Prognostic Utility of Nutritional Risk Index in Patients with Head and Neck Soft Tissue Sarcoma

Background: The nutritional risk index (NRI) is an excellent indicator of nutritional status and a significant prognostic factor in several malignancies, but the relationship between NRI and the prognosis of head and neck soft tissue sarcoma (HNSTS) patients remains unclear. The aim of this study was to investigate the role of NRI in patients with HNSTS. Methods: We retrospectively reviewed patients with HNSTS between 1990 and 2021. In order to determine the optimal cut-off value of NRI, the Maximally selected log-rank statistic was performed. We evaluated the effect of NRI on overall survival (OS) and progression-free survival (PFS) by using the Kaplan–Meier method and Cox regression analysis. Then, OS and PFS nomograms based on NRI were constructed. Results: In total, 436 HNSTS patients were included in this study. The optimal cut-off value of NRI was 99.34. Patients with low-NRI showed significantly worse OS and PFS than patients with high-NRI, respectively (5-year OS rate of 43.0 vs. 70.8%, 5-year PFS rate of 29.0 vs. 45.0%, all p < 0.05). In the multivariate analysis, distant metastasis, deep tumor depth, tumor grade, and NRI were prognostic factors for both PFS and OS, and treatment modality was associated with OS but not PFS. The concordance indexes (C-indexes) of OS and PFS nomograms were 0.794 (95% CI, 0.759–0.829) and 0.663 (95% CI, 0.626–0.700), respectively, which also performed well in the validation set. Conclusions: NRI is an independent predictor of OS and PFS in HNSTS patients. The validated nomograms based on NRI provide useful predictions of OS and PFS for patients with HNSTS.


Introduction
Head and neck soft tissue sarcoma (HNSTS) arising from mesenchymal cells is a rare and heterogeneous malignancy, accounting for only 1% of head and neck tumors and approximately 10% of soft tissue sarcomas (STS) [1][2][3]. Up to 50 pathological subtypes of HNSTS have been identified, characterized by a conglomerate of anatomical locations and heterogeneous subtypes [4]. HNSTS has more malignancy than STS originating from other body sites that might contribute to poor prognosis of patients [5], with the five-year overall survival (OS) range from 31% to 80% [6]. Traditional prognostic indicators such as histological grade, tumor site, tumor size, and the presence of metastasis have been considered significant predictors of outcome [7][8][9]. However, inaccuracy and inadequacy have been gradually exposed in clinical practice [10]. Therefore, the identification of novel prognostic biomarkers is essential for clinicians to further stratify patients with HNSTS for optimal treatments and improve clinical outcomes.
We retrospectively reviewed patients who underwent R0 resection for HNSTS at Sun Yat-sen University Cancer Center (SYSUCC) between January 1990 and January 2021 (n = 536). R0 resection was defined as the negative surgical resection margins of macroscopic disease. The inclusion criteria were as follows: (1) confirmed HNSTS by pathological examination; (2) received definitive surgery with R0 resection. The exclusion criteria were as follows: (1) patients who were aged younger than 18 years; (2) missing medical records of height or weight or albumin level; (3) follow-up duration less than 1 year or lost to followup; (4) patients receiving any prior treatment; (5) patients with secondary malignancy; (6) patients with any clinical evidence of acute infection or chronic active inflammatory disease or autoimmune disease. After exclusion, a total of 436 patients with confirmed HNSTS and adequate data were included in our study (Supplementary Figure S1). This study was approved by the Medical Ethics Committee of SYSUCC. Written informed consent was obtained from all patients in the hospital to offer related information.

