Definition of Clinically Meaningful Thresholds for the Brief Esophageal Dysphagia Questionnaire and Eckardt Score: A Multicenter Psychometric Study

ABSTRACT

Background

The Brief Esophageal Dysphagia Questionnaire (BEDQ) and Eckardt score (ES) are the most used patient‐reported outcomes (PROs) for dysphagia. However, their patient‐centered cutoffs, minimal important difference (MID), and test–retest reliability have not been established. We aimed to address these gaps and evaluate BEDQ treatment responsiveness.

Methods

This prospective multicenter study was conducted in tertiary centers in Spain and Latin America and enrolled adults referred for high‐resolution manometry. Three cohorts were derived: the cutoff cohort (subsequently subdivided into a derivation and a validation cohort), the test–retest cohort (consecutive subsample re‐administered PROs), and the treatment responsiveness cohort (patients treated for achalasia). Self‐perceived severity was evaluated using a 5‐point Likert Patient Global Impression of Severity (PGIS) scale and applied as PGIS < 2 (No/Mild symptoms (NMSS)) and ≥ 2.

Results

The cutoff cohort included 571 patients, subdivided into derivation (397) and validation (174) cohorts. A BEDQ cutoff of < 9 and ES < 4 optimally predicted NMSS. The validation cohort and bootstrapping confirmed the stability of these cutoffs. MID was 5 points for BEDQ and 1 point for ES, consistent across cohorts. BEDQ showed excellent test–retest reliability (intraclass correlation coefficient = 0.97), significantly higher than ES (0.85). Both PROs showed statistically significant reductions following treatment (p < 0.001 for both), with standardized response means of 2.14 and 1.98 for BEDQ and ES, respectively.

Conclusions

BEDQ < 9 and ES < 4 are clinically meaningful thresholds for identifying patients with NMSS. Changes exceeding 5 points for BEDQ and 1 point for ES represent clinically meaningful differences. BEDQ demonstrates superior psychometric properties compared to ES.

Keypoints

• The Brief Esophageal Dysphagia Questionnaire (BEDQ) and Eckardt score (ES) are the most used tools to grade dysphagia severity, but their patient‐centered cutoffs and test–retest reliability have not been established.

• We found that BEDQ < 9 and ES < 4 predict no or mild dysphagia. BEDQ showed superior test–retest reliability compared to the Eckardt score.

• BEDQ < 9 and ES < 4 can be used to define no/mild dysphagia. Changes exceeding 5 BEDQ points or 1 ES point represent clinically meaningful change. BEDQ has psychometric advantages over Eckardt.

Plain Language Summary

The BEDQ and Eckardt scores are extensively used to grade dysphagia. Nevertheless, clear thresholds to predict mild symptoms or define meaningful clinical changes are lacking. In a multicenter study we recruited over 500 patients from Spain and Latin America and performed an extensive psychometric evaluation of these scores. We found that scores below 9 for the BEDQ and 4 for the Eckardt score identify patients with no or minimal symptoms. We also found that a 5‐point improvement in the BEDQ and a 1‐point change in the Eckardt score represent a meaningful clinical change. Remarkably, the BEDQ showed significantly greater reliability than the Eckardt score for result stability over time. These cutoffs and values now can help clinicians with precise targets to monitor treatment success in patients with dysphagia.

This multicenter study identified BEDQ < 9 and Eckardt < 4 as thresholds for no/mild dysphagia. Minimal important differences were 5 (BEDQ) and 1 (Eckardt). BEDQ demonstrated superior reliability and responsiveness compared to the Eckardt score, providing validated, patient‐centered tools for monitoring clinical success in dysphagia patients.

1. Introduction

Dysphagia is highly prevalent, affecting 3% of the general population [1]. It impacts individual's nutritional status, quality of life, and overall health [2]. It is not possible to fully characterize its complete nature using only objective measurements, requiring the use of patient‐reported outcomes (PROs) [3].

A correct assessment of symptom severity drives better care. It informs every step—from the selection and timing of treatment to measuring efficacy, and even shaping broader public health strategies [4, 5]. On the other hand, the Chicago Classification version 4.0 (CC4.0) emphasizes that specific motility disorders should only be considered clinically relevant in the presence of compatible symptoms and supportive testing [6]. Therefore, a standardized tool for grading dysphagia is crucial for both clinical and research purposes.

