Salvage Surgery for Patients With Residual Diseases After Improper or Insu cient Treatment of Oral Squamous Cell Carcinoma: Can We Rectify These Mistakes?

Yue He Shanghai 9th Peoples Hospital A liated to Shanghai Jiaotong University School of Medicine Zhonglong Liu Shanghai 9th Peoples Hospital A liated to Shanghai Jiaotong University School of Medicine Surui Sheng Shanghai 9th Peoples Hospital A liated to Shanghai Jiaotong University School of Medicine Weijin Gao The First A liated Hospital of Wenzhou Medical University Xiao Tang Shanghai 9th Peoples Hospital A liated to Shanghai Jiaotong University School of Medicine Xiaoguang Li Shanghai 9th Peoples Hospital A liated to Shanghai Jiaotong University School of Medicine Ma Chunyue (  maxifama@163.com ) Shanghai 9th Peoples Hospital A liated to Shanghai Jiaotong University School of Medicine https://orcid.org/0000-0003-3165-646X

found in the outpatient clinical database. Speci c causes were summarized after reviewing the information.
Based on the varied admission status to our institution, all the included patients were rstly classi ed into three categories: patients with microscopic residual diseases (MiRDs); with macroscopic residual diseases (MaRDs) or with rapid recurrent diseases (RRDs). MiRDs were de ned as those with reports of initial positive margins, and suspected residual diseases via physical examination (palpable) and imaging (invisible or indeterminant), while MaRDs were those with both gross (palpable and visible) residual diseases and initial positive margins, while RRDs were de ned as those with initial negative margins but visible residual diseases right after prior operations (< 3 months). Apart from that, as far as the involved subsites were concerned, the studied cohort was further divided into three groups: with local residual, regional (cervical) residual, and both locoregional residual diseases. In consideration of the distinct residual tumor sizes, the extent of SS was tentatively graded according to the re-resections and reconstructions: 1 for simple re-resections, or simple re-neck dissections followed by direct wound closure or local aps coverage; 2 for radical re-resections with reconstructions with pedicled pectoralis major myocutaneous aps (PMMF) or free aps; 3 for SS involving craniofacial resection (skull base), carotid artery resection, total glossectomy, total maxillectomy with orbital exenteration, or hemi-mandibulectomy, with free-ap reconstructive procedures.
In order to analyze the e cacy of our immediate salvage treatment, these patients were also classi ed into two groups based on different presurgical treatment: patients received upfront SS (SS group); and those underwent adjuvant treatment regimens rst, followed by SS (AT-SS group). The SS complications were recorded as well. In addition, various adjuvant treatment modalities were applied to these patients.
For sake of statistical analysis, the postoperative adjuvant treatment (after SS) was further summarized into ve categories: none, radiotherapy, radio-chemotherapies, radio-chemotherapy with targeted therapy, and radio-chemotherapy without targeted therapies. None of the patients received simple postoperative chemotherapy with/without targeted therapies. The targeted drugs included epithelial growth factor receptor inhibitor (Nimotuzumab or Cetuximab), and vascular endothelial growth factor receptor inhibitor (Apatinib). However, due to the unavailability of PD-1 related pembrolizumab at that time, immunotherapies were not applied in any of these patients.
Overall survival (OS) time was calculated as the time from the start of SS to death/last outpatient visit in months. Salvage outcomes were recorded and compared between these patients with residual OSCCs. The Chi-square test and Fisher exact test were used to compare categorical variables. Univariate log-tank test was adopted to analyze survival time-dependent variables. Subsequently, Cox regression analysis was carried out on the variables that achieved univariate statistical signi cance. All statistical analyses were conducted via SPSS 21 for Windows (IBM, Chicago, Illinois).

