Introduction

Dysphagia is a disability affecting 8–16% of the general population and may concern more than 50% of the elderly population [1,2,3,4]. Many medical conditions are known to lead to dysphagia, i.e., head and neck cancers, cerebral vascular accidents, laryngopharyngeal reflux, laryngopharyngeal allergy, presbyphagia, etc. [5,6,7]. The prevalence of dysphagia in most of these conditions increases with age [8]. Patients with dysphagia often have substantial impairments of physical, emotional and social quality of life [9] and they may encounter several comorbidities such as respiratory infections, aspiration pneumonia, functional disability and frailty, malnutrition and death [4, 10]. These complications can be associated with substantial morbidity and mortality and need to be early identified. In this way, many clinical tools have been developed to detect dysphagia in patients at risk; ensuring appropriate precautions and interventions.

In 2008, Belafsky et al. developed Eating Assessment Tool (EAT-10) that is a self-administered, symptom-specific outcome instrument for dysphagia [11]. According to some clinical studies, EAT-10 is a valid clinical tool with moderate-to-high internal consistency, reliability and discriminative validity [12,13,14]. Currently, EAT-10 is used for the initial assessment of dysphagia [11, 15], the detection of aspirations [16, 17], and the follow-up of patients with dysphagia benefiting from rehabilitation [15, 16].

To date, there is no validated French version of the EAT-10 available for use in French-speaking countries, which include more than 400 million inhabitants. In this paper, we present a version of EAT-10 adapted for French speakers (Fr-EAT-10), and we assess its test–retest reliability, internal consistency, and clinical validity with the aim of providing the French-speaking community with an effective tool for the detection of dysphagia, aspirations and the monitoring of therapeutic approaches in patients with swallowing disorders.

Materials and methods

The study protocol was approved by the local ethics committees of CHU Saint-Pierre [Brussels, Belgium (no B076201733642)] and EpiCURA Hospital [Baudour, Belgium (no A2014/001)]. Patients were invited to participate, and investigators obtained informed consent from patients enrolled in the study. The first author of the study (JRL) obtained the permission to develop and publish Fr-EAT-10 from Dr. Peter C. Belafsky (The University of California, Davis, CA, USA), the creator of the original EAT-10.

Translation and development of Fr-EAT-10

A multidisciplinary team composed of an otolaryngologist, two psychologists, one statistician, two speech therapists and one linguist worked on the French adaptation of the American version of EAT-10 [11]. All members of the team were native French speakers. The team carefully analyzed misunderstandings and the Fr-EAT-10 has been improved to remain as the current version of Fr-EAT-10 (Fig. 1).

Fig. 1
figure 1

French version of Eating Assessment Tool (EAT-10). The French version of EAT-10. A multidisciplinary team mainly composed of French native speakers translated the American version of EAT-10

Full size image

Participants

Fifty-six patients with oropharyngeal dysphagia were enrolled from March 2017 to April 2018 from the Departments of Otolaryngology–Head & Neck Surgery of CHU Saint-Pierre (Brussels, Belgium) and EpiCURA Hospital (Ath, Belgium). Patients had a medical indication for an instrumental evaluation of swallowing and benefited from fiberoptic endoscopic evaluation of swallowing (FEES) and videofluoroscopy. There were 41 males and 11 females with a mean age of 66.4 ± 13.7 (ranged from 44 to 93 years old). Patients with dementia, severe neurological diseases limiting the understanding of the study protocol, severe respiratory disease and a high risk of superinfection, as well as those who were not native French-speaker were carefully excluded. Four patients were excluded due to Illiteracy. Fifty-two patients completed the study. The characteristics of patients included in this study are available in Table 1. The etiology of oropharyngeal dysphagia mainly included head and neck carcinoma, esophageal neoplasia, neurological disorders and psychological dysphagia (Table 1). Seventy-three healthy subjects composed the control group. The mean age was 27.0 ± 9.1 (ranged from 20 to 67 years old) and there were 24 females. Healthy subjects were recruited from the Université Libre de Bruxelles and they had no history of any swallowing disorders, laryngopharyngeal reflux disease (according to French version of Reflux Symptom Score [18]), smoking habits, medical history impacting swallowing or any history of head and neck surgery.

