “low grade,” grade 1 was assigned. In borderline cases, the higher grade was assigned. Because of the inherent difficulties of this method and the variability in the semantics used by pathologists, only 23 patients could conclusively be assigned a grade (Table 2). A total of 20 (29%) of patients had positive margins reported in their pathology reports, and 22 (32%) had documented perineural invasion, with 5 (22%) of these showing invasion of a named nerve.

Table 2. Distribution by Stage and Grade
StageT stage (n)AJCC stage (n)Grade (n)
  1. a

    AJCC = American Joint Committee for Cancer.

114137
21386
311131
411159

Treatment modalities included sole surgical resection, sole radiation therapy, combined therapy (surgery and radiation), or other (chemotherapy and radiation). The distribution of patients according to primary therapy is shown in Table 3. Of the 59 patients who had an operative procedure as part of their initial care, 15 (25%) had an adjunctive neck dissection. Field sizes (> 8 × 8 cm or less) and dose of radiation (> 60 Gy or less) were also recorded in cases receiving radiation. These cut-offs were chosen in light of previous reports showing improved control rates using these numbers [18, 19]. The mean total dose of radiation in those patients receiving radiation was 5,730 Gy. Ten patients had intraoperative brachytherapy as part of their initial treatment. Ten (14%) of patients developed a second primary after their ACC, two of which were squamous cell carcinomas of the oral cavity.

Table 3. Distribution by Therapeutic Modality
Primary modality of therapyFrequency
Surgery18
Radiation6
Surgery and radiation41
Radiation and chemotherapy2

All patient data, including demographics, staging, treatment, and outcome variables were tabulated and statistical analysis was performed using the Prodas Database Management System on a commercially available computer. The product limit method of Kaplan-Meier was used to construct survival curves, and the log-rank test was used to test differences between actuarial curves. Chi-square was used to test categorical associations.

Immunohistochemistry for HER-2/neu

An attempt to retrieve the paraffin-embedded tumor blocks of the patients described above was undertaken. A total of 29 specimens were found and serial sections, 4 microns thick, were prepared for immunohistochemistry. Sections were deparaffinized in xylene and rehydrated in graded alcohols. Endogenous peroxidases were blocked with 10% hydrogen peroxide in phosphate-buffered saline. We demonstrated clear membrane staining in our positive controls without any antigen retrieval and opted for this method in all of our cases. Slides were then incubated with 10% normal rabbit serum (NRS) to block nonspecific protein binding. The primary antibody employed was the murine monoclonal CB11 antibody (Novocastra, Burlingame, Calif.) diluted 1:40 in 10% NRS. After overnight incubation at 4°C, the secondary antibody, a rabbit anti-mouse antibody (Zymed, San Francisco, Calif.) was incubated for another hour at room temperature. Further treatment with the streptavidin-horseradish peroxidase complex (Vectastain kit, Vector, Burlingame, Calif.) and visualization with diaminobenzidine followed. Sections from a cell block of the SKBR3 cell line, known to overexpress the HER-2/neu oncoprotein, were used as positive controls. Negative controls were obtained by substituting the primary antibody incubation with 10% NRS. The slides were read by a senior pathologist and graded on a scale from 1+ to 4+, with 4+ being the most intense. Only membrane staining was considered positive.