The proper use of any PRO necessitates a prior assessment of its reliability and validity [7, 8, 9]. The US Food and Drug Administration (FDA) strongly advocates the use of PROs that have proven these qualities through psychometric analysis [10]. During the last 30 years, the Eckardt score (ES) [11] has been commonly used to grade obstructive esophageal symptoms in achalasia, as well as other esophageal motility disorders [12, 13, 14, 15]. Nevertheless, all studies to date have shown that its psychometric characteristics are far from optimal [16, 17]. The Brief Esophageal Dysphagia Questionnaire (BEDQ) [18] is a short PRO developed following current design and validations recommendations [10], suitable for grading dysphagia across different etiologies. Ranging from 0 to 40, it has shown strong psychometric properties in different populations [18, 19, 20, 21]. It has been increasingly used to grade dysphagia in different diseases [22, 23, 24, 25, 26, 27]. A threshold of 9–10 points has been used. This was the cutoff suggested by the original description study, determined as the best predictive value for major motor dysfunction in High Resolution Manometry (HRM) [18]. In the case of ES, a cutoff of 4 points was used in its description [11], empirically defining a score of 0–3 as clinical remission. Slone et al. determined an ES cutoff of 4 as the best predictor of clinical remission, defined by the review of medical records by two clinicians [28]. To our knowledge, there is no study specifically addressing the best cutoff using patients' perception as an anchor, for either BEDQ or ES.

Two other relevant components of the psychometric evaluation have not been evaluated for either of these PROs: the minimal important difference (MID), which is the smallest change in a score that patients perceive as beneficial, and the stability of their results over time (a key component of a PRO's reliability) using a test–retest evaluation.

Finally, as the BEDQ is a novel score, its responsiveness to treatment has not been formally evaluated.

Thus, the main objective of this study was to determine the best cutoff value for the BEDQ and ES from a self‐reported patient's symptomatic burden perspective. The secondary objectives were as follows: (i) to estimate the MID and test–retest stability for the BEDQ and ES, and (ii) to evaluate treatment responsiveness for the BEDQ in patients with achalasia.

2. Materials and Methods

From October 2023 to December 2025, adult patients were prospectively recruited among those referred for HRM at six specialized centers in Spain and Latin America (Clinica Alemana de Santiago (Chile); Hospital Clínico de la Pontificia Universidad Católica de Chile (Chile); Hospital San Ignacio‐Pontificia Universidad Javeriana de Bogotá (Colombia); Hospital Universitari Vall D′ Hebron‐Barcelona (Spain); Hospital Universitario, Fundación Favaloro‐Buenos Aires (Argentina); and Universidad Veracruzana, Medical Biological Research Institute‐Veracruz (Mexico)). From this registry, we derived three cohorts: the cutoff, test–retest, and treatment responsiveness cohorts (Figure 1).

FIGURE 1.

Flow diagram of patient recruitment and allocation for the cutoff, test–retest, and treatment responsiveness analyses.

The cutoff cohort consisted of patients with esophageal symptoms referred to HRM. The exclusion criteria were: (1) Previous esophagogastric surgery or procedures. (2) Patients referred for oropharyngeal dysphagia. (3) Incomplete fulfillment of the required information. Then, this cohort was randomly separated into two sub‐cohorts, which were used for derivation and subsequent validation of cutoff values, using a preestablished proportion of 70% and 30%, respectively [29].

The test–retest cohort consisted of a consecutive subsample drawn from patients enrolled in the cutoff cohort between September and December 2025, using the same inclusion/exclusion criteria. In centers that could logistically do so, patients recruited during this period were contacted by phone within 2–5 days of their index manometry visit. Patients were included in the test–retest analysis if they were successfully reached by phone, agreed to re‐administer the BEDQ and ES, and reported no clinical events that could account for a change in their dysphagia.

For the treatment response cohort, post‐treatment achalasia patients were simultaneously recruited during the September–December 2025 period. Patients were included if they had been treated for proven achalasia during the last 6 months. These patients described their post‐treatment symptomatic status using a dichotomous standardized adequate or inadequate relief question. In a subgroup of this cohort, pre‐treatment data were also available, and we used these patients to calculate the intraindividual change in the BEDQ score following treatment.

Before the HRM study, all patients completed clinical and epidemiological questionnaires, including the BEDQ, ES, Gastroesophageal Reflux Questionnaire (GERDQ), and a 5‐point patient global impression of severity (PGIS). We used validated Spanish versions of each questionnaire [17, 19, 30].

