Development and external validation of a prediction model for tube feeding dependency for at least four weeks during chemoradiotherapy for head and neck cancer

Open AccessPublished:November 24, 2021DOI:https://doi.org/10.1016/j.clnu.2021.11.019

      Summary

      Background & aims

      Patients who receive chemoradiotherapy or bioradiotherapy (CRT/BRT) for locally advanced head and neck squamous cell carcinoma (LAHNSCC) often experience high toxicity rates interfering with oral intake, causing tube feeding (TF) dependency. International guidelines recommend gastrostomy insertion when the expected use of TF exceeds 4 weeks. We aimed to develop and externally validate a prediction model to identify patients who need TF ≥ 4 weeks and would benefit from prophylactic gastrostomy insertion.

      Methods

      A retrospective multicenter cohort study was performed in four tertiary head and neck cancer centers in the Netherlands. The prediction model was developed using data from University Medical Center Utrecht and the Netherlands Cancer Institute and externally validated using data from Maastricht University Medical Center and Radboud University Medical Center. The primary endpoint was TF dependency ≥4 weeks initiated during CRT/BRT or within 30 days after CRT/BRT completion. Potential predictors were extracted from electronic health records and radiotherapy dose–volume parameters were calculated.

      Results

      The developmental and validation cohort included 409 and 334 patients respectively. Multivariable analysis showed predictive value for pretreatment weight change, texture modified diet at baseline, ECOG performance status, tumor site, N classification, mean radiation dose to the contralateral parotid gland and oral cavity. The area under the receiver operating characteristics curve for this model was 0.73 and after external validation 0.62. Positive and negative predictive value for a risk of 90% or higher for TF dependency ≥4 weeks were 81.8% and 42.3% respectively.

      Conclusions

      We developed and externally validated a prediction model to estimate TF-dependency ≥4 weeks in LAHNSCC patients treated with CRT/BRT. This model can be used to guide personalized decision-making on prophylactic gastrostomy insertion in clinical practice.

      Keywords

      1. Introduction

      Side effects of concurrent chemoradiotherapy or bioradiotherapy (CRT/BRT) often impair oral intake in patients with locally advanced (stage III/IV) head and neck squamous cell carcinoma (LAHNSCC), which may contribute to involuntary weight loss [
      • Jin S.
      • Lu Q.
      • Sun Y.
      • Xiao S.
      • Zheng B.
      • Pang D.
      • et al.
      Nutrition impact symptoms and weight loss in head and neck cancer during radiotherapy: a longitudinal study.
      ]. Weight loss has a detrimental effect on the risk of side effects, therapy tolerance, response rate, and survival [
      • Baptistella A.R.
      • Hilleshein K.D.
      • Beal C.
      • Brambatti J.S.
      • Caron R.
      • Baptistella S.F.
      • et al.
      Weight loss as a prognostic factor for recurrence and survival in oropharyngeal squamous cell carcinoma patients.
      ,
      • Cho Y.
      • Kim J.W.
      • Keum K.C.
      • Lee C.G.
      • Jeung H.C.
      • Lee I.J.
      Prognostic significance of sarcopenia with inflammation in patients with head and neck cancer who underwent definitive chemoradiotherapy.
      ,
      • Kwon M.
      • Kim R.B.
      • Roh J.L.
      • Lee S.W.
      • Kim S.B.
      • Choi S.H.
      • et al.
      Prevalence and clinical significance of cancer cachexia based on time from treatment in advanced-stage head and neck squamous cell carcinoma.
      ,
      • Matsuzuka T.
      • Kiyota N.
      • Mizusawa J.
      • Akimoto T.
      • Fujii M.
      • Hasegawa Y.
      • et al.
      Clinical impact of cachexia in unresectable locally advanced head and neck cancer: supplementary analysis of a phase II trial (JCOG0706-S2).
      ,
      • Meyer F.
      • Fortin A.
      • Wang C.S.
      • Liu G.
      • Bairati I.
      Predictors of severe acute and late toxicities in patients with localized head-and-neck cancer treated with radiation therapy.
      ]. In order to maintain sufficient nutritional intake, tube feeding (TF) has to be initiated in 37–74% of LAHNSCC patients undergoing CRT/BRT [
      • Brown T.E.
      • Getliffe V.
      • Banks M.D.
      • Hughes B.G.
      • Lin C.Y.
      • Kenny L.M.
      • et al.
      Validation of an updated evidence-based protocol for proactive gastrostomy tube insertion in patients with head and neck cancer.
      ,
      • Karsten R.T.
      • Stuiver M.M.
      • van der Molen L.
      • Navran A.
      • de Boer J.P.
      • Hilgers F.J.M.
      • et al.
      From reactive to proactive tube feeding during chemoradiotherapy for head and neck cancer: a clinical prediction model-based approach.
      ,
      • van der Linden N.C.
      • Kok A.
      • Leermakers-Vermeer M.J.
      • de Roos N.M.
      • de Bree R.
      • van Cruijsen H.
      • et al.
      Indicators for enteral nutrition use and prophylactic percutaneous endoscopic gastrostomy placement in patients with head and neck cancer undergoing chemoradiotherapy.
      ]. TF can be administered using a nasogastric tube (NGT) or a percutaneous gastrostomy, either placed radiologically (PRG) or endoscopically (PEG). The advantages of a gastrostomy compared to a NGT are increased physical mobility, less cosmetic disadvantage, and better quality of life. Patients fed via NGT experience more dislodgement and weight loss compared to patients with a gastrostomy tube [
      • Wang J.
      • Liu M.
      • Liu C.
      • Ye Y.
      • Huang G.
      Percutaneous endoscopic gastrostomy versus nasogastric tube feeding for patients with head and neck cancer: a systematic review.
      ].
      Previously, prophylactic gastrostomy insertion (before onset of side effects impairing oral intake) in all LAHNSCC patients undergoing CRT/BRT, used to be common in the majority of the clinical settings [
      • Beijer Y.J.
      • Koopman M.
      • Terhaard C.H.
      • Braunius W.W.
      • van Es R.J.
      • de Graeff A.
      Outcome and toxicity of radiotherapy combined with chemotherapy or cetuximab for head and neck cancer: our experience in one hundred and twenty-five patients.
      ,
      • Vrolijk L.
      • de Roij van Zuijdewijn C.L.M.
      • Slingerland M.
      • Wiggenraad R.G.J.
      • Verschuur H.P.
      • Jeurissen F.J.F.
      PEG tube dependency after prophylactic placement in 209 head and neck cancer patients treated with chemoradiotherapy or radiation with cetuximab.
      ,
      Nederlandse Werkgroep Hoofd-Hals Tumoren (NWHHT)
      Landelijke richtlijn larynxcarcioom, versie 3.0.
      ]. However, gastrostomy insertion is not a risk-free procedure; tube-related and infectious complications occur in 6–16% [
      • Strijbos D.
      • Keszthelyi D.
      • Bogie R.M.M.
      • Gilissen L.P.L.
      • Lacko M.
      • Hoeijmakers J.G.J.
      • et al.
      A systematic review and meta-analysis on outcomes and complications of percutaneous endoscopic versus radiologic gastrostomy for enteral feeding.
      ]. Therefore, new guidelines recommend that a prophylactic gastrostomy should only be inserted upon indication in LAHNSCC patients treated with CRT/BRT [
      • Leemans LES C.R.
      • Langendijk J.A.
      • de Boer J.P.
      • Terhaard C.H.J.
      • Roodenburg J.L.N.
      • Klomp F.W.J.
      • et al.
      Richtlijn hoofd-halstumoren.
      ]. It is generally agreed that when the expected use of TF exceeds four weeks, gastrostomy insertion should be considered [
      • Talwar B.
      • Donnelly R.
      • Skelly R.
      • Ye Y.
      • Huang G.
      Nutritional management in head and neck cancer: United Kingdom National Multidisciplinary Guidelines.
      ,
      • Arends J.
      • Bodoky G.
      • Bozzetti F.
      • Fearon K.
      • Muscaritoli M.
      • Selga G.
      • et al.
      ESPEN guidelines on enteral nutrition: non-surgical oncology.
      ,
      • Brown T.E.
      • Spurgin A.L.
      • Ross L.
      • Tripcony L.
      • Keller J.
      • Hughes B.G.
      • et al.
      Validated swallowing and nutrition guidelines for patients with head and neck cancer: identification of high-risk patients for proactive gastrostomy.
      ,
      • Garrow WPTJ J.S.
      • Ralph A.
      ,
      • Nunes G.
      • Fonseca J.
      • Barata A.T.
      • Dinis-Ribeiro M.
      • Pimentel-Nunes P.
      Nutritional support of cancer patients without oral feeding: how to select the most effective technique?.
      ]. Ideally, patients at risk of TF ≥ 4 weeks are identified prior to treatment, so they can be provided with a gastrostomy before the onset of side effects potentially complicating insertion, e.g. mucositis (painful insertion), neutropenia (infection risk), and ongoing weight loss (higher complication risk) [
      • Strijbos D.
      • Keszthelyi D.
      • Gilissen L.P.L.
      • Lacko M.
      • Hoeijmakers J.G.J.
      • van der Leij C.
      • et al.
      Percutaneous endoscopic versus radiologic gastrostomy for enteral feeding: a retrospective analysis on outcomes and complications.
      ].
      Until recently it remained challenging to predict for which patient prophylactic gastrostomy insertion would be appropriate. In a previously published study, we developed and internally validated a prediction model for calculating a patients’ individual probability of TF dependency ≥4weeks [
      • Willemsen A.C.H.
      • Kok A.
      • van Kuijk S.M.J.
      • Baijens L.W.J.
      • de Bree R.
      • Devriese L.A.
      • et al.
      Prediction model for tube feeding dependency during chemoradiotherapy for at least four weeks in head and neck cancer patients: a tool for prophylactic gastrostomy decision making.
      ]. New normal tissue complication probability (NTCP) models shed light on the potential additional value of RT doses on the pharyngeal constrictor muscles (PCM) and oral cavity (OC) in predicting swallowing outcomes [
      • Anderson Head M.D.
      Neck Cancer Symptom Working Group
      Beyond mean pharyngeal constrictor dose for beam path toxicity in non-target swallowing muscles: dose-volume correlates of chronic radiation-associated dysphagia (RAD) after oropharyngeal intensity modulated radiotherapy.
      ,
      • Wopken K.
      • Bijl H.P.
      • Langendijk J.A.
      Prognostic factors for tube feeding dependence after curative (chemo-) radiation in head and neck cancer: a systematic review of literature.
      ,
      • Christianen M.E.
      • Schilstra C.
      • Beetz I.
      • Muijs C.T.
      • Chouvalova O.
      • Burlage F.R.
      • et al.
      Predictive modelling for swallowing dysfunction after primary (chemo)radiation: results of a prospective observational study.
      ]. Therefore, we considered it worth investigating whether these RT parameters could increase the performance of the new model. The present study describes the development and external validation of a prediction model to identify patients at risk for TF dependency ≥4 weeks who would benefit from prophylactic gastrostomy insertion.

