SEARCH

SEARCH BY CITATION

Keywords:

  • head and neck neoplasms;
  • mouth neoplasms;
  • neoplasm staging;
  • weight loss;
  • prognosis;
  • mortality;
  • recurrence;
  • squamous cell neoplasms

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES

BACKGROUND

The prognosis of patients with recurrent tumors of the head and neck generally is considered poor. Better prediction of outcomes can help physicians counsel patients about the merits of additional treatment. The TNM system, which was created for patients with primary tumors, may not provide optimal information. Anatomic staging systems traditionally have ignored symptom-based variables, such as weight loss, despite their known prognostic value. The objectives of this study were 1) to measure the prognostic impact of weight loss, 2) to evaluate the prognostic value of the TNM staging system, and 3) to create a practical staging system capable of predicting survival after patients develop recurrent tumors of the oral cavity and oropharynx.

METHODS

A retrospective chart review was used to identify an inception cohort of patients seeking treatment for recurrent, persistent, and second primary tumors of the oral cavity and oropharynx at the University of Washington. The primary outcome variable was 1-year survival.

RESULTS

The 1-year survival rate for the cohort (n = 97 patients) was 38%, with a median survival of 0.7 years. Multivariate analysis (Cox regression) identified weight loss, previous radiation to the head and neck, and TNM stage of the recurrent tumor as factors that had a substantial impact on mortality. A second multivariate technique called conjunctive consolidation was used to determine the relative quantitative impact of each variable on survival and to develop a clinical staging system. Weight loss and previous radiation had the greatest influence, and the use of just these two variables resulted in a three-tiered staging system with 1-year survival rates of 62% (16 of 26 patients), 44% (18 of 41 patients), and 10% (3 of 30 patients). In contrast, the TNM staging system produced survival rates of 60% (patients with Stage I disease), 67% (patients with Stage II disease), 32% (patients with Stage III disease), and 32% (patients with Stage IV disease).

CONCLUSIONS

The authors found substantial variation in survival after patients developed recurrent tumors of the oral cavity and oropharynx. Two readily available clinical variables—weight loss and previous radiation—were combined to create a clinically practical staging scheme with more prognostic power than the TNM staging system. Until molecular markers can reliably used be to predict outcomes, greater attention needs to be given to the utility of simple, inexpensive, and surprisingly powerful clinical variables. Cancer 2002;95:553–62. © 2002 American Cancer Society.

DOI 10.1002/cncr.10711

Continued reliance on the TNM staging system for patients with head and neck carcinoma has been based on its simplicity; the relative ease of obtaining information about tumor, lymph node, and metastatic status; and its value as a predictive variable and marker of disease severity.1, 2 However, the originators of the TNM system recognized the limitations of using only morphologic variables,3 and the characterization of these anatomic variables has been reevaluated continuously since its inception.4 The addition of nonanatomic, clinical variables, such as symptoms and comorbidity, to TNM data substantially improved the prediction of mortality in patients with primary disease.5–9

The TNM system is less useful for patients with nonprimary tumors.10–12 It was not designed to evaluate patients with persistent disease. Recent studies suggest that the TNM system has limited ability to predict mortality for patients with recurrent disease of the oral cavity and oropharynx10, 11 and the larynx.12 Furthermore, surgical treatment of primary tumors may preclude TNM staging of recurrent tumors. For example, a laryngectomy makes TNM staging of recurrences impossible, because the anatomic subsites (e.g., epiglottis, false cords, and true cords) that are needed by the TNM system have been removed.