RESULTS

Survival

Mean survival in our cohort was 7.7 years. Overall survival rates at 5, 10, and 15 years were 72%, 44%, and 34%, respectively (Fig. 1). Cause-specific survival at 5, 10, and 15 years was 83%, 71%, and 55%, respectively. Despite these relatively favorable survival outcomes, recurrence statistics were not as optimistic. Recurrence-free survival rates were 65%, 52%, and 30% at 5, 10, and 15 years, with a total of 29 of 68 (43%) eventually suffering either a local or distant failure. Overall survival was significantly affected by T stage, with T stage 1 patients showing significantly better survival than T stages 2, 3, or 4 (P < 0.03). Similar trends in survival were seen among T stages 2, 3, and 4. Similarly, significant differences in survival were seen between AJCC stages 1 and 2 when compared with 3 and 4 (Fig. 2, P < 0.008). Tumor grades 3 and 4 were also associated with decreased survival when compared with grades 1 and 2, and this difference was highly significant (Fig. 3, P < 0.005). However, because of the small numbers compared, and the difficulty in grading these tumors retrospectively from pathology reports, caution must be exercised in interpreting this particular finding. Overall survival was also significantly poorer in primaries of the nasal cavity/paranasal sinuses, when compared with primaries in the major salivary glands or of the oral cavity/oropharynx, P < 0.03 and P < 0.02, respectively (Fig. 4). No significant relationships in survival could be demonstrated with margin status, perineural invasion, primary treatment modality, total radiation dose, or radiation field sizes.

Figure 1.

Actuarial survival for adenoid cystic carcinoma in the overall population.

Figure 2.

Actuarial survival for adenoid cystic carcinoma by stage groupings. Stages 1 and 2 versus 3 and 4: P < 0.01.

Figure 3.

Actuarial survival for adenoid cystic carcinoma by grade groupings. Grades 1 and 2 versus 3 and 4: P < 0.01.

Figure 4.

Actuarial survival for adenoid cystic carcinoma by site. Major glands versus nasal cavity: P < 0.02; oral cavity/pharynx versus nasal cavity: P < 0.03.

Recurrence

As mentioned above, recurrence-free survival rates were 65%, 52%, and 30% at 5, 10, and 15 years, with 43% (29 of 68) of patients experiencing a recurrence. There were a total of 21 (31%) local failure events and 16 (24%) distant failures. Six patients experienced both a local failure event and a distant failure. Mean time to local failure was 5.6 years, whereas mean time to distant failure was 7.6 years. The most common site of distant failure was the lungs (10 of 16, 62.5%), with brain and bone being other sites of distant failure. Total recurrence-free survival, local and distant recurrence rates were distinctly better in primaries of the oral cavity/oropharynx when compared with those in other locations. Five of 23 (22%) patients with tumors of the oral cavity/oropharynx recurred, whereas 25 of 45 (56%) of the rest of the group suffered either a local or distant recurrence (Table 4). Additionally, poorer distant recurrence-free survival was significantly associated with increasing T stage and AJCC stage. No other variables were predictive for recurrence. Diseased neck nodes were found in 4 (6%) of patients. Mean survival time after discovery of lung metastasis was 4.9 years, and the overall mean survival time was 12.8 years for these patients. After excluding three more patients who had both a local recurrence and pulmonary metastasis, the seven patients with sole pulmonary metastasis showed a mean survival time after discovery of metastasis of 6.8 years and a mean overall survival time of 13.9 years. On life table analysis, presence of pulmonary metastasis conferred no survival disadvantage when compared with the rest of the cohort or even to the recurrence-free subset.

Table 4. Tumor Site and Recurrence Status
SiteRecurrence-freeRecurrenceTotal (n)
  1. a

    P = 0.007.

Oral cavity/oropharynx18 (78%)5 (22%)23
All others20 (44%)25 (56%)45

Site

Site of primary was an important predictor of outcome in our series, as discussed above. Additionally, we found in our series that major gland tumors were more likely to show perineural invasion and invasion of a named nerve than minor gland tumors (12 of 26 vs. 10 of 42 and 4 of 26 vs. 1 of 42, respectively). However, the former trend did not reach significance (P = 0.056 and 0.045, respectively). Also, major gland primaries were significantly more likely to have an associated neck dissection (10 of 26) than minor gland primaries (5 of 42), P = 0.02. No other differences were detected among the major and minor gland subsets, including survival and recurrence variables. Nasal cavity/paranasal sinus tumors were more likely to be of higher T stage (P = 0.008) and AJCC stage (P = 0.01) at presentation, and were more often associated with perineural invasion (P = 0.002).