Table 1 Patient characteristics
Full size table

Questionnaires and swallowing examinations

Patients and controls completed Fr-EAT-10 twice over a 7-day period (Fr-EAT-10 d0 and Fr-EAT-10 d7). Moreover, patients and controls fulfilled French version of dysphagia handicap index (DHI), a validate self-estimated questionnaire constructed in French [19] at baseline and an the same time benefited from FEES and videofluoroscopy. FEES included a static and dynamic evaluation of the structures in the upper aerodigestive tract and an examination of laryngopharyngeal sensitivity (by directly stimulating the various pharyngeal–laryngeal areas with the tip of the rhinopharyngolaryngoscope). We used water colored with methylene blue and/or blue-dyed food for the assessment of swallowing. The videofluoroscopy was performed with the following material: Siemens Axion Luminos dRF (Siemens, Healthcare, GMBH, Erlangen, Germany). Full-face and profile video were analyzed by two experienced physicians in a blind manner using the Group for Learning Useful and Performant Swallowing (GLUPS) score (Appendix 1) and the Penetration–Aspiration Scale (PAS) of Rosenbek et al. [20] Questionnaires with missing items were not accepted.

Reliability was assessed through internal consistency and test–retest reliability for all subjects (N = 125). Internal consistency of Fr-EAT-10 was measured using Cronbach’s alpha for the 10 items of the questionnaire. Test–retest reliability between Fr-EAT-10 d0 and Fr-EAT-10 d7 was assessed for each item and for the total score using Spearman’s correlation coefficient. External validity was measured by a correlation study between EAT-10 d0 and DHI using Spearman’s correlation coefficient. Internal validity was measured by a comparison of Fr-EAT-10 d0 of patients and healthy controls using Mann–Whitney U test.

The normative value of Fr-EAT-10 was calculated on the data of the healthy individuals determining the superior threshold of two standard deviations below the mean (N = 73). FEES and videofluoroscopy were used to detect aspiration. To assess the ability of Fr-EAT-10 to detect aspiration, we firstly studied the association between Fr-EAT-10 (items and total score) and the occurrence of aspiration during videofluoroscopy according to the PAS (Spearman’s correlation test). A PAS score ≥ 4 was considered as abnormal [20]. Secondly, we performed a receiver operating characteristic (ROC) curve with regard to the sensitivity and the specificity of each potential threshold values of Fr-EAT-10. The area under the curve was measured to obtain the best threshold value associated with the higher sensitivity and specificity.

Statistical analyses

Statistical Package for the Social Sciences for Windows (SPSS version 22.0; IBM Corp, Armonk, NY, USA) was used to perform the statistical analyses. A level of significance of p < 0.05 was used. The difference in Fr-EAT-10 score between patients and controls was evaluated using the Mann–Whitney U test (internal validity). According to the distribution of data, Spearman’s correlation test was used to perform the different correlation analyses (test–retest reliability, external validity). ROC curve was performed by the same software.

Results

Cronbach’s alpha for the 10 items of Fr-EAT-10 d0 for patients and controls was 0.95 indicating high internal consistency. According to the Spearman correlation test, the test–retest reliability was high for total scores (rs = 0.921, p < 0.001) and moderate-to-high for all item scores (Table 2). External validity analyses reported a high correlation between Fr-EAT-10 total score and DHI (Table 3) indicating high external validity.

Table 2 Test retest reliability analysis
Full size table
Table 3 External validity analysis
Full size table

Concerning internal validity, the mean Fr-EAT-10 scores of patients and controls were 16.33 ± 10.90 and 0.55 ± 1.26 respectively (Table 4); the difference between group being significant (p < 0.001, Mann–Whitney U test). About the normative data, a cut-off score of 3 was considered to be reflective of abnormalities. The mean Fr-EAT-10 scores of patients with Head and Neck cancer (HNC) and patients without HNC were respectively 18.9 ± 10.5 and 11.3 ± 10.1 and aspiration prevalence were 40% and 9.1%, respectively.

Table 4 Internal validity analysis
Full size table

The characteristics of patients according to FEES and GLUPS score (videofluoroscopy) are described in Table 5. The mean Fr-EAT-10 score of patients with aspiration was 21.4 ± 9.93, while the mean Fr-EAT-10 score of those without aspiration was 14.2 ± 11.2. The correlation between Fr-EAT-10 score and the occurrence of aspiration is moderate but is still significant (rs = 0.327, p < 0.05). The correlation analysis between each item and aspiration reported that the stronger correlation concerns the item 9 (cough during the eating) of Fr-EAT-10 (rs = 0.506). According to the ROC curve, the cut-off value that exhibited better sensitivity and specificity for patients with aspiration is 17/40 (sensitivity: 82.35% and specificity: 57.1%). This indicates that aspirations need to be highly suspected for patients with Fr-EAT-10 ≥ 17 (Fig. 2). According to the prevalence of aspiration in our study (32.7%), the positive predictive value (PPV) and negative predictive value (NPV) were 48.3% and 86.9%, respectively (Table 6).

