p53 and Ki-67 as markers of radioresistance in head and neck carcinoma

Authors


Abstract

BACKGROUND

p53 and Ki-67 are regarded as potential interesting predictors of radioresistance, although their exact influence awaits confirmation on a large cohort of uniformly treated patients.

METHODS

In a retrospective cohort of 304 patients with squamous cell carcinoma of the head and neck who were treated with radical radiotherapy, the expression levels of p53 and Ki-67 were assessed by immunohistochemistry. Local control and survival curves were generated for p53 and Ki-67 using the Kaplan–Meier method. The difference between curves was calculated in univariate and multivariate analyses.

RESULTS

The overexpression of p53 was associated with local treatment failure (P = 0.01) but not with survival (P = 0.09). In a Cox analysis, p53 overexpression remained an independent predictor of local failure, with a relative risk of local failure of 1.5 (P = 0.05). Low proliferation (Ki-67 < 20%) was a significant factor in local failure for patients with tumors of the oral cavity only (P = 0.01). Patients with both unfavorable immunohistochemical markers (p53 overexpression and low proliferation) had a 45% rate of local control compared with a 67% rate for all other combinations (P = 0.002). This association was even more significant in patients with T1–T2 lesions (45% vs. 77%; P = 0.0002).

CONCLUSIONS

The results support the role of p53 as an independent predictor of local failure in patients with squamous cell carcinoma of the head and neck who are treated by radical radiotherapy, suggesting that it may predict radioresistance. Combined with p53, Ki-67 may help in the better selection of patients for radiotherapy, especially for patients with early-stage tumors. Prospective studies are now needed to confirm these results and to define better the role of these markers in the management of patients with head and neck carcinoma. Cancer 2002;94:713–22. © 2002 American Cancer Society.

DOI 10.1002/cncr.10232

One of the most challenging aspects of head and neck oncology is the selection of the most appropriate treatment for an individual patient with squamous cell carcinoma. Equivalent results have been obtained with radiotherapy and surgery1, 2 in terms of local control and survival, as confirmed in two recent series of patients with early-stage tumors3 and selected advanced tumors.4 These data, as well as the encouraging results from neoadjuvant chemotherapy and radiotherapy for patients with larger tumors,5, 6 call for the identification of new accurate predictors of response to radiotherapy so that the appropriate therapy can be tailored better for individual patients.

p53 and Ki-67 are regarded as good candidates for radioresistance markers. p53 is a nuclear protein that induces G1 cell cycle arrest in response to DNA damage, allowing cells to repair DNA before subsequent DNA synthesis. If DNA repair fails, then p53 is thought to induce apoptosis. In approximately 50% of squamous cell carcinomas of the head and neck, wild type p53 function is abrogated by mutations generating nonfunctional and more stable p53 proteins that accumulate in the nucleus of tumor cells. The hypothesis is that the accumulation of p53 protein induces radioresistance by reducing tumor cell apoptosis after DNA damage by radiotherapy. Tumors devoid of the p53 gene, when they are injected into immunocompromised mice, are more resistant to gamma radiation and contain fewer apoptotic cells compared with tumors that express the wild type p53 gene.7 Furthermore, the radioresistant phenotype of a head and neck carcinoma cell line (JSQ-3) that bears a mutant form of p53 can be suppressed in vitro and in vivo by treatment of the cells with a viral vector containing wild type p53.8 Ki-67 is a nuclear protein that is expressed in actively proliferating cells and is absent in nonproliferating cells (phase G0). In a hypoxic environment, which is a radioresistance factor per se,9 cells are preferentially in G0 phase.10 Therefore, low Ki-67 levels, by revealing the low proliferation rate of a tumor, are believed to reflect hypoxia and radioresistance.

Since Vokes et al.11 reviewed head and neck carcinoma in 1993, several groups have focused on the influence of p53 and Ki-67 as potential markers of radioresistance in patients with this type of neoplasm. The majority of these studies, which included patients with no uniform radiotherapy treatment, have been the source of numerous and conflicting reports and have not yielded an accurate scheme to predict the response to radiotherapy for individual lesions.12 The current study was undertaken to clarify the role of p53 and Ki-67 in radioresistance using a large cohort of patients who were treated uniformly with primary radiotherapy.

