Dynamic Variation and Prognostic Value of Weight Loss During Radiotherapy in Nasopharyngeal Carcinoma: A Large-Scale Cohort Study

Background: We aim to investigate the dynamic variation and prognostic value of weight loss among patients with nasopharyngeal carcinoma (NPC). Methods: A total of 1149 newly diagnosed NPC patients who received radical radiotherapy (RT) were retrospectively analyzed. Patients’ weights were measured at initiation of RT and every week during RT. Recursive partitioning analyses (RPAs) were utilized to determine cut-off value for rate of weight loss (RWL). Disease-free survival (DFS) was our primary endpoint. Secondary endpoints included locoregional relapse-free survival (LRRFS), distant metastasis-free survival (DMFS), and overall survival (OS). Results: RWLs were 0%, 0%, -1.54%, -2.86%, -4.11%, -5.98%, -6.56% at 1 st , 2 nd , 3 rd , 4 th , 5 th , 6 th , and 7 th week of RT, respectively. RWL optimal threshold with respect to DFS was -5.3% based on RPAs. Thus, a consistent threshold of -5% (> -5% versus ≤ -5%) was selected to classify NPC patients into low RWL and high RWL groups for survival analysis. Compared to high RWL ( ≤ -5%), patients with low RWL (> -5%) had signicantly better ten-year DFS (78.6% versus 61.2%; P< 0.001), OS (86.6% versus 70.1%; P< 0.001), and DMFS (88.5% versus 80.2%; P = 0.007). However, no difference was observed between LRRFS groups (91.7% versus 94.3%; P= 0.173). In multivariate analysis, high RWL was an independent risk factor for DFS (HR, 1.56; 95% CI, 1.19-2.03; P= 0.001), OS (HR, 1.54; 95% CI, 1.11-2.15; P= 0.011) and DMFS (HR, 1.47; NPC: Nasopharyngeal carcinoma; RT: Radiotherapy; IMRT: RWL: Rate of weight loss; Intensity-modulated radiation therapy; MRI: Magnetic resonance imaging; CT: Computed tomography; ECT: Whole body bone scan; PET-CT: 18F-uorodeoxyglucose positron emission tomography and CT; EBV: Epstein-Barr virus; PTV: Planning target volume; GTVnx: Nasopharyngeal gross tumor volume; GTVnd: GTV of the metastatic lymph nodes; CTV1: High-risk clinical target volume; CTV2: Low-risk clinical target volume; IC: Induction chemotherapy; CCRT: Concurrent chemoradiotherapy; LDH: Lactate dehydrogenase; hs-CRP: High sensitivity C-reactive protein; WPre-RT: Bodyweight before RT; DFS: Disease-free survival; DMFS: Distant metastasis free survival; LRRFS: Locoregional relapse free survival; OS: Overall survival; RPAs: Recursive partitioning analyses;


Background
Nasopharyngeal carcinoma (NPC), an epithelial malignancy that is distinguished from other head and neck cancers, is highly prevalent in southern China [1]. The main treatment for NPC is radiotherapy (RT) due to anatomical restrictions and radio-sensitivity. Over the past decade, advances in imagining techniques, chemotherapy, and radiation technology contributed to the improved NPC survival. However, 20-30% of patients still die as a result of NPC recurrence [2][3]. Therefore, efforts to identify modi able risk factors can potentially provide new insights on developing clinical intervention for increasing longterm survival.
Patients diagnosed with head and neck cancers often experience weight loss during treatment due to acute toxicity, such as mucositis, dysgeusia, xerostomia, and nausea [4][5][6][7]. Previous studies [8,9] have estimated weight loss to range from 40-90%, especially among NPC patients where rates are high.
Substantial weight loss during treatment was associated with poor survival in NPC [10][11][12][13]. Monitoring the decreasing weight during RT will allow for clinicians to evaluate the current treatment plan effectiveness for NPC [14]. Moreover, knowing the dynamic variation of patients' weight loss during treatment is helpful in selecting the optimal time for nutritional intervention and altering RT treatment.
However, bodyweight was only obtained at baseline visit and again at the end of treatment in previous studies [10][11][12], without considering the dynamic weight loss during treatment.
To ll the current gaps in knowledge and limitations of previous studies, we conducted a large-scale retrospective study of NPC patients treated with radical RT. The present study sought to (1) draw a dynamic map of weight variation during RT; (2) identify the weight loss prognostic value on survival outcomes; (3) demonstrate risk factors for weight loss; and (4) provide insight on individualized nutritional intervention and the timing of RT replanning for NPC patients.

