Impact of the peripheral blood inflammatory indices and modified nomogram-revised risk index on survival of Extranodal Nasal-Type Natural Killer/T-Cell lymphoma

BACKGROUND: At present, peripheral blood markers are easily accessible information and clinically valuable prognostic indicators in extranodal nasal-type natural killer/T-cell lymphoma (ENKTCL). Nevertheless, the role of its comprehensive score in ENKTCL remains to be determined. OBJECTIVE: Therefore, this study aimed to investigate the prognostic effect of the peripheral inflammation score on ENKTCL. METHODS: The retrospective study included 183 patients with ENKTCL. Univariate Cox regression analyses and least absolute shrinkage and selection operator (LASSO) Cox regression were used to construct the inflammation-related prognostic index named Risk. Univariate and multivariate Cox regression analyses and regression adjustment with propensity score matching (PSM) were used to evaluate the prognostic ability of risk. The performance of the modified nomogram-revised risk index (NRI) by integrating risk was evaluated with the area under the time-dependent receiver operating characteristic (ROC) curve (AUC), decision curve analysis (DCA), and integrated Brier score (IBS). RESULTS: The risk cut-off value, constructed by the lymphocyte count, platelet count, albumin level, LMR, and PNI, was -1.3486. Before PSM, multivariate analysis showed that risk was significantly associated with OS (HR = 2.577, 95% CI = 1.614–4.114, P< 0.001) and PFS (HR = 2.679, 95% CI = 1.744–4.114, P< 0.001). After PSM adjustment, risk was still an independent factor for OS (HR = 2.829, 95% CI = 1.601–5.001, P< 0.001) and PFS (HR = 2.877, 95% CI = 1.735–4.770, P< 0.001). With the NRI, the modified NRI by integrating risk increased the AUC and clinical net benefit and decreased the IBS. CONCLUSIONS: Risk is an easily accessible and inexpensive indicator that may be used as a prognostic marker and could improve NRI predictive power in patients with ENKTCL.


Introduction
Extranodal Nasal-Type Natural Killer/T-Cell Lymphoma (ENKTCL) is an aggressive extranodal lymphoma of natural killer (NK)-cell or T-cell lineage lymphoma characterized by vascular damage and destruction, prominent necrosis, a cytotoxic phenotype, and is associated with Epstein-Barr virus [1].ENK-TCL is more prevalent among Asian, Mexican, Central American, and South American indigenous populations.ENKTCL is more common in males than females, with a median age of 44-54 years at diagnosis [2].
The most common prognostic scores of ENKTCL include the Korean Prognostic Index (KPI), the International Prognostic Index (IPI), and the prognostic index of natural killer lymphoma (PINK) [3][4][5][6].However, these prognostic scores still have some limitations in clinical application.They cannot always accurately predict the prognosis, cannot help doctors customize the initial therapy, and cannot identify risk groups that are beneficial to prognosis in early patients [7].Yang et al. [8] developed and validated an easily used nomogram-revised risk index (NRI) that stratified patients based on age, Ann Arbor stage, Eastern Cooperative Oncology Group (ECOG) scores, lactate dehydrogenase (LDH) levels, and primary tumour invasion (PTI).The NRI can evaluate the outcome more accurately than the KPI, IPI, and PINK.The risk-adaptive therapy proposed based on its conclusions can effectively identify patients who will benefit from radiotherapy followed by chemotherapy [7][8][9][10].
However, the risk factors for NRI did not involve inflammatory biomarkers.Previous studies have shown that inflammation is closely related to the development and prognosis of malignant tumours and has been well validated in central nervous system lymphoma, small cell lung carcinoma, pancreatic cancer, and others [9][10][11].Albumin, monocytes, lymphocytes, platelets, neutrophil-to-lymphocyte ratio (NLR), derived neutrophil-to-lymphocyte ratio (dNLR), plateletto-lymphocyte ratio (PLR), lymphocyte-to-monocyte ratio (LMR), and prognostic nutritional index (PNI) have been reported to be associated with the prognosis of ENKTCL [12][13][14][15].Although these inflammatory biomarkers are essential, they provide limited predictive performance when evaluated separately.
This study aimed to evaluate the association between inflammatory biomarkers and ENKTCL patient prognosis and construct a peripheral blood inflammatory index named Risk.Furthermore, we assessed an additional improvement of combined Risk to NRI prediction performance.