MATERIALS AND METHODS

We studied a retrospective cohort of 304 patients who were treated with primary radical radiotherapy with curative intent (dose > 50 grays [Gy]) for squamous cell carcinoma of the head and neck at our institution between 1986 and 1997. Ninety patients from our pilot study12 were included, because we used a different anti-p53 antiserum for this study. Ki-67 immunohistochemistry performed in the pilot study also was reevaluated by a different observer (C.C.). Clinical data retrieved from patient charts included gender, age, tobacco use, alcohol use, clinical T and N classifications, and tumor site. The endpoints were local control and survival. Local treatment failure was defined as either presence of neoplasm after radiotherapy or the appearance of local recurrence at the irradiated site confirmed by histology.

All histology slides from initial pretreatment biopsies were reviewed by two authors (C.C. and B.T.) to confirm the diagnoses. A representative block was selected for each patient. Unstained sections (4 μm) were cut for routine staining with hematoxylin and eosin and for immunohistochemistry. Tumors were graded as well differentiated (Grade 1), moderately differentiated (Grade 2), or poorly differentiated (Grade 3). Immunostaining for p53 (p1801; Oncogene Science, Cambridge, MA; dilution, 1:500) and Ki-67 (MIB-1; Immunotech, Westbrook, ME; dilution, 1:50) was performed after epitope retrieval using the avidin-biotin-peroxidase complex method described by Hsu et al.13

Immunohistochemical staining was evaluated independently by two authors (C.C. and B.T.) who were blinded to the clinical data. Immunostaining for p53 was evaluated at intermediate magnification (×100) using a semiquantitative scale (0%, < 10%, 10–50%, or > 50% positive tumor cells). For statistical analysis, an accumulation or overexpression of p53 was considered present if ≥ 10% of the malignant cells showed nuclear positivity. For Ki-67, the number of positively staining nuclei was counted in 1000 tumor cells at high magnification (×400) and was expressed as a percentage of the total positive cells. Only unequivocal staining was interpreted as positive. For statistical analysis, based on our prior study,12 a threshold value of 20% Ki-67-positive tumor cells separated tumors with a high proliferation rate (Ki-67 ≥ 20%) from tumors with a low proliferation rate (Ki-67 < 20%).

Statistical analysis was performed using the Statistica software package (Statsoft, Tulsa, OK). First, local control and survival curves were generated for all clinical, histologic, and immunohistochemical data using the Kaplan–Meier method. Statistical differences between curves were calculated using log-rank or Wilcoxon tests. A multivariate analysis using the Cox regression model14 was then performed on all variables with significant prognostic influence in univariate analysis (P ≤ 0.05). The assumption of proportional hazards was tested using the time dependent covariate test.15 Comparisons between groups were made using Pearson or maximum-likelihood chi-square tests for categoric data and Student t tests for comparison of means. For the comparison of multiple means, an analysis of variance (ANOVA) or Kruskall–Wallis test was used.

RESULTS

Population

The cohort was composed of 241 men (79%) and 63 women (21%) with a mean age at the time of diagnosis of 60 years for men (range, 25–88 years) and 59 years for women (range, 39–79 years). The site of the primary tumor was the oropharynx in 104 patients (33%), the oral cavity in 77 patients (25%), the larynx in 90 patients (30%), the hypopharynx in 21 patients (7%), and the nasopharynx in 12 patients (4%) (Table 1). Distribution by T classification and N classification is shown in Table 1. The majority of patients had early T-stage tumors (63% T1–T2) and early N-stage lymph nodes (74% N0–N1).

Table 1. Univariate Analysis of Local Control and Survival at 5 Years After Diagnosis
VariableNo.%Local controlOverall survival
%P value%P value
  • a

    The number of patients for whom the value was not available for analysis. The total number of patients was 304.