Patient characteristics
The present study was a retrospective cohort study utilizing in-patient medical records from Sun Yat-Sen University Cancer Center consisting of rst diagnosis of histologically con rmed, non-disseminated NPC from January 2006 to October 2014. We included patients if they met the following criteria: (1) newly diagnosed non-disseminated NPC; (2) Karnosfky performance score ≥ 70; (3) no indication of distant metastases; (4) absent of secondary malignancy; (5) treated with radical intensity-modulated radiation therapy (IMRT); and (6) weekly assessment of weight. This study was conducted in compliance with institutional policy to protect patients' private information, and was approved by the Institutional Review Board of our center.
We included 1149 patients, and the baseline assessment included full physical examination, beroptic nasopharyngoscopy, neck and nasopharyngeal magnetic resonance imaging (MRI), computed tomography (CT), abdominal ultrasonography, biochemistry pro ling and hematology, whole body bone scan (ECT) or 18 F-uorodeoxyglucose positron emission tomography and CT (PET-CT). Real-time quantitative polymerase chain reaction was used to measure Epstein-Barr virus (EBV) DNA concentrations as previously described in detail [15]. Patients were staged based on the 7th edition of the American Joint Commission on Cancer staging system [16].

Data collection
Patients' age, height, weight, sex, pre-therapy laboratory counts of serum lactate dehydrogenase (LDH), high sensitivity C-reactive protein (hs-CRP), plasma EBV DNA, pathological types, clinical stage, and treatment type were extracted from medical records. Digital electronic scale (XiangShan, EB9871) was used to measure bodyweight to the nearest 0.1 kg in light garment and without shoes. We measured patients' bodyweight at initiation of RT and every week during RT. Bodyweight before RT (W Pre−RT ) was measured at initiation of RT, and W RT1,2,3,4,5,6,7 (body weight at 1st, 2nd, 3rd, 4th, 5th, 6th, and 7th week of RT) was measured at each week of RT. RWL 1,2,3,4,5,6,7 was calculated using the following equation: (W RT1,2,3,4,5,6,7 -W Pre−RT )/W Pre−RT × 100%. At time of study, all patients were on 100% oral intake, where no type of enteral feeding tube or total parental nutrition were used.

Follow-up and endpoints
Patients were examined every three months during the rst two years, and every six months for years three through ve, and annually thereafter until death. Disease-free survival (DFS) was our primary endpoint, de ned as time from diagnosis to documented recurrence of disease (either distant metastasis or locoregional disease recurrence) or mortality from any cause, whichever occurred rst. Secondary endpoints consisted of (1) distant metastasis free survival (DMFS) (no documented distant metastasis); (2) locoregional relapse free survival (LRRFS) (no documented locoregional recurrence); and (3) overall survival (OS).

Statistical methods
In this study covariates included host factors (e.g. age, gender, tobacco smoking status, hs-CRP, LDH, and plasma EBV DNA), treatment factors (e.g. treatment modality), and tumor factors (e.g. histology, T stage, and N stage). We classi ed categorical variables based on clinical ndings. Determined by routine cutoff points and ndings from prior studies, continuous variables were converted to categorical variables [17][18][19]. χ 2 test or Fisher's Exact test were used to compare clinicopathologic characteristics groups.
Kaplan-Meier method was used to calculate actuarial rates, where we compared differences by the logrank test. Optimal RWL threshold was identi ed by performing recursive partitioning analyses (RPAs) for the cohort. Univariable with a P < 0.05 was used to include covariates in the multivariable model.
Multivariate Cox proportional hazards model was applied to evaluate the impact of RWL on survival outcomes. Statistical tests were two-tailed and P < 0.05 was considered statistically signi cant. All statistical analysis and generate gures were performed by the rms package in R version 3.3.2 (http://www.r-project.org/).

Patient characteristics
Clinicopathological characteristics of the 1149 patients are shown in Table 1. Among all patients, 892 (77.6%) were male and majority (1058, 92%) of patients experienced weight loss during RT. The percentage of patients at stage I, II, III, and IVA-B were 2.1%, 9.9%, 53.7%, and 34.3%, respectively.

Dynamic variation of body weight during RT
The dynamic variation of weight loss during RT was analyzed. We observed that RWL remained largely unchanged in the rst two weeks of RT, and then began to drop at a relatively stable rate during weeks three to ve of RT (from-1.54% to -4.11%; at a rate of -1.3% weight loss per week). At the 6th week of RT, the RWL dropped faster than the prior week (from − 4.11% to -5.98%; nearly 2% weight loss in this week). RWL then slowed in the last week of RT (from − 5.98% to -6.56%; approximately 0.5% weight loss in this week). The median weight loss during RT was − 4.0 kg (IQR, − 2.0 to -6.0 kg) and the percentage for median weight loss was − 6.7% (IQR, − 3.5 to -9.7%).  Abbreviations: HR, hazard ratio; 95% CI, 95% con dence interval; LDH, lactate dehydrogenase; hs-CRP, high sensitivity C-reactive protein; RT, radiotherapy; CCRT, concurrent chemoradiotherapy; IC + CCRT, induction chemotherapy plus concurrent chemoradiotherapy.

Correlation between RWL and clinicopathological characteristics
a P values were calculated using an adjusted Cox proportional hazards model. b All variables were measured before treatment.