The BEDQ is an 8‐question PRO [18] that evaluates dysphagia burden. Each question is rated on a 6‐point Likert scale that includes avoidance behaviors. It ranges from 0 to 40, with higher scores indicating greater esophageal dysphagia severity.

The ES is a widely used 4‐question PRO that evaluates dysphagia burden in achalasia patients [11]. Each question is scored 0 to 3 points, and the total score corresponds to their sum.

The GERDQ is a 6‐question score that evaluates GERD‐related symptoms [31]. Each question is rated on a 4‐point Likert scale. The score is generated by summing four positive predictors (scored 0–3) and two reverse Likert scale items of negative predictors (scored 3–0). A cutoff of 8 points is used.

We used a 5‐point Likert PGIS (0 = None, 1 = Mild, 2 = Moderate, 3 = Severe, and 4 = Very severe). A spanish version was provided by the Mapi Research Trust, Lyon, France (https://eprovide.mapi‐trust.org) [32]. Simple PGISs have been recommended as a valid tool for the evaluation of symptomatic burden [10, 33, 34, 35] and have been previously used in the evaluation of esophageal conditions [36]. As patients with no or mild symptoms are usually not candidates for escalation therapy, we included these patients into one category, which we called NMSS (No/Mild symptom severity).

All centers used Medtronic devices. All HRM studies were performed, analyzed, and classified according to CC4.0 [6, 37]. Briefly, patients fasted for at least 4 h, and a solid‐state manometry catheter with 36 circumferential sensors was transnasally placed.

Ethical approval was obtained from each participating center and all patients signed provided consent before enrollment in the study.

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

2.1. Statistical Analysis

Sample sizes were calculated a priori for each analysis using established methods: 194 patients for ROC curve analysis (easyROC v1.3.1, AUC = 0.6, α = 0.05, 80% power) [38]; 31 for test–retest reliability (ICC = 0.6, α = 0.05, 80% power) [39]; 19 for between‐group treatment responsiveness (effect size = 0.7, α = 0.05, 80% power); and 9 for intra‐patient pre/post analysis (effect size = 1.5, α = 0.05, 95% power), the latter two using GPower 3.1.9.2.

All analyses were performed using SPSS v26 and R v4.4.1 (R Core Team, 2024) [40]. Kolmogorov–Smirnov test showed that both BEDQ and ES have a non‐normal distribution (D(397) = 0.16, p < 0.001; D(397) = 0.15, p < 0.001, respectively). For group comparisons, we used the Mann–Whitney U test for continuous variables and the chi‐square test for categorical variables. Kruskal–Wallis test was used for comparing three or more groups. Correlations were evaluated using Spearman's correlation coefficient, and the scores' reliability was described using Cronbach's α. Logistic regression was used to evaluate the predictive capacity of certain variables. In the ROC curve analysis, the accuracy was estimated using the area under the ROC curve (AUROC) and the best cutoff values were calculated using the Youden index. Comparisons between AUROCs were performed using the DeLong test. To estimate the stability of the cutoff values, non‐parametric bootstrapping was applied to the derivation cohort. In each of the 2000 iterations, patients were resampled with replacement from the derivation cohort, and the Youden optimal cutoff was recalculated on each resampled dataset. The distribution of these 2000 cutoff values was used to construct 95% bootstrap confidence intervals (CI). A cutoff was considered stable if the value from the derivation and validation cohorts remained within the 95% confidence interval (CI) of the bootstrapping analysis.

As different approaches to determine the Minimal Important Difference (MID) use different assumptions, current recommendations advocate for simultaneous application of multiple methods [10, 41, 42, 43]. Anchor‐based methods are recommended as the primary reference, and distribution‐based estimates should provide supportive bounds. The final MID should be determined by triangulation—that is, by selecting the value at which approaches converge [44, 45]. We applied three methods: the anchor‐based method used the mean difference in BEDQ and ES scores between patients with NMSS (PGIS 0–1) and those with PGIS ≥ 2 as the primary estimate of clinically meaningful change. The first distribution‐based method defined the MID as 0.5 standard deviations (SD) of the score at baseline [41]. The second distribution‐based method estimated the standard error of measurement (SEM) using SEM = SD√(1 − Cronbach's α) [41, 46], applying previously reported Cronbach's α values of 0.9 for BEDQ [18, 19, 20] and 0.6 for ES [17].