Patients
A total of 183 patients with a pathological diagnosis of ENKTCL admitted to Shanxi Provincial Cancer Hospital from January 2002 to December 2018 were retrospectively reviewed.The inclusion criteria were the same as those described in our previous study [11].This study was approved by the Ethics Committee of Shanxi Province Cancer Hospital, Shanxi Hospital Affiliated to Cancer Hospital, Chinese Academy of Medical Sciences, and Cancer Hospital Affiliated to Shanxi Medical University.The Review Committee exempted informed consent requirements.

Laboratory measurement
The laboratory indicators and their derived indicators included albumin, monocyte, lymphocyte, platelet, NLR, dNLR, PLR, LMR, and PNI within one week before treatment.The compound prognostic scores were calculated by Eq. (1).

Statistics
Overall survival (OS) was defined as the time from the date of diagnosis to the date of any cause of death or final follow-up.Progression-free survival (PFS) was measured from the date of diagnosis to the date of first recurrence, progression, death, or last follow-up.Maximally selected rank statistics determine the optimal cutoff value of continuous variables.Variables with a P value < 0.05 in univariate Cox regression analysis were included in the least absolute shrinkage and selection operator (LASSO) Cox regression analysis to construct Risk [16].Chi-square or Fisher's exact test was used for categorical variables.Propensity score matching (PSM) was performed to adjust the differences between the low-and high-risk groups.Univariate and AIC-based multivariate Cox regression analyses were used to evaluate the prognostic ability of Risk before and after PSM.Kaplan-Meier and log-rank tests were used to plot survival curves and detect the survival differences in strat-ified risk before and after PSM.The area under the time-dependent receiver operating characteristic (ROC) curve (AUC), decision curve analysis (DCA), and integrated Brier score (IBS) were used to evaluate the performance of the modified NRI, which combined Risk.The overall reclassification improvement was evaluated by the integrated discrimination improvement (IDI), continuous-net reclassification improvement (NRI), and median improvement in risk score.

Patient characteristics
The clinical characteristics of all patients are summarized in Table 1.The median age of the patients was 46 years (range 9-81 years).The number of patients receiving radiotherapy and L-asparaginase (L-ASP) was 61 and 83, respectively.The median OS was 137 months, and 86 patients (47.0%) died during the follow-up period.The 3-year and 5-year OS rates were 56.7% and 51.7%, respectively.The 3-year and 5-year PFS were 50.6% and 43.3%, respectively.

Risk construction and survival analysis
The cut-off values of inflammatory biomarkers are listed in Table S1.Higher PNI (P < 0.001), LMR (P = 0.002), ALB (P < 0.001), LYM (P = 0.003), and PLT (P = 0.027) were associated with better OS.Inflammatory biomarkers with P < 0.05 were included in the LASSO Cox regression analysis to construct the risk score.The LASSO Cox regression model (Eq.( 2)) was used to calculate each patient's risk score.According to the optimal cut-off value for the risk score (−1.3747), all patients were divided into high-and lowrisk groups.
Before PSM, the 5-year PFS and OS rates in the high-risk group were 24.0% vs. 33.1%,respectively, whereas those in the low-risk group were 58.6% vs. 66.5%, respectively (Fig. 1).The baseline characteristics distribution between high risk and low risk has significant differences.To better control the bias caused by retrospective clinical data, the PSM method was applied to adjust the differences between variables so that all characteristics were well-balanced (Table 2).After PSM, the 5-year PFS and OS rates in the high-risk group were 29.2% vs. 40.8%,respectively, whereas those in the low-risk group were 62.1% vs. 72.4%,respectively (Fig. 1).Then, we used univariate and multivariate analyses to evaluate the prognostic value of the included factors.The results are shown in Table 2. Before PSM, multivariate analysis showed that Risk was significantly associated with OS (HR = 2.577, 95% CI = 1.614-4.114,P < 0.001) and PFS (HR = 2.679, 95% CI = 1.744-4.114,P < 0.001).After PSM adjustment, Risk was still an independent factor for OS (HR = 2.829, 95% CI = 1.601-5.001,P < 0.001) and PFS (HR = 2.877, 95% CI = 1.735-4.770,P < 0.001) (Table 3).

