SEARCH

SEARCH BY CITATION

Keywords:

  • adenoid cystic carcinoma prognosis;
  • biomarkers;
  • p63 immunostaining;
  • salivary adenoid cystic;
  • computerized image analysis

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

BACKGROUND:

In a long-term retrospective immunohistochemical study of adenoid cystic carcinoma (ACC) of salivary gland, we investigated the relation of p63 immunodetection to prognosis. Although it is generally agreed that the solid pattern is the most aggressive pattern of growth, ACCs with predominantly cribriform or tubular patterns have an unpredictable clinical course, with a relatively favorable 5-year survival but a low 20-year survival.

METHODS:

Formalin-fixed paraffin sections from 35 cases of ACC showing a predominantly better differentiated histopathology, ie, cribriform or tubular patterns of growth, were immunostained for p63. Automated image analysis was used to quantify p63 positivity, using a modification of a previously developed algorithm.

RESULTS:

Patients alive for more than 10 years had a lower extent of p63 expression than those who died of disease. Kaplan-Meier analysis revealed that separation of patients with morbidity and mortality from those alive with no evidence of disease, could be achieved at a cutoff of 35% p63 positivity (P = .0031, log-rank test). Multivariate analysis using the Cox proportional hazard model revealed p63 and tumor stage to be independent predictors of survival (P = .012 and P = .0003, respectively).

CONCLUSIONS:

To our knowledge, the present study is the first to report prognostic significance of p63 in salivary gland ACC and the first report of a robust and well-studied immunohistochemical stain performable on routinely fixed and processed tissue with prognostic utility. Cancer 2010. © 2010 American Cancer Society.

Adenoid cystic carcinoma (ACC) of salivary gland comprises approximately 10% of all epithelial salivary gland tumors. It may involve the parotid, submandibular, or minor salivary glands.1-15 Typically, ACC is characterized by persistent slow growth, high rates of recurrence, and distant metastasis, resulting in poor patient survival.1-2 Five-year survival is approximately 75%, however, long-term survival is poor, with 25% alive after 20 years.3

The histopathological patterns of ACC can be categorized into three subtypes: cribriform, tubular, and solid.3-5 Most studies suggest that a predominately solid histopathologic growth pattern correlates most strongly with poor prognosis, compared with predominant cribriform or tubular patterns.3-5 Various parameters such as positive surgical margins, clinical stage, histopathological patterns and perineural invasion have been reported as relevant prognostic factors in patients with ACC.6-7 Unfortunately, these may not always be accurate predictors because they do not take into account the variable biology of individual tumors. The discovery of a reliable prognostic biomarker could facilitate the triaging of ACC patients into different treatment categories3, 8 as well as reassure a subset of patients with ACC of a lower risk of morbidity and mortality. The expression of p63, a widely studied marker of basal cells in normal salivary glands, breast, prostate, respiratory and squamous epithelia,16-18 and of tumor cells from various malignancies,19-33 has been the subject of relatively few studies in ACC to date.9-10 To our knowledge, the present study is the first to examine and establish a correlation of the extent of p63 expression with prognosis of ACC.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

Selection of Specimens

p63 expression was examined in archival tissue specimens of ACC of salivary gland from 35 patients for whom long-term follow-up data were available. Criteria used for inclusion in this study are as follows: 1) patients treated at our institution between the years 1981-2005 for whom detailed clinical follow-up could be obtained; 2) a minimum of 4.5 follow-up years for the 14 patients alive and with no evidence of disease, with 8 of 15 having greater than 10 years of follow-up time; 3) all patients failing treatment or dying of disease for whom a detailed medical record at our institution was examined; 4) histological sections showing only the better differentiated cases (cribriform and tubular variants as the predominant histopathologic pattern): cases with purely or mostly solid histopathology were excluded and 2 cases with about 65% well differentiated and 35% solid areas were included. The cases were staged according to the tumor, node, metastasis (TNM) system adopted by the American Joint Committee on Cancer Staging in 2002.34 Length of survival was calculated from the date of surgery to the date of the latest clinical follow-up, or to date of death by disease. Patients who died of postoperative complications and those dying from other causes unrelated to ACC were excluded from the study.