      2. Methods

      This study was conducted in accordance with the Declaration of Helsinki and approved by the institutional research ethics boards. We reported this study in accordance with Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) guidelines [
      • Collins G.S.
      • Reitsma J.B.
      • Altman D.G.
      • Moons K.G.
      Transparent reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement.
      ].

      2.1 Source of data

      The electronic health records of patients treated in four Dutch cancer centers were retrospectively reviewed to compile the development and validation dataset. For every center, data was collected by different independent researchers, in consultation with the executive researchers about the methods of data extraction and any uncertainties about the way of reporting.

      2.2 Populations

      The developmental dataset consisted of LAHNSCC patients treated between 2013 and 2016 in University Medical Center Utrecht (UMCU) and patients treated between 2014 and 2017 in Netherlands Cancer Institute (NCI). The external validation of the model was performed on data from patients treated between 2013 and 2016 in Maastricht University Medical Center + (MUMC+) and Radboud University Medical Center (RUMC).
      LAHNSCC patients were included when they were treated with primary or adjuvant concurrent CRT or BRT. Patients were excluded from the study in case of histology other than squamous cell carcinoma, esophageal tumor location, bilateral neck dissection with removal of submandibular glands (RT dose calculation on contralateral gland not possible), refusing TF despite the physician's strong recommendation, premature discontinuation of RT, switch to palliative treatment, or death during oncological treatment.
      Oncological treatment was previously described in detail [
      • Willemsen A.C.H.
      • Kok A.
      • van Kuijk S.M.J.
      • Baijens L.W.J.
      • de Bree R.
      • Devriese L.A.
      • et al.
      Prediction model for tube feeding dependency during chemoradiotherapy for at least four weeks in head and neck cancer patients: a tool for prophylactic gastrostomy decision making.
      ,
      • Driessen C.M.
      • Ham J.C.
      • Te Loo M.
      • van Meerten E.
      • van Lamoen M.
      • Hakobjan M.H.
      • et al.
      Genetic variants as predictive markers for ototoxicity and nephrotoxicity in patients with locally advanced head and neck cancer treated with cisplatin-containing chemoradiotherapy (the PRONE study).
      ,
      • Karsten R.T.
      • Al-Mamgani A.
      • Bril S.I.
      • Tjon A.J.S.
      • van der Molen L.
      • de Boer J.P.
      • et al.
      Sarcopenia, a strong determinant for prolonged feeding tube dependency after chemoradiotherapy for head and neck cancer.
      ]. In brief, patients treated with CRT received cisplatin (100 mg/m2 three weekly or 40 mg/m2 weekly) or carboplatin (1.5 AUC weekly) combined with RT. BRT treatment consisted of a loading dose of cetuximab (400 mg/m2), followed by a weekly dose of cetuximab (250 mg/m2) combined with RT. RT was given in 33–35 daily fractions of 2 Gy (CRT) or 30 to 34 fractions of 2 Gy (BRT). All patients were counseled by a dietitian.

      2.3 Outcome

      The primary endpoint of this study was the use of TF ≥ 4 weeks initiated during CRT/BRT or within 30 days after CRT/BRT completion. TF was initiated when oral nutritional intake was insufficient in meeting nutritional requirements according to the Dutch guideline on malnutrition [
      • Stuurgroep Ondervoeding
      ] as described earlier [
      • Willemsen A.C.H.
      • Kok A.
      • van Kuijk S.M.J.
      • Baijens L.W.J.
      • de Bree R.
      • Devriese L.A.
      • et al.
      Prediction model for tube feeding dependency during chemoradiotherapy for at least four weeks in head and neck cancer patients: a tool for prophylactic gastrostomy decision making.
      ].

