Page 116 - Functional impairment and cues for rehabilitation of head and neck cancer patients -

114
Chapter 5
Voice and speech outcomes
In order to assess observer-rated voice and speech outcomes, audio recordings were made of patients performing a set of speech tasks which included respectively reading aloud a 149 word long Dutch reading text called ‘Tachtig dappere fietsers’ (Eighty brave cyclists), a word list, and sustained vowels (/a/, /i/, and /u/). All recordings were analyzes using the PRAAT program (30). The primary observer-rated speech outcome was the vowel space area, a measure of articulation, for which the read text was used, or the word list if the text was not available. It was calculated as a percentage of the maximum total area of the vowel triangle (31). In this study, values below 80% were used to indicate abnormal articulation.
The primary patient-rated speech outcome was the Speech Handicap Index (SHI). This is a thirty- item speech-related quality of life questionnaire on which a patient indicates the frequency of problems experienced on a five-point scale: never (= 0), almost never (= 1), sometimes (= 2), almost always (= 3), and always (= 4). The score can range from 0–120 with higher scores indicating more speech-related problems. A psychosocial and a speech function subscale can be calculated from these thirty questions. The SHI also includes one global question indication the overall speech quality (excellent (= 0), good (= 30), average (= 70), and bad (= 100)) (32, 33).
Secondary speech outcomes were the articulation rate in syllables per second, which was measured from the reading text using a script in PRAAT (34). The voice outcome measure was the acoustic voice quality index (AVQI), which was determined using a combination of 3 seconds of the sustained /a/ and 4 seconds of the read text (35, 36). If no 3 seconds of /a/ was available, a combination of the sustained vowel records was used. If the read text was not present, 4 seconds of the word list was used. This outcome ranges from 1 to 10, with 1 being most equal to normal and 10 least equal to normal. A value of the AVQI less than 2.95 was considered a good voice quality (37).
Statistical analysis
Analyses were performed using IBM® SPSS® Statistics 25.0. Baseline characteristics were presented using descriptive statistics. To test whether patient and tumor characteristics of the patients at t0, t1 and t2 were different, the Kruskal-Wallis test was used for continuous data and the linear-by-linear approximation of the Pearson’s Chi-square test (exact two-sided p value) for dichotomous and ordinal data. To test differences in baseline characteristics of included patients and patients who were excluded because they either had only data at t0 available or did not have data at t0 available, the Mann Whitney U test for continuous data was used, the linear-by-linear approximation of the Pearson’s Chi-square test (exact two-sided significance) for ordinal data and the Fisher’s exact test for dichotomous data. Proportions and percentages were used to describe dichotomous outcomes and the median and range were used to describe all continuous outcomes. Differences between three timepoints were statistically analyzed by means of paired tests (i.e., Friedman test for continuous or ordinal data and a Cochran’s Q for dichotomous data) as well as the differences between two timepoints (i.e., Wilcoxon signed rank test for continuous or ordinal data and the McNemar test for dichotomous data).