In this retrospective study, we included a total of 100 patients with HNSCC who had received initial treatment for primary HNSCC such as cancer of the oropharynx, hypopharynx, and larynx between February 2012 and March 2016. Exclusion criteria involved patients with missing data and undergoing palliative therapy only. All clinical data were collected using electronic medical records. This study was approved by the appropriate institutional research ethics committee (reference numbers: 2019-29) and was conducted according to the tenets of the Declaration of Helsinki. The requirement for informed consent was waived owing to the retrospective nature of the analysis.

Patients were treated with surgery, radiotherapy (RT) alone, and concurrent chemoradiotherapy (CCRT) considering various factors such as age, stage of disease, risk factors, performance status, and comorbidities. The listed treatments were initiated according to the guidelines of the National Comprehensive Cancer Network. Briefly, T1 and T2 cases prefer RT alone (total of 60–66 Gy) or transoral surgery, while T3 and T4 cases were administered CCRT (Cisplatin; 80 mg/m², infused on days 1, 22, and 43, RT; a total of 66 Gy) or surgery (total laryngectomy or pharyngolaryngectomy) based on several factors of the patients. In advanced cases, surgery was preferred.

In all 100 cases, cervical computed tomography (CT) imaging was obtained before the treatment. According to a previously described method by Swarz et al. (11), SMM was determined in each patient. Briefly, a single axial CT slide at level C3, displaying the entire vertebral arc, was selected first, when the C3 vertebra was scrolled from a caudal to a cephalic direction. The paravertebral muscle and both sternocleidomastoid muscle segments were highlighted in red and traced manually using the ImageJ software ( Figure 1 ). We calculated the sum of the delineated areas of both the paravertebral and sternocleidomastoid muscles at the level of C3 vertebrae, defined as the cross-sectional muscle area (CSA) at level C3. Next, we estimated CSA at level L3 using the prediction method previously described by Swartz et al. (11). The estimated CSA at level L3 was normalized for the height by dividing it by the squared height, defined as the lumber skeletal muscle index (LSMI, cm²/m²).

Sarcopenia is characterized by an LSMI below 43.2cm²/m² (12) according to the international consensus. In our study, although only low LSMI was used for sarcopenia definition, albumin and BMI were also collected as indicators reflecting nutritional status. Based on institutional criteria, the cut-off values of albumin and BMI were set at 3.5 and 18.5, respectively.

The blood cell counts of the patients were measured within one week of treatment administration. NLR, PLR, and LMR were calculated by dividing the neutrophil count by the lymphocyte count, the platelet count by the lymphocyte count, and the lymphocyte count by the monocyte count, respectively. Low values of LMR and high values of NLR and PLR suggest high inflammatory status. We developed Receiver Operating Characteristic (ROC) curves for the NLR, PLR, and LMR using OS as the primary endpoint. NLR ≥ 2.180 was defined as high with an area under the curve (AUC), sensitivity, and specificity of 0.62, 60.0%, and 62.0%, respectively. A high PLR was defined as ≥ 112.8 with an AUC, sensitivity, and specificity of 0.59, 42.0%, and 84.0%, respectively. A lower LMR was defined as ≤ 4.118 with an AUC, sensitivity, and specificity of 0.62, 68.0%, and 58.0%, respectively.

Continuous variables are shown as the median (or mean) and range, while we presented categorical variables as frequencies. For comparisons between groups, we analyzed continuous data using the Mann–Whitney U test, while we performed categorical data analysis using Fisher’s exact test or the Chi-square test. We defined OS as the time from diagnosis until the last follow-up or death from any cause. We defined progression-free survival (PFS) as the time from diagnosis until the detection of the first detection of disease progression, the last follow-up, or death from any cause. We also compared how potential risk factors (age, sex, primary site, T or N classification, treatment type, anemia, BMI, sarcopenia, NLR, PLR, or LMR) could affect OS and PFS using the log-rank test and analyzed by generating Kaplan–Meier survival curves. We used Cox hazard regression analysis to perform multivariable analysis on the variables with P-values of p < 0.05 in the univariate analysis and clinically important OS and PFS predictors. HRs and their corresponding 95% confidence intervals are presented. We compared the combined prognostic factor of sarcopenia and systemic inflammation markers (NLR, PLR, and LMR) according to the ROC curve and also calculated the AUC. All statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria). More precisely, the referred interface is a modified version of R commander designed to add statistical functions that are frequently used in biostatistics. P < 0.05 was considered statistically significant.