Data Collection
Clinical and pathological information of the patients was collected from the hospital information system, including patient demographics (age, gender, history of smoking exposure, laboratory data, and physical assessment data), pathologic characteristics (tumor site, tumor size, lymph node metastasis, distant metastasis, tumor depth, tumor grade, and TNM stage), treatment modality and follow-up data (follow-up duration, survival, and progression). The laboratory and physical assessment data included serum ALB, height, and weight, which were collected from all subjects in the week preceding surgery. Tumor locations were categorized as scalp or face, neck, nasal cavity or paranasal sinus, oral cavity, pharynx or larynx, and other sites, and were further divided into low-risk sites and highrisk sites depending on the previous studies [5,18]. The TNM stage was classified according to the 7th edition of the American Joint Committee on Cancer (AJCC) [19] and tumor grade was evaluated on the basis of the French Federation of Cancer Centers Sarcoma Group (FNCLCC) systems [7]. All pathological diagnoses of HNSTS were performed by at least two independent experienced pathologists at SYSUCC. NRI value was calculated by this formula: NRI = 1.487 × serum ALB concentration (g/L) + 41.7 × preoperative weight/ideal body weight (kg) [20]. In this cohort, we defined ideal body weight as 22 × height (m) 2 [21]. BMI was evaluated as weight (kg)/height (m) 2 [22].

Follow Up
In the first 2 years, all patients were followed up with physical examination and/or magnetic resonance imaging (MRI) every 6 months and then checked annually. Patients with suspicious loco-regional progression on physical examination or MRI underwent a biopsy to confirm pathology. If distant metastasis is suspected, biopsy or serial follow-up imaging should be carried out.
The primary and secondary endpoints were OS and PFS, respectively. OS was defined as the time from the diagnosis to death from any cause or last follow-up and PFS was defined as the time from the diagnosis to local, regional, or distant progression or death from any cause or latest clinical observation.

Statistical Analysis
Data for continuous variables are expressed as mean (SD) or median and interquartile range (IQR) while data for categorical variables are expressed as absolute numbers and percentages. The Student's t-test was used to compare continuous variables, and the Chi-square test or Fisher's exact test was used to compare categorical variables. To investigate differences between NRI and baseline characteristics, the optimal cut-off value of NRI was determined by Maximally selected log-rank statistics to categorize the groups. Kaplan-Meier curves with the log-rank test were generated to assess OS and PFS among the whole cohort and different subgroups. Multivariable Cox proportional hazards models were constructed with a univariable inclusion criterion of p < 0.05 to identify prognostic factors for mortality and disease progression. Then, we established two nomograms based on the results of multivariate COX regression models with endpoints of 3-and 5-year OS and PFS. The accuracy of predictions was evaluated by estimating the model's calibration with 1000 bootstrap samples, and discrimination was measured by the concordance index (C-index). Receiver operating characteristic (ROC) curves were used to evaluate the predicting ability of OS and PFS nomograms by measuring the area under the curve (AUC). All statistical tests were conducted using SPSS version 26.0 (SPSS, Chicago, IL, USA) and R version 4.1.2 (http://www.rproject.org, accessed on 10 October 2022). p < 0.05 were considered statistically significant for the statistical analyses.

The Association between NRI and Clinical Characteristics
A maximally selected log-rank statistic was applied to establish the optimal NRI cut-off value of prognostic significance. The highest log-rank statistic coincided with a NRI level of 99.34 (Supplementary Figure S2). Based on the cut-off value, the whole cohort was divided into two subgroups: the low-NRI group (NRI ≤ 99.34) and the high-NRI group (NRI > 99. 34). The relationship between the NRI and various clinicopathological features were shown in Table 1. There were no differences between the two subgroups in gender, history of smoking, tumor site, tumor size, the status of lymph node metastasis, and treatment modality; however, patients with lower NRI were significantly associated with older age (p = 0.003), lower BMI (p < 0.001), distant metastasis (p = 0.018), deeper depth of tumor (p = 0.015), higher tumor grade (p = 0.003) and advanced TNM stage (p < 0.001). We randomly divided the entire cohort into training and validation sets in the ratio of 7:3, and the baseline characteristics of the training and validation sets are listed in Table 2. The training and validation sets were similar with respect to most variables, and 61 (19.8%) and 29 (22.7%) patients were assigned to the low-NRI groups, respectively. The Kaplan-Meier survival analysis showed that the low-PNI group was significantly associated with shorter OS relative to the high-NRI group (p < 0.001, Figure 1A) and shorter PFS (p = 0.003, Figure 1B). The 5-year OS and PFS rates for the high-NRI cohort were 70.8% and 45.0%, while those in the low-NRI cohort were merely 43.0% and 29.0%, respectively. In addition, compared with high-NRI patients, the median (IQR) survival time decreased from 84 (27-149) months to 43 (17-90) months for OS and from 33 (9-96) months to 17 (6-51) months for PFS in the whole cohort. The Kaplan-Meier survival analysis showed that the low-PNI group was significantly associated with shorter OS relative to the high-NRI group (p < 0.001, Figure 1A) and shorter PFS (p = 0.003, Figure 1B). The 5-year OS and PFS rates for the high-NRI cohort were 70.8% and 45.0%, while those in the low-NRI cohort were merely 43.0% and 29.0%, respectively. In addition, compared with high-NRI patients, the median (IQR) survival time decreased from 84 (27-149) months to 43 (17-90) months for OS and from 33 (9-96) months to 17 (6-51) months for PFS in the whole cohort.

