The design of this study was a population-based, prospectively collected, retrospectively reviewed cohort study. Participants were recruited from the Manitoba Cancer Register, which is a comprehensive and accredited population-based registry [3]. Participants were recruited consecutively, completing supervised self-reported questionnaires and clinical assessment by a speech and language pathologist (SLP) at 3, 6, 12, 24, 36, and ≥ 60 months post-treatment appointments. In the first year, assessments could vary by as much as ± 1 month. At 24 and 36 months visits could vary ± 3 months. Follow up in the ≥ 60 month period ranged from 5 to 12 years. Acquisition of this data was part of the regular clinical follow-up of all patients treated for upper aerodigestive tract cancer within CancerCare Manitoba, the solitary tertiary cancer centre for the Province of Manitoba, Canada. The study was reviewed and approved by the Research Ethics Board, Office of Research Ethics and Compliance, University of Manitoba (Approval #HS19561). This study was exempt from individual-level consent as patients were receiving standard of care therapy and this was considered a low-risk, noninterventional, observational study.

Pre-treatment was defined as prior to initial radiation. Post-treatment was defined as the period following the last day of radiation. Electronic medical records were reviewed retrospectively for demographic, oncologic, selected functional outcome, and late effects (complications of radiation) data for patients treated between 2008 and 2016 inclusive. Patients were included in this study if they had American Joint Committee on Cancer (AJCC VII) stage I-IV (M0) nasopharyngeal carcinoma and were treated with curative intent with IMRT alone or with chemo-radiation. Patients were excluded if they underwent previous head and neck irradiation to the primary tumour, palliative intent defined by a radiation dose < 50 Gy, or had a previous head and neck cancer within 5 years prior to diagnosis.

Data from patients who suffered recurrent disease was censored at the time of the recurrence. All patients were treated within the same tertiary care, academic health network: CancerCare Manitoba and the University of Manitoba teaching hospitals. All patients had their management reviewed at the multidisciplinary case conference of the Head and Neck Disease Site Group.

The Manitoba Cancer Registry (MCR) is a comprehensive and accredited population-based registry for the province of Manitoba, Canada; population 1.2 million. The MCR is a member of the North American Associated of Central Cancer Registries which administers a program that review member registries for their ability to produce complete, accurate, and timely data. Fields included diagnoses coded using the International Classification of Diseases 10th revision for Canada (ICD-10-CA), age, and stage. Health care is freely provided by the government and non-participation in the plan is rare. Therefore, the registry and any derived cohort can be considered complete and population based. All incident registry cases of nasopharyngeal carcinoma from 2008 to 2016 inclusive were reviewed to ensure all eligible cases were included.

CancerCare Manitoba established prospective collection of selected functional outcomes following treatment of head and neck cancer as part of routine follow up in 2003. The protocol was intended to capture functional outcomes consistent with the International Classification of Functioning, Disability and Health (ICF) [4, 5]. The outcomes chosen for this study included 7 instruments which have proven inter-rater reliability and validity and are widely used in head and neck cancer patient assessments.

Two validated clinician-rated general performance status scales were included: The Karnofsky Performance Status (KPS) scale and the Eastern Cooperative Oncology Group (ECOG) toxicity and response criteria scale are standard tools used to assess performance in activities of daily living [6].

Five validated clinician-rated scales of swallowing and communication were included: The Performance Status Scale for Head and Neck Cancer Patients (PSS-HN) which describes performance on three subscales: normalcy of diet (maximum oral diet texture tolerated), eating in public, and understandability of speech; and the Royal Brisbane Hospital Outcome Measure for Swallowing (RBHOMS) which is a measure of swallowing and tube feed dependency [7].

To provide clinical context to the reader, a brief summary of the 3 PSS-HN scales are as follows: (1) “Normalcy of Diet” is a diet texture scale in increments of 10, from 0 to 100. Score 0: NPO; 10–30 = liquid or pureed diet; 40–50 = soft diet; 60–90 = increasingly solid diet but some limitations; 100 = normal diet, no restrictions. (2) “Understandability of Speech” is scaled in increments of 25, from 0–100. Score 0 = never understood; 25 = difficult to understand; 50 = usually understood with face-to-face conversation; 75 = usually understood in all circumstances; 100 = always understood. (3) “Eating in Public” is scaled in increments of 25, from 0 to 100. Score 0 = always eat alone; 25 = only eats at home in presence of selected people; 50 = eats in front of selected people in selected locations; 75 = no restriction to place, but modified diet; 100 = no restrictions to diet or place [8].