      2.4 Predictors

      The potential predictors of TF dependency were based on existing literature and included: age [
      • Sachdev S.
      • Refaat T.
      • Bacchus I.D.
      • Sathiaseelan V.
      • Mittal B.B.
      Age most significant predictor of requiring enteral feeding in head-and-neck cancer patients.
      ], gender [
      • Mortensen H.R.
      • Overgaard J.
      • Jensen K.
      • Specht L.
      • Overgaard M.
      • Johansen J.
      • et al.
      Factors associated with acute and late dysphagia in the DAHANCA 6 & 7 randomized trial with accelerated radiotherapy for head and neck cancer.
      ,
      • Sanguineti G.
      • Rao N.
      • Gunn B.
      • Ricchetti F.
      • Fiorino C.
      Predictors of PEG dependence after IMRT+/-chemotherapy for oropharyngeal cancer.
      ], tobacco use [
      • Setton J.
      • Lee N.Y.
      • Riaz N.
      • Huang S.H.
      • Waldron J.
      • O’Sullivan B.
      • et al.
      A multi-institution pooled analysis of gastrostomy tube dependence in patients with oropharyngeal cancer treated with definitive intensity-modulated radiotherapy.
      ], alcohol use, Body Mass Index (BMI) at baseline [
      • Strom T.
      • Trotti A.M.
      • Kish J.
      • Rao N.G.
      • McCaffrey J.
      • Padhya T.A.
      • et al.
      Risk factors for percutaneous endoscopic gastrostomy tube placement during chemoradiotherapy for oropharyngeal cancer.
      ,
      • Wermker K.
      • Jung S.
      • Huppmeier L.
      • Joos U.
      • Kleinheinz J.
      Prediction model for early percutaneous endoscopic gastrostomy (PEG) in head and neck cancer treatment.
      ], pretreatment weight change [
      • Kanayama N.
      • Kierkels R.G.J.
      • van der Schaaf A.
      • Steenbakkers R.
      • Yoshioka Y.
      • Nishiyama K.
      • et al.
      External validation of a multifactorial normal tissue complication probability model for tube feeding dependence at 6months after definitive radiotherapy for head and neck cancer.
      ], texture modified diet at baseline (e.g. ground, minced or liquid) [
      • Mortensen H.R.
      • Overgaard J.
      • Jensen K.
      • Specht L.
      • Overgaard M.
      • Johansen J.
      • et al.
      Factors associated with acute and late dysphagia in the DAHANCA 6 & 7 randomized trial with accelerated radiotherapy for head and neck cancer.
      ], Eastern Cooperative Oncology Group performance status (ECOG PS) [
      • Matuschek C.
      • Bolke E.
      • Geigis C.
      • Kammers K.
      • Ganswindt U.
      • Scheckenbach K.
      • et al.
      Influence of dosimetric and clinical criteria on the requirement of artificial nutrition during radiotherapy of head and neck cancer patients.
      ], tumor site [
      • Mortensen H.R.
      • Overgaard J.
      • Jensen K.
      • Specht L.
      • Overgaard M.
      • Johansen J.
      • et al.
      Factors associated with acute and late dysphagia in the DAHANCA 6 & 7 randomized trial with accelerated radiotherapy for head and neck cancer.
      ,
      • Wermker K.
      • Jung S.
      • Huppmeier L.
      • Joos U.
      • Kleinheinz J.
      Prediction model for early percutaneous endoscopic gastrostomy (PEG) in head and neck cancer treatment.
      ], T classification [
      • van der Linden N.C.
      • Kok A.
      • Leermakers-Vermeer M.J.
      • de Roos N.M.
      • de Bree R.
      • van Cruijsen H.
      • et al.
      Indicators for enteral nutrition use and prophylactic percutaneous endoscopic gastrostomy placement in patients with head and neck cancer undergoing chemoradiotherapy.
      ,
      • Mortensen H.R.
      • Overgaard J.
      • Jensen K.
      • Specht L.
      • Overgaard M.
      • Johansen J.
      • et al.
      Factors associated with acute and late dysphagia in the DAHANCA 6 & 7 randomized trial with accelerated radiotherapy for head and neck cancer.
      ], N classification [
      • Mortensen H.R.
      • Overgaard J.
      • Jensen K.
      • Specht L.
      • Overgaard M.
      • Johansen J.
      • et al.
      Factors associated with acute and late dysphagia in the DAHANCA 6 & 7 randomized trial with accelerated radiotherapy for head and neck cancer.
      ,
      • Wermker K.
      • Jung S.
      • Huppmeier L.
      • Joos U.
      • Kleinheinz J.
      Prediction model for early percutaneous endoscopic gastrostomy (PEG) in head and neck cancer treatment.
      ] (AJCC 7th edition TNM staging system [
      • Edge S.B.
      • Compton C.C.
      The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM.
      ]), disease stage, p16 status [
      • Brown T.E.
      • Wittholz K.
      • Way M.
      • Banks M.D.
      • Hughes B.G.
      • Lin C.Y.
      • et al.
      Investigation of p16 status, chemotherapy regimen, and other nutrition markers for predicting gastrostomy in patients with head and neck cancer.
      ] (immunohistochemically as a surrogate marker for human papillomavirus (HPV), treatment setting (primary or adjuvant) [
      • Wermker K.
      • Jung S.
      • Huppmeier L.
      • Joos U.
      • Kleinheinz J.
      Prediction model for early percutaneous endoscopic gastrostomy (PEG) in head and neck cancer treatment.
      ], type of systemic therapy (platinum based or cetuximab) [
      • Setton J.
      • Lee N.Y.
      • Riaz N.
      • Huang S.H.
      • Waldron J.
      • O’Sullivan B.
      • et al.
      A multi-institution pooled analysis of gastrostomy tube dependence in patients with oropharyngeal cancer treated with definitive intensity-modulated radiotherapy.
      ] and neck irradiation (non or unilateral versus bilateral) [
      • van der Linden N.C.
      • Kok A.
      • Leermakers-Vermeer M.J.
      • de Roos N.M.
      • de Bree R.
      • van Cruijsen H.
      • et al.
      Indicators for enteral nutrition use and prophylactic percutaneous endoscopic gastrostomy placement in patients with head and neck cancer undergoing chemoradiotherapy.
      ]. The dosimetric parameters extracted from electronic health records were: mean RT dose (in Gy) to the contralateral submandibular and parotid gland, swallowing muscles (PCM), and oral cavity (OC). The contours for the PCM and the OC were not available in all cases in the radiation treatment planning system and were delineated for the purpose of this study. All organs at risk were contoured according to Brouwer et al. [
      • Brouwer C.L.
      • Steenbakkers R.J.
      • Bourhis J.
      • Budach W.
      • Grau C.
      • Gregoire V.
      • et al.
      CT-based delineation of organs at risk in the head and neck region: DAHANCA, EORTC, GORTEC, HKNPCSG, NCIC CTG, NCRI, NRG Oncology and TROG consensus guidelines.
      ] and added to the database.

      2.5 Sample size

      As a rule of thumb, at least ten events should be included for each candidate predictor to minimize the risk of overfitting [
      • Kubben P.
      • Dumontier M.
      • Dekker A.
      Fundamentals of clinical data science.
      ]. The least frequent outcome is defined as an event. In our study, receiving TF < 4 weeks was the least frequent outcome and was therefore defined as an event. For the external validation set, at least 100 events and 100 non-events are recommended [
      • Collins G.S.
      • Ogundimu E.O.
      • Altman D.G.
      Sample size considerations for the external validation of a multivariable prognostic model: a resampling study.
      ].

      2.6 Missing data

      Missing data were imputed using stochastic regression imputation with full conditional specification, while considering the following covariates: age, gender, tobacco use, alcohol use, BMI at baseline, pretreatment weight change, texture modified diet at baseline, ECOG PS, tumor site, T classification, N classification, disease stage, p16 status, treatment setting, systemic therapy, mean RT dose to the contralateral submandibular and parotid gland, mean RT dose to the PCM, mean RT dose to the OC, and TF ≥ 4 weeks. Values to be imputed were drawn using predictive mean matching.

      2.7 Statistical analysis methods

      All potential predictor variables underwent screening through univariable logistic regression. Factors with p < 0.30 were selected as potentially relevant predictor variables and were entered in a multivariable logistic regression model. Stepwise backward elimination was used to omit all predictors from the model that did not contribute substantially, using a p-value for selection of 0.10. Model performance was quantified as the model's ability to correctly discriminate between those who will and those who will not develop TF dependency ≥4 weeks using the area under the receiver operating characteristic curve (AUC).
      For external validation, we applied the model to our validation dataset. For evaluating the performance, the AUC was computed. The Hosmer and Lemeshow goodness-of-fit test was used to assess the agreement between predicted and observed probabilities. A significant p-value would denote significant deviation from a good model [
      • Hosmer D.W.
      • Lemeshow S.
      Applied logistic regression.
      ].
      All statistical analyses were performed using IBM SPSS Statistics for Windows, version 25 (IBM, Armonk, NY) [
      Computing RFfS: R
      A language and environment for statistical computing.
      ].