Table 5 Videofluoroscopy and FEES characteristics of patients
Full size table
Fig. 2
figure 2

ROC curve. The grayed value is the Fr-EAT-10 score showing the higher sensitivity and specificity; indicating that 17 is the most appropriate cut-off value for the detection of aspiration

Full size image
Table 6 Contingency table
Full size table

PPV and NPV especially for item 9 (cough during eating) of Fr-EAT-10 were 46.67% and 86.3%, respectively.

Discussion

Dysphagia is frequently encountered in otolaryngology and is associated with significant rates of morbidity and mortality. Early screening is, therefore, recommended to reduce the related risk of complications. The initial version of the EAT-10 reported high validity and reliability for the screening of dysphagia, aspiration and for the evaluation of treatment efficiency [11, 21]. EAT-10 is used worldwide and other translated versions of EAT-10 reported high validity and reliability [12, 13, 33].

In the present study, we sought to develop a French version of EAT-10 and to assess its internal consistency, test–retest reliability, clinical validity and its ability to detect aspirations.

Internal consistency was excellent (Cronbach’s alpha 0.95) which is consistent with other versions of EAT-10 (Table 7) [11,12,13, 22,23,24,25]. Considering that the value of the coefficient should be above 0.70 to show sound reliability, our results reported a high reliability of the Fr-EAT-10.

Table 7 Comparison of internal consistency of the different version of the EAT-10
Full size table

The correlation coefficients for the test–retest reliability were high (rs = 0.921) for both total and item scores, indicating a good external reliability. Our results are consistent with those described in other studies of EAT-10 validation, which were all ≥ 0.85 [12, 22,23,24,25,26]. External validity analysis reported a high correlation between Fr-EAT-10 total score and the DHI indicating high external validity. In other words, similar to DHI, Fr-EAT-10 appears to be able to detect and evaluate the severity of swallowing disorders. However, the comparison with other studies is difficult because no similar procedure of external validity assessment using DHI was conducted. In the literature, two authors assessed the external validity of EAT-10 throughout a statistical correlation analysis with another patient-reported outcome questionnaire [22, 24]. In the Portuguese version of EAT-10, Nogueria et al. found a significant correlation between EAT-10 and Quality of Life instrument EuroQoL (EQ-5D) [24]. Similarly, the Italian version of EAT-10 seemed to have high external validity in regard to a significant correlation between EAT-10 and Dysphagia Outcome Severity Scale (DOSS) [22]. The internal validity has been demonstrated using a statistical comparison between patient and control mean of Fr-EAT-10 scores. As expected, the analysis reported that Fr-EAT-10 score was significantly higher in patients in comparison with controls that are consistent with the results of previous studies [11, 13, 22, 24,25,26].

About the normative data, a score of Fr-EAT-10 > 3 has been identified as abnormal. During the development of the Hebrew version EAT-10, Abu-Ghanem et al. found a sensitivity of 92.3% and a specificity of 97.3% when EAT-10 score of 3 was used as the cutoff for dysphagia [25]. In both Swedish and Italian versions of EAT-10, authors also identified a threshold ≥ 3 to distinguish patients and controls [12, 22]. Finally, in a large cohort of patients with oropharyngeal dysphagia with or without aspiration, Giraldo-Cadavil et al. found a cut-off value ≥ 2 with a sensitivity of 93.6% and a specificity of 36.4% [13]. Overall, the majority of studies has identified a cut-off ≥ 3 as abnormal.