Staging systems largely have ignored valuable prognostic variables, such as weight loss, even though the association between weight loss and mortality has been well established.13–15 Less attention has been given to the prognostic value of weight loss in patients with carcinoma of the head and neck, although there is evidence that weight loss bodes poorly in patients with primary disease.16–18 We previously showed that weight loss may have a strong effect on mortality in patients with persistent and recurrent tumors of the oral cavity and oropharynx.10 Our objectives for this study were 1) to determine whether the strong prognostic impact of weight loss observed in a previous study could be reproduced in a new cohort of patients, 2) to evaluate the prognostic value of the TNM staging system for recurrent tumors of the oral cavity and oropharynx, and 3) to create an improved and practical staging system that predicts survival after the development of recurrent tumors of the oral cavity and oropharynx.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES

Study Population

A retrospective chart review was used to identify an inception cohort of patients who sought treatment for recurrent, persistent, and second primary tumors of the oral cavity and oropharynx at the University of Washington Medical Center between January 1, 1993 and December 31, 1999. The medical records were extracted from January to June, 2001, so that the length of follow-up ranged from 1 year to 8 years (median, 4.9 years). Documentation of prior treatment for primary squamous cell carcinoma of either the oral cavity or the oropharynx was required for inclusion, but patients who received palliative care for their primary tumors were excluded. In patients with more than one recurrence, the initial recurrence was considered the index tumor.

Below, we will use the term second tumor to refer to three separate subcategories of tumors: second primary tumors, persistent tumors, and truly recurrent tumors. Usually, little distinction is made between these types of tumors, and they all commonly are called recurrent tumors; in fact, we adopted this loose terminology in the title of this article. However, to avoid ambiguity, we will adopt this more precise terminology for the remainder of this article. The specific distinctions between recurrent, persistent, and second primary tumors are detailed below.

We used the registry of the Department of Otolaryngology to identify all patients who were treated for squamous cell carcinoma of the oral cavity and oropharynx during the study period. The paper and electronic charts of all patients were reviewed to identify patients who presented with second tumors. Second tumors included both mucosal instances of tumor as well as isolated cervical and/or distant metastases. Of 270 patients who were identified in the registry, 157 patients had primary disease only. Insufficient information (e.g., missing medical records) prevented the inclusion of 11 other patients, and registry errors (e.g., tumor pathology not squamous carcinoma or site of tumor not in the oral cavity or oropharynx) eliminated 5 additional patients. The exclusion of these 173 patients resulted in a cohort of 97 patients with second oral cavity and oropharyngeal tumors.

Data Collection and Management

Data from paper and electronic medical records were extracted onto a standardized form. The general methods of extraction and classification of archival data for patients with malignant disease have been discussed previously.19–21 In addition, study variables and potential values were defined explicitly in a handbook to ensure reproducibility. A separate coding guide was used to convert each data point into a numerical value for data analyses. Discrepancies about clinical variables in the medical record were addressed with the following conventions: symptoms or findings were recorded as present if they were noted by at least one member of the medical or nursing staff. When dimensional data (e.g., the amount of weight loss, greatest tumor dimension) were discrepant, the largest number was recorded.

The first 5% of charts were extracted by both authors to provide training for the first author (T.V.N.). A second 5% of medical records were then extracted separately by both authors. Analysis of these extractions demonstrated good interextractor agreement, so the remainder of the records were extracted independently by the first author. Data were entered into a Microsoft Access data base (Microsoft Corp., Redmond, WA) with double-entry verification techniques to ensure accuracy.

Data Classification

Zero time

Zero time (baseline) was defined as the start of the treatment for the index (i.e., second) tumor. In patients who received no treatment, the date of the decision not to treat was used as zero time. If this information was not available, then the date of diagnosis was designated as zero time.

The prezero time data (pertaining to the primary tumor) included the TNM stage, histologic grade, and treatment (for details, see below). Zero time data included the TNM stage and histologic grade of the second tumor, symptoms, physical findings, risk factors, functional status, comorbidity, laboratory data, and treatment information. Postzero time data included disease and vital status.

Treatment

Treatment data included the types and dates of surgical procedures, dates and dosages of radiation therapy (including external beam radiotherapy and brachytherapy), and dates and types of chemotherapy. Treatment was deemed palliative if the record explicitly stated so or if the documented intent was limited to comfort measures or the relief of pain. In the absence of such documentation, it was presumed that treatment was curative.