One or two paraffin tissue blocks from the primary tumor excisional specimen were selected for each case based upon review of hematoxylin and eosin stained sections. Tissue blocks containing normal surrounding salivary gland tissue were used whenever possible, allowing for internal positive control for p63 staining and comparison to normal salivary gland staining. Histological grading of ACC was performed retrospectively in accordance with the criteria described by Szanto et al.5 Twenty-nine cases were classified as grade 1 (Fig. 1), 4 cases as grade 2 and two cases with >30% solid histology as grade 3.

thumbnail image

Figure 1. Classic grade I adenoid cystic carcinoma demonstrating cribriform and tubular pattern of growth and absence of solid pattern (H & E magnification, ×10).

Download figure to PowerPoint

Immunohistochemistry

Immunohistochemical staining was performed as follows: 4 uM thick sections were deparaffinized, treated with 3% hydrogen peroxide to block endogenous peroxidase activity, then treated with 0.01M citric acid (pH 6.0) for 5 minutes at 100°C, and followed by slow cooling for antigen retrieval. Slides were then incubated overnight at room temperature with anti-p63 monoclonal antibody 4A4 (1:600, Santa Cruz Biotechnology, Santa Cruz, Calif) stained using either a streptavidin-biotin kit (Biogenex, San Ramon, Calif, catalog no.QP900-9L)) or DAKO Envision plus reagents, followed by exposure to diaminobenzidine as chromogen and hematoxylin as counterstain.

Immunohistochemical staining for p63 was evaluated by 2 authors (NR & DB) who were blinded to clinical data. p63 expression was considered positive only if distinct nuclear staining was present.

Initially a semiquantitative assessment of p63 expression was performed in tumor cells (data not shown) as follows: 1 = positive p63 nuclear staining in fewer than 50% of cells, 2 = positive nuclear p63 in 50%-75% of cells and 3 = p63 nuclear positivity in >75% cells. The intensity of immunostaining was also evaluated as absent (0), weak (−), and moderate to strong (+).

Computerized Image Analysis

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

To more precisely establish the proportions of negatively and positively stained tumor cells, an image analyzing algorithm was developed using conditional morphological operations, which represents a modification of a previously reported technique.35 Two different grey-scale images were composed from the color components of the original image such that each of the grey-scale images had relatively lower average intensities in the regions of either negative or positive cells than those in the rest of the image. In each of the composed images, the pixels whose intensities were higher than the global median of the image were reset to the median intensity. A low-pass filter with the Gaussian transfer function was applied to smooth each of the flattened images. An iterative approach of conditional morphological operations was applied and sparsely spiked images were produced. Large spikes indicated the locations of cells. Twenty or more images from each case were taken randomly to test for reproducibility of initial results and to simulate everyday practice (Fig. 2). The slides were assigned a random number so that investigators could work blinded to clinical data.

thumbnail image

Figure 2. p63 immunohistochemical stain illustrating p63 positive-negative staining tumor cells (magnification, ×20).

Download figure to PowerPoint

Patient Cohort

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

A computerized search of the Department of Pathology files at the Mount Sinai Medical Center during the years 1993-2005 revealed 32 patients who had definitive surgical excision of ACC of salivary gland or head and neck origin at our institution, for which a detailed medical record could be obtained, and who met other criteria described above. Three additional patients presenting with recurrence during the search years were found to have primary tumors dating as far back as 1981 and had available medical records and histopathologic blocks. Thirty-five patients in total were included in this study, which comprised 11 men and 24 women and mean age was 51.5 years (range, 18-82 years). Thirteen cases involved the major salivary glands and 22 involved the minor glands. Sites of origin were parotid (n = 9), sinonasal (n = 10), palate/maxilla (n = 6), submandibular gland (n = 4), tongue (n = 4), and larynx (n = 2). All 35 specimens were excisional specimens. Nine patients were TNM stage I, 18 patients stage II, 6 patients stage III, and 2 patients stage IV (Table 1). All 35 patients had surgery and postoperative radiation as initial disease treatment. Two patients later received additional chemotherapy for lung and lymph node metastases. Fifteen patients had tumor in or near (< 1 mm) surgical margins. Two male patients with T2 (2.5 cm) sinonasal tumors received a second surgical treatment 6 weeks after initial surgery; in both patients, negative margins were recorded in the second operative record and pathology report (Table 1).

