Apples and oranges: Comparison of two surgical procedures with different indications for the management of oral squamous cell carcinoma

To the Editor, We read the audit by Stoop et al. with interest comparing disease‐ specific survival (DSS) and locoregional recurrence rate (LRR) between two approaches to mandibulectomy for mandibular bone invading oral squamous cell carcinomas (OSCC) in a retrospective cohort of 210 patients. As oncologists at a busy academic university centre in India treating nearly 1000 cases of OSCC every year by a combined multidisciplinary team approach, we have specific queries regarding the manuscript. The distribution of patients between the two groups was lopsided, with nearly three times the number of patients in the segmental resection arm. At the outset, this increases the Type II error in the study. The table describing the patient and tumour variables is inconveniently represented, with each subgroup considered as 100% (e.g., Brown Types I, Ic, II, III, and IV are separately considered). We tried to reorganise the table considering each arm of the study separately as discrete groups (marginal mandibulectomy, n = 59 and segmental mandibulectomy, n = 151), and our results are shown in Table 1. What is clear from the original and the reorganised tables is that the two arms were entirely dissimilar in terms of tumour size (p = .001), depth of infiltration (p < .001), and the Brown classification of mandibular defects (p < .001). For example, 54% of patients in the marginal resection arm had a Class I Brown defect as compared to only 23% in the segmental resection arm. With such differences in baseline characteristics (relatively “bad” disease in the segmental arm, relatively “good” disease in the marginal arm), it is not clear if the conclusion of no significant differences between DSS and LRR can be of any significance. The classification of surgical margins as either ≤1mm or >1mm is inappropriate. It is also not clear whether it is a bony margin, or soft tissue margin, or both. Taking a 1mm surgical margin is a violation of the principles of oncological resection and the proportion of patients having a margin ≤ 1mm is unacceptably high (47% in marginal resection and 36% in segmental resection). Safe margin in oral cancers should be a minimum of 1 cm clinically and 5mm pathologically. We fail to understand how a 1mm margin for oral cancers can have any clinical significance, even though there may be some statistical significance as mentioned by the authors that a surgical margin of less than 1mm entailed a poorer DSS by Cox regression analysis. The indications of postoperative radiotherapy depend on several factors, and it is not clear which patients were selected for radiotherapy. All patients with close resection margins should receive postoperative radiotherapy to reduce locoregional recurrences and improve survival, but postoperative radiotherapy cannot compensate for inadequate oncological resection. In the case of microscopically involved margins, chemoradiotherapy (CRT) is the standard of care. Only four patients in the marginal resection arm and 19 patients in the segmental resection arm received CRT; however, 28 and 54 patients in the respective arms had a margin of ≤1 mm. All patients had mandibular bone invasion (T4) and in our opinion should have been eligible for radiotherapy given Stage IV disease. Nevertheless, 13 patients in the marginal resection arm and 38 patients in the segmental resection arm did not receive any adjuvant therapy. Therefore, the multimodality treatment appears to be inadequate. The authors conveniently report 3 and 5‐year disease‐ specific survival and locoregional recurrence in the manuscript. For oral cancers, published literature usually reports DSS and LRR at 2 and 5 years. A likely reason for reporting 3‐year DSS could be that the Kaplan–Meier (K–M) plots for DSS are separated at a period of 2 years of follow‐up and then converge at 3 years. The higher DSS at 2 years in the marginal resection arm reiterates the baseline differences between the two arms. Consequently, it could be possible that the actual cause of the K–M curves converging at 3 years is due to patient deaths in the marginal resection arm between 2 and 3 years owing to either inadequate treatment in terms of lower than adequate surgical margin or the failure to add CRT as explained earlier. Reporting follow‐up time in days instead of months reduces the readability of the K–M curve. Anyway, the most relevant endpoint in oncology is overall survival (OS) which includes death due to any cause, and since more than 50% of the population had already had events at the time of analysis, OS should have been reported. DSS, although important, is only a surrogate endpoint. In conclusion, given the differences in the baseline tumour characteristics, the eventual similar LRR and DSS reflects a more unsatisfactory outcome for the relatively “better” tumours in the marginal resection group. Marginal mandibulectomy and segmental mandibulectomy have different indications, and while segmental resection can be performed where marginal resection is indicated, the converse is not valid. As such, the comparing groups in this study seem improper, and the equivalence of marginal mandibulectomy in terms of LRR and DSS can only be established by comparing the two