In a total of 100 patients, 94 were men and the median age at diagnosis was 69 years (range, 39-92 years). Of these, 12, 24, and 64 patients suffered from oropharyngeal, hypopharyngeal, and laryngeal cancer, respectively. The patients were divided into non-sarcopenia and sarcopenia groups, according to the LSMI cut-off described in the Material and Methods section. Table 1 presents the patient characteristics in the non-sarcopenia (n = 29, 29%) and sarcopenia (n = 71, 71%) groups. Patients in the sarcopenia group were older, at a more advanced T- and TNM stage, and displayed lower BMI compared to those in the non-sarcopenia group while values of albumin did not show significant differences. Concerning the inflammatory markers, NLR and PLR were significantly higher in the sarcopenia than in the non-sarcopenia group (p = 0.021 and 0.031, respectively) and LMR was significantly lower in the sarcopenia than in the non-sarcopenia group (p = 0.040).

Twenty-eight patients died over a median follow-up duration of 78 months (range, 1–138 months). The 3-year OS and PFS rates among all 100 patients were 79 and 77%, respectively. Our univariate analysis revealed that NLR was associated with OS, but not with PFS ( Table 2 ). Moreover, the univariate analysis showed that T classification, sarcopenia, and LMR were associated both with OS and PFS. Multivariate analysis using factors that showed significant differences in univariate analysis revealed that only sarcopenia was a significant predictor of both OS and PFS. According to the Kaplan–Meier analysis, the patients with sarcopenia had poorer OS (log-rank test: p = 0.002; Figure 2A ) and PFS (log-rank test: p = 0.005; Figure 2B ) than those with non-sarcopenia.

Neither systemic inflammation index alone was a significant predictor in multivariate analysis, but we also evaluated whether their utility could be improved by combining them with sarcopenia. We created ROC curves and compared the AUC values to assess the discrimination ability of each prognostic score. The AUC values of the sarcopenia alone, sarcopenia plus NLR, sarcopenia plus PLR, and sarcopenia plus LMR for OS were 0.678 (95%CI = 0.570–0.786), 0.756 (95%CI = 0.662–0.850), 0.727 (95%CI = 0.624–0.829), and 0.752 (95%CI = 0.655–0.848), respectively. Although the combination of sarcopenia and NLR, PLR or LMR showed significantly greater AUC values than sarcopenia alone, there were no significant differences between sarcopenia plus NLR, sarcopenia plus PLR and sarcopenia plus LMR.

To compare the utility of three different combination patterns of sarcopenia and systemic inflammatory indices, we stratified patients into sarcopenia plus high inflammatory status and sarcopenia plus low inflammatory status using NLR, PLR, and LMR. Patients with sarcopenia plus high NLR had a worse OS than patients with sarcopenia plus low NLR (5-year OS; 66.7% vs. 39.4%, log-rank test: p = 0.043; Figure 3A ). Similarly, patients with sarcopenia plus low LMR had significantly worse OS than those with sarcopenia plus high LMR (5-year OS; 67.6% vs. 37.8%, log-rank test: p = 0.012; Figure 3C ). However, the combined index of sarcopenia and PLR did not show significant differences between two groups (5-year OS; 80.0% vs. 55.7%, log-rank test: p = 0.377; Figure 3B ). On the other hand, regarding PFS, the combination of sarcopenia and LMR is the only indicator that showed significant differences between two groups (5-year PFS; 61.8% vs 32.4%, log-rank test: p = 0.033; Figure 4C ), while sarcopenia plus NLR (5-year PFS; 60. 6% vs. 34.2%, log-rank test: p = 0.070; Figure 4A ) and sarcopenia plus PLR (5-year PFS; 60.0% vs. 52.9%, log-rank test: p = 0.876; Figure 4B ) showed no statistical differences, suggesting that the combination of sarcopenia and LMR is the most reliable prognostic index. We thus used this parameter in the subsequent subgroup analysis. The group of patients with sarcopenia plus low LMR had a higher percentage of oro-hypopharyngeal cancer (51.4% vs. 38.2%) and Stage III-IV cancers (62.2% vs. 44.1%) compared to those with sarcopenia plus high LMR. As for the treatment, the group of patients with sarcopenia plus low LMR tended to perform surgery compared to those with sarcopenia plus high LMR (37.8% vs. 20.6%), but there were no significant differences concerning treatment choice.

We then performed a multivariate analysis including sarcopenia plus LMR status, T classification, and albumin which showed significant differences in univariate analysis. The results revealed that sarcopenia plus low LMR was a significantly poor prognostic factor both for OS and PFS with greater HR values than sarcopenia alone ( Table 3 ).