Results:
Demographic information During the 5-year interval (from 2013 to 2017), a total number of 1761 patients with recurrent malignancies had received SS in our institution, according to the chart database. Within these patients, 103 (5.84%) met our inclusion criteria of patients with "residual OSCCs". Of these, 68 (66%) were men and the rest were women (n = 35, 34%); The average age reached 56.3 years, of whom 42.7% were smokers. Most patients (n = 36, 35.0%) were with initial diagnoses of tongue cancers, followed by buccal (20.4%), lower gingival (20.4%) and oor-of-mouth (20.4%). The mean follow-up reached 31.1 months (range, 4-65 months). The detailed demographics were summarized in Table 1.

Initial treatment and possible causes for failures
With regards to the primary disease status, most patients (n = 74, 71.8%) in our series were diagnosed with primary T3-T4 stages. Within these, DOI > 10 mm was con rmed in 45 (43.7%) patients after pathological review, resulting in the increased 21 (20.4%) cases of T3 stages. The patterns for primary N stages were different, with merely 20 (19.4%) patients with higher nodal metastasis grades (N2-3). The pathological grades of the primary OSCCs were as follows: highly differentiated grade in 16 (15.5%) cases, and poorly differentiated in 22 (21.4%). Neoadjuvant chemotherapy was applied in 14 (13.6%) patients, while most received upfront primary surgery. As for the extent of the initial surgery, a total number of 59 (57.3%) received ipsilateral or bilateral neck dissections. When comparing the surgical procedures, direct wound closure and local aps were mostly frequently used for defect coverages (72.8%). The important postoperative reports for surgical margins revealed that 77.7% (n = 80) patients were positive, within whom an astounding number of 53 (51.5%) were without any reports of frozen intraoperative margins according to the referral information. Judging from the initial reports, the primary resections were quite arbitrary, with 26 cases with two or more positive margins. Unfortunately, of these residual OSCC patients, a large proportion (n = 37, 35.9%) were with insu cient deep margins, implying illconsidered surgical decisions or incomplete resections. Within the whole group, 23 (22.3%) patients were with reports of "negative" margins while they still suffered from the persistent symptoms (mostly pain) after prior surgical treatment. A closer inspection of these initial "negative" margins revealed that 14 (13.6%) were suspected for close margins (< 5 mm), while 9 (8.7%) were found with intraoperative re-resections due to rstly positive-margin reports. All these "negative" margins were later doubted due to the con rmed biopsies prior to SS. The initial radiotherapy was administered to 5 patients, yet all ceased and resort to retreatment in our institution. Thus, most patients in this study were radiation-naïve. (Table 1) Referral status and initial failure analysis Within these patients with unsuccessful prior treatment, most (n = 78, 75.7%) were referred by other institutions in the rst place, while some were initiated by patients themselves. After communication with all the transfer-applying doctors, most of the referred patients (74, 71.8%) received initial surgical treatment in institutions of low-volume OSCC cases loads (< 50 cases per year), while others were from institutions of high-volume OSCC cases loads. Nineteen (18.4%) patients were with elderly ages over 70, while comorbidities were found in 41.7% (n = 43) of the entire cohort. Interesting, we found insu cient surgical margin information (≤ 3 margins taken/patient) in almost half (45.6%) of the referral reports, indicative of inadequate margin analyses. Besides, the clinical diagnoses were mistaken for other pathologies in 17 (16.5%) patients, while even 23.3% patient received initial operations without con rmations by preoperative biopsies. Within those who received preoperative biopsies, improper delays (> 2 months) between biopsy and surgery were found in 16 patients (15.5%), due either to patients' or iatrogenic reasons. Interesting, the counselling records with patients revealed that a striking number (n = 41, 39.8%) of patients in this study were reluctant or terri ed to receive radical tumor resections with freeap reconstructions in the initial treatment settings. Besides, from a surgical standpoint, treatment design loopholes were plentiful in these cases, with insu cient margins of depth (35.9%), mismatch between lesion sizes and resection/reconstruction methods (31.1%). Unstandardized operative practices were also suspected for residual lymph nodes found in the cervical basin (n = 24, 23.3%) after initial neck dissections. On the other hand, non-en-bloc (non-continuity) resections were applied in T4 cases (n = 11, 10.7%) involving tongue, oor of mouth, low gingiva, resulting in possible residual lesions in the middle zones. (Table 2)

Clinicopathological data for residual OSCCs
According to the previously mentioned classi cation, most patients in our study were in the MaRD group.
In terms of size, most of the residual OSCCs (n = 69, 67.0%) were not larger than 4 cm, suggestive of the curable local conditions. Within these, 45.7% were with gross (> 2 cm) residual lesions. Upper and lower vital-structure (skull base, orbit, carotid artery) involvement were found in approximately 13% of the cases, for whom SS was even more challenging. After SS, pathological differentiation upgrades were found in 23 (22.3%) patients with residual OSCCs. Besides, neurovascular invasions were con rmed in 19.4% of the cases. As for the salvage treatment, eighty-ve (82.5%) cases received upfront SS, while others (n = 18, 17.5%) were in the AT-SS group. The extent of SS was more extensive than the initial treatment, with 78.5% (n = 84) receiving radical resections and ap reconstructions. Within these patients, complicated wide-excision surgery, such as craniofacial skull base surgery, total glossectomy, total maxillectomy or hemi-glossectomy were not rare (25.2%). Despite our SS, post-salvage margin reports still revealed positive margins in 4 (3.9%) cases, of whom most were with larger (> 4 cm) residual OSCCs, or lesions, extending near or through vital structures. In addition, 37 patients experienced complications in postoperative settings. Most were minor wound infections or lung infections. In addition to SS, adjuvant therapies based on radiotherapy or radio-chemotherapy, were offered to most cases (n = 84, 81.6%), while targeted therapies were to 17 (16.5%) ( Table 3). The correlation between the parameters regarding initial treatment and residual retreatment were analyzed. Both primary T stage (p < 0.001) and neoadjuvant therapy (p = 0.024) were found to be related to residual lesion size. Primary T stage was also correlated with the residual subgroup (p < 0.001), and vital-structure involvement (0 = 0.017).

Survival outcomes and statistical analyses
The OS rate reached 60.2%, with 41 death within the whole group. When it comes to the speci c death causes, locoregional re-recurrences were found in 26 (25.2%) cases, while both recurrences and distant metastases in 11 (10.7%), representing the two major reasons for our salvage failures. The univariate logrank analyses of the initial treatment data revealed that sex (p = 0.046), primary T stage (p = 0.011) and neoadjuvant therapies (0.006) were related to the patients' prognosis. As for the residual OSCC data, the univariate analyses showed signi cances in residual subgroups (p = 0.005), size of the residual OSCC (p < 0.001), vital-structure involvement (p = 0.049), bone invasion (p = 0.023), salvage margin status (P < 0.001), salvage treatment combinations (0.033) and salvage resection and reconstruction extent (0.001).
Among all the variables, both primary T stage (p = 0.003), and residual lesion size (p < 0.001) were signi cantly associated with OS, based on the nal Cox multivariate analysis (Table 5, Fig. 1).

Discussion:
It is well known that the best opportunity to cure patients with OSCC is through the delivery of fast and appropriate therapy at rst presentations [8][9] . Theoretically speaking, management of "recurrence" after prior treatment is a challenging clinical situation, with decreased chances of cure by retreatment 10 .
Although there is no standard criteria or consensus of a "true recurrent" OSCC, most still consider "recurrences" as those with similar pathological pro ling, involving nearby anatomic structures (< 3 cm) and within 3 years of follow-up 11 . In literature, such "recurrences" were only divided by years, as either rapid or late recurrences, irrespective of detailed primary treatment [10][11][12] . As far as we are concerned, initial treatment, primary surgical margin and postsurgical symptom (pain) should all be taken into consideration when differentiating true "recurrent" and "residual" OSCCs, as some "recurrences" were in fact residual lesions (without intermittent remission of symptoms) 13 . These OSCCs become residual due more to improper initial treatment or insu cient resections, rather than to oncological aggressiveness of OSCCs. Determining the optimal retreatment regimens for this special group is very important, as most patients are extremely anxious about the likelihood of rapid and curative salvage re-resections 14 .
According to the referral/admission report, the report of positive margins, along with the unrelieved painful symptoms, always encroached on the retreatment con dence in the primary treatment centers, given the fact that a high proportion (24.3%) of referrals were requested by patients. As occasionally encountered with these referrals, we tried to answer the question of whether these patients with residual OSCCs could still be rescued with SS-based treatment, as controversy for such decisions still exists.
Such residual OSCC problems were caused by a lot of factors, which however has long been underevaluated. To a large extent, initial (primary) treatment status will negatively in uence the survival outcomes 16 . Firstly, the factors of surgeons should not be downplayed. According to the referral reports and patients' statements, the initial surgical treatment was carried out in some patients with unproven preoperative biopsies, which violated the principles of National Comprehensive Cancer Network (NCCN) guidelines 17 . Such condition was mostly due to false biopsy practice or lack of experiences for OSCC diagnosis. Besides, sometimes the variety of clinical presentations of OSCC and other premalignant oral lesions will also confuse the clinical diagnosis 18 . From a baseline diagnostic perspective, single or multiple incisional biopsies are required for large and non-homogenous lesions to con rm the OSCC diagnosis preoperatively 18 . The other mistake was the surgical completeness. Mismatch between primary OSCC stages and resection/reconstructive methods were abundant in our series, as some locally advanced lesions (n = 32, 31.3%) were even resected and reconstructed with direct closure or local aps.
Thus, the radicality of initial treatment was seriously questioned in these cases. In addition, a fairly large number of the cases in our study were with initial positive deep margins, implying possible awed intraoperative resection regarding the tumor depth, which will nally compromise the treatment e cacies 20,21 . Due to the terrible margin status in most patients, we advocate that en-bloc, or even compartment surgeries should be strongly recommended to ensure margin safety, particularly for adequate deep margins in advanced primary cases 23,24 . Interestingly, even in some cases with primary early-stage OSCCs, residual lesions were still found in the tumor basins. We gured that such iatrogenic mistakes, which could have been avoided, were mostly due to unprepared preoperative surgical plans. For example, for cases with tongue cancers, the para-glossal resections should not be overly conversed for lingering fear of oro-cervical communications. The removal of sublingual gland and oor of mouth mucosa should also be advocated for a clear middle-zone eradication [22][23][24] . For cases with buccal cancers, especially those in the anteromedial buccal subsites, thorough-and-thorough resections should be attempted despite possible cosmetic dis gurement. For retromolar and lower buccal lesions, the resections of medial, sometimes lateral pterygoid muscles, marginal medial mandibulectomy should always be highlighted in those with clinically presentations of seemingly "early-stage" diseases, with true invasive fronts regarding the tumor depths [25][26][27] . Anatomically speaking, these parapharyngeal structures are adjacent, or in direct connection with the oral epithelial tissues, where improper surgical practice will result in positive margins 26 . Considering the treatment outcomes of these residual lesions, it is better to "err on the safe side" for extending the margins a bit wider, and to prepare intraoperative ap reconstructions, especially for some clinically T2-3 cases 26,27 . Besides, the existence of cervical residual OSCCs were, in our opinion, partly due to unstandardized or improper resections or neck dissections, and to higher primary N grades which had also been cited in other studies as the reasons for regional (cervical) recurrences after neck dissections 9,28 . We consent to the recent Clinical Practice Guideline issued by American Society of Clinical Oncology for establishing preliminary recommendations on the preliminary criteria of a high-quality neck dissection 9 . The anatomic hallmarks, levels and lest number of nodal specimens should also be emphasized for the best practice of surgical care for OSCC patients.
Apart from the surgical problems, as re ected in Table 2, other clinical factors should also be cautiously evaluated for avoiding treatment malpractice. Firstly, as is re ected in our series, 41.7% of the cases were with comorbidities, which might cause hesitations of aggressive surgical treatment from the patients' and doctors' perspectives 29 . Besides, the competencies of surgeons for such OSCC treatment should be assessed 30 , as 33.1% of the patients in our study received their initial treatment from junior consultants, or even surgeons from other non-relating specialties. Besides, patients who received surgical treatment from low-volume peripheral institutions tend to have improper or low-quality treatment practice in our series, with more chances of positive margins and lower likelihood of providing care adherent to guidelines [31][32] . However, such view was refuted by Eskander for con icting evidence comparing the quality of care between high-and low-volume institutions 33 . For us, the ample experiences of treating OSCC on a regular basis made difference between different institutions and surgeons. In addition, the adverse survival relationships of "delays between biopsies and treatment" was consistent with the reports of others 34 . Due to such varied negligence in primary treatment, we call for strictly adhering to the treatment and diagnosis guidelines otherwise it may cause tremendous disaster to the patients.
For the treatment of resectable residual diseases, there were still unsettled controversies about the role and outcomes of SS, with vastly con icting survival outcomes ranging from 8.3-62.5% 6,10,36,37 . Most of these studies were with both residual and recurrent OSCC cases, which were further complicated by a higher proportion of patients with histories of prior radio-or radio-chemotherapies 6,37 . We came up with the rst report for the outcomes of immediate SS-based treatment against residual OSCCs, who were mostly radiation-naive. The answer of salvage likelihood for residual OSCCs was partially answered in our study, as the survival outcomes diversi ed among these patients. According to us, careful case selections for immediate SS should be emphasized based on both the initial and residual status. In the current study, patients with both smaller primary and residual OSCC sizes were mostly salvageable under a sound retreatment. However, for cases with larger residual disease burdens, the prognosis was generally unfavorable with a meager survival of 15.4%. The involvement of vital structures in residual OSCCs were also found to decrease the likelihood of rescue. As for the treatment designs, we found a slight advantage of survival for the SS group over AT-SS group. A stronger association was also found for the salvage resection and reconstruction extent, as most patients with wide margin re-resections and freeap (including PMMF) reconstructions enjoyed better survival outcomes. Adjuvant radio-or radiochemotherapy following SS should be considered in patients with residual OSCCs for a 10-20% survival advantage, which was also reported in other studies for recurrent OSCCs 38,39 . As for other treatment combinations, the effects of targeted (EGFR or VEGF-based) therapies fell short of expectations as the trends of treatment outcomes reversed despite such added treatment regimens. We owed this phenomenon to both the treatment toxicities, and to the more advanced disease status of those who were inclined to receive such combinations. As far as we are concerned, routine postoperative radiotherapy or radio-chemotherapy was able to reach a similar, or even better outcome without the supplement of molecular targeted therapies, judging from our statistics.
Some limitations were inherent in the present study. Firstly, our results were obtained in a retrospective cohort in a single institution. Secondly, the treatment bene ts for advanced residual cases were unable to summarize due to the small number in this investigation. Most patients were also irradiation-naïve in the primary treatment. In addition, the case selection for curative SS were quite subjective. Lastly, the effects of immunotherapies were elusive given the absence of such treatment at that time.

Conclusions:
When encountered with primary OSCC cases, a well-round, evidence-based surgical plan, together with an able surgical expertise, is mandatory for the ultimate treatment success. Cases with residual OSCCs were mostly due to mistakes which could have been avoided if the guidelines and practice codes were strictly followed. SS for cases with both smaller residual radiation-naïve OSCCs is still feasible with acceptable outcomes, when carefully designed and performed.