Treatment Modality

No significant differences were found in overall, recurrence-free, local recurrence-free, or distant recurrence-free survival among the different treatment modalities. Of the six patients treated by radiation alone, none were alive at 10 years. Five of these patients were poor surgical candidates due to comorbid conditions, and were treated with curative intent, while one patient was treated palliatively. In the former group, there were four local and one distant recurrence and all patients died with disease. Adverse radiation effects were limited to skin changes. In the subset treated with sole surgical resection, four patients with documented uninvolved margins survival at 20 years was 100%, in contrast to three patients with documented involved margins in whom survival was 33% at 5 years and 0% at 10 years. Similarly, in the surgery plus radiation group, one of six patients (17%) with negative margins expired, whereas 10 of 23 (43%) with positive or close margins died (P = 0.03). This observation draws attention to the importance of margin status in these tumors, although it is difficult to make a true statistical judgment because of the small numbers compared. Dose of radiation and field sizes did not correlate with outcome variables.

Nerve Invasion and Margin Status

Also of interest was the finding that perineural invasion was significantly correlated with margin status with 15 of 20 patients with positive margins showing evidence of perineural invasion, while only 5 of 17 patients with negative or close margins showed nerve invasion (P = 0.02) (Table 5).

Table 5. Margin Status and Perineural Invasion
Margin statusPerineural invasionNo perineuralTotal (n)
  1. a

    P = 0.02.

Positive15 (75%)5 (25%)20
Negative5 (29%)12 (71%)17

Her-2/neu Expression

Of the 29 specimens retrieved, 2 (7%) demonstrated positive membrane staining, and another 2 (7%) demonstrated equivocal membrane staining. Cytoplasmic staining was observed in a number of cases, but was disregarded. No association with any of the other tumor variables, including survival, recurrence, or tumor size, could be demonstrated.

DISCUSSION

Our data concur with previous reports on ACC in terms of survival and recurrence statistics. Table 6 presents a comparison of our series with similar size series reported in the literature. Two larger series are included in Table 7 as reference studies. Stage of disease and site of primary were clearly important determinants of outcome in the present series. We confirm the contention of Spiro et al. [21] that staging these tumors using AJCC criteria is simple, reproducible, and provides a valuable and standardized method of assessing tumor burden and potential outcome. Stage has been reported as an important predictor of outcome in several studies [3, 6–8]. Grade may also serve a role in decision making, as was seen in this series. However, the significance of grade as a marker of outcome has been contested and remains unresolved, with both proponents [8, 22] and opponents [1]. In a large series of more than 200 cases reported by Spiro et al. [4], lower grade showed a trend toward longer survival, but this trend was not significant. In a later report by the same authors [3], shorter disease-free intervals in the first year of follow-up were noted for higher grade lesions. Perhaps a sensible approach given the present information would be to let grade play an adjunctive role in patient management.

Table 6. Comparison with Previous Studies
AuthorDateNSurvival (%)Recurrence-free survivalRecurrence rate (%)Factors significant for outcomeComments
  1. a

    NED = no evidence of disease; RT = radiation therapy.