Gender
 Male24179620.40490.02
 Female63217173
Tobacco use
 Never166730.18860.005
 Stopped55227661
 Active182726349
Alcohol use (55a)
 Never3916780.20820.03
 Stopped50206957
 Active160646354
Site
 Larynx9030740.01750.07
 Oropharynx104346645
 Oral cavity77255754
 Nasopharynx1245745
 Hypopharynx2174845
T classification (1a)
 T16321860.00006540.00008
 T2127426861
 T357195147
 T456184330
N classification (1a)
 N015451690.10560.0002
 N170235950
 N251175443
 N32897243
Grade (1a)
 110536590.30560.20
 2139476256
 350176546
p53 (2a)
 < 10%16956690.01570.09
 ≥ 10%133445951
Ki-67 (1a)
 < 20%12541630.76590.43
 ≥ 20%178596550

Treatment

All patients were treated with primary radiotherapy: Two hundred forty-six patients (81%) received external radiotherapy alone, and 58 patients (19%) received external radiotherapy plus a boost of brachytherapy (usually for T3–T4 lesions). The total dose of radiation was higher for patients with advanced T classification tumors (P = 0.001; ANOVA). The mean total doses and 95% confidence intervals (95% CI) for patients with T1–T4 tumors, respectively, were as follows: 61 Gy (95% CI, 59–63 Gy), 64 Gy (95% CI, 62–65 Gy), 67 Gy (95% CI, 65–70 Gy), and 64 Gy (95% CI, 61–66 Gy). The total dose was often adjusted in accordance with the response of the tumor. The dose per fraction was 2 Gy. None of the patients was treated with chemotherapy, radiochemotherapy, or surgery for the primary tumor. Thirty-six patients (11%) underwent cervical dissection as part of their initial management: 24 patients in association with external radiotherapy and brachytherapy and 12 patients with external radiotherapy alone.

Follow-Up

There were 99 local treatment failures (33%) and 145 deaths (48%), of which 98 deaths (68%) were caused by disease. The follow-up without local recurrence ranged from 2 months to 120 months (median, 48 months). Only 33 of 205 patients (16%) with local control had < 2 years of follow-up, and the median follow-up for these patients was 55 months.

Immunostaining

Nuclear accumulation of p53 protein by ≥ 10% of tumor cells was present in 133 tumors (44%). Many specimens also contained nuclear staining in nontumoral squamous epithelium that was either dysplastic or hyperplastic, usually in the basal layer, although this staining was not counted. Nuclear staining of Ki-67 ranged from 1% to 90% of tumor cells. Based on a 20% threshold value, there were 125 tumors (41%) with a low proliferation rate and 178 tumors (59%) with a high proliferation rate.

Univariate Analysis

Traditional clinical parameters, as expected, remained of paramount influence. T classification was the strongest predictor of local control (P = 0.00006) and overall survival (P = 0.00008). N classification also had a very significant association with survival (P = 0.002). Local control was influenced by the anatomic site of the primary tumor (P = 0.01), with the larynx, the oropharynx, the oral cavity, the nasopharynx, and the hypopharynx the most favorable sites, in decreasing order. Survival was better in women (P = 0.02), nonsmokers (P = 0.05), and nondrinkers (P = 0.03). We found no influence for other clinical parameters—age, histologic grade, or radiation dose—on either local control or survival. Overexpression of p53 was correlated significantly with local treatment failure (Fig. 1A). Two years after diagnosis, patients with p53 positive tumors had a local control rate of only 60% compared with a rate of 75% for patients with p53 negative tumors. At 5 years after diagnosis, the difference remained at 59% versus 69%, respectively. The same trend, although it was not significant, was observed for survival (P = 0.09). The proliferation rate was not associated with either local control (P = 0.76) or survival (P = 0.43) when the entire cohort was considered. Neither p53 nor Ki-67 was associated with any of the clinical variables.

Figure 1.

(A) Local control curve for categorized values of p53 using the Kaplan–Meier method. The statistical difference between curves is indicated in the top right corner. (B) Local control curves for categorized values of Ki-67 using the Kaplan–Meier method (oral cavity only). The statistical difference between curves is indicated in the top right corner. (C) Local control curve for categorized values of combined p53 and Ki-67 using the Kaplan–Meier method (T1–T2 lesions only). The statistical difference between curves is indicated in the top right corner.