      3. Results

      3.1 Patient sample

      The development cohort consisted of 409 patients. The validation cohort included 334 patients. Characteristics of both datasets are displayed in Table 1. Of note is the difference between the cancer centers with regard to the tube insertion protocol: In both UMCU and MUMC + gastrostomies were placed prophylactically in the majority of patients, NCI placed reactive gastrostomies and the RUMC prefers insertion of a NGT, instead of a gastrostomy tube. Details on tube insertion and TF use per cancer center are shown in Supplemental Table 1.
      Table 1Frequency distribution of patient, tumor, and treatment characteristics of the developmental and validation cohort.
      Development cohort UMCU and NCI, n = 409
      Original data (not imputed) presented as mean ± SD for continuous variables or absolute n (%) for categorical variables.
      (%)
      Validation cohort MUMC+ and RUMC, n = 334
      Original data (not imputed) presented as mean ± SD for continuous variables or absolute n (%) for categorical variables.
      (%)
      p-value
      Patient characteristics
      Age (mean ± SD)60.2 ± 8.158.5 ± 8.10.003
      Male274 (67.0)222 (66.5)0.880
      Female135 (33.0)112 (33.5)
      History of tobacco use220 (53.8)292 (87.4)0.383
      No history of tobacco use39 (9.5)42 (12.6)
      Missing150 (36.7)0 (0.0)
      Alcohol consumption ≥1/day145 (35.5)196 (58.7)0.510
      No alcohol consumption114 (27.9)138 (41.3)
      Missing150 (36.7)0 (0.0)
      BMI at baseline (kg/m2) (mean ± SD)24.4 ± 4.624.9 ± 4.90.120
      Weight change baseline (%) (mean ± SD)−4.4 ± 7.0−2.9 ± 5.50.003
      No modified diet at baseline246 (60.1)230 (68.9)0.014
      Texture modified diet
      Texture modified diet includes ground, minced, liquid, or full tube feeding without oral intake.
      at baseline
      163 (39.9)104 (31.1)
      ECOG PS 0142 (34.7)85 (25.4)<0.001
      ECOG PS 1180 (44.0)224 (67.1)
      ECOG PS 232 (7.8)24 (7.2)
      ECOG PS 32 (0.5)1 (0.3)
      Missing53 (13.0)0 (0.0)
      Tumor characteristics
      Oral cavity85 (20.8)41 (12.3)<0.001
      Nasopharynx/sinus35 (8.6)29 (8.7)
      Oropharynx174 (42.5)156 (46.7)
      Hypopharynx56 (13.7)49 (14.7)
      Larynx29 (7.1)54 (16.2)
      Unknown primary13 (3.2)5 (1.5)
      Synchronous tumors9 (2.2)0 (0.0)
      Neck recurrence9 (2.0)0 (0.0)
      T classification (TNM)
       T020 (4.9)8 (2.4)0.233
       T132 (7.8)38 (11.4)
       T278 (19.1)64 (19.2)
       T3101 (24.7)83 (24.9)
       T4178 (43.5)141 (42.2)
      N classification(TNM)
       N069 (16.9)77 (23.1)0.106
       N153 (13.0)35 (10.5)
       N2269 (65.8)213 (63.8)
       N318 (4.4)9 (2.7)
      Disease stage
       Stage I0 (0.0)1 (0.3)
       Stage II12 (2.9)6 (1.8)
       Stage III47 (11.5)49 (14.7)
       Stage IV350 (85.6)278 (83.2)
      p16 expression in oropharynx only
       p16+74 (42.5)87 (55.8)0.017
       p16-92 (52.9)74 (47.4)
       Missing8 (4.6)5 (3.2)
      Treatment characteristics
      Primary treatment324 (79.2)291 (87.1.)0.005
      Adjuvant85 (20.8)43 (12.9)
      Systemic therapy
       Platinum-based313 (76.5)264 (79.0)0.413
       Cetuximab96 (23.5)70 (21.0)
      Neck irradiation
       Unilateral47 (11.5)22 (6.6)0.040
       Bilateral333 (81.4)308 (92.2)
       No neck RT29 (7.1)4 (1.2)
      Mean RT dose to contralateral
      submandibular gland (Gy) (mean ± SD)44.4 ± 17.446.6 ± 15.40.060
      Missing4 (1.0)0 (0.0)
      Mean RT dose to contralateral0.279
      parotid salivary gland (Gy) (mean ± SD)20.6 ± 9.921.3 ± 10.7
      Missing5 (1.2)0 (0.0)
      Mean RT dose to PCM (Gy) (mean ± SD)52.6 ± 15.053.1 ± 11.40.480
      Missing7 (1.8)0 (0.0)
      Mean RT dose to OC (Gy) (mean ± SD)42.6 ± 16.139.1 ± 16.30.010
      Missing6 (1.5)0 (0.0)
      Tube type
       Gastrostomy256 (62.6)132 (39.5)<0.001
       Nasogastric tube38 (9.3)86 (25.7)
       No feeding tube115 (28.1)116 (34.7)
       Missing0 (0.0)0 (0.0)
      Tube feeding use274 (67.0)200 (59.9)0.040
      No tube feeding use135 (33.0)134 (40.1)
      Tube feeding use ≥4 weeks261 (63.8)176 (52.7)0.003
      No tube feeding use ≥4 weeks148 (36.2)158 (47.3)
      Abbreviations: BMI, body mass index; OC, oral cavity; PCM, pharyngeal constrictor muscles; RT, radiotherapy; ECOG PS, Eastern Cooperative Oncology Group performance status; TNM-classification, tumor, node, metastasis classification according to the 7th edition; Gy, Gray. Bold values denote statistical significance at the level of p<0.05.
      1Independent samples t-test. 2Pearson’s chi-square test.
      a Original data (not imputed) presented as mean ± SD for continuous variables or absolute n (%) for categorical variables.
      b Texture modified diet includes ground, minced, liquid, or full tube feeding without oral intake.
      In the development cohort, 261 out of 409 patients (64%) required TF ≥ 4 weeks, whereas in the validation cohort, 176 out of 334 (53%) required TF ≥ 4 weeks, p = 0.003.
      In the development cohort, 36% (n = 148) remained on a total oral diet or used TF < 4 weeks. The risk of overfitting is minimized if no more than fourteen predictors are included in the model. Regarding the 36% without TF or TF < 4 weeks, we aimed to compile an external validation set of at least 278 subjects (100/36∗100%). With 158 patients (47%) receiving TF < 4 weeks and 176 patients (53%) receiving TF ≥ 4 weeks, our validation dataset meets the criteria of at least 100 events and 100 non-events.

      3.2 Model development

      Univariable regression analysis revealed p < 0.30 for the following variables in the development cohort: tobacco use, BMI at baseline, pretreatment weight change, texture modified diet at baseline, ECOG PS, tumor site, T classification, N classification, disease stage, p16 status, treatment setting, neck irradiation, mean RT dose to the contralateral submandibular and parotid gland, mean RT dose to the PCM, and mean RT dose to the OC (Table 2).
      Table 2Results of univariable logistic regression analysis of potential predictors for tube feeding for at least four weeks.
      ORCI-95%p value
      lowerupper
      Age (years)0.9880.9631.0130.341
      Male gender0.9470.6171.4520.801
      Tobacco use1.5230.7513.0910.244
      Alcohol consumption one or more per day0.9440.5541.6100.834
      BMI at baseline (kg/m2)0.9500.9090.9930.023
      Baseline weight change (%)0.9430.9110.9760.001
      Texture modified diet
      Texture modified diet includes ground, minced, liquid, or full tube feeding without oral intake.
      at baseline
      1.9811.2913.0400.002
      ECOG PS ≥ 12.1241.4003.223<0.001
      Oral cavity, oropharynx, and hypopharynx0.6890.4191.1330.143
      T classification ≥ T2 (TNM)1.4720.8172.6520.198
      N classification ≥ N2 (TNM)1.9841.2853.0620.002
      Disease Stage IV2.2051.2633.8490.005
      p16 + oropharynx0.6990.4241.1510.159
      Primary treatment setting0.7650.4691.2470.283
      Cetuximab0.9850.6121.5840.949
      Bilateral neck irradiation2.3151.3973.8370.001
      RT dose to contralateral submandibular glands (Gy)1.0221.0101.034<0.001
      RT dose to contralateral parotid glands (Gy)1.0461.0221.070<0.001
      RT dose to PCM (Gy)1.0271.0131.041<0.001
      RT dose to OC (Gy)1.0281.0151.041<0.001
      Abbreviations: BMI, body mass index; CI, confidence interval; Gy, Gray; OC, oral cavity; OR, Odds ratio; PCM, pharyngeal constrictor muscles; RT, radiotherapy; TNM-classification, tumor, node, metastasis classification according to the 7th edition; ECOG PS, Eastern Cooperative Oncology Group performance status. Bold values denote statistical significance at the level of p<0.05.
      a Texture modified diet includes ground, minced, liquid, or full tube feeding without oral intake.

      3.3 Model specification

      In the multivariable regression analysis tobacco use, BMI at baseline, T classification, disease stage, p16 status, treatment setting, neck irradiation, mean RT dose to the contralateral submandibular and PCM did not yield a p-value <0.10 and were therefore eliminated from the model. Pretreatment weight change, texture modified diet at baseline, ECOG PS, tumor site, N classification, mean RT dose to the contralateral parotid gland and OC were significant predictors of risk of TF use ≥4 weeks. Table 3 shows the regression coefficients for all predictors included in the final multivariable regression model.
      Table 3Regression coefficients in the model for predicting tube feeding use for at least four weeks.
      Regression coefficientsS.E.p-valueOR (95%CI)
      Model intercept−1.4190.001
      Pretreatment weight change(%)−0.0380.0200.0540.963 (0.926–1.001)
      Texture modified diet at baseline
       No modified diet (reference)0.4480.2470.0701.565 (0.965–2.538)
       Texture modified diet
      Texture modified diet includes ground, minced, liquid, or full tube feeding without oral intake.
      ECOG PS
        0 (reference)0.6740.2320.0041.963 (1.246–3.092)
       >0
      Tumor site
       Others (reference)−0.7930.2860.0060.452 (0.258–0.792)
       Oral cavity, oropharynx, and hypopharynx
      N classification (TNM)
       N0, N1 (reference)0.6460.2460.0091.908 (1.179–3.088)
       N2, N3
      Mean RT dose to contralateral parotid gland (Gy)0.0270.0080.0381.027 (1.001–1.054)
      Mean RT dose to the OC (Gy)0.0220.0130.0041.022 (1.007–1.037)
      Abbreviations: BMI, body mass index; CI, confidence interval; Gy, Gray; OC, oral cavity; OR, Odds ratio; RT, radiotherapy; S.E. standard error; TNM-classification, tumor, node, metastasis classification according to the 7th edition [
      • Matuschek C.
      • Bolke E.
      • Geigis C.
      • Kammers K.
      • Ganswindt U.
      • Scheckenbach K.
      • et al.
      Influence of dosimetric and clinical criteria on the requirement of artificial nutrition during radiotherapy of head and neck cancer patients.
      ]; ECOG PS, Eastern Cooperative Oncology Group performance status.
      a Texture modified diet includes ground, minced, liquid, or full tube feeding without oral intake.
      The individual probability for TF ≥ 4 weeks can be calculated as: P(TF ≥ 4 weeks) = 1/(1 + e−LP), in which LP is the linear sum of all predictor values multiplied by the regression coefficients, as shown in Fig. 1.
      Fig. 1
      Fig. 1Example calculation and flow chart for the use of the model in clinical practice.
      The formula is accessible via the online supplemental material (Supplemental File 1) and invites the reader to use the prediction model in clinical practice, as suggested in Fig. 1.

