Study design and setting
This was a two-centre study with an intention-to-treat, prospective, randomized, open-label trial with blinded evaluators (PROBE) design. Information on the study design can be found at ClinicalTrials.gov, identifier: NCT04164420. The study was conducted according to the recommendations for Interventional Trials and the CONSORT guidelines for randomised controlled trials (Appendix 1). Patients with a first-ever stroke and persistent swallowing dysfunction four weeks post-stroke according to the timed water-swallow test (TWST) [10-12], treated at Umeå University Hospital or Hudiksvall Hospital, Sweden, were offered to participate in the interventional study. The participants were randomized into an intervention group that received oral neuromuscular training for five weeks or a control group. A randomization list was constructed by the statistician and trial manager and was stored in envelopes locked in a cupboard. The randomization list did not include any stratification or minimization. For ethical reasons, both groups continued to receive orofacial sensory-vibration stimulation in parallel, initiated before inclusion, if the patients showed dysphagia one week after their stroke. Evaluations were made at baseline (four weeks after stroke onset), at the end of treatment (after five weeks treatment), and at the 12-month follow-up after a period without treatment to determine if any lasting positive training effect was present. At the end of treatment and 12-month follow-up, the evaluating assessors were blinded to what treatment the participants had received.
Inclusion criteria for participation were the following: first-time stroke and a pathological TWST [10-12] 4 weeks after stroke onset. Exclusion criteria were: inability to cooperate, percutaneous endoscopic gastrostomy (PEG), neurological diseases other than stroke, known history of dysphagia prior to the stroke, prominent horizontal overbite (contra-indication due to the oral device’s design), or hypersensitivity to the acrylate.
Oral neuromuscular training
An oral device (Muppy®; Dr. Hinz Dental, Herne, Germany) was used for oral neuromuscular training (see Figure 1) that aims to stimulate sensory input and strengthen the facial, oral, and pharyngeal muscles . During the training, the device was placed pre-dentally behind closed lips, and the patient sat with the body and head in a strictly upright position. The participants were instructed to hold the device against a gradually-increasing horizontal pulling force for 5–10 s, whilst trying to resist the force by tightening the lips . The participants were instructed to achieve a pulling force that was as high as possible without losing hold of the device with the lips. The exercise was performed three times per session, and three times daily before eating. If the participant was unable to hold the oral device, relatives or ward staff were instructed to assist in providing traction at a right angle to the mouth. Training was also possible when sitting in a bed with the head supported firmly. Verbal, practical, and written instructions about training at home were given to the participants, relatives, or care assistants. The participant or care giver reported the training in a log book that was evaluated after a 5-week training period. The log book was placed visible for the participant and care givers as a reminder of keeping up the training, and it was considered reliable.
Orofacial sensory-vibration stimulation
All participants in the intervention group and the control group self-administered or were assisted by relatives or ward staff in orofacial sensory-vibration stimulation (see Figure 3) by using an electric toothbrush with a round and rotating head (Oral-B®, Procter & Gamble, Cincinnati, USA) . Instructions were given on how to stimulate the buccinator mechanism [4, 7], lips, external floor, and the tongue, three times daily before a meal.
The primary outcome was changes in swallowing rate measured by TWST [10-12] at the end of treatment compared with baseline and 12 months post-treatment, within each group and comparing the intervention group with the control group. Secondary outcomes were changes in lip force measured by a lip-force test  at the end of treatment compared with baseline and 12 months post-treatment, comparing the intervention group with the control group. Swallowing dysfunction according to VFS, including the presence of aspiration-penetration as sign of unsafe swallowing as well as premature spillage and pharyngeal residue as signs of ineffective swallowing, was assessed for comparison with the TWST results.
For both study groups, the swallowing rate (TWST), lip force, and swallowing function (VFS) were assessed. A nurse, speech-and-language pathologist, or physician performed the assessments, except the VFS examinations, which were performed by a general or an oral-and-maxillofacial radiologist. A flow-chart of the included participants and the study process is shown in Figure 4.