Test–retest reliability was evaluated using the intraclass correlation coefficient (ICC), using a two‐way random‐effects, absolute agreement, single measures model. ICC values were interpreted according to the guidelines proposed by Koo and Li: values less than 0.50, between 0.50 and 0.75, between 0.75 and 0.90, and greater than 0.90 were indicative of poor, moderate, good, and excellent reliability, respectively [47].

BEDQ and ES treatment responsiveness was evaluated in patients with achalasia (a separate cohort) in two ways: (1) Using the Mann–Whitney test, we compared the post‐treatment values in the self‐reported adequate vs. inadequate relief groups. (2) Pre‐ and post‐treatment scores were compared using the Wilcoxon signed‐rank test for paired samples. The standardized response mean (SRM) was reported, and previously suggested thresholds were applied (small = 0.2, medium = 0.5, large = 0.8) [48, 49].

To account for multiple hypothesis testing, Bonferroni correction was applied within three pre‐specified inferential families: (1) primary cutoff analyses (logistic regression, Mann–Whitney comparisons, and ROC curve significance for BEDQ and ES; α = 0.0083); (2) secondary discriminant analyses for severe symptoms and abnormal HRM (α = 0.025); and (3) treatment responsiveness analyses (Mann–Whitney and Wilcoxon tests; α = 0.0167). Descriptive and model‐comparison statistics (Kolmogorov–Smirnov, DeLong test, ICC) were not subject to correction as they do not constitute hypothesis tests within an inferential family.

3. Results

3.1. Cutoff Determination Cohort

Between October 2023 and December 2025, 795 patients were referred for HRM. After applying the exclusion criteria, 571 patients were included in the final analysis (Figure 1). This cohort was subsequently randomly separated into a derivation (397 patients, 70%) and validation cohort (174 patients, 30%).

3.1.1. Derivation Cohort

Two hundred and thirty‐eight (59.9%) were female and the mean age was 53 ± 12 years. Patients showed a moderate self‐perceived symptom severity predominant profile (median ES 3 (IQR 2–5) and median BEDQ 8.5 (IQR 1–18)). 47 patients (11.8%) had achalasia. Other clinical and epidemiological characteristics are displayed in Supplementary S1.

The BEDQ showed excellent reliability, with a Cronbach's α = 0.88. The ES showed a Cronbach's α = 0.59. No significant correlation was found between the BEDQ and GERDQ scores. In contrast, the ES showed a moderate significant correlation with GERDQ (Spearman Rho = 0.21. p < 0.01). Patients with abnormal HRM had significantly higher BEDQ scores than those with normal HRM (median 10.5 (IQR 2–21) vs. 5.0 (IQR 0–14), U = 63.13, z = 4.50, p < 0.001).

Ninety‐eight (24.7%) patients reported a PGIS of < 2 (NMSS). These patients had 63.5% normal HRM and significantly lower ES and BEDQ than the PGIS ≥ 2 group (median 2.5 (IQR 2–5) vs. 4.0 (IQR 2–5), U = 651.5, z = 3.55, p < 0.001; and 5.5 (IQR 2–15) vs. 11.0 (IQR 5–17), U = 707.0, z = −3.41, p < 0.001, respectively).

The cohort had a median BEDQ of 8.5 (IQR 1–18), with 246/397 (61.96%) with BEDQ < 9. BEDQ levels differed among the PGIS categories (Kruskal–Wallis H(4) = 14.87, p = 0.005; Figure 2). Pairwise comparisons showed no difference in the BEDQ levels between the PGIS categories 0 and 1 (p = ns) or between 2 and 4 (p = ns) (Figure 2). The same was observed when comparing the ES levels. In logistic regression models, both BEDQ and ES significantly predicted PGIS ≥ 2. Higher BEDQ scores were associated with increased odds of PGIS ≥ 2 (OR = 1.112 per point, 95% CI: 1.040–1.189, p = 0.002; Nagelkerke pseudo‐R ² = 0.157). Similarly, higher ES scores were associated with increased odds of PGIS ≥ 2 (OR = 1.504 per point, 95% CI: 1.350–1.675, p < 0.001; Nagelkerke pseudo‐R ² = 0.14).

FIGURE 2.

BEDQ levels ± 95% confidence intervals. There was a significant difference in BEDQ levels comparing PGIS 0–1 (NMSS, blue) to 2–4 (red). p ns = non‐significant.