Subgroup analysis
Before PSM, among 122 patients who received radiotherapy, the 5-year OS rates of the high-and lowrisk groups were 42.4% and 69.6%, respectively (P = 0.002), and the 5-year PFS rates of the high-and lowrisk groups were 29.6% and 65.0%, respectively (P < 0.001).Among the 83 patients who received L-ASPbased chemotherapy, the 5-year OS rates of the highand low-risk groups were 47.5% and 72.5%, respectively (P = 0.016), and the 5-year PFS rates of the highand low-risk groups were 36.8% and 63.5%, respectively (P = 0.004).After PSM adjustment, the 5-year OS rates of the high-and low-risk groups who received radiotherapy were 45.5% and 82.3%, respectively (P = 0.001), and the 5-year PFS rates of the high-and lowrisk groups were 31.8% and 76.8%, respectively (P < 0.001).The 5-year OS rates of the high-and low-risk groups who received L-Asp-based chemotherapy were 52.1% and 79.6%, respectively (P = 0.016), and the 5-year PFS rates of the high-and low-risk groups were 39.8% and 71.3% (P = 0.010) (Fig. 2).Except for subgroups with a small number of patients, such as females, extranodal sites greater than 2, non-UADT, RT alone, and gemcitabine, other subgroup analyses showed OS and PFS differences between high-risk and low-risk groups (Fig. S1).UADT, upper aerodigestive tract NK/T-cell lymphoma; LN, Regional lymph node involvement; RT, radiotherapy; L-Asp, L-asparaginase; CT, chemotherapy.

Modified NRI by integrating risk
At 12, 24, 36, 48, and 60 months, the revised NRI predicted AUCs for OS of 0.707, 0.720, 0.760, 0.760, and 0.800, respectively, while the NRI predicted AUCs for OS of 0.671, 0.684, 0.735, 0.736, and 0.784, respectively (Fig. 3).The 5-fold cross-validation also showed that the AUC of the modified NRI in predicting the prognosis of ENKTCL at 0-60 months was higher than that of the NRI (Fig. S2).The decision curve showed that when predicting 5-year OS, the NRI had more clinical net benefit than "none" or "all" in the threshold probability from 0.246 to 0.801.In contrast, the modified NRI had a higher clinical net benefit in the threshold probability from 0.203 to 0.876 (Fig. 4).The IBSs in the NRI were 0.100, 0.139, 0.162, 0.173, and 0.179, respectively, and 0.096, 0.133, 0.154, 0.164, and 0.169 in the modified NRI, respectively (Fig. S3).The integrated discrimination improvement (IDI), continuous-net reclassification improvement (NRI), and median improvement in risk score were used to assess additional prog-nostic discrimination of the modified NRI (Table S2), and significant predictive performance improvements were observed compared with the NRI within 12, 24, 36, 48 and 60 months.