Type of second tumor

The distinction between truly recurrent, persistent, and second primary tumors is defined poorly. For example, when an undetectable focus of tumor persists, the subsequent clinical reappearance of the tumor may be deemed a persistence (as it truly is) or a recurrence (as it may appear). For this study, we used the following criteria to classify second tumors: Persistent tumors showed continued evidence of the original tumor in the form of symptoms or morphology (physical examination, histopathology, and computed tomography reports). Tumors were classified as persistent, regardless of symptoms or morphology, if the second tumor was diagnosed within 3 months after the completion of therapy for the original tumor. Recurrent tumors had documented resolution of symptoms and morphologic signs of the primary tumor for at least 3 months. Second primary tumors involved a different oral site.

Symptoms and weight loss

Symptoms associated with the second tumor had to be distinguished from those associated with previous treatment or coexisting disease. Our ability to do so was limited given the retrospective nature of the data, so we attributed the symptoms to the second tumor if they were consistent with the manifestations of the disease rather than with the antineoplastic treatment or comorbidity. For example, a patient with prior radiation therapy who presented with xerostomia and oral bleeding would have xerostomia attributed to treatment and bleeding attributed to tumor.

Both localized symptoms and systemic symptoms were catalogued. Local symptoms included oral pain, otalgia, oral bleeding, dysphagia, odynophagia, hemoptysis, hoarseness, and stridor. Systemic symptoms included fatigue and anorexia. Weight loss could represent either local problems or systemic problems. The amount of weight loss sometimes was quantified at zero time by a clinician. When this was not done, we calculated weight loss as the difference between the patient's weight during the most recent tumor free period and zero time. When the amount of weight loss was discrepant, the highest recorded value was used. When no information regarding weight loss was present, we assumed that there was no weight loss.

Physical examination and TNM staging

Because determinations about T stage in the oral cavity and oropharynx are based on size criteria (and not anatomic details, as in laryngeal tumors),22 we applied TNM staging criteria for primary tumors to the second tumors in this cohort. If the descriptions of the physical examinations of different physicians were discrepant, then we relied on the examination of the more senior clinician. We also catalogued the presence of trismus, tumor fixation, cranial nerve paralysis, and skin involvement. Like the symptoms, specific physical findings were considered present when they were noted by at least one provider.

Risk factors, functional status, and comorbidity

The highest sustained exposure to tobacco and alcohol was recorded. The patient's functional status was classified as dependent, partly dependent, or independent. Functional eating status was recorded as a trichotomized variable: regular solid diet, soft/liquid diet, and tube feedings. Both the method of Charlson et al.23 and the method of Kaplan and Feinstein24 were used to quantify comorbidity in all patients. American Society of Anesthesiologists (ASA) scores were extracted when available (usually from the anesthetic record if the patient underwent surgery).

Histopathology, laboratory, and imaging results

When it was available, the histologic grade of the tumor was noted. In the absence of an official pathology report, the highest reported grade (i.e., the most anaplastic) was used. The results of complete blood counts, electrolytes, albumin, calcium, and liver function tests were recorded at zero time. When repeated results were available, the results nearest to zero time but still preceding treatment were selected. Radiology reports were reviewed, but the original films were not re-evaluated.

Follow-up and outcome

Disease status (presence or absence of tumor), vital status (alive or dead), and most recent date of follow-up were recorded.

Data Analysis

Bivariate statistical analyses, including t tests, chi-square tests, chi-square tests for linear trend, and analyses of variance, were performed with the Statistical Analysis Software system (version 8.0; SAS Institute, Inc, Cary, NC). Variables that had a significant impact on 1-year survival or were clinically significant were entered in a multivariate model. Multivariable analyses included stepwise proportional hazards regression (SAS) and conjunctive consolidation.5, 25 The latter technique was used to combine prognostically important variables into a practical staging system to predict mortality.

To identify prognostic factors and create a staging system, all patient outcomes were analyzed regardless of the subsequent treatment. This conventional strategy made use of the therapeutic nil hypothesis.26–29 The assumption allowed the development of the staging system for all patients, regardless of therapy; the effects of treatment can be explored subsequently for patients in similar prognostic stages.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES

The inception cohort of 97 patients included 54 men and 43 women with a median age of 64.6 years (range, 23–91 years). The 1-year survival rate was 38%. Although the mean duration of survival was 1.5 years, the distribution of survival was skewed, so the median survival of 0.7 years is a better summary statistic.