Factors Predicting OS and PFS Outcomes
We carried out univariate Cox analyses to identify the significant prognostic factors (Supplementary Table S1), and multivariate Cox analyses were performed to assess the independent prognostic factors affecting OS and PFS (Table 3). Age, BMI, tumor site, lymph node metastasis, distant metastasis, tumor depth, tumor grade, TNM stage, treatment modality, and NRI were potential prognostic factors for OS, and these factors without age, BMI, and lymph node metastasis were prognostic factors for PFS (p < 0.05 for all).

Nomogram Development and Validation
We further constructed two nomograms for OS and PFS that integrated independent prognostic factors in the training cohort ( Figures 2B and 3B). To estimate 3-and 5-year OS and PFS, the scores for each independent predictor were plotted and serially summed up.
Good calibration curves were observed for the 3-and 5-year OS and PFS in the validation cohort (Supplementary Figure S3A,B). In the validation cohort, the OS model had a C-index of 0.866 (95% CI, 0.817-0.915) and the PFS model had a C-index of 0.648 (95% CI, 0.583-0.713). As shown in Supplementary Figure S4A 1.132-2.210, p = 0.007) were independent prognostic factors for PFS.

Nomogram Development and Validation
We further constructed two nomograms for OS and PFS that integrated independent prognostic factors in the training cohort ( Figures 2B and 3B). To estimate 3-and 5-year OS and PFS, the scores for each independent predictor were plotted and serially summed up.

Subgroup Analysis of Common Clinical Variables According to NRI Groups
As shown in Figure 4, we investigated the association of PNI with OS and PFS in subgroups of patients with common clinical characteristics. The effect of low-NRI on OS was consistently unfavorable (all hazard ratios, >1.0) across all prespecified subgroups, and the effect of low-NRI on PFS was mostly unfavorable (hazard ratios > 1.0), except for the subgroup of G1. Further Kaplan-Meier curves of clinically meaningful subgroup analysis showed that lower NRI was still significantly related to an increased risk of death in patients with age ≤ 50 or not ( Figure 5A,B), in males and females ( Figure 5C,D), in patients with low-risk site tumor or high-risk ( Figure 5E,F), and in patients with surgery alone or surgery plus adjuvant therapy ( Figure 5G,H). Similar results were observed for PFS in Supplementary Figure S5. the subgroup of G1. Further Kaplan-Meier curves of clinically meaningful subgroup analysis showed that lower NRI was still significantly related to an increased risk of death in patients with age ≤ 50 or not ( Figure 5A,B), in males and females ( Figure 5C,D), in patients with low-risk site tumor or high-risk ( Figure 5E,F), and in patients with surgery alone or surgery plus adjuvant therapy ( Figure 5G,H). Similar results were observed for PFS in Supplementary Figure S5.

