2.1 | Patients
In this retrospective single-center chart review, we identified a database of 310 consecutive UICC stage III–IVB HNSCC patients (except for nasopharyngeal and sinonasal sites) treated between 2007 and 2012 with primary or adjuvant chemo-RT with a curative intent. Ethics committee approval (Ref.-Nr. KEK-BE: 289/2014) was obtained for this study and it has been conducted in full accordance with ethical principles, including the World Medical Association Declaration of Helsinki (version 2002) and the additional requirements.
2.2 | Treatment and follow-up
Treatment strategies were based on institutional policies following the case-based multidisciplinary tumor board decision, as previously published12,13. Patients who were first diagnosed before 2010 (UICC 6th edition) were re-staged according to the 7th edition during data acquisition. During this period, there was no algorithm as to which patient should be recommended for a pPEG. Prophylactic PEG placement was recommended to all patients based on the subjective evaluation of their general condition, expected radiation volume and side effects by the attending radiation oncologists. The cases in which a patient rejected a pPEG and the reasons of rejection were not systematically assessed. The planning and delivery of RT as well as the definitions of clinical target volume and planned target volume followed international recommendations14-18. The RT was administered with 2 Gy daily fractions using a volumetric modulated arc technique up to a total dose of 72 Gy for macroscopically detectable tumor, 66 Gy for postoperative positive or narrow resection margin(s), and the lymph node region(s) with pathological extracapsular extension. Elective nodal regions received 54 Gy. Sequential boosts were performed.
2.3 | Percutaneous endoscopic gastrostomy placement
Percutaneous endoscopic gastrostomy placements (pPEG and rPEG) were performed according to the pull method described by Ponsky et al19. Antibiotic prophylaxis with Amoxicillin/Clavulanate 1.2 g intravenously and a Freka PEG gastral 15 Ch/Fr EnFit (Fresenius Kabi, Switzerland) were routinely used.
2.4 | Definition of unplanned hospitalizations and follow-up
All complications leading to a hospitalization from the initial histopathological diagnosis to the last follow-up were recorded. Emergency or unplanned admissions were defined as unplanned hospitalizations (UHs). However, elective hospitalizations, including those due to socially or logistically difficult circumstances (e.g. long journey, initially poor general condition, etc.), were not analyzed. If an elective hospitalization was associated with a complication and therefore an extension to the planned length of stay, the time from that complication to discharge was defined as an UH. UHs were classified as being related to comorbidities, index HNSCC and recurrences, or cancer treatment. When a UH was related to cancer treatment, it was defined as a treatment-related UH (TRUH). In order not to have more than three endpoints and to enable a sound and simple statistical methodology, we analyzed only the first two UHs and thus only the first two consecutive TRUHs in patients who had multiple UHs. In the case of externally UH, additional information was obtained from the hospital where the emergency took place.
2.5 | Toxicities and the course of body weight
Toxicities and the course of body weight from the initial histopathological diagnosis to the last follow-up were recorded and graded according to CTCAE (version 5.0)6. The patient’s weight was recorded before, during, and after therapy. Changes were graded by CTCAE: grade 1, 5% – <10% from baseline, intervention not indicated; grade 2, 10 – ≤20% from baseline, nutritional support indicated; grade 3 ≥20% from baseline, tube feeding, or total parenteral nutrition indicated. Symptoms of pain, dermatitis, mucositis, dysphagia, xerostomia, and osteoradionecrosis were assessed. Acute and late toxicities were defined as post-treatment-related complications occurring during and/or within 3 months or ≥3 months after commencing chemo-RT, respectively. Baseline pre-treatment tumor-related morbidity using the same criteria were also assessed.
2.6 | Statistical analysis
Patients were grouped as pPEG and nPEG. Patients who received rPEG were included in the nPEG group according to the intention-to-treat principle. The endpoints were defined as: first TRUH (TRUH1), second consecutive TRUH (TRUH2), and overall survival (OS). The start date of the first and second TRUH, and the date of death, were counted as the corresponding time points, respectively. Cox’s proportional hazard model was used to evaluate time-to-event endpoints, calculated from the date of histopathological diagnosis of the initial HNSCC. For multivariate analyses, backwards stepwise elimination was performed by including variables yielding p values ≤0.05 in univariate analyses. Actuarial time to event rates were depicted by Kaplan–Meier methodology. The chi-squared test was used to compare categorical variables. All tests were two-tailed. No adjustment was done for multiple testing. Due to the lack of concrete evidence or consensus regarding pre-treatment risk factors for feeding tube requirement to calculate and assign propensity scores, no matched-pair analyses were performed. Statistical analyses were performed with JMP (version 14.2.0; SAS Institute, Cary, NC, USA).