      3.4 Model performance

      Fig. 2, Fig. 3 show the performance of the prediction model. The receiver operating characteristic (ROC) curve of the model yielded an AUC of 72.8% before external validation. The Hosmer–Lemeshow test statistics showed a p-value of 0.46, indicating a good model calibration. External validation in the combined MUMC+ and RUMC sample showed an AUC of 62.4%. External validation in the MUMC + sample only showed a considerably higher AUC of 70.8%, whereas external validation in the RUMC sample only showed an AUC of 55.3%. The calibration plot shows a good agreement between predicted probability and the observed use of TF ≥ 4 weeks.
      Fig. 2
      Fig. 2Receiver operating characteristic curve of the prediction model before external validation (A); after external validation in MUMC+ and RUMC combined (B); after external validation in MUMC + only (C); and after external validation in RUMC only (D).
      Fig. 3
      Fig. 3Calibration plot with the actual probability of the use of tube feeding for at least four weeks by predicted probability. The triangles indicate quantiles of patients with a similar predicted probability of the use of tube feeding for at least four weeks.

      3.5 Sensitivity and specificity

      The positive and negative predictive value for a risk of 90% or more of TF dependency ≥4 weeks were 81.8% and 42.3%, respectively. Specifications of sensitivity and specificity at different cut-off values are shown in Supplemental Table 2.

      4. Discussion

      In the current study we developed and externally validated a prediction model to identify LAHNSCC patients who are expected to use TF ≥ 4weeks and thus would benefit from prophylactic gastrostomy insertion. According to our knowledge, this is the first external validation study in a large multicenter retrospective cohort (n = 409 and n = 334). The model includes the following predictors: pretreatment weight change, texture modified diet at baseline, ECOG PS, tumor site, N classification, and mean RT dose to the contralateral parotid gland and OC.
      Remarkably, RT dose to the PCM was not a significant predictor of TF dependency in the model. Previous studies described a significant relationship between increasing RT dose to the PCM and the rising incidence and duration of TF dependency and long-term dysphagia [
      • Wopken K.
      • Bijl H.P.
      • Langendijk J.A.
      Prognostic factors for tube feeding dependence after curative (chemo-) radiation in head and neck cancer: a systematic review of literature.
      ,
      • Christianen M.E.
      • Schilstra C.
      • Beetz I.
      • Muijs C.T.
      • Chouvalova O.
      • Burlage F.R.
      • et al.
      Predictive modelling for swallowing dysfunction after primary (chemo)radiation: results of a prospective observational study.
      ,
      • Jackson J.E.
      • Anderson N.J.
      • Wada M.
      • Schneider M.
      • Poulsen M.
      • Rolfo M.
      • et al.
      Clinical and dosimetric risk stratification for patients at high-risk of feeding tube use during definitive IMRT for head and neck cancer.
      ]. An explanation for these different outcomes might be that we used total RT dose to all PCM, while other studies often used RT dose per PCM subtype; superior, middle, and inferior PCM, with dose to the superior PCM being highly predictive for dysphagia [
      • Christianen M.E.
      • Schilstra C.
      • Beetz I.
      • Muijs C.T.
      • Chouvalova O.
      • Burlage F.R.
      • et al.
      Predictive modelling for swallowing dysfunction after primary (chemo)radiation: results of a prospective observational study.
      ]. Although dose to the PCM is not a predictor in our multivariable model, it does not mean that minimizing dose to the PCM in radiotherapy planning is not useful. Indeed our univariable results indicate that dose to the PCM is associated with the risk of TF ≥ 4 weeks. The association between OC dose and TF dependency may be explained by the fact that the OC has an important function in salivation, taste, chewing, and bolus transport. In a recent study by Van de Bosch et al. on the dosimetric effects of organs at risk, the oral cavity was involved in several toxicity-related effects including dysphagia [
      • Van den Bosch L.
      • van der Schaaf A.
      • van der Laan H.P.
      • Hoebers F.J.P.
      • Wijers O.B.
      • van den Hoek J.G.M
      • et al.
      Comprehensive toxicity risk profiling in radiation therapy for head and neck cancer: a new concept for individually optimised treatment.
      ].
      Previous studies have also shown that dosimetric variables were statistically dependent, particularly dose to the PCM and OC, the latter being a predictor in our model. Inclusion of such a dependent variable might make the other variable non-significant following correction in the statistical model [
      • Jackson J.E.
      • Anderson N.J.
      • Wada M.
      • Schneider M.
      • Poulsen M.
      • Rolfo M.
      • et al.
      Clinical and dosimetric risk stratification for patients at high-risk of feeding tube use during definitive IMRT for head and neck cancer.
      ].
      In addition, dysphagia, toxicity-related nausea and severe taste alterations (dysgeusia) causing food aversion can also negatively affect oral intake leading to TF requirement. Up to now, it remains difficult to predict which patients will experience dysgeusia during CRT/BRT.
      In contrast to our previously published model, BMI at baseline, disease stage, type of systemic therapy and mean dose to the contralateral submandibular gland were not included into this new model as they did not yield a p < 0.10 in the multivariable analysis. We included RT dose to the contralateral salivary glands as potential predictors as the remaining saliva production will correlate with the dose on the spared gland [
      • Houweling A.C.
      • Dijkema T.
      • Roesink J.M.
      • Terhaard C.H.
      • Raaijmakers C.P.
      Sparing the contralateral submandibular gland in oropharyngeal cancer patients: a planning study.
      ]. Although one study previously reported mean RT dose to the contralateral submandibular gland to have a predictive value for TF at six months [
      • Gensheimer M.F.
      • Nyflot M.
      • Laramore G.E.
      • Liao J.J.
      • Parvathaneni U.
      Contribution of submandibular gland and swallowing structure sparing to post-radiation therapy PEG dependence in oropharynx cancer patients treated with split-neck IMRT technique.
      ], this was not a significant predictor in our model. This could be explained by the different endpoints of both studies: TF initiation during CRT/BRT versus TF dependency at six months. Mean dose to the parotid gland was a significant predictor in accordance with our previously published model [
      • Willemsen A.C.H.
      • Kok A.
      • van Kuijk S.M.J.
      • Baijens L.W.J.
      • de Bree R.
      • Devriese L.A.
      • et al.
      Prediction model for tube feeding dependency during chemoradiotherapy for at least four weeks in head and neck cancer patients: a tool for prophylactic gastrostomy decision making.
      ].
      It should also be noted that potential predictors not included in our final model could still have predictive value. However, the current combination of predictors presented the strongest prediction model.