When EAT-10 was first established, one of the purposes of the questionnaire was to predict aspiration risk in patients with dysphagia [11, 21]. Our analysis reports a moderate but significant correlation between the Fr-EAT-10 and the occurrence of an aspiration (rs = 0.327). Moreover, our ROC curve analysis suggests that aspirations should be highly suspected for patients with Fr-EAT-10 ≥ 17; this cut-off value exhibiting the higher sensitivity (82.3%) and specificity (57.1%). Two groups of authors performed similar procedures and obtained relatively similar results, respectively, rs = 0.273 [21] and rs = 0.660 [27]. The lower cut-off value of PAS score adopted in the study of Arrese et al. was substantially different from our (PAS ≤ 2 and PAS ≥ 3 versus PAS < 4 and PAS ≥ 4) that could explain the differences found between our correlation coefficient values. In a retrospective study, Kendall et al. did not find significant correlation (rs = 0.03) between EAT-10 and aspiration [28]. Giraldo-Cadavil et al. also investigated the ability of EAT-10 to detect aspiration [13]. Using ROC curve, these authors support that EAT-10 ≥ 4 is associated with a sensitivity of 94.3% and a specificity of 49.5%. The etiology of dysphagia, the severity of the diseases and the related mean value of EAT-10 score substantially vary between studies that may explain some differences in our respective results. Thus, the proportion of patients with head and neck cancer in both the study of Kendall et al. [28] and in the present report was, respectively, 1.4% and 58%, limiting the comparison. In the same way, the cohort of Giraldo-Cadavil et al. [13] is mainly composed of patients with cerebrovascular disease who are characterized by different pathophysiological mechanisms underlying the development of dysphagia and aspiration in comparison with patients with head and neck cancers [29, 30]. In the context of very different populations, the comparison between studies focusing of the ability of EAT-10 to detect aspiration is still difficult and can lead to unclear conclusion. For this reason, the future establishment of cut-off values associated with a higher risk of aspiration could take into consideration the etiology of dysphagia, the pathophysiological mechanism of the disease, the clinical course of the disease and the related risks of aspiration.

The prevalence of aspirations in this study was 32.7%; PPV and NPV were, respectively, of 48.3% and 86.9%. If we look at the PPV and NPV using only the item 9 of the Fr-EAT-10, 46.6% and 86.3%, respectively, we found out that this item, “cough during eating”, is a very strong item and has a huge impact on the detection of aspirations in our studied population. As exhibited in Table 8, these PPV and NPV were relatively close to those of the study of Cheney et al. but are slightly inferior to those of Rofes et al. [21, 31]. However, the cohort study of Cheney et al. is mainly composed of patients with head and neck cancer as our cohort; while the cohort of Rofes et al. is mainly composed of patients with neurodegenerative disorders and gastroesophageal reflux disease [31]. In addition to the potential impact of the type of the patient population on the establishment of cut-off value, it has been demonstrated that the method used to detect aspiration also has a significant impact on prevalence, sensitivity, specificity, positive and negative predictive values [32, 33]. In the present study, we used both videofluoroscopy and FEES for the detection of aspirations, while Rofes et al. only used videofluoroscopy. It is therefore possible that the only use of videofluoroscopy led to an underestimation of the aspiration; that could be highlighted by the different prevalences of aspiration reported in our two respective studies (18.9% versus 81%).

Table 8 Comparison of the studies using EAT-10 as screening tool for aspirations
Full size table

In that respect, the main strength of our study is the use of both FEES and videofluoroscopy to detect aspirations. Indeed, on 17 aspirations, only one-third was detected by both videofluoroscopy and FEES, one-third by FEES, and the last third by videofluoroscopy. In practice, we did not find substantial difference between both methods in the detection of aspiration and they are complementary. We support that it is still important to perform both examinations to ensure a high detection rate of aspiration according to the physician experience, local availability and cost of videofluoroscopy because both methods are characterized by different sensitivity and specificity in the detection of aspiration [32, 33]. In summary, videofluoroscopy assesses oro-pharyngeal disorders in terms of coordination and provides information about esophageal disorders, while FEES provides information about the anatomical lesions, the sensitivity of the upper aerodigestive tract mucosa and the occurrence of potential anatomical abnormalities [32,33,34].

The present study has some limitations. Firstly, although some previous studies are characterized by similar number of patients [25], the low number of patients reduces the statistical power. Secondly, the large proportion of patients with head and neck cancer could limit the comparison of some analyses, (i.e., the establishment of a cut-off value for the detection of aspiration) with other studies that are characterized by patients with different etiologies of dysphagia.

Conclusion

The Fr-EAT-10 is a valid and reliable self-administered survey for the detection of dysphagia and aspiration. Fr-EAT-10 seems to be highly reproducible, with a good construct-based and criterion-based validity. Fr-EAT-10 > 3 can be considered as abnormal and suggestive of swallowing disorder. Particularly in patients with head and neck cancers, a cut-off value of Fr-EAT-10 of 17 or higher could be associated with a substantial risk of aspiration. However, the establishment of these thresholds should be influenced by the disease underlying the dysphagia and the characteristics of the patient population. Future studies should take into consideration these two parameters in the determination of adequate thresholds.