Uncontrolled local-regional disease is a serious problem in patients with squamous cell carcinoma of the head and neck (1). Tumor stage and site are currently the major clinical determinants of treatment selection. However, identification of better predictive biologic markers would improve the making of decisions regarding treatment and might lead to intervention strategies that target patients most likely to benefit.
Therefore, we analyzed the data from a series of patients with cancer of the upper aerodigestive tract; these patients were enrolled in an ongoing, multicenter, phase III placebo-controlled trial of 13-cis-retinoic acid for the prevention of second primary tumors. Our purpose was to evaluate the associations among treatment failure (disease recurrence), baseline sociodemographic characteristics, and an in vitro marker of susceptibility, i.e., bleomycin-induced chromosomal sensitivity, which we previously showed is an independent risk factor for cancers of the upper aerodigestive tract (2).
The eligibility criteria for the chemoprevention trial included a confirmed diagnosis of squamous cell carcinoma of the upper aerodigestive tract (oral cavity, pharynx, or larynx), within the previous 3 years, stage I or II disease (according to American Joint Committee Staging criteria), and definitive prior local therapy. Eligible patients were enrolled at The University of Texas M. D. Anderson Cancer Center, Houston, and by the Radiation Therapy Oncology Group and Clinical Community Oncology Program during the period from January 6, 1992, through October 19, 1995. This research was approved by all relevant review boards and in accord with an assurance filed with and approved by the U.S. Department of Health and Human Services. Patient consent forms were obtained.
Questionnaires that comprehensively determined cigarette and alcohol consumption at baseline and before diagnosis were administered to the patients at trial entry by a research nurse. Ten milliliters of blood for the bleomycin mutagen sensitivity assay, described previously (3), was drawn from the patients before randomization (except for 98 patients whose first blood samples were not available for the assay because of logistic reasons, such as shipping delay, and had blood redrawn within either 1 day of randomization (n = 76) or 1 week of randomization (n = 22). For each sample, the frequency of spontaneous and induced chromosome breaks in 50 metaphases was counted, and the results were averaged to determine the number of induced chromosome breaks per cell. Lee et al. (4) have demonstrated that scoring a minimum of 50 metaphases is statistically valid.
The end points considered for this evaluation were local–regional or distant disease recurrence (histologically confirmed) and recurrence-free survival in months to recurrence. Time-torecurrence measurements began with the date of randomization, but time from completion of definitive primary treatment to randomization was considered as a covariate. Kaplan–Meier cumulative incidence curves associated with the mutagen-sensitive phenotype were used to estimate time to recurrence (5). Proportional hazards models were constructed with multiple covariates (including mutagen sensitivity, sex, age, tumor site, previous treatment, duration in months from initial primary treatment to randomization, disease stage, and smoking and alcohol status) and by use of a backward elimination procedure. All P values are two-sided.
The study included 397 men and 95 women with a median follow-up of 33.6 months. Thirty-five patients (26 male and 9 female) had documented local– regional recurrences, and an additional three developed distant metastases. The treatment failure rate in patients with an initial pharyngeal cancer (22.2%) (P = .07) or oral cavity cancer (11.3%) (P = .03) was higher than that in patients with laryngeal cancer (5.7%) (control) (Table 1). A recurrence developed in 7.1% of patients with stage I disease and in 9.4% of patients with stage II disease. Current smokers (and recent quitters) were more likely (11.5%) than longterm quitters (5.0%) and never smokers (6.6%) to develop recurrences. There was a similar trend with respect to alcohol consumption. Overall, 11.5% of mutagen- sensitive patients (⩾1.0 mean breaks per cell) but only 5.3% of nonsensitive patients (<1.0 break per cell) developed recurrences (P = .02). Current smokers and current alcohol users (irrespective of mutagen-sensitive phenotype) had the highest risk of recurrence.
As reported recently in the same group of patients (6), there were no statistically significant differences in the distribution of mutagen sensitivity with respect to sex, age, ethnicity, or tumor site (data not shown). However, the values for mean breaks per cell (± standard deviation) were 1.06 (±0.41) for the 38 patients with disease recurrence and 0.89 (±0.40) for 454 patients who remained recurrence free (P = .01, t test).
The 5% cumulative rate of disease recurrence occurred at 5 months among mutagen-sensitive patients and at 20 months among the nonsensitive group. Shorter time to recurrence for mutagensensitive patients was consistently found, regardless of tumor site, disease stage, previous treatment, or sociodemographic variables. Mutagen-sensitive patients who were current alcohol users and long-term quitters of smoking exhibited the shortest time to recurrence.
At 3 years, the overall recurrencefree survival rates were 0.95 for nonsensitive patients and 0.88 for sensitive patients (P = .02). These differences were statistically significant in the subgroups of men (P = .02), younger patients (P = .08), patients with oral cavity cancer (P = .02), patients with stage II disease (P = .02), long-term former smokers (P = .01), and current alcohol users (P = .02) (Table 2).
The major prognostic factors for disease recurrence were evaluated in a proportional hazards model. The final model included mutagen sensitivity (odds ratio [OR] = 2.56; 95% confidence interval [CI] = 1.34–4.91), oral cavity site (OR = 2.12; 95% CI = 1.10–4.09), and duration in months from initial primary treatment to randomization (OR = 0.89; 95% CI = 0.84–0.95).
Smoking has been implicated as a determinant of prognosis (7–9). Browman et al. (8) demonstrated a significantly lower complete response rate in patients who continued to smoke during radiation therapy than in former smoking patients and never smokers (45% versus 74%). In our study, current smokers were more likely than never smokers to develop disease recurrences and to have shorter disease-free survival.
Our data suggest that in vitro sensitivity to bleomycin, a radiomimetic agent, may predict risk of disease recurrence. Data are also accumulating on molecular techniques for predicting tumor recurrence. Brennan et al. (10) have shown that molecular analysis of tumor margins shown to be histopathologically negative for tumor cells can identify clonal populations of tumor cells that harbor p53 mutations and are associated with a high risk of local recurrence. Shin et al. (11) and Koch et al. (12) reported that patients with p53-positive tumors had significantly lower overall survival and shorter time to tumor recurrence than patients with p53-negative tumors. These findings complement our data on mutagen sensitivity. Gallo et al. (13) reported a statistically significant correlation between p53 protein expression and bleomycin-induced mutagen sensitivity in 25 patients with multiple head and neck cancers (P<.01).
The underlying mechanism of mutagen sensitivity associated with cancer proclivity is still being explored but may reflect, in part, an altered repair process. Pandita and Hittelman (14) suggested that the mutagen-sensitivity phenotype may involve an inherent chromatin alteration that increases the efficiency of translating DNA damage into chromosome damage after mutagen exposure. Using a host–cell reactivation assay in parallel with the mutagen-sensitivity assay, Wei et al. (15) examined DNA repair capacity in 16 established lymphoblastoid cell lines (including three from head and neck cancers). These investigators noted that reduced cellular DNA repair capacity correlated with increased frequency of mutagen-induced chromatid breaks, suggesting that repair fidelity may be impaired in hypersensitive persons.
In summary, our data suggest that the mutagen-sensitive phenotype may predict a higher likelihood of treatment failure in early stage cancer of the upper aerodigestive tract. This assay may prove useful in predicting not only cancer susceptibility but also long-term response to therapy.