Bivariate Analyses

Before multivariate analyses began, individual baseline variables (at zero time) were examined for their bivariate relation to survival. Table 1 shows the effect of selected variables. Age, gender, comorbidity (data shown only for the Charlson method), and functional status had no statistically significant effect on 1-year survival rates in bivariate analyses. TNM staging had limited ability to stratify survival rates. Patients with Stage I disease fared worse than patients with Stage II disease, and there was no difference in survival rates for patients with Stage III disease and patients with Stage IV disease. The type of second tumor also was associated with differences in survival, because patients who had persistent tumors appeared to fare substantially worse compared with patients who had truly recurrent tumors or second primary malignancies.

Table 1. Bivariate Analyses of One-Year Survival for Baseline Variables (at zero time)
VariableNo.One-year survival (%)P value (chi-square test)
  • a

    P value < 0.05.

Age (yrs)   
 ≤ .5936310.252
 60–693135 
 ≥ 703050 
Gender   
 Male54350.501
 Female4342 
Comorbidity   
 None/mild (Charlson score ≤ 1)53400.727
 Moderate (Charlson score ≥ 2)1429 
 Unknown3040 
Functional status   
 Independent91380.313
 Partly dependent520 
 Completely dependent1100 
TNM stage of second tumor (recurrence)   
 I10600.096
 II967 
 III232 
 IV5632 
Type of second tumor   
 Recurrent tumor46410.027a
 Persistent tumor2719 
 Second primary2454 
Weight loss   
 None5754< 0.001a
 ≤ 10%2317 
 > 10%1712 
Anorexia or fatigue   
 No90410.031a
 Yes70 
Histologic grade   
 Well differentiated13770.004a
 Moderately well differentiated2245 
 Poorly differentiated3324 
 Missing data2931 
Prior treatment   
 No radiation36530.023a
 Radiation6130 
 No surgery16310.534
 Surgery8140 
Total9738 

Weight loss had a strong prognostic impact on 1-year survival, with the best survival rates in patients who had no weight loss, poorer survival in patients who had > 10% weight loss, and the worst survival rates in patients who had < 10% weight loss (P < 0.001; chi-square test). Several other variables had a statistically significant impact on 1-year survival rates, but their utility in developing a prognostic model was hindered by rare occurrences (e.g., systemic symptoms like anorexia and fatigue) or missing data (e.g., the histologic grade of the second tumor often was not available when tumors were treated nonsurgically). No other symptoms had a statistically significant impact on survival rates (data not shown). Similarly, no significant correlations were identified for disease risk factors, physical findings, or laboratory data.

The type of prior treatment used for the previous primary tumor also proved to be important. Although the prior use of surgery did not have an impact on survival after the second tumor, patients who received radiation therapy (with and without surgery) for the primary oral tumor had higher mortality rates compared with patients who did not receive prior radiation.

Multivariate Analyses

To identify variables that had the greatest independent impact on survival, we used two separate multivariate approaches. In each model, we included variables that had a statistically significant impact on survival in bivariate analyses (type of second tumor, weight loss, anorexia/fatigue, prior radiation therapy, and histology) as well as variables of clinical importance, even if they were not statistically significant (TNM stage, age).

Cox regression analysis

We began with a proportional hazards (Cox) multivariate regression. We coded weight loss in several formats: as a dichotomous variable (no weight loss vs. any weight loss), as the trichotomous variable illustrated in Table 1, and as a continuous variable (percent weight loss). In each patient, weight loss remained in the model. When weight loss was coded as a dichotomous or trichotomous variable, it was the most important predictive variable. We chose to use a dichotomous weight loss variable in the conjunctive consolidation (see below); thus, the results for our Cox regression are shown using the dichotomous variable as well (Table 2). Other variables with independent impact on survival included previous radiation to the head and neck and TNM stage of the second tumor. The impact of the type of second tumor on survival lost statistical significance in the Cox regression analysis. Because histologic grade was missing from 29 of 97 patients, we left it out of the final model.