Discussion
To our knowledge, this is the first investigation to identify and validate the prognostic value of NRI in HNSTS patients with curative resection. A low NRI was associated with a shorter OS and PFS in patients with HNSTS. NRI is not only a good indicator of nutritional status in patients but also an independent predictor of mortality and disease progression in HNSTS. Furthermore, the nomograms incorporated NRI were constructed to predict the OS and PFS of patients, which are useful for the personalized management of HNSTS.
Malnutrition is a complex condition, which is characterized by reduced protein reserves, weakened immune defenses, and calorie breakdown [23]. Approximately 20-30% of deaths in patients with cancer were due to malnutrition, rather than cancer itself [24]. Accumulating evidence suggests that malnutrition is related to disease progression and is a major cause of poor therapeutic outcomes [25,26]. The incidence of malnutrition has been reported to be high not only in cancer patients but also in sarcoma patients, and nutritional assessment can effectively reflect the prognosis of patients [27][28][29][30]. Actually, patients with HNSTS are more likely to suffer from the functional restrictions of chewing and swallowing after surgery, because of the special complicated anatomical location of HNSTS and postoperative complications. Additionally, nausea and vomiting symptoms are more often seen in head and neck malignancies, especially for the patients who received adjuvant therapy, which would exacerbate any malnutrition [31]. In the subgroup analysis of patients with surgery and adjuvant therapy, low-NRI continued to be predictive of shorter OS, demonstrating its good clinical applicability. Overall, malnutrition is more prevalent among HNSTS patients. Hence, NRI, as an excellent indicator of the nutritional status of surgical patients [32], is becoming more and more crucial for improving the quality of HNSTS patients, and it deserves sufficient consideration. Of note, it provides a long-term trend of the nutritional status of patients by integrating body weight, height, and serum ALB, which can remain stable in the short term [33]. Interestingly, a previous study based on weight loss, BMI, and NRI was reported by Righini et al., revealing that approximately 49% of patients with HNSC were at the status of malnutrition, and patients with oral or oropharyngeal tumor sites were more likely to suffer from malnutrition [34]. However, among the tumor sites of HNSTS, our study dividing the nutritional status based on NRI showed that the scalp or face and neck sites were the most common sites. The malnourished proportion of oral and oropharyngeal HNSTS was only 26.5% and 34.5%, and there was no statistical significance with NRI according to tumor sites, probably because HNSTS has the heterogeneous malignancy with different subtypes and distinct biological behaviors.
In the present study, we found that a low NRI level was significantly associated with older age, lower BMI, distant metastases, deeper depth of tumor, higher tumor grade, and advanced TNM stage, indicating that there is a potential correlation between aggressive and advanced tumors and the reduction of nutrient levels. Similar to our finding, several reports also found that patients with malnutrition are at a higher probability of aggressive head and neck cancer [35,36]. In addition, Nakayama et al. showed that patients with low NRI levels were at higher risk of advanced clinical stage and mortality [37]. The reason for the relationship between low NRI levels and advanced tumors might be due to the various cytokines released by the advanced tumor [38][39][40]. Moreover, several studies have shown that elevated inflammatory factors and poor nutrient absorption are associated with increased catabolism, leading to anorexia and malnutrition [41][42][43]. Older patients with lower BMI are more likely to have a low level of NRI, highlighting the importance of attention to the nutritional status of older patients.
This study also showed that distant metastasis, deep tumor depth, tumor grade, and NRI were prognostic factors for both PFS and OS, and treatment modality was associated with OS but not PFS. Of note, after multivariate adjustment for prognostic factors, NRI remained the independent prognostic factor both for OS and PFS. Similarly, Jung et al. conducted related research and found that age, tumor grade, tumor size, and initial distant metastasis were independent predictors for OS, and tumor size and tumor grade were independent predictors for PFS [44]. Moreover, a large-scale prospective study for the nutritional assessment of oral cancer demonstrated that patients with low levels of NRI were associated with worse prognosis in oral cancer and the prognostic performance of NRI was superior to BMI or serum ALB [17]. Nevertheless, there are minor differences with our study, probably because of the genomic heterogeneity between these malignancies, but our study with longer follow-up time and larger sample size allowed us to better identify the risk assessment of NRI and conventional factors for OS and PFS outcomes.