      4.1 Performance of the model

      The model has good accuracy (AUC on internal validation 0.73 and after external validation 0.62 and 0.71 depending on the composition of the validation cohort), but there was a remarkable difference between the two cancer centers participating in the external validation process. While the AUC did not differ much in the MUMC + validation cohort, a marked decrease of AUC was seen in the pooled cohort of MUMC+ and RUMC together. Despite adherence to national guidelines on when to initiate TF, individual and institutional preferences in feeding tube insertion policy might have affected the external validity outcome. RUMC had fewer patients receiving TF ≥ 4 weeks compared to the three other centers (43% versus 70%, 61% and 54% for RUMC and UMCU, MUMC+ and NCI respectively). This difference might be explained by the variations in patient characteristics. Also the effect of the cisplatin administration protocol, weekly in RUMC versus three weekly in all other cancer centers, cannot be ruled out as additional explanation for the differences in TF prevalence. High level evidence for best treatment regimen in primary setting in terms of toxicity and survival is lacking [
      • Kiyota N.
      • Tahara M.
      • Fujii H.
      • Yamazaki T.
      • Mitani H.
      • Iwae S.
      • et al.
      Phase II/III trial of post-operative chemoradiotherapy comparing 3-weekly cisplatin with weekly cisplatin in high-risk patients with squamous cell carcinoma of head and neck (JCOG1008).
      ,
      • Szturz P.
      • Wouters K.
      • Kiyota N.
      • Tahara M.
      • Prabhash K.
      • Noronha V.
      • et al.
      Low-dose vs. High-dose cisplatin: lessons learned from 59 chemoradiotherapy trials in head and neck cancer.
      ]. Another remarkable difference that should be highlighted is the significantly lower number of gastrostomy insertions in the validation cohort versus the developmental cohort (39.5% and 62.6%). This is the result of a different policy in the RUMC regarding prophylactic gastrostomy insertion where reactive NGT insertion is preferred with only 5% of the RUMC patient sample receiving a gastrostomy.
      To our clinical experience, prophylactic gastrostomy insertion could lower the threshold for TF initiation. Studies have shown that reactive NGT insertion is associated with a shorter duration of TF use [
      • Wang J.
      • Liu M.
      • Liu C.
      • Ye Y.
      • Huang G.
      Percutaneous endoscopic gastrostomy versus nasogastric tube feeding for patients with head and neck cancer: a systematic review.
      ,
      • van den Berg M.G.
      • Kalf J.G.
      • Hendriks J.C.
      • Takes R.P.
      • van Herpen C.M.
      • Wanten G.J.
      • et al.
      Normalcy of food intake in patients with head and neck cancer supported by combined dietary counseling and swallowing therapy: a randomized clinical trial.
      ,
      • Soria A.
      • Santacruz E.
      • Vega-Pineiro B.
      • Gion M.
      • Molina J.
      • Villamayor M.
      • et al.
      Gastrostomy vs nasogastric tube feeding in patients with head and neck cancer during radiotherapy alone or combined chemoradiotherapy.
      ]. This was also reflected in our study population, as the median TF duration in RUMC (reactive NGT) was 23 days versus 85 and 82 days in UMCU and MUMC + respectively (prophylactic gastrostomy). It has been argued that (prophylactic) gastrostomies might be related to long term swallowing dysfunction based on the ‘use-it-or-lose-it’ paradigm of dysphagia rehabilitation, but the literature remains controversial on this side effect [
      • Axelsson L.
      • Silander E.
      • Nyman J.
      • Bove M.
      • Johansson L.
      • Hammerlid E.
      Effect of prophylactic percutaneous endoscopic gastrostomy tube on swallowing in advanced head and neck cancer: a randomized controlled study.
      ,
      • Prestwich R.J.
      • Teo M.T.
      • Gilbert A.
      • Williams G.
      • Dyker K.E.
      • Sen M.
      Long-term swallow function after chemoradiotherapy for oropharyngeal cancer: the influence of a prophylactic gastrostomy or reactive nasogastric tube.
      ,
      • Sethugavalar B.
      • Teo M.T.
      • Buchan C.
      • Ermis E.
      • Williams G.F.
      • Sen M.
      • et al.
      Impact of prophylactic gastrostomy or reactive NG tube upon patient-reported long term swallow function following chemoradiotherapy for oropharyngeal carcinoma: a matched pair analysis.
      ,
      • Goff D.
      • Coward S.
      • Fitzgerald A.
      • Paleri V.
      • Moor J.W.
      • Patterson J.M.
      • et al.
      Swallowing outcomes for patients with oropharyngeal squamous cell carcinoma treated with primary (chemo)radiation therapy receiving either prophylactic gastrostomy or reactive nasogastric tube: a prospective cohort study.
      ]. The present study did not evaluate long-term swallowing function after CRT/BRT with or without gastrostomy insertion. Differences in feeding tube policy between the cancer centers, as shown by our nationwide survey [
      • Kok A.
      • van der Lugt C.
      • Leermakers-Vermeer M.J.
      • de Roos N.M.
      • Speksnijder C.M.
      • de Bree R.
      • et al.
      Nutritional interventions in patients with head and neck cancer undergoing chemoradiotherapy: current practice at the Dutch Head and Neck Oncology centres.
      ], could be considered a limitation of the current study. However, we decided to accept this heterogeneity in patient populations to validate our model, since this reflects real world inter-center heterogeneity. An explanation for the diverse policies is the existence of regional differences in hospital logistics, but also differences in the sociocultural background of patients and health professionals and the lack of high-quality evidence in the literature regarding the indication for prophylactic gastrostomy insertion. These findings emphasize the challenge of standardizing gastrostomy insertion management nationwide. This study was not designed to investigate the best approach for TF initiation and feeding tube insertion. Differences in the effect of reactive versus prophylactic feeding tube insertions on oncological therapy outcome, weight loss and quality of life cannot be evaluated here.

      4.2 Generalizability of the model (external validity)

      We suggest that in case the model estimates a probability >90% for TF dependency, a prophylactic gastrostomy insertion should be recommended. In case of a probability >70%, a prophylactic gastrostomy insertion should be discussed with the patient. For patients’ comfort and to reduce the risk of side effects, we recommend prophylactic gastrostomy insertion in high-risk patients before or within the first two weeks of oncological treatment when mucositis and neutropenia have not developed yet [
      • Trotti A.
      • Bellm L.A.
      • Epstein J.B.
      • Frame D.
      • Fuchs H.J.
      • Gwede C.K.
      • et al.
      Mucositis incidence, severity and associated outcomes in patients with head and neck cancer receiving radiotherapy with or without chemotherapy: a systematic literature review.
      ,
      • Rawat S.
      • Srivastava H.
      • Ahlawat P.
      • Pal M.
      • Gupta G.
      • Chauhan D.
      • et al.
      Weekly versus three-weekly cisplatin-based concurrent chemoradiotherapy as definitive treatment in head and neck cancer- where do we stand?.
      ]. This data-driven model indicates that in case of a probability >90%, approximately 18.2% of the patients with a prophylactic gastrostomy insertion will not develop TF dependency ≥4 weeks. However, that does not mean that these 18.2% patients do not benefit from a gastrostomy. They may still need TF but for a period <4 weeks or they may use their gastrostomy for supplemental fluid administration to prevent nephrotoxicity. In 57.7% of the patients with a probability <90%, a reactive feeding tube insertion will be necessary.

      5. Conclusion

      We developed and externally validated a prediction model to estimate TF-dependency ≥4 weeks in LAHNSCC patients treated with CRT/BRT. This model can be used to guide personalized decision-making on prophylactic gastrostomy insertion in clinical practice.

      Funding statement

      This study has been partially funded by the Nutrim Graduate Programme.

      Author contribution

      Anna C.H. Willemsen: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing - Original Draft, Visualization, Funding acquisition. Annemieke Kok: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Writing - Original Draft, Visualization. Laura W.J. Baijens: Writing – Review & Editing, Supervision. Jan Paul de Boer: Writing – Review & Editing. Remco de Bree: Conceptualization, Writing – Review & Editing, Supervision. Lot A. Devriese: Writing – Review & Editing. Chantal M. L. Driessen: Writing – Review & Editing. Carla M. L. van Herpen: Writing – Review & Editing. Frank J.P. Hoebers: Conceptualization, Methodology, Resources, Writing – Review & Editing. Johannes H.A.M. Kaanders: Resources, Writing – Review & Editing. Rebecca T. Karsten: Investigation, Writing – Review & Editing. Sander M.J. van Kuijk: Conceptualization, Methodology, Formal analysis, Writing – Review & Editing. Roy I. Lalisang: Conceptualization, Writing – Review & Editing. Arash Navran: Investigation, Resources, Writing – Review & Editing. Susanne R. Pereboom: Investigation, Writing – Review & Editing. Annemie M.W.J. Schols: Writing – Review & Editing, Supervision. Chris H.J. Terhaard: Conceptualization, Methodology, Resources, Writing – Review & Editing. Ann Hoeben: Conceptualization, Writing – Review & Editing, Supervision.

      Disclaimers

      The views expressed in this article are our own and is not an official position of the institution or funder. This study has been performed with great care. However, the authors do not take any responsibility and are not liable for any damage caused by the use of the prediction model.

      Conflict of interest

      Anna C.H. Willemsen, Annemieke Kok, Jan Paul de Boer, Remco de Bree, Chantal M.L. Driessen, Johannes H.A.M. Kaanders, Rebecca T. Karsten, Sander M.J. van Kuijk, Roy I. Lalisang, Arash Navran, Susanne R. Pereboom, Annemie M.W.J. Schols, Chris H.J. Terhaard, and Ann Hoeben declare that they have no conflict of interest.
      Laura W.J. Baijens.
      Consulting or advisory role: Phagenesis Limited, member of the Independent FEES Review Committee for the PhINEST study.
      Lot A. Devriese.
      Consulting or advisory role: MSD, Bristol Myers Squibb.
      Frank J.P. Hoebers.
      Consulting or advisory role: Bristol Myers Squibb.
      Carla M.L. van Herpen.
      Consulting or advisory role: Bayer, Bristol Myers Squibb, MSD, Regeneron, TRK Fusion Cancer Medical.
      Research Funding: Astra Zeneca, Bristol Myers Squibb , MSD , Merck , Ipsen , Novartis , Sanofi, France .