Timed water-swallow test (TWST)
The participant first received a teaspoon of water, repeated three times consecutively. In case of a total inability to swallow, or if a swallow was misdirected (e.g., there were clinical signs of aspiration such as cough or wet/gurgly voice), the swallowing rate was graded as zero and the TWST was not performed. If the swallowing was safe during the teaspoon test, a TWST was completed.
The TWST is a timed test of a person’s ability to swallow water [10-12]. It was performed with the participants sitting in upright position. The participants were instructed to swallow water (150 mL) as quickly as possible without pausing. The time was recorded from the start of drinking until the last swallow was completed. Any remaining water in the glass was measured. The swallowing rate was calculated as the amount of swallowed water (in millilitres) divided by time (seconds). A swallowing rate of 10 mL/s was regarded as the lower normal limit [10-12]. The TWST has been demonstrated to have high intra-rater, inter-rater, and test-retest reliability and to be a valid index for assessing swallowing dysfunction [10-12, 14].
Lip force was recorded in newtons using a LF100 (MHC1 AB Detector, Gothenburg, Sweden)  (see Figure 2). The patient was instructed to hold a pre-dentally placed oral device (Muppy®) (see Figure 1) for as long as possible by tightening the lips and pressing the head backwards as the investigator applied traction at a right angle to the patient’s mouth with increasing force for 10 seconds, or until the subject lost their grip of the oral device. Each subject was measured three times, with a 2-minute rest between attempts. The maximum lip-force value was recorded and blinded for the investigator. The lower limit for normal lip force was 15 newtons .
During VFS in lateral projection, the participant initially swallowed 5 mL of iodine contrast medium (Visipaque® 320 mg I/mL) delivered with a spoon, to assess the presence of aspiration (before, during, or after swallowing). If severe aspiration was present, the examination was concluded. If no or minor aspiration was found, a bolus of 10 mL iodine contrast medium was tested. If there was no aspiration, the examination proceeded with 10 and 20 mL liquid barium sulphate contrast medium (Mixobar® Colon 1 mg/mL) delivered with a spoon, and a free amount of liquid barium sulphate from a cup. If the patient was still free of aspiration, this was followed by assessment in the frontal projection with 5 mL, 20 mL, and a free amount of barium sulphate contrast medium. To assess correspondence with TWST, oral and pharyngeal residue (bolus retention after the bolus has passed through the pharynx), premature spillage (bolus entering the pharynx without eliciting the swallowing reflex, with the airway remaining open with increased risk of aspiration), penetration (bolus entering the larynx and proceeding below the vocal cords), and aspiration (bolus entering the airway below the vocal cords) with and without cough were recorded. The improvement in swallowing function was assessed according to the penetration-aspiration scale (the PAS is scored 1–8, where score 1 means no bolus enters the airway, and scores 2–8 are different degrees of penetration or aspiration into the airway) . The PAS scores were dichotomised into ≤2 versus >2. All assessments were performed individually by two radiologists. The inter-rater agreement was calculated. The assessments were compared, and, in case of disagreement, consensus was reached in a joint reassessment that was used as the final result.
A professional statistician and a data manager from the clinical research centre at Uppsala University were involved in planning the study according to good clinical practice (GCP). In order to detect a critical change with a power of 80%, a sample size of 44 participants (22 in the intervention group and 22 in the control group) that fulfilled the study protocol was required for a type I error of 5%. Wilcoxon’s signed-rank test was used to evaluate possible changes within a group between different test occasions. Spearman´s rank test was used for correlation analysis of different test methods. The Mann-Whitney U test and Fisher´s exact test were used to compare data from the intervention group and the control group. When non-parametric tests were used, confidence intervals are presented to indicate the precision of the estimates. Inter-rater reliability was estimated using Cohen’s Kappa (κ). A value of P < 0.05 was considered significant. All statistical analyses were performed using SAS version 9.1 (SAS Institute Inc., Carey, NC, USA).