Table 2. Results of Cox Multivariable Analysis for One-Year Survival
VariableParameter estimateStandard errorWald chi-square testaP valueHazard ratio
  • a

    Overall Wald chi-square test, 29.2 (P < 0.0001).

Weight loss (dichotomous)1.040.2419.6< 0.00012.84
Radiation for primary tumor0.500.225.20.02241.66
TNM stage of second tumor0.290.117.70.00561.34
Conjunctive consolidation

Although the Cox regression analysis establishes the statistical contribution of each variable, it is difficult to appreciate each variable's relative quantitative contribution (clinical importance) to the model. Furthermore, clinically practical staging systems are hard to create with standard multivariate approaches. An alternative multivariate technique called conjunctive consolidation5, 25 addresses both weaknesses of the Cox technique. The effects on survival of two variables at a time are examined simultaneously with a cross-tabulation (conjunction). If each variable has a statistically distinct effect on prognosis, then the cells of the table will show a double gradient, which occurs when the row variable produces a gradient in the cells of each column, and the column variable produces a gradient in the cells of each row.5 When this occurs, a new composite variable is created (consolidation). The process is repeated until no new variables contribute to the model. The challenge of using conjunctive consolidation is that it relies on clinical judgment, because no automated data processing has been possible.

In the current study, we used conjunctive consolidation to examine individual baseline variables (at zero time) with a clinically significant, bivariate correlation with survival, which included, but was not limited to, the variables that were entered into the Cox regression model. Our starting point was weight loss, which had the strongest bivariate impact on survival. We dichotomized weight loss to increase the number of patients in each cell. We dropped histology as a variable, because it frequently was missing. Several variables were dropped because of poor univariate distributions. For example, although the conjunctive consolidation of weight loss and the systemic symptoms of fatigue and anorexia produced a double gradient and created a composite variable with a strong impact on survival, the presence of fatigue and anorexia occurred too infrequently to make the resulting prognostic model useful in many patients.

We found that the cross-tabulation of weight loss and previous radiation therapy to treat primary tumors also produced a double gradient, which indicated that both weight loss and previous radiation had a statistically independent impact on survival (Table 3). This confirmed our findings with the Cox regression analysis. The shading indicates how the cells were consolidated into clinical Stages A, B, and C (shown in Table 4). Excellent stratification was created with just two variables.

Table 3. 
Thumbnail image of
Table 4. Clinical Stage: Composite of Weight Loss and Prior Radiation
Clinical stageDescriptionOne-year survival (%)a
  • a

    The numerator in each cell represents the number of patients alive after 1 year. The denominator represents the number of patients in each category at zero time. The 1-year survival rate is shown in parentheses.

ANeither weight loss nor prior radiation16/26 (62)
BEither weight loss OR prior radiation18/41 (44)
CBoth weight loss AND prior radiation3/30 (10)

We then cross-tabulated TNM stage with clinical stage to determine whether TNM staging provided any additional prognostic value (Table 5). We noted that the double gradient was less evident, especially when moving along each row (controlling for weight loss and prior radiation). This suggests that TNM has less statistical and quantitative impact on 1-year survival rates. A combined composite variable could have been created (by using the shading in Table 5), but it would have led to relatively little improvement over clinical stage alone and a much more complex staging system.

Table 5. 
Thumbnail image of
Table 6. Staging System incorporating Clinical and TNM Stage
StageDescriptionOne-year survival (%)a
  • a

    The numerator in each cell represents the number of patients alive after 1 year. The denominator represents the number of patients in each category at zero time. The 1-year survival rate is enclosed in parentheses.