Table 1. Clinical Comparison Between Clinical Pathologic Factors, P63 Staining, and Clinical Outcome Among 35 Adenoid Cystic Carcinoma Patients
Age, ySexLocationGrade, Stage, TNM Margin Status% p63 PositivityStatusLocal RecurrenceMetastasisFailure Time, moOverall Survival Time, mo
  • + indicates positive or close tumor surgical margin; A&NED, alive and no evidence of disease; DOD, dead of disease; submand, submandibular; 0 month, initial presentation; A&WD alive with disease.

  • *

    Had second surgical procedure for positive margin.

74Msinonasal1,3+0.31A&NED62
51Msinonasal1,2*1.32A&NED140
45Fparotid1,111.19A&NED102
54Fparotid1,111.67A&NED156
55Fpalate MX3,2+15.64A&NED138
47Mparotid1,2+16.48A&NED132
53Fsinonasal1,225.14A&NED54
38Fsinonasal1,2+28.5A&NED91
71Mtongue1,232.46A&NED114
69Mpalate1,134.53A&NED81
62Fpalate MX1,236.48DODyes3838
36Fsinonasal2,2+36.55DODyes72102
45Fpalate MX2,T3N0M1+36.72DODlung, 0 mo11
64Mpalate1,1+38.46DODyes101141
32Fsubmand1,146.13A&NED156
61Fparotid1,146.93DODyesliver, 24mo2445
47Msinonasal1,347.28DODyes3232
31Ftongue1,250.09DODyeslymph node, 32mo32114
43Ftongue1,T3N150.81DODlymph node, 0 mo82
69Fsinonasal1,1+51.78A&WDyes115134
18Msinonasal2,252.69DODyeslung, 84 mo84231
52Msinonasal1,2*53.92A&WDyeslung,137 mo40144
26Fparotid1,254.57DODyeslung, 132mo69146
65Flarynx1,2+56.97DODyeslung, 89 mo8991
50Msubmand1,257.77DODyes1848
33Fparotid1,2+58.19DODyeslymph node, 25mo2565
66Fparotid1,160.14DODyeslung ,72mo60122
82Flarynx1,T3N1M1+60.83DOD lung & lymph node, 0 mo10
44Fpalate MX1,261.24A&WDyes24141
51Fparotid1,162.87A&WD3 times2556
56Fsinonasal3,T2N1+64.78DOD lymph node, 0 mo31
42Fsubmand2,271.37DODyes101101
46Fparotid1,2+72.71DODyes5656
50Msubmand1,T3N174.67DODlymph node, 0 mo21
77Ftongue1,T3N176.36DODlymph node, 0 mo30

Clinical Course

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

Follow-up revealed 15 of 35 patients were alive. Eleven patients were without evidence of active disease at the end of follow-up, which ranged from 54 to 157 months (mean 111 months or 9.2 years) and 4 were alive with disease, 3 with recurrence or local failure and 1 with both local recurrence and distant lung metastasis. The follow-up range for the 4 patients alive with disease was 58-144 months with a mean of 121.7 months or 10.1 years. Twenty patients died of ACC, 13 from metastatic disease, and 7 from uncontrolled local recurrent or persistent disease at primary tumor site. Fourteen patients had metastases involving 15 metastatic sites that included 7 lung, 1 liver, and 7 regional lymph nodes.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

p63 was detected in all 35 cases and was manifested as well-delineated nuclear staining. When analyzed for prognostic significance, increased p63 expression in tumor cells correlated with poorer survival, compared with tumors with relatively reduced p63 expression. Kaplan-Meier analysis revealed that a cutoff of 35% of tumor cells with p63 positivity separated patients with morbidity and mortality from those with long-term survival (P = .0031, log-rank test) (Fig. 3). Immunodetection of p63 was less than 35% in 10 patients and 35% or greater in 25 patients. Of the 24 patients failing treatment, all scored >35% for p63 staining; 4 (4/25) were alive with disease and 20 (20/25) died of disease. Figure 4 demonstrates the distribution of percentages of p63 positivity among the 3 groups of patients (those dying of disease [n = 20], those alive with disease [n = 4], and those alive with no evidence of disease [n = 11]).