      Acknowledgments

      The authors would like to thank Marlies Granzier, laboratory technician at Maastro Clinic, and students Roos Kuipers and Rengin Sabaoglu for their effort and time invested in reviewing and delineating CT scans.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Jin S.
        • Lu Q.
        • Sun Y.
        • Xiao S.
        • Zheng B.
        • Pang D.
        • et al.
        Nutrition impact symptoms and weight loss in head and neck cancer during radiotherapy: a longitudinal study.
        BMJ Support Palliat Care. 2021; 11: 17-24
        • Baptistella A.R.
        • Hilleshein K.D.
        • Beal C.
        • Brambatti J.S.
        • Caron R.
        • Baptistella S.F.
        • et al.
        Weight loss as a prognostic factor for recurrence and survival in oropharyngeal squamous cell carcinoma patients.
        Mol Clin Oncol. 2018; 9: 666-672
        • Cho Y.
        • Kim J.W.
        • Keum K.C.
        • Lee C.G.
        • Jeung H.C.
        • Lee I.J.
        Prognostic significance of sarcopenia with inflammation in patients with head and neck cancer who underwent definitive chemoradiotherapy.
        Front Oncol. 2018; 8: 457
        • Kwon M.
        • Kim R.B.
        • Roh J.L.
        • Lee S.W.
        • Kim S.B.
        • Choi S.H.
        • et al.
        Prevalence and clinical significance of cancer cachexia based on time from treatment in advanced-stage head and neck squamous cell carcinoma.
        Head Neck. 2017; 39: 716-723
        • Matsuzuka T.
        • Kiyota N.
        • Mizusawa J.
        • Akimoto T.
        • Fujii M.
        • Hasegawa Y.
        • et al.
        Clinical impact of cachexia in unresectable locally advanced head and neck cancer: supplementary analysis of a phase II trial (JCOG0706-S2).
        Jpn J Clin Oncol. 2019; 49: 37-41
        • Meyer F.
        • Fortin A.
        • Wang C.S.
        • Liu G.
        • Bairati I.
        Predictors of severe acute and late toxicities in patients with localized head-and-neck cancer treated with radiation therapy.
        Int J Radiat Oncol Biol Phys. 2012; 82: 1454-1462
        • Brown T.E.
        • Getliffe V.
        • Banks M.D.
        • Hughes B.G.
        • Lin C.Y.
        • Kenny L.M.
        • et al.
        Validation of an updated evidence-based protocol for proactive gastrostomy tube insertion in patients with head and neck cancer.
        Eur J Clin Nutr. 2016; 70: 574-581
        • Karsten R.T.
        • Stuiver M.M.
        • van der Molen L.
        • Navran A.
        • de Boer J.P.
        • Hilgers F.J.M.
        • et al.
        From reactive to proactive tube feeding during chemoradiotherapy for head and neck cancer: a clinical prediction model-based approach.
        Oral Oncol. 2019; 88: 172-179
        • van der Linden N.C.
        • Kok A.
        • Leermakers-Vermeer M.J.
        • de Roos N.M.
        • de Bree R.
        • van Cruijsen H.
        • et al.
        Indicators for enteral nutrition use and prophylactic percutaneous endoscopic gastrostomy placement in patients with head and neck cancer undergoing chemoradiotherapy.
        Nutr Clin Pract. 2017; 32: 225-232
        • Wang J.
        • Liu M.
        • Liu C.
        • Ye Y.
        • Huang G.
        Percutaneous endoscopic gastrostomy versus nasogastric tube feeding for patients with head and neck cancer: a systematic review.
        J Radiat Res. 2014; 55: 559-567
        • Beijer Y.J.
        • Koopman M.
        • Terhaard C.H.
        • Braunius W.W.
        • van Es R.J.
        • de Graeff A.
        Outcome and toxicity of radiotherapy combined with chemotherapy or cetuximab for head and neck cancer: our experience in one hundred and twenty-five patients.
        Clin Otolaryngol. 2013; 38: 69-74
        • Vrolijk L.
        • de Roij van Zuijdewijn C.L.M.
        • Slingerland M.
        • Wiggenraad R.G.J.
        • Verschuur H.P.
        • Jeurissen F.J.F.
        PEG tube dependency after prophylactic placement in 209 head and neck cancer patients treated with chemoradiotherapy or radiation with cetuximab.
        Clin Otolaryngol. 2019;
        • Nederlandse Werkgroep Hoofd-Hals Tumoren (NWHHT)
        Landelijke richtlijn larynxcarcioom, versie 3.0.
        2010
        • Strijbos D.
        • Keszthelyi D.
        • Bogie R.M.M.
        • Gilissen L.P.L.
        • Lacko M.
        • Hoeijmakers J.G.J.
        • et al.
        A systematic review and meta-analysis on outcomes and complications of percutaneous endoscopic versus radiologic gastrostomy for enteral feeding.
        J Clin Gastroenterol. 2018; 52: 753-764
        • Leemans LES C.R.
        • Langendijk J.A.
        • de Boer J.P.
        • Terhaard C.H.J.
        • Roodenburg J.L.N.
        • Klomp F.W.J.
        • et al.
        Richtlijn hoofd-halstumoren.
        2014
        • Talwar B.
        • Donnelly R.
        • Skelly R.
        • Ye Y.
        • Huang G.
        Nutritional management in head and neck cancer: United Kingdom National Multidisciplinary Guidelines.
        J Laryngol Otol. 2016; 130: S32-S40
        • Arends J.
        • Bodoky G.
        • Bozzetti F.
        • Fearon K.
        • Muscaritoli M.
        • Selga G.
        • et al.
        ESPEN guidelines on enteral nutrition: non-surgical oncology.
        Clin Nutr. 2006; 25: 245-259
        • Brown T.E.
        • Spurgin A.L.
        • Ross L.
        • Tripcony L.
        • Keller J.
        • Hughes B.G.
        • et al.
        Validated swallowing and nutrition guidelines for patients with head and neck cancer: identification of high-risk patients for proactive gastrostomy.
        Head Neck. 2013; 35: 1385-1391
        • Garrow WPTJ J.S.
        • Ralph A.
        10th Human nutrition and dietetics. Churchill Livingstone, Edinburgh2000: 490
        • Nunes G.
        • Fonseca J.
        • Barata A.T.
        • Dinis-Ribeiro M.
        • Pimentel-Nunes P.
        Nutritional support of cancer patients without oral feeding: how to select the most effective technique?.
        GE Port J Gastroenterol. 2020; 27: 172-184
        • Strijbos D.
        • Keszthelyi D.
        • Gilissen L.P.L.
        • Lacko M.
        • Hoeijmakers J.G.J.
        • van der Leij C.
        • et al.
        Percutaneous endoscopic versus radiologic gastrostomy for enteral feeding: a retrospective analysis on outcomes and complications.
        Endosc Int Open. 2019; 7: E1487-E1495
        • Willemsen A.C.H.
        • Kok A.
        • van Kuijk S.M.J.
        • Baijens L.W.J.
        • de Bree R.
        • Devriese L.A.
        • et al.
        Prediction model for tube feeding dependency during chemoradiotherapy for at least four weeks in head and neck cancer patients: a tool for prophylactic gastrostomy decision making.
        Clin Nutr. 2019;
        • Anderson Head M.D.
        • Neck Cancer Symptom Working Group
        Beyond mean pharyngeal constrictor dose for beam path toxicity in non-target swallowing muscles: dose-volume correlates of chronic radiation-associated dysphagia (RAD) after oropharyngeal intensity modulated radiotherapy.
        Radiother Oncol. 2016; 118: 304-314
        • Wopken K.
        • Bijl H.P.
        • Langendijk J.A.
        Prognostic factors for tube feeding dependence after curative (chemo-) radiation in head and neck cancer: a systematic review of literature.
        Radiother Oncol. 2018; 126: 56-67
        • Christianen M.E.
        • Schilstra C.
        • Beetz I.
        • Muijs C.T.
        • Chouvalova O.
        • Burlage F.R.
        • et al.
        Predictive modelling for swallowing dysfunction after primary (chemo)radiation: results of a prospective observational study.
        Radiother Oncol. 2012; 105: 107-114
        • Collins G.S.
        • Reitsma J.B.
        • Altman D.G.
        • Moons K.G.
        Transparent reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement.
        J Clin Epidemiol. 2015; 68: 134-143
        • Driessen C.M.
        • Ham J.C.
        • Te Loo M.
        • van Meerten E.
        • van Lamoen M.
        • Hakobjan M.H.
        • et al.
        Genetic variants as predictive markers for ototoxicity and nephrotoxicity in patients with locally advanced head and neck cancer treated with cisplatin-containing chemoradiotherapy (the PRONE study).
        Cancers. 2019; 11
        • Karsten R.T.
        • Al-Mamgani A.
        • Bril S.I.
        • Tjon A.J.S.
        • van der Molen L.
        • de Boer J.P.
        • et al.
        Sarcopenia, a strong determinant for prolonged feeding tube dependency after chemoradiotherapy for head and neck cancer.
        Head Neck. 2019; 41: 4000-4008
        • Stuurgroep Ondervoeding
        Kiezen toedieningsweg.
        2019 (Retrieved from)
        • Sachdev S.
        • Refaat T.
        • Bacchus I.D.
        • Sathiaseelan V.
        • Mittal B.B.