Analysis of Subgroups

First, we found that, when the group of patients with tumors of the oral cavity was evaluated separately (n = 77 patients), patients who had tumors with a high proliferation rate (Ki-67 ≥ 20%) responded much better to radiotherapy, with a local control rate at 5 years of 76% compared with a rate of only 46% for patients who had tumors with a low proliferation rate (Ki-67 < 20%; P = 0.01) (Table 2, Fig. 1B). Here again, there was no influence of Ki-67 on prognosis (P = 0.4). Second, the influence of p53 was studied for T classification subgroups (Table 3). Excluding patients with T4 tumors substantially increased the strength of the association between p53 overexpression and treatment failure (P = 0.007). Also, poor disease specific survival (but not overall survival) was associated with p53 overexpression, even when patients with T4 tumors were included (P = 0.05). Third, when the combined influence of p53 and Ki-67 was assessed (Table 4), the simultaneous presence of p53 overexpression and high proliferation emerged as a major factor for local treatment failure that translated into a 22% drop of the control rate (from 67% to 45%; P = 0.002) when lesions that had both unfavorable markers were compared with lesions that had only one or none (Fig. 1C). When only patients with T1–T2 lesions were included, the difference in the local control rate reached 32% (77% vs. 45%; P = 0.0002). Such correlations were not observed in patients with T3–T4 lesions. Finally, with regard to p53, a trend was observed between the combination of abnormal p53 and low Ki-67 with decreased overall survival (P = 0.09), whereas a significant association was found with disease specific survival (P = 0.02).

Table 2. Association of Ki-67 with Local Control and Survival (Univariate analysis of site subgroup) at 5 Years After Diagnosis
SiteKi-67 (%)Local controlOverall survivalDisease specific survival
%P value%P value%P value
Oral cavity< 20460.01480.60640.40
Oral cavity≥ 20765973
Table 3. Association of p53 with Local Control and Survival (Univariate analysis of T-stage subgroups) at 5 Years After Diagnosis
T classificationp53 (%)Local controlOverall survivalDisease specific survival
%P value%P value%P value
All< 10690.01570.08720.05
All≥ 10595164
T1–T3< 10740.007600.17760.02
T1–T3≥ 10606067
T4< 10500.97400.70500.90
T4≥ 10501650
Table 4. Association of Combined p53 and Ki-67 with Local Control and Survival (Univariate analysis of T-stage subgroups) at 5 Years After Diagnosis
T classificationp53 and Ki-67aLocal controlOverall survivalDisease specific survival
%P value%P value%P value
  • a

    Both unfavorable: p53 ≥ 10% and Ki-67 < 20%.

AllBoth unfavorable450.002540.09550.02
AllAll others674571
T1–T2Both unfavorable450.0002510.10560.006
T1–T2All others776380
T3–T4Both unfavorable500.80530.90330.80
T3–T4All others505239

Multivariate Analysis

All variables that were correlated with local control or survival in the univariate analysis were analyzed using three Cox regression models (Table 5). In the first model (local control; all patients), the variables T classification (P = 0.00002), tumor site (P = 0.01), and p53 expression (P = 0.05) remained independent factors for local control. The adjusted risk of local failure for tumors that overexpressed p53 was 1.5 compared with tumors that had no p53 overexpression. It is noteworthy that the risk of failure reached 1.9 for tumors with the combination of p53 overexpression and low proliferation (Ki-67 < 20%; P = 0.01). When T4 tumors were excluded (second Cox model: local control; T1, T2, and T3 tumors), the risk increased to 1.7 with p53 overexpression (P = 0.02) and to 2.2 with both p53 overexpression and low proliferation (P = 0.005). In the third model (survival; all patients), the variables gender (P = 0.03), T classification (P = 0.002), and N classifications (P = 0.006) remained associated with survival. There was also a trend toward an unfavorable prognosis with p53 overexpression, but it did not reach statistical significance (P = 0.09).

Table 5. Multivariate Analysis
Variable:RatioAdjusted relative risk (of local failurea or of dyingb)95% confidence interval of relative riskP value
  • a

    First and second models.

  • b

    Third model.

  • c

    Both unfavorable: p53 ≥ 10% and Ki-67 < 20%.