thumbnail image

Figure 3. Kaplan-Meier analysis relation between p63 immunostaining and survival. Dotted line: >35% tumor cells are p63 positive; continuous line: >35% p63 positive.

Download figure to PowerPoint

thumbnail image

Figure 4. Distribution of mean percentage rates of p63 immunodetection among 35 patients (20 high-powered fields/slide). Patients are categorized as alive with no evidence of disease (n = 11), alive with disease (n = 4), and those dying of disease (n = 20).

Download figure to PowerPoint

Multivariate analysis using the Cox proportional hazard model revealed p63 as well as tumor stage to be independent predictors of survival (P = .012 and P = .0003, respectively). A Kaplan-Meier curve (Fig. 5) generated by combining stage and p63 variables demonstrates the comparison of survival among 3 groups. In Group A (p63 > 35% and low stage), all 9 patients are alive and with no evidence of disease at the end of study as compared with Group C (p63 > 35% and high stage) in which all 7 patients are dead of disease, with median survival time = 898 days. In Group B (other combinations; low p63/ high stage, high p63 / low stage), 13 of 19 remaining patients have died of disease with median survival time 3437 days. Other variables—age, gender, histologic grade, location, and margin status—had no predictive value. Unlike survival, no variable was found to be a predictor of likelihood of recurrence.

thumbnail image

Figure 5. Kaplan-Meier curve comparison of survival times among three groups. Group A: (n = 9) p63 < 35% and low stage; Group B: 13 of 19 patients dead of disease (combined p63 > 35% and high stage, p63 > 35% and low stage); Group C: (n = 7) p63> 35% and high stage.

Download figure to PowerPoint

By using a Student t test, 2 groups of patients were compared as follows: 1 group comprised all patients alive for more than 10 years (n = 11) and the other group comprised all patients dead at ≤10 years (n = 17). Seven patients alive for less than 10 years were excluded from this test. Equal variances were not assumed for the 2 groups. The 2 groups had significantly different average percentages for p63 positivity (P = .049.)

Associations between categorical variables were tested using the chi-square test or the Fisher exact test as needed. Significant association was noted between p63 score and patients failing treatment (P > .00001). Likewise significant association was observed with p63 score and death by disease (P = .000016).

Follow-up of the 2 patients that received additional initial surgery for positive tumor margins revealed 1 (Case 2, Table 1) to be alive with no evidence of disease at 11 years 8 months and the other (Case 22, Table 1) with local recurrence at 40 months and distant lung metastases 11 years 5 months after initial treatment. Of interest, p63 positivity in the former was 1.32%; in contrast, in the latter, p63 positivity was 53.9% (Table 1).

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

ACC is an uncommon epithelial tumor constituting 10% of all epithelial salivary gland tumors.1, 11 Classically, ACC has been described as a tumor with indolent but persistent and recurrent growth and late onset of metastases to lungs, bone, and liver,13 which eventually leads to death.12 Although the 5-year survival is deceptively high (75%), long-term survival is poorer, with only 25% of patients alive at 20 years.3 In one large series, 5-, 10-, 15-, and 20-year overall survival rates for all head and neck sites of ACC were 72%, 50%, 38%, and 23%, respectively.2 Clinically, however, all ACC are considered to have an uncertain prognosis and an unpredictable course.1-3, 14, 15

Higher grade corresponds to worse prognosis.4-5 Multivariate analysis of the present study did not find histologic grade to be an independent predictor of survival, although only 2 of our series had grade 3.

