Individualized Treatment Selection in Patients With Head and Neck Cancer: Do Molecular Markers Meet the Challenge?

Abstract

Integrated chemotherapy and radiation therapy have become the standard of care for most patients with advanced stage laryngopharyngeal cancers. The concurrent administration of these two modalities is the approach recommended by most experts at present. This strategy achieves higher rates of locoregional control compared to when chemotherapy is given as induction treatment before, or as an adjuvant after, radiation therapy. However, the benefits from concomitant chemoradiotherapy treatment are tempered by higher rates of treatment-related sequelae, especially in the short term. This issue is of particular concern in patients that fail to respond and who have to endure the adverse effects of treatment. Based on observations that response to chemotherapy predicts radiation response, initial combined modality trials used response to induction chemotherapy to select patients for subsequent organ preservation treatment with definitive radiation, reserving laryngectomy for chemoresistant patients. However, with a shift to concomitant administration of chemotherapy and radiation therapy, the opportunity for treatment selection based on initial response was lost. To maximize the potential value of induction chemotherapy in terms of patient selection while minimizing the delay in the start of concomitant treatment, Urba et al has promulgated an approach in larynx cancer in which one cycle of neoadjuvant chemotherapy is delivered to select patients for subsequent concomitant chemoradiotherapy. Their published work suggests that this approach offers advantages for survival improvement over historical controls. The group recently expanded their observations to oropharyngeal cancers, reporting 3-year overall survival of 67% (95% CI, 53.5% to 80.6%) and diseasefree survival of 80.5% (95% CI, 68.9% to 92.1%) that was superior to historical controls. The use of induction chemotherapy primarily in select patients for subsequent therapy is not without issues. For example, the results of Radiation Therapy Oncology Group trial 91-11 demonstrated that among patients who had less than a partial response at the primary site to induction chemotherapy but refused recommended total laryngectomy, durable disease control was still possible with radiation therapy alone. Similarly, it can be difficult to assess response after just one cycle of chemotherapy, raising the possibility that patients who would have done well with radiation-based treatment alone are triaged to unnecessary surgery. Finally, newer triplet induction chemotherapy regimens combining cisplatin and fluorouracil with a taxane have proven to be more efficacious than cisplatin and fluorouracil alone. By focusing on a short course of induction therapy for patient selection purposes, the potential therapeutic benefit of a more prolonged course on distant control may be compromised. Not surprisingly, there has been a growing interest in defining molecular subgroups that predict disease behavior to select treatment. The work from Kumar and colleagues reported in this issue of the Journal aims to define pretreatment molecular markers that predict response to chemotherapy in a clinical trials data set. This well designed study has the clear benefit of identifying markers that may be clinically applicable. The investigators chose to look at established markers individually and in combination based on biologic principles, leading to the identification of prognostically significant molecular predictors. The genetic analysis of cancers reached a crescendo with the completion of the Human Genome Project along with the development of high throughput genome-wide analytic techniques. These analyses have helped to shape our understanding of human malignancies generally, and head and neck cancers specifically. Global genomic analyses have identified molecular subsets of head and neck squamous cell carcinoma (HNSCC), which may have prognostic implications. The challenge has been to reproduce prognostic gene signatures. For example, two independent groups have reported gene signatures that predict outcome in breast cancers using gene arrays, one including 70 genes and the other 76 genes. Interestingly, although both sets of genes have been validated in independent patient cohorts, there is relatively little overlap between the gene sets. Similarly, the use of individual and combined markers to predict outcome in HNSCC has shown conflicting results, as is well illustrated by the variable correlation between p53 status and outcome reported by different investigators. Divergent findings are not uncommon when assessing individual molecular markers, reflecting the genetic complexity of HNSCC. To overcome these issues, several studies have attempted to define molecular prognostic groups based on global genomic profiles. Several studies have used comparative genomic hybridization (CGH) to categorize prognostic subgroups in HNSCC. Three of these studies define JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 26 NUMBER 19 JULY 1 2008