Discussion
To the best of our knowledge, this study is the longest follow-up analysis for dynamic variation of weight loss during RT. Results of this present study revealed body weight remained largely unchanged during RT for the rst two weeks, and dropped fastest at the sixth week of RT. Further analysis revealed that weight loss ≥ 5% during RT was associated with signi cantly inferior ten-year DFS, OS, and DMFS for NPC.
Additionally, older age and chemotherapy were predictive of greater weight loss for NPC.
Numerous studies have demonstrated that weight loss is correlated with poor prognosis among individuals diagnosed with head and neck cancer [20,21], especially among NPC patients [10,11].
Results from the present study aligned closely with prior studies. There are several potential reasons for these ndings. First, critical weight loss may result in loosening of posture xation, inaccurate radiation eld, and signi cant dosimetric change during IMRT [22]. Second, weight loss potentially causes reduction in treatment tolerance and radiotherapy breaks, thus in uencing therapeutic e cacy [23,24].
Third, weight loss is often used as a tool for assessment of newly developed malnutrition, which contributes to weakness in immunity defense mechanism, such as phagocyte function, anatomic barriers, cellular and humoral immunity. Therefore, increasing infection susceptibility and reduced response to malignancy [25,26].
Previous studies reported that weight loss ≥ 5% during RT was associated with poor survival [12,23]. Our results showed that RWL ≤ -5% was associated with poor DFS, OS, and DMFS, aligning with prior studies. However, Du et al. [11] assessed weight loss during the entire treatment procedure and observed weight loss ≥ 10% was an indicator for likelihood of metastasis and survival. One reason underlying this inconsistency cutoff value of RWL may partly be due to potential selection bias in the study population.
Collectively, the impact of weight loss on prognosis of NPC can be determined in the present study.
Previous literatures [27,28] reported approximately 90% local control rates for NPC in the IMRT era. Due to recent advances in RT technology, no observed associated differences in LRRFS between patients in high and low RWL groups were found in the present study. These nding are reasonable as excellent locoregional control is expected for IMRT, therefore actual impact of weight loss on LRRFS would be limited.
Numerous factors may in uence weight loss among cancer patients. We observed that patients treated with chemotherapy suffered more weight loss, similar with ndings by Qiu et al. [9] and Du et al. [11].
Although encouraging results attained by multimodal therapy for NPC, acute toxicities are more likely to occur during aggressive chemoradiotherapy [9], including severe oral mucositis, nausea, and vomiting.
Additionally, we observed that patients aged > 60 years had a greater likelihood of suffering higher weight loss. We speculate that elderly patients' reduced organ reserve, multiple comorbidities, and poor treatment tolerance explain the reason for greater weight loss. Other risk factors including radiation technique, segmentation model, and prescribed dose are partly relevant to oral mucositis and weight loss. Because this research adopts the uni ed radiation technique, segmentation model, and prescribed dose, we did not include the above factors for analysis at last.
Since weight loss is common among NPC patients, it is necessary to assess weight change over RT time.
This is due to potentially providing a more complete understanding on the relationship between bodyweight and survival among NPC patients. However, prior researches primarily evaluated weight loss at a single time point, usually pre-or post-treatment. For this reason, limited knowledge exists about the dynamic change of weight loss during RT. Our results indicated that bodyweight remained generally unchanged in the rst two weeks of RT, and then began to drop relatively stable the next three weeks, though fastest in the sixth week of RT. The following reasons may explain the observed results. First, the oral mucous membrane reaction of patients is mild and diet is less affected in the rst two weeks of RT.
With the increasing number of RT, weight loss is gradually accelerated due to oral mucositis, aggravated swallowing pain, and decreased treatment tolerance [4].
There are some limitations should be noted. First, we lacked detailed information on dietary habit and food intake, as well as data on nutritional status was unavailable for further analysis of weight loss.
However, during the study period, no standard criteria for nutritional support in patients undergoing RT has been established. Second, the data used in this study derived from only one institution in an endemic area, where a large proportion of physicians have expertise in diagnosing and treating of NPC. Future studies that incorporate external validation is needed.

Conclusions
In conclusion, our ndings revealed weight loss decreased fastest during the sixth week of RT. The optimal threshold for RWL adversely impacting NPC prognosis was − 5%. More efforts are needed to be given towards limiting weight loss during RT under 5% in clinical practice as a result of the detrimental impact of RWL on survival outcomes. In addition, use of chemotherapy and old age were attributed to the development of high weight loss. These ndings would be helpful in selecting optimal time for nutritional intervention and RT replanning.
(C2018063). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Availability of data and materials
The datasets analysed during the current study are available in the Research Data Deposit (RDD) public platform www.researchdata.org.cn, with the approval RDD number of RDDA2019001296. If someone need to access the data, he/she should obtain our consent, and have to explain the source of the data in their study.

Authors' contributions
Data collection and writing original draft was performed by JYN, XTL, and MDM. Formal analysis, reviewing and editing was done by YJJ and JC. JYN and XTL participated in project administration and study design. Study design was conceived and designed by XLP and HWZ. Prior to submission of this manuscript, all authors have reviewed and approved.

Ethics approval and consent to participate
This study was conducted in compliance with institutional policy to protect patients' private information, and was approved by the Institutional Review Board of Sun Yat-sen University Cancer Center. As the current study was a retrospective assessment of routine data, the ethics committee of our Cancer Center waived the need for individual informed consent.

Consent for publication
Not applicable.