A New Ultrasound Score for Swallowing Disorders Diagnosis in Difficult-to-wean Tracheostomised Patients

Intensive Care Unit (ICU) stays often lead to Swallowing Disorders (SD). Physiological and biomechanical changes due to tracheostomy increase the risk of SD, which can have serious consequences for quality of life and overall health.

Experts recommend systematic bedside screening. Most algorithms include the water swallow test followed by expert comprehensive swallowing assessments of screening-positive patients. Early intervention strategies could greatly benefit, highlighting the importance of assessments identifying which specific structures require intervention and enabling a targeted rehabilitation approach. However, clinical examinations lack precision.

Complementary examinations, such as videofluoroscopy or Fiberoptic Endoscopic Evaluation of Swallowing (FEES), are irradiating or invasive, consume numerous resources, and are difficult to access. In this context, ultrasound imaging has garnered increased interest for assessing swallowing disorders. Recognized for its non-invasive nature and its ability to overcome conventional constraints, ultrasound holds promise in integrating morphological and dynamically swallowing-related structures evaluation. This approach, enabling both qualitative and quantitative analysis of multiple structures, would enhance precision in targeting structures for rehabilitation.

A recent study shows strong potential for an ultrasound prediction model for diagnosing swallowing disorders in outpatients. The researchers intend to create an ultrasound diagnostic model for difficult-to-wean tracheostomized patients. Additionally, they aim to evaluate the reliability and accuracy of this model. Hypothesis: The investigators hypothesize that an ultrasound predictive model is valid and reliable for diagnosing swallowing disorders in difficult-to-wean tracheostomized patients.

Main objective: The primary aim is to develop an ultrasound predictive model based on the most relevant ultrasound measurements of the aerodigestive tract.

Secondary objectives:

• To evaluate the validity and reliability of a new ultrasound score for diagnosing swallowing disorders.
• To assess the reliability and validity of tongue movement in diagnosing swallowing disorders.
• To evaluate the reliability and validity of hyoid bone movement in diagnosing swallowing disorders.
• To assess the reliability and validity of digastric muscle ultrasound evaluation in diagnosing swallowing disorders.
• To evaluate the reliability and validity of genohyoid muscle ultrasound evaluation fraction in diagnosing swallowing disorders.

Study design:

This prospective observational multicenter study will adhere to the ethical standards of the Declaration of Helsinki and will be reported according to the standards for transparent reporting of a multivariable prediction model for individual prognosis or diagnosis guidelines (TRIPOD).

Population:

One hundred and nineteen difficult-to-wean tracheostomized patients will be enrolled. All consecutive patients with difficult-to-wean tracheostomies will be screened, and those meeting the eligibility criteria will be enrolled. Patients will provide oral consent to participate in accordance with French clinical research laws. Data collection, Fiberoptic Endoscopic Evaluation of Swallowing (FEES), and ultrasound (US) evaluations will be conducted within a two-hour timeframe.

Data collection:

Data will be recorded using an online case report form (Datacapt). Demographic data, comorbidities, medical, and surgical history will be recorded at enrollment. On-site monitoring is planned monthly.

Ultrasound examination:

Ultrasonography will be performed using both linear and convex probes. The ultrasonographer will be blinded to the patient's status and prior clinical or imaging assessments. All measurements will be performed in a semi-recumbent position (30°) with the neck in a neutral position. Subjects will be asked to swallow saliva to assess active movements and muscle contractions. Parameters such as tongue thickness, movement, and echointensity, hyoid bone movement, and suprahyoid muscle characteristics (thickness, cross-sectional area, and echointensity) will be evaluated. Suprahyoid muscles will be assessed both at rest and during swallowing, and muscle contraction ratios will be calculated accordingly. Detailed information on the ultrasound procedure is described in the research protocol.

To assess inter-examiner reliability, during the initial ultrasound evaluation, tests will be conducted by two different operators on a consecutive sample of patients. To limit the examination time for each patient, another consecutive sample of patients will be used for intra-examiner reliability. For the patients involved in the reliability assessment, an additional 10 minutes will be allocated.

FEES:

They will be performed by the attending physician in the patient's room. A physiotherapist or speech-language therapist will be present to assess swallowing and the involved structures systematically.

Evaluators will examine the integrity and functionality of structures such as the arytenoids or vocal cords, and assess laryngeal and hypopharyngeal sensitivity. The examination will conclude with swallowing trials using various textures to observe swallowing, enabling identification of any inhalation. Different bolus consistencies, viscosities, and volumes will be used during the assessment to evaluate the patient's full range of dietary options, based on preswallow assessments. The Yale Pharyngeal Residue Severity Rating Scale will be used to characterize the severity and localization of residues observed after each swallowing trial.

Statistical plan:

Descriptive statistics will summarize quantitative variables with means (± standard deviation) or medians [minimum-maximum], while qualitative variables will be described using numbers and percentages. Univariate comparisons will employ standard statistical tests, such as Chi-square, Fisher's test, t-test, ANOVA, Wilcoxon, or Kruskal-Wallis tests, with a significance level of 5% and 95% confidence intervals. Association measures between variables will use Pearson's or Spearman's correlation coefficients for quantitative data and Chi-square or Fisher's tests for qualitative data. Agreement assessment will utilize the Bland-Altman method for quantitative variables and the Kappa coefficient for qualitative variables.

To establish an ultrasound diagnostic model for diagnosing swallowing disorders, variables will be dichotomized based on a linear relationship hypothesis tested using ROC curve analysis. Patients will be categorized into positive (presence of swallowing disorders) and negative groups (absence of swallowing disorders) using FEES diagnosis. Ultrasound variables differing significantly between these groups (p < 0.20) will be included in multivariate logistic regression, guided by Akaike's information criterion (AIC) for variable selection while considering multicollinearity. The multivariate logistic regression scores will weight each ultrasound variable. Researchers will transform the model into an easy-to-use scoring system.

Validity will be determined by ROC curve analysis to establish the ultrasound score threshold for diagnosing swallowing disorders. Sensitivities, specificities, positive and negative predictive values, and diagnostic odds ratios will assess the score's diagnostic accuracy. Reliability, assessed with a correlation coefficient (ρ=0.80), requires a sample of 20 patients for a desired correlation coefficient (ρ0=0.50), ensuring sufficient power (α=0.05, β=0.20). Inter- and intra-examiner reproducibility will be evaluated using the Kappa coefficient.

Sample size calculation:

According to Harrell's recommendations on the construction of multivariate models, a maximum of 1 variable per 10 events should be included. The prevalence of swallowing disorders in our patient population is highly variable in the literature (e.g., 11-93% according to Skoretz or 50-83% according to Garuti et al). Based on our PMSI data, we estimate the prevalence of swallowing disorders at 65%. We aim to include a maximum of 4 variables in the multivariate model, i.e., 114 patients are required.