First Cox model: Local control (all patients)
 Site
  Oropharynx/larynx1.21.0–1.40.01
  Oral cavity/larynx1.51.3–1.8
  Nasopharynx/larynx1.81.6–2.9
  Hypopharynx/larynx2.32.0–2.7
 T classification
  T2/T11.51.2–1.90.00002
  T3/T12.31.9–2.8
  T4/T13.52.9–4.3
 p53
  ≥ 10%/< 10%1.51.0–2.20.05
 p53 and Ki-67
  Both unfavorablec/all others1.91.2–3.00.01
Second Cox model: Local control (T1, T2, and T3 only)
 p53
  ≥ 10%/< 10%1.71.1–2.70.02
 p53 and Ki-67
  Both unfavorablec/all others2.21.4–3.50.005
Third Cox model: Survival (all patients)
 Gender
  Male/female1.41.0–3.00.03
 N classification
  N1/N01.31.1–1.50.006
  N2/N01.61.4–1.9
  N3/N02.11.7–2.5
 T classification
  T3/T1–T21.41.1–1.80.002
  T4/T1–T22.11.6–2.5
 p53
  ≥ 10%/< 10%1.30.9–1.90.09

DISCUSSION

Three important results stem from this study. First, nuclear accumulation of p53 protein is associated strongly with local treatment failure in patients with squamous cell carcinoma of the head and neck region who are treated with primary radiotherapy, suggesting that p53 overexpression may help predict radioresistance. Second, patients with tumors of the oral cavity with low proliferation (Ki-67 < 20%) have a poorer response to radiotherapy than patients with tumors that have high proliferation. Third, the combination of p53 accumulation and low proliferation (Ki-67 < 20%) is associated with poor local control in patients with early T-stage lesions. For patients in all three situations, a trend was observed for a worse prognosis, but it did not reach statistical significance, although p53 overexpression alone and the combination of p53 overexpression and low proliferation (Ki-67 < 20%) was correlated significantly with disease specific survival.

The numerous reports on p53 and Ki-67 as potential markers of radioresistance or prognosis in patients with carcinoma of the head and neck have yielded conflicting results. The majority either used limited numbers of patients or studied patients who received to different therapeutic modalities, such as radiochemotherapy or postoperative radiotherapy.16–19 Definitive conclusions could not be drawn even from cohorts of patients who were treated uniformly with primary radiation, although p53 mutation/overexpression and low Ki-67 expression seem to have negative effects on local control or survival (Table 6).

Table 6. Reports of p53 and Ki-67 as Radioresistance or Prognosis Markers in Cohorts of Patients with Head and Neck Carcinoma Treated with Primary Radiotherapy
StudyNo. of patientsTumor siteFindings
  1. a/w: associated with.

Negative p53 studies
 Wilson et al.2899Head and neckp53 overexpression not predictive of outcome
 Awwad et al.2979Head and neckp53 overexpression not predictive of prognosis
 Kokoska et al.3070Larynxp53 overexpression not predictive of local control
 Pai et al.3186Glottic (T1–T2N0M0)p53 overexpression not predictive of clinical outcome
 Hirvikoski et al.32172Larynxp53 overexpression not predictive of prognosis
Positive p53 studies
 Raybaud-Diogène et al.12101Head and neckp53 overexpression a/w lower control and worse prognosis
 Narayana et al.3367Glottic (T1N0M0)p53 overexpression a/w lower control (case–control study)
 Gallo et al.3485Head and neckp53 mutation a/w lower control and worse prognosis
 Narayana et al.35102Glottic (T1N0M0)p53 overexpression a/w lower control
 Raybaud et al.3656Oral, pharynxp53 overexpression a/w lower control
 Obata et al.3735Oropharynxp53 mutation a/w lower control and worse prognosis
 This study304Head and neckp53 overexpression a/w lower control and worse prognosis
Negative Ki-67 study
 Hirvikoski et al.32172LarynxLow proliferation not predictive of prognosis
Positive Ki-67 studies
 Raybaud-Diogène et al.12101Head and neckLow proliferation a/w lower control and worse prognosis
 Kropveld et al.3836Larynx (T2)Low proliferation a/w lower control and worse prognosis
 Raybaud et al.3656Oral, pharynxLow proliferation a/w lower control and worse prognosis
 This study77OralLow proliferation a/w lower control and worse prognosis

