Impact of Center Size and Experience on Outcomes in Head and Neck Cancer

The study reported by Wuthrick et al in the article that accompanies this editorial provides additional evidence that patients with advanced head and neck cancer (HNC) should be treated in highvolume HNC centers for optimal survival outcomes. This study was a retrospective subgroup analysis of the impact of treatment center expertise on the overall survival (OS) of patients with oropharyngeal cancer and known human papillomavirus and smoking status who were treated as part of the Radiation Therapy Oncology Group (RTOG) 0129 randomized trial. In this study, high patient accrual to previous RTOG trials was used as a surrogate for HNC expertise. There were a total of 471 patients: the majority (321 patients) were treated at one of 88 historically lowaccrual centers (HLACs), whereas 150 patients were treated at one of 13 historically high-accrual centers (HHACs). Looking back over a 15-year history of RTOG HNC trials, the authors defined HHACs as the top tertile of accrual centers, with an average of more than 42 patients accrued per center. Patients treated at HLACs had inferior outcomes, with 5-year locoregional failure rates of 36% compared with 21% for patients treated at HHACs, and 5-year OS rates of 51% compared with 69% for patients treated at HHACs. There was a 91% increased risk of death for patients treated at HLACs after adjusting for age, T and N classification, performance status, smoking, and human papillomavirus status. Sensitivity analysis showed the results to be consistent when the cutoff for high accrual was decreased from 42 patients (top 5% of centers) to 25 patients (top 10% of centers) or when accrual was treated as a continuous variable. The authors showed that unacceptable radiotherapy (RT) protocol noncompliance was higher in HLACs compared with HHACs (11% v 5%; P .04), but in multivariable analysis, the impact of this on OS was only approximately 20% of the total impact of accrual volume on OS. This discrepancy likely reflects the requirement for expertise across the gamut of diagnostic, therapeutic, and support services to ensure optimal patient outcomes. However, it is also likely that RT quality could have contributed more to the observed difference, given that the quality assurance (RT QA) analysis was limited to total dose, overall treatment time, and field borders. There was no review of the diagnostic imaging and accuracy of gross tumor volume delineation, and no assessment of radiation dosimetry to the gross tumor volume or planning target volumes. There were no significant differences between HLACs and HHACs with respect to acute or late grade 3 to 5 treatment toxicities, but specific data on treatment-related deaths would have been of interest. The overall message from this study is remarkably similar to that from the Trans Tasman Radiation Oncology Group (TROG) 02.02 substudy published in 2010 that investigated the impact of radiation protocol compliance and quality in a randomized trial evaluating the addition of the hypoxic cytotoxin tirapazamine to chemoradiotherapy. It was the first HNC trial to include comprehensive RT QA and was able to quantify the impact of major RT protocol violations on patient outcomes. RT was found to be protocol noncompliant in 208 of 820 patients (25%). A secondary review was conducted of 206 of the 208 noncompliant plans, with 97 of 206 (47%) predicted to have an adverse impact on outcome. Indeed, these patients, compared with patients with protocolcompliant plans, had double the 2-year locoregional failure rate (46% v 22%) and an absolute reduction in 2-year OS of 20% (50% v 70%). Importantly, the probability of receiving poor-quality RT was most highly correlated with the number of patients who were enrolled at centers. The centers with the lowest accrual numbers ( 5 patients) had a disproportionate number of RT protocol violations compared with centers with large accrual numbers ( 20 patients): 30% versus 5% (P .001). As Wuthrick et al have highlighted in their article, the effect of center experience and expertise may be much greater with intensity-modulated RT (IMRT), which was not used in the RTOG 0129 and TROG 02.02 studies but is now the standard of care. Although the conformality of IMRT provides many advantages, specifically, the reduced dose to normal tissues (salivary, mandible, and pharyngeal muscles), it can also increase the risk of a geographic miss. This may occur if there is inaccurate patient assessment and/or tumor voluming which is more likely to result in a geographic miss than with previous large-volume, treat-everything RT fields. Hence, the potential for even greater disparity in HNC outcomes in low-volume centers could be much higher in the current IMRT era. The other factor to note is that at least the RTOG HLACs were enrolling some patients in clinical trials. There is another layer of HNC work that is being performed around the world in community centers that rarely or never enroll their patients onto clinical trials and rarely publish their treatment outcomes. So these patients could quite possibly have worse outcomes than the patients in this study who were seen at HLACs. This lack of data from such HNC centers contributes to the difficulties in JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L