Discussion
ENKTCL is usually characterized by necrosis and inflammation.However, little is known about the effects of systemic inflammatory markers in ENKTCL.We used univariate Cox regression and LASSO Cox to integrate independent prognostic inflammatory markers proven by previous studies and constructed a comprehensive inflammatory index named Risk.We found that Risk was a prognostic factor for PFS and OS in ENK-TCL patients.Moreover, the combination of Risk and NRI improves the performance of the NRI.
Systemic inflammation and malnutrition are important components of cancer [17,18].Malnutrition is responsible for nearly one-fifth of patients dying [19].In  addition to liver function, albumin is also one of the indicators of nutritional status and has been proven to be associated with the prognosis of various malignant tumours.Li et al. [20] found that albumin was independent of PFS (RR = 1.308, 95% CI: 1.037-1.649,P = 0.023) and OS (RR = 1.64, 95% CI: 1.284-2.094,P < 0.001).In this study, albumin was associated with OS and PFS, and patients with lower albumin had a poor prognosis.Lymphocytes can strengthen the immune surveillance ability and suppress the proliferation, invasion, and metastasis of tumour cells and have been used to predict the prognosis of patients with many types of tumours, such as oropharyngeal cancer, melanoma, and skin cancer [21][22][23][24].Serum monocyte levels can reflect the formation of tumour-associated macrophages (TAMs) in the tumour microenvironment [25].TAMs can promote tumour initiation and growth by stimulating proliferation and angiogenesis, such as in primary hepatocellular cancer, gastric cancer, and neuroblastoma [26][27][28].Lymphocytes and monocytes calculate the LMR ratio.As an inflammation-and immunityrelated biomarker, it is an important prognostic marker for several tumours.Decreased LMR is considered to be closely related to Epstein-Barr virus DNA positivity, and the infection is closely related to the occurrence and development of ENKTCL [14,29,30].Previous studies have found that decreased platelets are an adverse in-dependent factor in diffuse large B-cell lymphoma, peripheral T-cell lymphoma, and ENKTCL [31][32][33][34].Our study also found that patients with low PLT levels had poor prognoses.However, how platelets affect the prognosis of ENKTCL patients is still unclear.Researchers have suggested that patients with low platelet levels in blood system diseases may have a higher tumour burden [34].
In our study, the Risk based on six inflammatory biomarkers effectively stratified OS and PFS in ENK-TCL patients.High-risk patients benefit less from L-Asp-based chemotherapy and radiotherapy.Moreover, even if age, sex, extranodal site of involvement, ECOG score, Ann Arbor Staging System, B symptoms, regional lymph node involvement, primary site, and radiotherapy were adjusted by PSM analysis and multivariate analysis, the OS and PFS of patients with high risk were worse than those of patients with low risk.Furthermore, after combining Risk into the NRI, the modified NRI performance was improved, which can be demonstrated by AUC, IBS, and clinical net benefit.In the modified NRI performance at 12, 24, 36, 48, and 60 months, the AUC increased by 0.036, 0.036, 0.025, 0.024, and 0.016, respectively.IBS decreased by 0.004, 0.004, 0.008, 0.009, and 0.01, respectively.DCA shows that the modified NRI has favourable clinical application across a wider range of threshold probabilities.Adding Risk to the NRI also improved IDI, continuous NRI, and median improvement in risk score.
Although this study showed that integrating Risk could significantly improve the predictive performance of the NRI, there are still some limitations.First, this was a single-centre retrospective study with a small sample size and a lack of external validation.Second, because C-reactive protein was not counted, Glasgow's prognostic score was not included in the analysis.Third, the optimal cut-off value of the inflammatory index was determined according to the maximally selected rank statistics method, which may be inconsistent between different centres.In conclusion, further multicentre large-scale studies are needed to confirm the conclusions of this study.

Conclusions
Risk is routinely available and inexpensive and can predict the prognosis of ENKTCL.Combining Risk and NRI can effectively improve the prediction performance.

Fig. 1 .
Fig. 1.The survival curves of OS and PFS for risk groups before (A, C) and after PSM (B, D).

Fig. 2 .
Fig. 2. The survival curves of OS and PFS for high-and low-risk patients who received L-asp before (A, B) and after PSM (E, F); The survival curves of OS and PFS for high-and low-risk patients who received RT before (C, D) and after PSM (G, H).

Table 2
Distribution of clinical characteristics of risk before and after PSM