Present series199969Overall survival: 72%, 44%, and 34% (5, 10, and 15 years)Recurrence-free survival: 65%, 52%, and 30% (5, 10, and 15 years). Mean time to local failure: 5.6 years. Mean time to distant failure: 7.6 years.Total failure rate: 43%. Local failure: 31%. Distant failure: 24%.Higher stage, grade and sinonasal origin adverse factors. Oral cavity/oropharynx lesions favorable.Positive margins and perineural invasion positively correlated. 7% Her2 positivity, no correlation with clinical variables.
Prokopakis et al. [12]199958Median time interval until local failure: 25 months (2.1 years).Total failure rate: 59%. Local failure: 40%. Distant failure: 19%.Nasopharynx, nasal cavity, maxillary sinus associated with greater risk of local failure. Positive margins, greater risk. No association with minor/major glands, perineural invasion, adjunctive RT, RT dose.Nonsignificant trends toward higher rates of local failure with increasing T stage and grade.
Miglianico et al. [1]1986102Overall survival: 67%, 51.4%, and 32.3% (5, 10, and 15 years)NED (recurrence- free) survival: 46.1%, 36.5%, and 26.8% (5, 10, and 15 years).Local recurrence rate: 40.2%. Distant recurrence rate: 31.4%.Superior local control with postoperative RT. Poor local control with bone involvement.Long interval of clinically asymptomatic lung metastases. 20% lymph node metastases. 12% had a second malignancy.
Matsuba et al. [13]198671Median survival for group-10 years.Total recurrence rate: 70%. Local recurrence rate: 52%. Distant recurrence rate: 52%.Improved survival in patients treated with radiation plus surgery. Solid histology type associated with decreased survival. No association with perineural invasion or margin status.Median survival after onset of distant metastasis, 4 years. Perineural invasion, 85%.
Nascimento et al. [14]198561Survival for 40 patients with either follow-up > 5 years or status “dead with disease” 37.5, 9, 3.Local recurrence rate: 37%. Distant recurrence rate: 50%.Advancing stage, solid histology, minor glands associated with decreased survival. No association with perineural or vascular invasion.67% (41) of patients in stages 3 and 4. Surgery with negative margins best modality. Adjunctive RT prolongs survival time but not survival rate.
Matsuba et al. [15]198476Median survival for group-10 years.Mean time to local failure: 4.6 years.Local failure rate: 40%. Distant failure rate: 50%.Better local control with combined modality therapy (surgery plus RT), and with dose greater than 6000 rad. No association with major vs minor gland primary sites.Lymph node metastasis seen in 3 of 76 patients (4%). Median survival time after distant metastasis, 40 months.
Table 7. Reference Studies
AuthorDateNSurvival (%)Recurrence-free survivalRecurrence rate (%)Factors significant for outcomeComments
  1. a

    NED = no evidence of disease; RT = radiation therapy.

Garden et al. [6]1994198 Local control rates: 95%, 86%, and 79% (5, 10, and 15 years).Local failure rate: 12%.Decreased rates of local control with positive margins, and invasion of a major (named) nerve.Patients treated exclusively with surgery and postoperative radiation. 69% perineural invasion. Distant metastasis most common site of failure.
Spiro et al. [3]1992184Overall survival rates: 73%, 45%, and 35% (5, 10, and 15 years).Median time to recurrence, 3 years.Total failure rate: 62%. Local failure rate: 47%. Distant failure rate: 37%.Advancing stage associated with reduced survival and poorer local and distant control. No impact of tumor grade on survival, but shorter interval to recurrence with higher grade tumors.Lungs involved in 80% of distant failures. Highest local recurrence rate (63%) in nasal cavity/paranasal sinus tumors. 15% cervical node metastases.

In our series, major gland tumors were more likely to show perineural invasion, likely due to the proximity of these the glands to major nerves, yet this did not confer a survival disadvantage. Sinonasal primaries were associated with higher stage at diagnosis, more perineural invasion, and poorer survival. This finding may be due to later presentation and more difficult surgical anatomy in this region. Similarly, discouraging results for tumors in this region have been reported by others [3, 4, 23]. Tumors of the oral cavity and oropharynx, including palatal lesions, demonstrated more favorable recurrence-free outcomes, a finding that has been shown in previous reports as well [8, 22]. Neck disease is rare in ACC, and this finding has been common [1, 3, 24]. Prolonged survival after the development of pulmonary metastasis was seen in the present series and in other reports [21]. As such, the presence of pulmonary disease should not preclude vigorous efforts to control local disease, and the two events can almost be treated as separate entities.