Treatment planning for ACC, like squamous cell carcinoma, strictly depends on the accepted TNM staging system, which is also useful for determining prognosis. Unfortunately, the TNM system may not always be accurate, because it does not take into account the tumor biology of individual cases.22, 36

Margin status has been reported to be prognostically significant.7 However, multivariate analysis revealed no statistical significance of margin status in our series of cases.

Considerable interest has focused on p63, a gene related to p53, which may give rise to 6 isoforms as a result of initiation of transcription from alternative promoters and alternate splicing,16-18 Because of overexpression in epithelial cells and various solid tumors, 1 isotype, ΔNp63, has been suggested to have an oncogenic role in the regulation of proliferation and differentiation in many neoplasms.20

A direct role for p63 in tumorigenesis has not been demonstrated to date, although amplification of the 3q27 region has been detected in several tumors including squamous cell carcinoma.37 This is suggestive of a putative role of at least a subset of p63 isoforms as oncoproteins rather than as tumor suppressors.17 We previously demonstrated that some adenoid cystic carcinomas show a distinct pattern of p63 staining at the periphery of the cell nests.9 On the basis of its staining patterns, p63 is useful in the distinction of ACC from basaloid squamous cell carcinoma9 and high-grade neuroendocrine carcinoma.21

A recent microarray analysis of lung adenocarcinoma ranked p63 as one of the 100 genes whose mRNA expression is most closely associated with poor patient survival.23, 24

There are several studies demonstrating the correlation of p63 overexpression with prognosis. High p63 expression was associated with a more aggressive phenotype and poor prognosis in a study with 94 oral squamous cell carcinoma cases.20 Recently, p63 expression was shown to be a reliable indicator of histological grading and an early marker of poor prognosis in oral squamous cell carcinoma patients.22 In patients with head and neck squamous cell carcinoma, diffuse p63 expression, a characteristic of poorly differentiated neoplasms, correlated with poorer survival rates compared with the group with reduced p63 expression.20

A positive correlation was found between p63 expression and increased meningioma grade: 95% of grade 1 meningiomas lacked nuclear p63 expression and none exhibited cytoplasmic staining. Ninety-two percent of grade 2 meningiomas showed nuclear p63 expression and 31% also showed cytoplasmic expression. Seventy-five percent of grade 3 meningiomas showed nuclear expression and all grade 3 meningiomas expressed cytoplasmic p63.33 p63 positivity was also found to have prognostic significance in Merkel cell carcinoma, with positive cases demonstrating a more aggressive clinical course as compared with cases that did not express p63.19 Conversely, in a study of bladder carcinoma, lower p63 expression was significantly associated with a poor prognosis.29, 30 In esophageal squamous cell carcinoma, the 5-year overall survival was significantly longer in p63-positive compared with p63-negative tumors as defined by a specific cutoff value for p63 expression.28

We demonstrate herein significantly higher mortality rates in ACC patients whose tumors had higher p63 expression compared with those with lower p63 expression. Kaplan-Meier analysis revealed statistical significance could be achieved at a cutoff of 35% positivity of tumor cells. Statistical analysis comparing immunodetection of p63 among 11 patients alive for 10 or more years and 17 patients dead by 10 years demonstrated a significant difference among the percentage mean of rates of p63 immunodetection between the 2 groups. Multivariate analysis using the Cox proportional hazard model revealed p63 as well as stage of disease to be independent predictors of survival.

Taken collectively, our results suggest that the extent of p63 positivity may be a useful predictor of clinical outcome. Although staging continues to play a pivotal role in the management of patients with ACC, high p63 expression could warrant a more aggressive mode of therapy. Conversely, low p63 expression might identify a patient population more likely to survive both short-term and long-term.

In summary, the present findings suggest quantitative assessment of p63 expression may be useful in predicting prognosis and clinical course in ACC of salivary gland origin.

CONFLICT OF INTEREST DISCLOSURES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES

This work was supported by a generous bequest from the Estate of Hilda Leveen (to David E. Burstein).

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. Computerized Image Analysis
  5. Patient Cohort
  6. Clinical Course
  7. RESULTS
  8. DISCUSSION
  9. CONFLICT OF INTEREST DISCLOSURES
  10. REFERENCES