Commentary (Chen/Feigal): Inclusion of Comorbidity in a Staging System for Head and Neck Cancer

Publication
Article
OncologyONCOLOGY Vol 9 No 9
Volume 9
Issue 9

The tumor, node, metastases (TNM) cancer staging system is widely accepted by physicians as a predictor of prognosis and as a guide to therapy. Multiple national and international organizations, including the American Joint Committee on Cancer and the TNM Committee of the International Union Against Cancer have periodically evaluated and revised this international staging system since it was first proposed over four decades ago [1].

The tumor, node, metastases (TNM) cancer staging system is widely accepted by physicians as a predictor of prognosis and as a guide to therapy. Multiple national and international organizations, including the American Joint Committee on Cancer and the TNM Committee of the International Union Against Cancer have periodically evaluated and revised this international staging system since it was first proposed over four decades ago [1].

The statistical goal of staging cancer patients is to classify them into several stages so that within each stage the responses (survival) are homogeneous and the differences in survival between stages are large. The stages are defined through independent variables, which, in the case of TNM, include tumor histology, primary site, presence or absence of regional node involvement, and distant metastases.

The six major head and neck cancer sites are oral cavity, pharynx, larynx, maxillary sinus, salivary glands, and thyroid gland. The T classification indicates the extent of the primary tumor. The N classification indicates the extent of tumor spread to the regional lymph nodes. The M classification indicates the presence or absence of distant metastases. The staging system is based on the estimate of the extent of disease prior to initial treatment.

In this issue of ONCOLOGY, Piccirillo describes a method to add comorbidity, age, and symptom status to the TNM stage to improve prognostic precision. The sequential sequestration and conjunctive consolidation procedure (SSCC) [2] was used. There are two other similar statistical procedures: Discriminant analysis is used when class membership, for example responder/nonresponder, is determined before data analysis. Cluster analysis is used when there is no response variable, and a cluster consists of members close to each other measured by all the independent variables.

Sequential Sequestration/Conjunctive Consolidation

The sequential sequestration (SS) process is used to assigned a patient to a unique category of one independent variable, and create a gradient in average response among categories. For example, in a group of lung cancer patients, the clinical symptoms include categories of distant, quasimetastatic, mediastinal, regional, systemic, pulmonic, and asymptomatic. Since patients can have symptoms of multiple categories, they will be assigned to their own worst symptom category. The order of severity of symptoms is determined by the data at hand through a sequential procedure.

Next, the conjunctive consolidation (CC) process is used to merge several independent variables two at a time into a single staging variable. This merging process is done through a two-way (variable) tabulation with response average in each cell, examination of the significant main effects on the response, and merging the cells in this two-way table to create one variable with good gradient pattern. After this consolidation, this new variable will be examined with another independent variable (another CC) and combined into a new variable again. Finally one single staging variable will emerge. This whole process is carried out by clinical judgment and numerical inspection, without an objective mathematical criterion to guide the process.

Since the purpose and end result of the classification procedure used in this paper are the same as those of the classification and regression tree (CART) procedure (or recursive partitioning) [3], a comparison between them can be made. CART is a computer-intensive process with objective score criteria. The response variable could be binary (success or failure) or continuous (survival time). The splitting into classes is done recursively and every further split should decrease impurity score or increase purity score. These scores are defined by a mathematical formula. Since there are objective criteria to guide the process, cross-validation by using the data at hand can be done to evaluate the classification result [4]. For the objective criteria, we can use Akaike's information criteria [5] or others as a purity score. This purity score will be penalized by too many unnecessary classes.

Both methods (SSCC, CART) divide the space of independent variables into rectangles (all classes are rectangular shape; see Piccirillo's Table 4). There could be other possibilities, such as the use of Boolean algebra ("or" in addition to "and"), to define classes; this will create L-shaped classes. Also multiple-regression techniques, such as Cox's or logistic regressions, can generate different types of classes.

A Reasonable Approach

This paper presents a reasonable approach to incorporating comorbidity into the staging of cancer. The methods used can be improved upon by many new sophisticated and computer-intensive statistical techniques. The reliability of the new staging should be verified by cross-validation through the existing data and by the confirmation of other future data.

Should comorbidity be incorporated into staging? Since the major risk factors for many head and neck cancers are tobacco and alcohol, comorbid conditions, including lung, cardiovascular, and liver disease, are prevalent. When comorbidity and TNM stages are considered together, Piccirillo shows that the presence of comorbidity, even in TNM stage 1, predicts 5-year survival that is worse than TNM stage 4 without comorbidity.

Implications of the New System

There are several implications. Clinically, physicians already factor these comorbid conditions into the equation when evaluating a patient's prognosis and treatment options. Comparisons of treatment results from different centers or studies may be misleading if the prevalence of comorbidity varies. Research protocols may exclude patients with comorbidity since their survival has less to do with their malignancy than with their other medical problems.

TNM staging, however, will continue to have a central role in head and neck cancer simply because the therapy options are determined by anatomic extent of disease. A patient with early localized disease with comorbidity may have the same prognosis as a patient with metastatic disease without comorbidity, but the therapeutic options are very different. The combined TNM and comorbidity staging may not be very helpful for clinical trial stratification.

Finally, in considering how to improve staging, other factors, including biochemical, molecular, genetic and immunologic markers should be considered. In other diseases, hormone receptors, chromosomal translocations, and oncogenes are examples of factors related to tumor biology. Future progress may require having the data to be able to evaluate the role of anatomy, comorbidity, and tumor biology together.

References:

1. Denoix PF, Schwartz D: Regeles generales de classification des cancers et de presentation des resultants therapeutics. Acad Chirurg 85:415-424, 1959.

2. Feinstein AR, Wells CK: A clinical-sever-ity staging system for patients with lung cancer. Medicine 69:1-33, 1990.

3. Breiman L, Friedman JH, Olshen RA, et al: Classification and Regression Trees, pp 1-265. Belmont, CA, Wadsworth, 1984.

4. Stone M: Cross-validatory choice and assessment of statistical predictions (with discussion). J R Stat Soc B 36:111-147, 1974.

5. Akaike H: Information theory and an extension of the maximum likelihood principal, in Petrov BN, Csaki F (eds): 2nd International Symposium on Information Theory, pp 267-281. Budapest, Akademiai Kiado, 1973.

This work has been produced by the authors in their capacities as a Federal Government employee, as part of their official duties, and hence this work is in the public domain and is not subject to copyright.

Recent Videos
Treatment with toripalimab does not yield the same vascular toxicity seen with pembrolizumab in patients with advanced or metastatic nasopharyngeal carcinoma, according to Barbara Burtness, MD.
Overall survival also appears to improve with toripalimab compared with chemotherapy among patients with metastatic or advanced nasopharyngeal carcinoma.
head and neck cancer
head and neck cancer
head and neck cancer
Related Content