αAny TNM stage WITHOUT weight loss or radiation treatment16/26 (62)
βAny TNM stage WITH either weight loss or radiation treatment OR TNM Stage I or II AND weight loss and radiation treatment21/45 (47)
γTNM Stage III or IV AND weight loss and radiation treatment0/26 (0)

To compare the relative performance of both staging systems, we performed Kaplan–Meier analyses of TNM stage alone (Fig. 1) (P > 0.05; log-rank test) and clinical stage alone (Fig. 2) (P < 0.0001; log-rank test). The TNM staging system did not stratify well; in fact, Stages I and II had reversed effects. Conversely, clinical stage not only showed excellent stratification, but its impact also extended beyond 1-year survival.

thumbnail image

Figure 1. Kaplan–Meier plot of survival by TNM stage (I–IV) of second (recurrent) tumors in patients with carcinoma of the oral cavity and oropharynx. Note the reversal of survival curves for patients with Stage I and II disease (P = 0.6; log-rank test).

Download figure to PowerPoint

thumbnail image

Figure 2. Kaplan–Meier plot of survival by clinical stage (A–C) of second (recurrent) tumors in patients with carcinoma of the oral cavity and oropharynx (P = 0.0001; log-rank test).

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES

We documented substantial variation in survival after second or recurrent tumors of the oral cavity and oropharynx. This finding is in contrast with the common perception that patients with recurrent tumors of the head and neck, especially large tumors, have uniformly poor prognoses. We found that weight loss and previous radiation therapy to the head and neck were predictive of mortality, and that a simple combination of these two variables produced a clinical staging scheme with more prognostic power than TNM stage.

Although patients with early (recurrent TNM Stages I and II) and late (recurrent TNM Stages III and IV) disease had very different survival rates, there were nearly indistinguishable differences between patients with recurrent Stage I and Stage II disease (1-year survival rate, 67–60%) and between patients with recurrent Stage III and IV disease (both with 1-year survival rates of 32%). We previously noted relatively poor stratification between Stage I and II second tumors and, as a result, adopted a modified TNM staging system.10 Lacy et al.11 also did not use TNM staging of the second tumor in their study, choosing to quantify tumor burden with an extent of recurrence variable (local, regional, or distant). We do not discount the importance of tumor morphology as a prognostic variable. Rather, we argue that, because TNM staging was developed for primary tumors (which are biologically different from recurrent and persistent tumors), a new approach may be needed for second tumors.

Prior Studies on Weight Loss

Our findings support the well-established link between weight loss and cancer mortality in general. In a series of seminal articles in the 1960s, Feinstein and collegues demonstrated the prognostic impact of weight loss on mortality in malignancies of the lung and rectum.13–15, 30, 31 More recently, DeWys et al.32 documented that weight loss and performance status were associated strongly with median survival in over 3000 patients enrolled in 12 Eastern Cooperative Oncology Group protocols, including patients with lymphoma, sarcoma, and carcinoma of the lung, gastrointestinal tract, breast, and prostate. Our finding that anorexia has prognostic value replicates the results of Reuben et al., who demonstrated a link between both weight loss and anorexia with mortality in patients with terminal disease.33 Other authors have found similar associations between weight loss and mortality in patients with carcinomas of the esophagus34 and lung,5, 35 tumor sites that share common histologies and epidemiologic risk factors with carcinomas of the head and neck.

Weight Loss in Patients with Carcinoma of the Head and Neck

The association between weight loss and mortality was reported in patients with head and neck carcinoma by Brookes.16 Of 114 patients with untreated primary squamous tumors, the patients without weight loss had a 2-year survival rate of 57.5%, whereas only 7.5% of the patients with weight loss survived for 2 years. Those findings were substantiated recently by van Bokhorst-de van der Schueren et al.,18 who found that men with a preoperative weight loss > 5% had worse survival. Regueiro et al. showed that weight loss, in conjunction with tumor size and lymph node status, significantly influenced disease free survival in patients with primary oropharyngeal squamous cell carcinoma.17 More specifically, we previously demonstrated that weight loss in patients with recurrent, persistent, and second primary tumors of the oral cavity and oropharynx was highly predictive for mortality.10