More importantly, this is the first study to demonstrate the correlation of NRI level with OS and PFS in HNSTS. The Kaplan-Meier survival analysis showed that the 5-year OS rate of patients with low-NRI was nearly 30% lower than those of patients with high-NRI (43.0% vs. 70.8%) and the 5-year PFS rate of patients with low-NRI was nearly 16% lower than those of patients with high-NRI (29.0% vs. 45.0%), suggesting that patients who have low levels of NRI should require greater scrutiny and closer follow-up after initial treatment. However, the mechanism of NRI influencing the prognosis of patients with HNSTS is not clear.
In our subgroup analysis, evidence of the higher risk of mortality and disease progression in patients with low-NRI, as compared with the high-NRI group, was that the effects of low-NRI were consistently unfavorable with each subgroup. Based on these findings, further Kaplan-Meier curves also demonstrated that patients with low-NRI, grouped by age, gender, tumor site, and treatment modality, tended to have shorter OS and PFS, which increase the clinical applicability of NRI. Considering that this was a subgroup analysis, the results should be interpreted with caution and confirmed with further large sample studies.
Based on NRI and other independent risk factors, we established and validated novel prognostic nomograms to predict the 3-and 5-year OS and PFS for HNSTS patients treated with primary therapy. Notably, OS and PFS nomograms showed good discrimination and calibration in both training and validation cohorts, indicating excellent discriminative ability and general applicability. As compared with the previously published nomograms for HNSTS [45], which had the C-indexes of 0.748 for OS and 0.644 for PFS, our nomograms have superior effectiveness and predictive accuracy. Most importantly, our nomograms integrating significant clinicopathologic characteristics, treatment modality and nutritional status can be used to predict personalized estimates of OS and PFS, which can help stratify HNSTS patients precisely and provide discussions of prognosis.
As the NRI is a valid tool for the assessment of the nutritional status of patients, our study demonstrated that NRI is an independent prognostic factor in patients with HNSTS, which is consistent with several previous studies in liver cancer [46], breast cancer [47], gastric cancer [48] and esophageal squamous cell cancer [49]. While the conclusions of these studies were similar with respect to NRI, they differed slightly in the cut-off value. The possible reasons for the difference in cut-off value are the unique biological characteristics of different malignant tumors and different statistical methods for calculating the optimal cut-off point value. The findings of this study suggest that the comprehensive nutritional assessment with NRI may be essential for the individualized management of HNSTS patients. Therefore, in addition to the routine treatment, we should pay more attention to the levels of NRI for HNSTS patients.
There are some limitations in this study. First, due to the rarity of HNSTS, the study was conducted retrospectively with a limited number of patients from a single center, making it susceptible to the inherent biases of such a study format, although our study represents one of the largest cohorts of patients with HNSTS. Missing data and loss to follow-up may have introduced inherent biases, which is a limitation of our study. Therefore, future prospective multicenter studies are required to corroborate our findings. Second, numerous predictive biomarkers have been validated in many malignancies, such as the controlling nutritional status (CONUT) score [50], Glasgow prognostic score (GPS) [51], and prognostic nutritional index (PNI) [52], which could be further evaluated in the future study. Third, because of the histopathologic diversity, especially for the diverse heterogeneous subtypes of HNSTS, our cohort did not allow us to stratify the analysis of histopathologic subtypes. Nevertheless, our study enhances the clinical utility of NRI in rare malignancies.

Conclusions
In conclusion, NRI is an independent predictor of OS and PFS in HNSTS patients, which could be utilized for predicting the prognosis of patients with HNSTS in daily clinical practice. Moreover, the validated nomograms provide useful predictions of OS and PFS for patients with HNSTS, guiding clinicians to assess prognosis and develop a personalized treatment strategy.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu15030641/s1, Figure S1: Flowchart of the patient selection process. HNSTS, head, and neck soft tissue sarcoma; SYSUCC, Sun Yat-sen University Cancer Center; Figure S2: The maximum of the standardized log-rank statistics for NRI cut-off value; Figure S3: Calibration plot of OS (A) and PFS (B) at 3-and 5-year in the validation cohort; Figure S4  Institutional Review Board Statement: The present retrospective study was approved by the Institutional Review Board of our institution (No. B2021-314-01), and the need for informed consent was waived by the ethics committee.
Informed Consent Statement: Patient consent was waived due to this being a retrospective study using medical records/biological specimens obtained in previous clinical practices.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

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