Evidence implicating overexpression of the HER-2/neu oncoprotein, a member of the epidermal growth factor, tyrosine-kinase receptor family, in the pathogenesis of ACC is scarce. Shintani et al. [10] stained a series of 16 cases with the CB11 antibody, and reported some level of immuno-staining in all cases. Eight of these were described as having “high grade expression.” No further details regarding inclusion criteria for positivity (membrane versus cytoplasmic) were provided. Concordance between HER-2 and HER-3 expression was reported, as well as with tubular/cribriform histology. Giannoni et al. [11] evaluated a series of palatal salivary gland neoplasms, of which 24 were ACCs, for HER-2/neu. Nine cases had a score of 2 or more on a scale from 0 to 3, and the authors reported a significant association between overexpression and poor clinical outcome among the ACC patients. In contrast, Kernohan et al. [12] stained a series of 131 salivary gland tumors, of which 19 were ACCs. Only one of these exhibited positive membrane staining for HER-2/neu. Four other specimens from the other histologies were similarly positive, and no association with clinical behavior was found. Likewise, Cho et al. [13] found focal membrane positivity in 5 of 30 cases of ACC. Staining was predominantly in tubular and cribriform areas, and was associated with shorter disease-free intervals, although the authors question the significance of the finding due to the small numbers of positive cases and the focal nature of the reactivity. Two other studies [14, 15] looking at HER-2/neu overexpression in salivary neoplasms could not demonstrate any membrane staining at all among the ACCs subsets in their series. The first group [14] had 56 ACCs in their study, whereas the second study [15] had 8. The reasons for these differences are not completely apparent at this time. Possibilities include variations in protocols, antibody sensitivities, and criteria for positivity. Nonetheless, immunohistochemistry remains a relatively efficient and simple method to evaluate archival tumor specimens, and the search for predictive markers should continue. The evidence for p53 and PCNA positivity on immunohistochemistry among ACCs has been more encouraging [13, 16] than that for HER-2/neu.

We could not demonstrate any differences among the primary modalities of therapy, perhaps because of the nonrandomization of patients into the various treatment tracks and the inclusion of palliative cases. The group receiving radiation alone was clearly a less favorable subset, representing those patients with unresectable tumors or poor functional status and inability to tolerate an extensive surgical procedure. Conceivably, the decision to add radiation to surgery (combined therapy) was also taken in those patients who were deemed “less favorable” or more “at risk.” Other studies [1, 5, 19] have demonstrated a clear outcome advantage in adding RT, and this seems prudent at present, especially in cases where margins are positive. Adequate RT dose and generous field sizes have been recommended [5, 18, 19]. Surgery alone in cases of small primaries and clear margins may be adequate. Also, the role of sole, primary radiation for favorable lesions cannot be ruled out at present. In fact, Parsons et al. [6] report good outcomes in patients with ACC treated with radiation alone, irrespective of tumor stage. Also, the superiority of neutron radiotherapy over conventional photon treatment in inoperable or recurrent salivary gland tumors has been demonstrated in a multiinstitutional, randomized study by the Radiation Therapy Oncology Group (RTOG) and the Medical Research Council (MRC) [25], which showed significant improvement in local-regional control in the neutron group. Another study [26] has shown similarly favorable outcomes more specifically in locally advanced ACC of minor salivary glands. Still, the exact role of radiation and systemic chemotherapy, both in the adjuvant setting and for palliation, needs to be defined and optimized. Perineural invasion, dose and field size did not prove to be important factors in our experience and in other reports. Larger studies have shown these factors to be important [5, 23]. Others have questioned the importance of perineural invasion [22, 24]. Our finding of a significant correlation between positive margins and perineural invasion has been reported by others [27]. This may reflect more biologically aggressive and infiltrative tumors, thus making it more difficult to obtain clear margins. As with other reports regarding this disease, our study is subject to the usual limitations of a retrospective study. Nonetheless, these data do provide insights into the natural history and progress of this disease, which may be useful in clinical decision-making. Clearly, multi-center prospective studies are needed to define the optimal treatment of ACC of the head and neck.

Ancillary