        Age most significant predictor of requiring enteral feeding in head-and-neck cancer patients.
        Radiat Oncol. 2015; (10:93)
        • Mortensen H.R.
        • Overgaard J.
        • Jensen K.
        • Specht L.
        • Overgaard M.
        • Johansen J.
        • et al.
        Factors associated with acute and late dysphagia in the DAHANCA 6 & 7 randomized trial with accelerated radiotherapy for head and neck cancer.
        Acta Oncol. 2013; 52: 1535-1542
        • Sanguineti G.
        • Rao N.
        • Gunn B.
        • Ricchetti F.
        • Fiorino C.
        Predictors of PEG dependence after IMRT+/-chemotherapy for oropharyngeal cancer.
        Radiother Oncol. 2013; 107: 300-304
        • Setton J.
        • Lee N.Y.
        • Riaz N.
        • Huang S.H.
        • Waldron J.
        • O’Sullivan B.
        • et al.
        A multi-institution pooled analysis of gastrostomy tube dependence in patients with oropharyngeal cancer treated with definitive intensity-modulated radiotherapy.
        Cancer. 2015; 121: 294-301
        • Strom T.
        • Trotti A.M.
        • Kish J.
        • Rao N.G.
        • McCaffrey J.
        • Padhya T.A.
        • et al.
        Risk factors for percutaneous endoscopic gastrostomy tube placement during chemoradiotherapy for oropharyngeal cancer.
        JAMA Otolaryngol Head Neck Surg. 2013; 139: 1242-1246
        • Wermker K.
        • Jung S.
        • Huppmeier L.
        • Joos U.
        • Kleinheinz J.
        Prediction model for early percutaneous endoscopic gastrostomy (PEG) in head and neck cancer treatment.
        Oral Oncol. 2012; 48: 355-360
        • Kanayama N.
        • Kierkels R.G.J.
        • van der Schaaf A.
        • Steenbakkers R.
        • Yoshioka Y.
        • Nishiyama K.
        • et al.
        External validation of a multifactorial normal tissue complication probability model for tube feeding dependence at 6months after definitive radiotherapy for head and neck cancer.
        Radiother Oncol. 2018; 129: 403-408
        • Matuschek C.
        • Bolke E.
        • Geigis C.
        • Kammers K.
        • Ganswindt U.
        • Scheckenbach K.
        • et al.
        Influence of dosimetric and clinical criteria on the requirement of artificial nutrition during radiotherapy of head and neck cancer patients.
        Radiother Oncol. 2016; 120: 28-35
        • Edge S.B.
        • Compton C.C.
        The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM.
        Ann Surg Oncol. 2010; 17: 1471-1474
        • Brown T.E.
        • Wittholz K.
        • Way M.
        • Banks M.D.
        • Hughes B.G.
        • Lin C.Y.
        • et al.
        Investigation of p16 status, chemotherapy regimen, and other nutrition markers for predicting gastrostomy in patients with head and neck cancer.
        Head Neck. 2017; 39: 868-875
        • Brouwer C.L.
        • Steenbakkers R.J.
        • Bourhis J.
        • Budach W.
        • Grau C.
        • Gregoire V.
        • et al.
        CT-based delineation of organs at risk in the head and neck region: DAHANCA, EORTC, GORTEC, HKNPCSG, NCIC CTG, NCRI, NRG Oncology and TROG consensus guidelines.
        Radiother Oncol. 2015; 117: 83-90
        • Kubben P.
        • Dumontier M.
        • Dekker A.
        Fundamentals of clinical data science.
        Springer, 2019
        • Collins G.S.
        • Ogundimu E.O.
        • Altman D.G.
        Sample size considerations for the external validation of a multivariable prognostic model: a resampling study.
        Stat Med. 2016; 35: 214-226
        • Hosmer D.W.
        • Lemeshow S.
        Applied logistic regression.
        2nd. Wiley, New York2000
        • Computing RFfS: R
        A language and environment for statistical computing.
        R Core Team, Vienna, Austria2018 (ed 3.5.1)
        • Jackson J.E.
        • Anderson N.J.
        • Wada M.
        • Schneider M.
        • Poulsen M.
        • Rolfo M.
        • et al.
        Clinical and dosimetric risk stratification for patients at high-risk of feeding tube use during definitive IMRT for head and neck cancer.
        Tech Innov Patient Support Radiat Oncol. 2020; 14: 1-10
        • Van den Bosch L.
        • van der Schaaf A.
        • van der Laan H.P.
        • Hoebers F.J.P.
        • Wijers O.B.
        • van den Hoek J.G.M
        • et al.
        Comprehensive toxicity risk profiling in radiation therapy for head and neck cancer: a new concept for individually optimised treatment.
        Radiother Oncol. 2021; 157: 147-154
        • Houweling A.C.
        • Dijkema T.
        • Roesink J.M.
        • Terhaard C.H.
        • Raaijmakers C.P.
        Sparing the contralateral submandibular gland in oropharyngeal cancer patients: a planning study.
        Radiother Oncol. 2008; 89: 64-70
        • Gensheimer M.F.
        • Nyflot M.
        • Laramore G.E.
        • Liao J.J.
        • Parvathaneni U.
        Contribution of submandibular gland and swallowing structure sparing to post-radiation therapy PEG dependence in oropharynx cancer patients treated with split-neck IMRT technique.
        Radiat Oncol. 2016; 11: 151
        • Kiyota N.
        • Tahara M.
        • Fujii H.
        • Yamazaki T.
        • Mitani H.
        • Iwae S.
        • et al.
        Phase II/III trial of post-operative chemoradiotherapy comparing 3-weekly cisplatin with weekly cisplatin in high-risk patients with squamous cell carcinoma of head and neck (JCOG1008).
        J Clin Oncol. 2020; 38: 6502
        • Szturz P.
        • Wouters K.
        • Kiyota N.
        • Tahara M.
        • Prabhash K.
        • Noronha V.
        • et al.
        Low-dose vs. High-dose cisplatin: lessons learned from 59 chemoradiotherapy trials in head and neck cancer.
        Front Oncol. 2019; 9: 86
        • van den Berg M.G.
        • Kalf J.G.
        • Hendriks J.C.
        • Takes R.P.
        • van Herpen C.M.
        • Wanten G.J.
        • et al.
        Normalcy of food intake in patients with head and neck cancer supported by combined dietary counseling and swallowing therapy: a randomized clinical trial.
        Head Neck. 2016; 38: E198-E206
        • Soria A.
        • Santacruz E.
        • Vega-Pineiro B.
        • Gion M.
        • Molina J.
        • Villamayor M.
        • et al.
        Gastrostomy vs nasogastric tube feeding in patients with head and neck cancer during radiotherapy alone or combined chemoradiotherapy.
        Nutr Hosp. 2017; 34: 512-516
        • Axelsson L.
        • Silander E.
        • Nyman J.
        • Bove M.
        • Johansson L.
        • Hammerlid E.
        Effect of prophylactic percutaneous endoscopic gastrostomy tube on swallowing in advanced head and neck cancer: a randomized controlled study.
        Head Neck, 2017
        • Prestwich R.J.
        • Teo M.T.
        • Gilbert A.
        • Williams G.
        • Dyker K.E.
        • Sen M.
        Long-term swallow function after chemoradiotherapy for oropharyngeal cancer: the influence of a prophylactic gastrostomy or reactive nasogastric tube.
        Clin Oncol. 2014; 26: 103-109
        • Sethugavalar B.
        • Teo M.T.
        • Buchan C.
        • Ermis E.
        • Williams G.F.
        • Sen M.
        • et al.
        Impact of prophylactic gastrostomy or reactive NG tube upon patient-reported long term swallow function following chemoradiotherapy for oropharyngeal carcinoma: a matched pair analysis.
        Oral Oncol. 2016; 59: 80-85
        • Goff D.
        • Coward S.
        • Fitzgerald A.
        • Paleri V.
        • Moor J.W.
        • Patterson J.M.
        • et al.
        Swallowing outcomes for patients with oropharyngeal squamous cell carcinoma treated with primary (chemo)radiation therapy receiving either prophylactic gastrostomy or reactive nasogastric tube: a prospective cohort study.
        Clin Otolaryngol. 2017;
        • Kok A.
        • van der Lugt C.
        • Leermakers-Vermeer M.J.
        • de Roos N.M.
        • Speksnijder C.M.
        • de Bree R.
        • et al.
        Nutritional interventions in patients with head and neck cancer undergoing chemoradiotherapy: current practice at the Dutch Head and Neck Oncology centres.
        Eur J Cancer Care. 2021; e13518
        • Trotti A.
        • Bellm L.A.
        • Epstein J.B.
        • Frame D.
        • Fuchs H.J.
        • Gwede C.K.
        • et al.
        Mucositis incidence, severity and associated outcomes in patients with head and neck cancer receiving radiotherapy with or without chemotherapy: a systematic literature review.
        Radiother Oncol. 2003; 66: 253-262
        • Rawat S.
        • Srivastava H.
        • Ahlawat P.
        • Pal M.
        • Gupta G.
        • Chauhan D.
        • et al.
        Weekly versus three-weekly cisplatin-based concurrent chemoradiotherapy as definitive treatment in head and neck cancer- where do we stand?.
        Gulf J Oncolog. 2016; 1: 6-11