Effect of Previous Radiation Therapy

Our finding that previous radiation therapy boded poorly is supported by the existing literature on second tumors of the oral cavity and oropharynx. However, this finding has not been highlighted previously. In our previous report,10 patients with and without previous radiation therapy had 1-year survival rates of < 40% and nearly 60%, respectively. However, in our multivariate analyses, we only quantified prior treatment in terms of the number of modalities of treatment received (surgery, external beam radiation, chemotherapy, and brachytherapy); therefore, the effects of prior radiation alone were not tested. Lacy et al.11 also found a substantial effect: Patients with and without previous radiation therapy had 2-year survival rates of 15% and 37%, respectively. These intriguing results also were not retained in their multivariate analysis, but it is unclear how this variable was represented in their model. We speculate on four possible explanations why previous radiation may increase mortality after a second tumor. First, patients with more virulent tumors may have been more likely to have radiation in the first place; second, the use of previous radiation therapy reduces the arsenal of treatment available for the second, or recurrent, tumor; third, tumors resistant to radiation are more likely to recur and be lethal; and, fourth, radiation may induce genetic or molecular changes that make the tumor more virulent.

Other Variables

Lacy et al.11 found that TNM stage of the primary tumor predicted survival after treatment for the second tumor. Although we also found that this was true (data not shown), we omitted this variable from our practical staging system, because details about previous treatment often are difficult to obtain. Patients can report weight loss and previous radiation treatment, but they are unlikely to know about prior TNM staging data. Comorbidity has a significant impact on survival in patients with primary tumors of the oral cavity9 and oropharynx,8 but it is unlikely to have the same impact in recurrent and persistent tumors.10–12 Second tumors are much more likely to cause mortality and, thus, outweigh the impact of most common comorbid illnesses.

Increasing attention has been focused on prognostication based on genetic or molecular markers,36–47 because many promising associations with particular chromosomal deletions, loss of heterozygosity, genetic mutations, and the presence of oncoproteins have been uncovered. However, these tests are still under intense investigation and will require validation and then standardization before they can be considered clinically useful. Even then, these tests are likely to be expensive. Conversely, clinical variables are abundant, straightforward to collect, and inexpensive (free). Feinstein and Wells argued convincingly for the value of clinical data and showed that soft clinical data are reproducible and reliable.5, 48 Given the significant prognostic impact and convenience of clinical data, clinicians should be encouraged to consider their utility in formal staging systems.33

Limitations

We would have liked to explore the intriguing relation between tumor differentiation and mortality, because it may have more prognostic impact in second tumors than in primary tumors. Unfortunately, complete histologic data were not available when treatment was not surgical, and outside biopsies were not available in the archival data. The relatively small size of the inception cohort may have limited statistical power to detect other prognostic variables. In addition, our analysis of second tumor TNM stages was limited because, only 20% of the tumors represented early disease (Stage I or II). The retrospective nature of data collection also prevented accurate measure of actual weight loss, because we were reliant on archival data. In addition, although the majority of the patients in this study developed their second tumor more than 6 months after the completion of radiation therapy, we could not determine conclusively that the cases of documented weight loss were related directly to the second tumor rather than to prior treatment. Therefore, our findings provide support for the importance of weight loss in general, rather than weight loss exclusively associated with the second tumor. Prospective evaluation of the importance of weight loss and previous radiation treatment is now needed to determine the role of these variables in predicting outcomes after patients develop second tumors of the oral cavity and oropharynx.

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES

In the absence of a staging system designed specifically for patients with recurrent head and neck tumors, the TNM staging system, which was designed originally for primary tumors, has served as the default model. However, the behavior of recurrent, persistent, and second primary tumors may be markedly different from that of primary tumors. Factors that are important in primary disease, such as comorbidity, may have significantly less influence in second tumors. In addition, TNM staging may not be possible if the surgical field has been altered substantially. The need to develop alternative methods for staging second tumors is clear. In this study, two readily available variables—weight loss and previous radiation—were combined in a clinically practical staging scheme that predicts survival better than TNM stage. Until molecular markers can be used reliably to predict outcomes, paying attention to the simple, inexpensive, and yet surprisingly powerful clinical variables will help physicians to counsel patients about the merits of additional treatment.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES

The authors thank John Olson for his invaluable assistance with data entry and article preparation.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES