Effect of Pre-Existing Sarcopenia on Oncological Outcomes for Oral Cavity Squamous Cell Carcinoma Undergoing Curative Surgery: A Propensity Score-Matched, Nationwide, Population-Based Cohort Study

Simple Summary Although sarcopenia during cancer diagnosis is an independent prognostic factor for poor overall survival in patients with various cancers, whether pre-existing sarcopenia is an independent risk factor for oral cavity squamous cell carcinoma (OCSCC) remains unclear. Therefore, we conducted a head-to-head propensity score matching (PSM) study to estimate the oncological outcomes of pre-existing sarcopenia in patients with OCSCC undergoing curative surgery. Both univariate and multivariate Cox regression analyses indicated that pre-existing sarcopenia was associated with poor survival than nonsarcopenia. Old age, male sex, advanced pT, advanced pN, differentiation grade II–III, margin-positive cancer, lymphovascular invasion, and CCI ≥ 1 were significant poor prognostic factors for survival in the patients with OCSCC undergoing curative surgery. Abstract Purpose: The effect of pre-existing sarcopenia on patients with oral cavity squamous cell carcinoma (OCSCC) remains unknown. Therefore, we designed a propensity score-matched population-based cohort study to compare the oncological outcomes of patients with OCSCC undergoing curative surgery with and without sarcopenia. Patients and Methods: We included patients with OCSCC undergoing curative surgery and categorized them into two groups according to the presence or absence of pre-existing sarcopenia. Patients in both the groups were matched at a ratio of 2:1. Results: The matching process yielded 16,294 patients (10,855 and 5439 without and with pre-existing sarcopenia, respectively). In multivariate Cox regression analyses, the adjusted hazard ratio (aHR, 95% confidence interval [CI]) of all-cause mortality for OCSCC with and without pre-existing sarcopenia was 1.15 (1.11–1.21, p < 0.0001). Furthermore, the aHRs (95% CIs) of locoregional recurrence and distant metastasis for OCSCC with and without pre-existing sarcopenia were 1.07 (1.03–1.18, p = 0.0020) and 1.07 (1.03–1.20, p = 0.0148), respectively. Conclusions: Pre-existing sarcopenia might be a significant poor prognostic factor for overall survival, locoregional recurrence, and distant metastasis for patients with OCSCC undergoing curative surgery. In susceptible patients at a risk of OCSCC, sarcopenia prevention measures should be encouraged, such as exercise and early nutrition intervention.

Sarcopenia, characterized by the loss of muscle mass, strength, and performance [12][13][14], can occur not only in overweight and underweight individuals but also in those with normal weight [15]. Unlike cachexia, sarcopenia does not require the presence of an underlying illness [16]. In addition, although most people with cachexia are sarcopenic, most individuals with sarcopenia are not considered cachectic [16]. Sarcopenia is associated with increased functional impairment, disability, fall, and mortality rates [17]. The causes of sarcopenia are multifactorial and include disuse, endocrine function alteration, chronic diseases, inflammation, insulin resistance, and nutritional deficiencies [14]. Therefore, sarcopenia and cancer cachexia-related sarcopenia are distinct conditions. Pre-existing sarcopenia can be prevented, whereas cancer-related sarcopenia cannot be prevented but can be treated.
Sarcopenia is associated with increased mortality for most cancers, except hormonerelated cancers (endometrial, breast, ovarian, and prostate cancers) and hematopoietic cancers [18][19][20][21], thus making it a major prognostic factor for poor overall survival and mortality in patients with cancer [18][19][20][21]. Sarcopenia-related cancer mortality might be a consequence of treatment-related toxicity [22,23]. However, whether pre-existing sarcopenia is an independent risk factor for different cancers, including OCSCC, remains unclear. A propensity score matching (PSM)-based design can resolve this issue by maintaining balance among the confounding factors of the case and control groups-all in the absence of bias [24][25][26]. Moreover, PSM is currently the recommended standard tool for estimating the effects of covariates in studies where any potential bias may exist [24][25][26]. Therefore, we conducted a head-to-head PSM study to estimate the oncological outcomes of pre-existing sarcopenia in patients with OCSCC undergoing curative surgery.

Study Population
We selected patients with OCSCC who had undergone curative surgery-tumor resection and neck dissection-between 1 January 2007 and 31 December 2017 from the Taiwan Cancer Registry Database (TCRD). The follow-up period was from the index date (i.e., date of surgery) to 31 December 2018. The types and indications of neck dissection were as follows: supraomohyoid neck dissection for clinically N0 tumors [27], modified neck dissection for ipsilateral clinically positive nodes [28], and bilateral neck dissection for contralateral metastases or tumors cross the midline [29]. Adjuvant treatments indicated for patients with OCSCC were based on the National Comprehensive Cancer Network (NCCN) guidelines and patients' tolerance [30]. The TCRD contains detailed cancer-related data of patients, including the clinical stage, cigarette smoking habit, treatment modalities, pathologic data, and grade of differentiation [5,[8][9][10]31]. The study protocols were reviewed and approved by the Institutional Review Board of Tzu-Chi Medical Foundation (IRB109-015-B).
The diagnoses of the enrolled patients were confirmed after reviewing their pathological data, and patients who were newly diagnosed as having OCSCC were confirmed to have no other cancers or distant metastasis (DM). All patients with OCSCC underwent curative-intent surgery. The inclusion criteria were as follows: being aged ≥20 years, having a diagnosis of pathologic stage I-IVB OCSCC without metastasis according to the American Joint Committee on Cancer criteria (AJCC, 7th edition), and undergoing tumor resection and neck dissection. Patients were excluded if they had a history of other cancers before the index date, an unknown pathological stage, missing sex data, unclear differentiation of tumor grade, or a nonsquamous cell carcinoma pathologic type.

Interventions/Exposures
Our definition of sarcopenia is according to the previous study from the Taiwan NHIRD [32]. In order to diminish the selection bias of the definition of sarcopenia, we only recorded the sarcopenia from the rehabilitation specialists, orthopedics, or family physicians. We have also added the sensitivity analysis of the recorded sarcopenia from the rehabilitation specialists, orthopedics, and family physician with/without other specialties (including endocrinology department) (Supplementary Table S2). In Taiwan, the coding of sarcopenia was based on a previous Taiwan study [33]; sarcopenia was defined as the skeletal muscle mass index (SMI) of 2 standard deviations (SDs) or more below the normal sex-specific means for young persons. Patients diagnosed as having sarcopenia after OCSCC diagnosis and those with sarcopenia diagnosed within 1 year before OCSCC diagnosis (excluding cancer treatment-related and cancer cachexia-related sarcopenia) were excluded. We also supplied the sensitivity analysis for the comparison of washout time intervals of one year and two years (Supplementary Table S1).

Comparisons
We categorized the patients into two groups depending on whether they had sarcopenia before OCSCC diagnosis: Group 1 (nonsarcopenic OCSCC) and Group 2 (pre-existing sarcopenic OCSCC). In addition, we estimated oncological outcomes (all-cause mortality, locoregional recurrence [LRR], and DM) associated with sarcopenia. Comorbidity was assessed using the Charlson comorbidity index (CCI) [6,34]. Only comorbidities which appeared 12 months before the index date were included and they were coded and classified according to the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) or International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes at the first admission or after >2 appearances of a diagnostic code at outpatient visits.

Outcomes
The oncologic outcomes were defined as all-cause death, LRR, and DM according to the previous oncologic studies [35][36][37]. All-cause mortality was the primary endpoint in both the groups. The secondary endpoints were LRR and DM.

Design Setting
To reduce the effects of potential confounders when comparing all-cause mortality between patients without and with sarcopenia, we performed 2:1 PSM with a caliper of 0.2 for the following variables: age, sex, years of diagnosis, AJCC pathologic stages, pathologic tumor stages (pT), pathologic nodal stage (pN), differentiation grade, surgical margin, lymphovascular invasion (LVI), adjuvant treatments, CCI scores, cigarette smoking, alcohol use, and betel nut chewing. These variables are potential prognostic factors for all-cause mortality for patients with OCSCC undergoing curative surgery. A Cox proportional hazards model was used to regress all-cause mortality in patients with OCSCC with a robust sandwich estimator used to account for clustering within matched sets [38]. Potential confounding factors for all-cause mortality for OCSCC were controlled in the PSM (Table 1). After well-matched PSM, the actual real-world data can indicate the oncological outcomes of pre-existing sarcopenia in patients with OCSCC undergoing curative surgery.

Statistical Analysis
The aforementioned variables might be independent prognostic factors for all-cause mortality with residual imbalance after PSM [39,40]. Therefore, multivariate Cox regression analyses were performed to calculate hazard ratios (HRs) to determine whether pre-existing sarcopenia is an independent predictor of all-cause mortality.
After adjustment for confounders, all statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). In a two-tailed Wald test, p < 0.05 was considered significant. OS, LRR, and DM were estimated using the Kaplan-Meier method and betweengroup differences were compared using the stratified log-rank test (stratified according to matched sets) [41].

Study Cohorts before and after PSM
We identified 45,219 patients with OCSCC undergoing curative surgery (39,775 without and 5445 [12.04%] with pre-existing sarcopenia) before PSM (Supplementary Table  S1). Compared with the patients without pre-existing sarcopenia, those with sarcopenia were older; were predominantly women; had higher CCI scores; more likely received the diagnosis in 2015-2017; had more advanced pT and pN stages; had more poor differentiation, margin positivity, and LVI-positive tumors; and received more adjuvant concurrent chemoradiotherapy (CCRT), higher radiotherapy (RT) doses, and higher cumulative platinum doses. PSM yielded 16,294 patients (10,855 without and 5439 with sarcopenia) who were eligible for further analysis and their characteristics are summarized in Table 1. Age, sex, years of diagnosis, cancer subtypes, AJCC pathological stages, pT, pN, differentiation, surgical margin, lymphovascular invasion, adjuvant treatments, CCI scores, cigarette smoking, alcohol use, and betel nut chewing were balanced between the cohorts (all p > 0.05). After PSM, the crude all-cause mortality, LRR, and DM were significantly higher in the patients with sarcopenia than in those without sarcopenia (Table 1).

Cox Proportional Hazard Models of All-Cause Mortality
According to multivariate Cox regression analysis, pre-existing sarcopenia was a significant predictor of all-cause mortality ( Table 2). Both univariate and multivariate Cox regression analyses indicated that sarcopenia was associated with poorer OS than nonsarcopenia. The HR for the univariate model was similar to that for the multivariate Cox regression analysis. Old age, male sex, advanced pT, advanced pN, differentiation grade II/III, margin positivity, LVI positivity, and CCI ≥ 1 were significantly poor prognostic factors for OS in the patients with OCSCC. In multivariate Cox regression analyses, the adjusted hazard ratio (aHRs, 95% confidence interval [CI]) of all-cause mortality for OCSCC with and without pre-existing sarcopenia was 1.14 (1.10-1.19, p < 0.0001). The aHRs (95% CIs) of mortality for male sex, age 50-59 years, age ≥ 60 years, pT2, pT3, pT4A, pT4B, pN1, pN2, pN3, differentiation grades II and III, margin positivity, LVI positivity, CCI ≥ 1, cigarette smoking, alcohol use, and betel nut chewing compared with female sex, age < 50 years, pT1, pN0, differentiation grade I, margin negativity, LVI negativity, CCI = 0, no cigarette smoking, no alcohol use, no betel nut chewing were 1. 28 Table 2 were used in multivariate analysis.

Cox Proportional Hazard Models of LRR and DM
Both univariate and multivariate Cox regression analyses indicated that pre-existing sarcopenia was associated with higher risk of LRR and DM than nonsarcopenia (Tables 3 and 4). In the multivariate Cox regression analysis, the aHRs (95% CIs) of LRR and DM for OCSCC with and without pre-existing sarcopenia were 1.07 (1.03-1.18, p = 0.0020) and 1.07 (1.03-1.20, p = 0.0148), respectively. In addition, poor prognostic factors for LRR and DM were similar with those of mortality, except old age and CCI scores. The multivariable Cox model revealed that male sex, advanced pT, advanced pN, differentiation grade II-III, margin positivity, LVI positivity, cigarette smoking use, alcohol use, and betel nut chewing use were independent poor prognostic factors for LRR and DM (Tables 3 and 4).     Figure 1 and Supplementary Figures S1 and S2 present survival curves for OS, LRR, and DM plotted using the Kaplan-Meier method for the PSM sarcopenia and nonsarcopenia OCSCC groups who underwent curative surgery. The OS curve for nonsarcopenic OCSCC was higher than that for sarcopenic OCSCC (Figure 1, p < 0.001). The 5-year OS was 56.03% and 48.93% for the patients with OCSCC without and with pre-existing sarcopenia, respectively. Moreover, the cumulative LRR and DM rates were significantly higher for sarcopenic OCSCC than nonsarcopenic OCSCC in the log-rank test (Supplementary Figures S1 and S2, p values were all <0.0001 for LRR and DM, respectively).  Figure 1 and Supplementary Figures S1 and S2 present survival curves for OS, LRR, and DM plotted using the Kaplan-Meier method for the PSM sarcopenia and nonsarcopenia OCSCC groups who underwent curative surgery. The OS curve for nonsarcopenic OCSCC was higher than that for sarcopenic OCSCC (Figure 1, p < 0.001). The 5-year OS was 56.03% and 48.93% for the patients with OCSCC without and with pre-existing sarcopenia, respectively. Moreover, the cumulative LRR and DM rates were significantly higher for sarcopenic OCSCC than nonsarcopenic OCSCC in the log-rank test (Supplementary Figures S1 and S2, p values were all <0.0001 for LRR and DM, respectively).

Discussion
Sarcopenia is an independent prognostic factor for poor survival in patients with HNC undergoing surgery, RT, or CCRT [20,[42][43][44][45][46][47]. However, these studies included heterogeneous definitions of sarcopenia, inconsistent treatments for HNCs, different HNC subtypes, inhomogeneous HNC stages, very small sample sizes, and inconsistent cancer subtypes including oropharyngeal, hypopharyngeal, oral cavity, and laryngeal cancers [20,[42][43][44][45][46][47]. None of these studies differentiated between sarcopenia as pre-existing or that related to cancer cachexia. Accordingly, their result that sarcopenia is a poor prognostic factor for survival outcomes might be due to cancer-related cachexia-induced sarcopenia or cancer treatment-related sarcopenia instead of pre-existing sarcopenia [20,[42][43][44][45][46][47]. However, sarcopenia is different from cancer cachexia [14,16,17]. The causes of sarcopenia are multifactorial [14] and include muscle disuse, changes in endocrine function, chronic diseases, inflammation, insulin resistance, and nutritional deficiencies; many of these conditions can be detected early on and corrected through measures such as exercise or nutrition to prevent sarcopenia progression [48][49][50][51]. Therefore, we estimated the oncological outcomes of pre-existing sarcopenia in the patients with OCSCC undergoing curative surgery to determine the effect of pre-existing sarcopenia on OCSCC. To our knowledge, this is the first head-to-head PSM, largest, and longest follow-up study evaluating the effect of pre-existing sarcopenia on patients with OCSCC undergoing curative surgery. Our data indicated that pre-existing sarcopenia is an independent poor prognostic factor for mortality, LRR, and DM.
The definition of sarcopenia has been inconsistent in previous studies [20,[42][43][44][45][46][47]. In patients with HNC receiving RT or CCRT, sarcopenia has been reported to be associated with poor OS and disease-free survival outcomes [42][43][44][45]47]. Only one report including patients with HNC receiving surgical excision demonstrated that sarcopenia appears to be a significant negative predictor of long-term OS in patients with HNC undergoing major surgery [43]. Stone et al. defined sarcopenia by using cross-sectional abdominal imaging performed within 45 days prior to surgery [43]. However, this definition precluded the differentiation of pre-existing sarcopenia from cancer cachexia-related sarcopenia [43]. This renders any results on the effect of sarcopenia unclear [43] and does not affect clinical practice in patients with HNC because cachexia is a well-known poor prognostic factor for OS in HNCs [52,53]. Our study is the first to present a clear definition of pre-existing sarcopenia (diagnosed ≥1 year before the diagnosis of OCSCC) in a homogenous group of patients with the same subtype of HNC (OCSCC) undergoing curative surgery. Therefore, our finding that pre-existing sarcopenia is the poor prognostic factor for OS, LRR, and DM might encourage the implementation of early screening for sarcopenia and intervention such as resistance exercise, protein supplementation, and vitamin D for patients at a high risk of OCSCC (betel nut chewing, cigarette smoking, or alcohol abuse) [48][49][50][51]. These valuable outcomes would provide references for the health government to establish health policies to correct, interrupt, or prevent the progression of pre-existing sarcopenia, particularly in the susceptible population.
Performing a randomized controlled trial (RCT) to evaluate oncological outcomes in patients with OCSCC undergoing curative surgery with and without pre-existing sarcopenia is difficult because sarcopenia cannot be treated using a tangible intervention [54]. Traditionally, striking a balance among the confounding factors of mortality in patients with OCSCC with and without sarcopenia (i.e., the case and control groups, respectively)-a main requirement of the RCT design-is impossible [54]. Although the main advantage of the PSM methodology is the more precise estimation of the covariate effect, PSM cannot control for factors not accounted for in the model. Moreover, PSM is predicated on an explicit selection bias of those who could be matched; in other words, individuals who could not be matched are not part of the scope of inference.
The mechanism through which pre-existing sarcopenia serves as a poor prognostic factor for OS, LRR, and DM might be associated with multiple factors including the metabolic processes of insulin resistance and systemic inflammation [14,16,17]. Patients with sarcopenia might have systemic inflammation that reduces liver cytochrome activities and drug clearance and metabolic processes, leading to a poor therapeutic effect [60]. In addition, inflammation by sarcopenia can cause a decrease in skeletal muscle density. A decreased muscle density is related to intramuscular lipid accumulation and favored by systemic inflammation, thus leading to a vicious cycle [60]. Therefore, early intervention to break this cycle is critical in patients with sarcopenia [48][49][50][51]. According to an epidemiological study in Taiwan, the incidence of oral cancer was 123-fold higher in patients who smoked, consumed alcohol, and chewed betel quid than in abstainers [2]. Patients with sarcopenia with risk factors for OCSCC [60] are the susceptible population for poor OS. Early screening for and treatment of sarcopenia for the susceptible population might improve survival outcomes in case they develop OCSCC.
This study has several limitations. First, the cohort derived from an Asian population in Taiwan. Although no evidence indicating a significant difference in survival of OCSCC between Asian and non-Asian populations has been reported, the current results should be cautiously extrapolated to non-Asian populations. Second, this study was performed on a big database and thus it is a real challenge to rule out an ecological bias (attributed to confounding or risk factors). PSM cannot control for factors not accounted for in the model and is predicated on an explicit selection bias of the variables that were matched. Third, patients with antecedents of other cancers were excluded. The field cancerization theory is well accepted on this anatomical area, i.e., a patient with oral cancer has a higher risk to develop future aerodigestive carcinomas (and vice versa) [4,61,62]. However, the primary endpoint in the current study is the all-cause death between sarcopenia and nonsarcopenia OCSCC, OCSCC patients combined with other cancers will have higher mortality attributed to more aggressive treatments or more advanced stages on the other cancers, whatever synchronous or metachronous cancers [4,61,62]. In order to decrease the bias of all-cause death from the other cancers in the OCSCC patients, patients with antecedents of other cancers were excluded. Fourth, the diagnoses of all comorbid conditions were based on ICD-9-CM or ICD-10-CM codes in this study. Nevertheless, the Taiwan Cancer Registry Administration reviews charts and interviews of beneficiaries in the TCRD to verify the accuracy of the diagnoses, and it audits hospitals with outlier chargers or practices and subsequently heavily penalizes them if it identifies any malpractice or discrepancies. However, to obtain precise population specificity and disease occurrence data, a large-scale RCT carefully comparing patients with OCSCC with or without sarcopenia is warranted, but such RCTs may be difficult to execute.
Despite these limitations, a major strength of our study is the use of a nationwide population-based registry with detailed baseline information. The TCRD is linked with Taiwan's National Cause of Death Database; thus, in the current study, we could perform a lifelong follow-up for most patients. Moreover, this study is the first, largest, and longest follow-up comparative cohort study to estimate the primary endpoint of OS in patients with OCSCC with and without pre-existing sarcopenia undergoing curative surgery. The covariates between the two groups were homogenous and any bias between the two groups was removed through PSM (Table 1). Considering the magnitude and statistical significance of the observed effects in the current study, the limitations are unlikely to have affected our conclusions.

Conclusions
Our results indicate that pre-existing sarcopenia is a significantly poor prognostic factor for OS, LRR, and DM in patients with OCSCC undergoing curative surgery. Individuals with a high risk of OCSCC, such as those who have a habit of betel nut chewing, alcohol, or smoking, should be screened for sarcopenia and intervention in terms of exercise and nutrition should be promoted.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cancers14133246/s1, Figure S1: Kaplan-Meier overall cumulative locoregional recurrence curves for the propensity score-matched sarcopenia and nonsarcopenia groups (controls); Figure S2: Kaplan-Meier overall cumulative distant metastasis curves for the propensity score-matched sarcopenia and nonsarcopenia groups (controls); Table S1: Sensitivity analysis of washout time-intervals of one year and two years for definition of preexisting sarcopenia; Table S2: Sensitivity analysis of preexisting sarcopenia recorded by special specialties and all specialties. Informed Consent Statement: Not applicable.

Data Availability Statement:
The data sets supporting the study conclusions are included in the manuscript. We used data from the National Health Insurance Research Database and Taiwan Cancer Registry database. The authors confirm that, for approved reasons, some access restrictions apply to the data underlying the findings. The data used in this study cannot be made available in the manuscript, the supplementary files, or in a public repository due to the Personal Information Protection Act executed by Taiwan's government, starting in 2012. Requests for data can be sent as a formal proposal to obtain approval from the ethics review committee of the appropriate governmental department in Taiwan. Specifically, links regarding contact info for which data requests may be sent to are as follows: http://nhird.nhri.org.tw/en/Data_Subsets.html#S3 and http://nhis.nhri.org.tw/ point.html (accessed on 5 February 2021).

Conflicts of Interest:
The authors declare no conflict of interest.