The cohort used in this study was both uniform in treatment and large in size. In fact, this was the largest cohort of patients with carcinoma of the head and neck who were treated uniformly with primary radiotherapy and were studied by immunohistochemistry. Surprisingly, uniformity of treatment—primary radiotherapy in this study—is a neglected issue in most studies. Studying patients with a pure radiotreated phenotype avoids the confounding effects of chemotherapy or surgery. Large cohort size confers robustness to our results, especially because it translates into a good representation of all anatomic sites, T classifications, and N classifications, a feature not always found in other studies. This is important considering that, as we also showed, clinical parameters like T classification and N classification have more predictive power than immunohistochemical factors. Thus, a large cohort size certainly helps with detecting minimal differences. Large cohort size and representation of all disease sites and T classifications also made possible the interesting analysis of subsets of patients stratified for anatomic site (Ki-67 in the oral cavity) or T classification (combination of p53 and Ki-67 for early T-classification lesions).

Our subgroup analysis (Tables 3 and 4) revealed a striking difference in response between different T classifications. Expression of p53 and the p53/Ki-67 combination were correlated well with the response to radiotherapy in patients with early-stage lesions but not in patients with locally advanced disease, especially in the T4 subgroup. If p53 and Ki-67 indeed are predictors of radioresistance, then the poor performance of patients with T4 lesions may be explained in part by the fact that a small biopsy specimen taken from a large tumor is not representative of the whole tumor and may potentially contain different clones with mutations on genes other than p53. Larger tumors also often contain mitotically active areas as well as less active areas.

In the current study, only disease specific survival reached statistical significance with p53 expression and the p53/Ki-67 combination, whereas overall survival did not. This may be explained by the high prevalence of concurrent medical problems related to alcohol or tobacco, such as heart, lung, and liver diseases, which are estimated to account for 30% of deaths among patients with carcinoma of the head and neck.20 In our cohort, 32% of deaths were not related to disease. In a retrospective study, we believe that disease specific survival, although it is questionable from a statistical standpoint, is an appropriate methodologic approach to control for the very high frequency of these noninformative deaths, which are an inherent characteristic of patient populations with head and neck carcinoma. In this respect, it is noteworthy that both overall survival and disease specific survival followed the same statistical trend and that there was a strong association between local treatment failure and poor overall survival rates (23% vs. 63%, respectively; P < 0.00001).

Although they were not as strong as traditional clinical parameters, abnormal p53 expression and low Ki-67 expression by immunohistochemistry were independent factors for radioresistance in this study. This suggests that these markers may assist in the better selection of treatment options for patients with carcinoma of the head and neck. An especially promising result is the combination of both p53 and Ki-67 expression adding predictive power to T classification for patients with early-stage tumors. This suggests that p53 and Ki-67 can identify tumors that, despite their apparently favorable T classification, will be resistant to radiotherapy. This finding has the potential to substantially modify the clinical management of patients who present with early-stage tumors who represent approximately 35% of newly diagnosed patients11 and for whom the choice between curative radiotherapy or surgery probably is the most challenging. Other very interesting possibilities for the use of these markers include decreasing salvage surgery and its major complications,21, 22 preventing unnecessary surgical mutilation that can result in poor quality of life,23, 24 and controlling radiosensitive tumors by standard or lower doses of radiotherapy rather than with more aggressive (and more toxic) modalities, such as radiochemotherapy25 or accelerated radiotherapy.26

Although it was not the main objective of this article, it should be emphasized that immunohistochemical markers offer many advantages over other methods, such as DNA sequencing and polymerase chain reaction analysis. Immunohistochemistry is simple to perform and to evaluate. It is also cheap, fast, and can be used readily on the same small biopsies that are used to diagnose head and neck tumors. It is also the only method that allows distinction between the p53 abnormalities confined to the tumor component that probably are more significant than the abnormalities located in the adjacent nontumoral mucosa.

In conclusion, this study suggests that immunohistochemical markers such as p53 and Ki-67 may play a key role in the clinical management of patients with head and neck carcinoma by helping to predict which tumors are more likely to be eradicated by radiotherapy. Despite the improvement of chemotherapy and radiotherapy protocols, the advances in conservative surgery, and the advent of gene therapy, the prognosis for patients with head and neck carcinoma has remained the same for almost 3 decades,27 and, considering the precious information that widely available immunohistochemical markers can reveal at low cost, it is our opinion that prospective studies should be undertaken in the near future to confirm our results.

Ancillary