The validated clinician-rated Royal Brisbane Hospital Outcome Measure for Swallowing (RBHOMS) was used to evaluate relative feeding tube dependency. This 10-point scale measures relative feeding tube dependence and dysphagia, with higher scores indicating better function. RBHOMS scores of 1–5 indicate that the patient is tube feed dependent, either totally (i.e. patient NPO; scores 1–3) or partially (scores 4–5). RBHOMS scores of 6–10 indicate that the patient can sustain themselves nutritionally and is not feeding tube dependent. A RBHOMS scores of 6–7 cover liquid diets to introduction of solid foods; a score of 8 indicates that solids have become a consistent part of the diet but the diet is still substantially modified, 9 indicates diet is near normal solid diet with minor modifications [9].

Finally, the Voice Handicap Index-10 (VHI-10) was used [10]. This patient-rated scale has 10 questions with answers ranging from 0 (never) to 4 (always) for a total of 40 possible points with a higher score indicating worse self-perceived voice handicap. The VHI-10 has been collapsed into 3 ranges, 0–9 (never to almost never severe), 10–19 (sometimes severe), and > 20 (more than sometimes severe). Ranges are based on clinically relevant points in functional outcome scales for analysis, limited by the low number of patients with scores > 20.

In addition to tracking functional outcomes, other late effects of treatment were documented as part of the scheduled pre- and post-treatment SLP visits. The validated modified Edmonton Symptom Assessment Scale (mESAS) is a patient-rated visual analog scale used to measure the severity of multiple symptoms of advanced cancer. Symptoms are scored from 0 to 10 with a score of 10 indicating the worst possible symptoms. The mESAS was used to collect the patient-reported symptom of xerostomia and dysphagia [11]. SLP also documented the presence of the following problems: trismus; neck flexion deformity; cranial nerve 12 palsy; whether dysphagia was bad enough to require a videofluoroscopic swallowing study in the preceding interval, and whether aspiration was evident; treatment for aspiration pneumonia; presence of muscle spasms; presence of tracheostomy; and feeding tube dependency.

Patient demographic and oncologic data collection included age, gender, AJCC (7th Edition) Prognostic Stage, T classification, N classification, radiation dose and volume, chemotherapy details, date and location of recurrence, and date and cause of death. Radiation volumes collected included: Gross tumour volume (GTVp); gross nodal volume (GTVn); planning target volume receiving at least 70 Gy (PTV70); planning target volume receiving at least 59 Gy (PTV59); and sum of all target volumes receiving a therapeutic dose (PTVall); and mean dose to left and right parotid (MD parotid L,R).

Overall survival (OS) and disease-free survival (DFS) were determined using standard Kaplan–Meier methodology, whereas disease-specific survival (DSS) was determined using cumulative incidence with competing risk for death. Times for assessments were measured from the date of the last radiation treatment. Patients with recurrence were followed and assessed until death or end of study follow-up. Patients were censored at the time of death, with or without recurrence. There were two components to the late effects analysis: (1) A descriptive analysis of function and side effects over time focusing on the nature and severity of late effects and whether there was a trend to worsening function and side effects in later time intervals, and (2) A regression analysis of factors that predict poor functional outcome.

Regression models were used to predict outcomes and examine the relationship between baseline predictors and outcomes. Logistic mixed models were used to predict binary outcomes with repeated measurements, and the output was marginalized. Quantile regression with a clustered variance was used to predict ordinal outcomes with repeated measurements and reported median differences. Regression models analyzed data from 1 year and onward, which excluded the acute post-treatment effects. Regression models were used to examine the association between predictors and outcomes, as well as provide estimated marginalized probabilities (for binary outcomes) and predicted median values (ordinal outcomes) over time to describe functional health during follow-up. The relationship between continuous predictors and outcomes were demonstrated with predicted values over time for the 10th, 50th, and 90th percentile values of the continuous predictor. Natural splines were used to account for non-linear time effects in regression models. R version 4.03 was used to analyze the data with the following packages: cmprsk, GLMMadaptive, quantreg, splines, and survival.