Eligibility criteria and study design
This study targeted adult patients treated or followed for thyroid cancer between January 2015 and January 2019 at the University Institute of Face and Neck, Nice, France. All the contacted patients had a thyroid cancer, were 18 years or older and had a social security coverage. They all signed a written consent after being giving informations about the study. Patients were excluded if they had a pathology seriously impairing their quality of life, if they could not speak French, if they had a severe psychopathological disorder and if they were vulnerable people. All questionnaires were sent by post. We conducted a retrospective chart review on responding patients.
Data collection
Patient’s clinical and pathological data, retrieved retrospectively from an electronic patients records management system (Clinicom® computerized record software), were collected on a database managed using Ennov Clinical® software. Questionnaires responses were integrated to the same database. Incomplete questionnaires were handled as per the THYCA-QoL and EORTC QLQ-C30 developer’s recommendation [13, 19]. All data were controlled and validated by a Clinical Data Manager.
Questionnaires package description
The questionnaire package included a coversheet explaining the study, an informed consent and three questionnaires: the French version of the EORTC QLQ-C30 (version 3.0) [19], the French version of the THYCA-QoL (F-THYCA-QoL) [17], and a sociodemographic and clinical characteristics questionnaire. A self-addressed postage paid envelope was provided for the return of the questionnaires and the signed patient consent. The permission of developers was obtained for the use of the questionnaires when appropriate.
The F-THYCA-QoL questionnaire assesses thyroid-specific symptoms and was developed according to the EORTC’s guidelines [13], for use with the EORTC QLQ-C30 questionnaire. The F-THYCA-QoL comprises 24 items translated by our institution into French and was already published [17]. It evaluates seven scales (neuromuscular, voice, concentration, psychological, sympathetic, throat/mouth and sensory problems) and six single items. Items are scored on a Likert four-points scale from 1,“not at all” to 4, “very much”. The evaluation time frame is the previous week to the questionnaire reception, except for the sexual item, for which the time frame is set to the four previous weeks. Scores were transformed in a linear scale from 0 to 100. A High score for a functional scale or a single item represents a high HRQoL’s impairment.
The French version of the EORTC QLQ-C30 (version 3.0) questionnaire includes 30 items and measures five functional scales (physical, role, emotional, cognitive and social functioning), global health status (GHS), financial difficulties and eight symptom scales (fatigue, nausea and vomiting, pain, dyspnea, insomnia, appetite loss, constipation, diarrhea). The evaluation time frame is the previous week to the questionnaire reception. All items are scored on a Likert four point response ranging from 1, “not at all” to 4, “very much”, with the exception of the global health status scale including 2 items scored on a seven point modified linear analogue scale ranging from 1, “very poor” to 7, “excellent” [19]. For each dimension, one score from 0 to 100 is generated according to the EORTC scoring manual procedure [20].
Sample size calculation and statistical analysis
To validate a questionnaire, it is recommended to recruit a minimum of 10 patients per item [18]; with 24 items, the F-THYCA-QoL validation would need at least 240 patients. Moreover, exploratory factor analysis (EFA) to investigate the dimensionality of a questionnaire requires a minimum of 30 patients per scale; with 7 response scales, the F-THYCA-QoL validation would necessitate a minimum of 210 patients [21]. So, to ensure the robustness of the statistical quantitative validation of the psychometric properties of the F-THYCA-QoL questionnaire, we planned to include 240 patients.
We analyzed acceptability, floor and ceiling effects, reliability, convergent and discriminant validities, multi-dimensionality and criterion validity of the F-THYCA-QoL.
Acceptability of the questionnaire was assessed by the percentage of missing data (missing items and missing forms). A high proportion of missing data may indicate poor acceptability of the instrument [22].
Floor and ceiling effects were estimated for each item. As in the original study, floor or ceiling effects were considered to be present if more than 80% of the responders achieved the lowest or highest possible score respectively [13].
Reliability of the questionnaire was assessed using Cronbach’s α coefficient (and confident interval 95%) for internal consistency. Cronbach’s α was estimated for each subscale, and it was expected to be greater than 0.70 [23].
Construct validity was evaluated in terms of Convergent and discriminant validities and Multi-dimensionality. Convergent and discriminant validities was analyzed using multitrait scaling analysis [21] conducted separately for the seven scales. The convergent validity of each item was assessed using Spearman’s correlation coefficient between each item and its own scale score, computed without including the corresponding item. The convergent validity was considered satisfactory if the absolute value of the correlation coefficient was greater than 0.40 [21]. For the discriminant validity, the correlation between each item and its own scale score was expected to be greater than the correlation between this item and the other scale scores. Multi-dimensionality of the questionnaire was assessed by a seven factors EFA with orthogonal rotation. All items were integrated into the EFA, the variance explained by the seven factors was reported and the resulting factors were interpreted [24].
Criterion validity was assessed using a correlation matrix between F-THYCA-QoL scale summary scores and the QLQ-C30 scale scores. The correlation between the F-THYCA-QoL and the QLQ-C30 scores assessing the same HRQoL domain was expected to be greater or equal than 0.4, in absolute value [21]. Conversely, the correlation between each THYCA-QoL scale score and other scales of the QLQ-C30 assessing other HRQoL domains was expected to be lower than 0.4.
All analyses were performed using R software version 3.0.2. All tests were two-sided, and the alpha risk was set as 5%.