ROC curve analysis of the BEDQ for predicting the NMSS showed acceptable discrimination, with an AUROC of 0.70 (95% CI 0.59–0.80; p < 0.05) (Figure 3). The Youden Index suggested that a cutoff of 9 points had the best discrimination capacity, and bootstrapping analysis confirmed the stability of this cutoff, as the 95% CI was 4–13. The best cutoff was 15 for severe symptoms (PGIS ≥ 3) (AUROC 0.65) and abnormal HRM (AUROC 0.61). The use of BEDQ ≥ 9 was associated with higher rates of PGIS 3–4 (33.3% vs. 15.4%, χ ²(1) = 4.56, p = 0.03) and abnormal HRM (51.3% vs. 32.3%, χ ²(1) = 3.69, p = 0.05) (Supplementary S2); however, neither reached statistical significance after Bonferroni correction for secondary discriminant analyses (α = 0.025). ROC curve analysis for ES predicting NMSS showed similar results, with an AUROC = 0.72 (95% CI 0.62–0.82; p < 0.05) (Figure 3) and the best cutoff value was 4 (bootstrapping 95% CI 1–5). The best cutoff was 6 for both severe symptoms (PGIS ≥ 3) (AUROC 0.69) and abnormal HRM (AUROC 0.63).

FIGURE 3.

ROC curves of BEDQ (A) and ES (B) for predicting NMSS (PGIS < 2) in the derivation (blue line) and validation (red line) cohorts. There was no difference in AUROCs between cohorts (p ns).

To eliminate the probability that GERD symptoms could be a relevant determinant of PGIS, we performed a subgroup analysis excluding patients with a GERDQ > 8. A total of 313 patients were included in this analysis. In this subgroup analysis, the BEDQ cutoffs were consistent with those of the primary analysis: the best cutoffs were 9 (AUROC 0.67) and 15 (AUROC 0.68) for predicting NMSS and severe symptoms (PGIS ≥ 3), respectively.

Table 1 shows the MID determined using the three described methods. In summary, all yielded consistent estimates (BEDQ: 5.01, 4.96, and 6.14 points; ES: 1.20, 1.25, and 1.58 points), supporting a final MID of 5 points for BEDQ and 1 point for ES.

TABLE 1.

Three different methods for MID determination in derivation and validation cutoff cohorts.

Score	0.5 SD	SEM	Anchor method
Derivation cutoff cohort
BEDQ	4.96	6.14	5.01 (1.20–8.85)
ES	1.25	1.58	1.2 (0–2.52)
Validation cutoff cohort
BEDQ	5.00	4.16	6.02 (4.62–7.41)
ES	1.27	1.38	1.68 (1.22–2.03)

Note: Anchor method: Calculated mean score difference between patients with PGIS < 2 (no or mild symptom severity) and PGIS ≥ 2. Values in parentheses correspond to 95% confidence intervals.

Abbreviations: BEDQ, Brief Esophageal Dysphagia Questionnaire; ES, Eckardt score; MID, minimal important difference; PGIS, patient global impression of severity; SD, standard deviation; SEM, standard error of measurement.

3.1.2. Validation Cohort

One hundred and seventy‐four patients constituted this cohort, with 106 (60.9%) being female and a mean age of 55 ± 11 years old. This cohort showed no statistically significant difference to the derivation cohort in terms of center of derivation, symptoms perception severity, and manometric diagnosis (see Supplementary S1 for details). For predicting NMSS, the best cutoffs were 8 and 2 points for BEDQ and ES, respectively. Table 2 shows operating characteristics of different cutoffs for predicting PGIS ≥ 2. The BEDQ AUROC in the validation cohort was 0.67 (95% CI 0.62–0.71). The observed difference of 0.03 between AUROCs was not statistically significant in the bootstrap comparison (p = 0.563). Similar findings were observed for ES: The AUROC was 0.72 (95% CI 0.67–0.76) in the validation cohort which was not significantly different from the derivation cohort (p = 0.887).

TABLE 2.

Operating characteristics at different cutoffs predicting moderate and severe symptoms (non‐NMSS).

PRO cutoff	Sensitivity	Specificity	Positive predictive value	Negative predictive value
BEDQ ≥ 8	0.553	0.741	0.869	0.348
BEDQ ≥ 9	0.533	0.747	0.867	0.341
ES ≥ 2	0.839	0.473	0.832	0.484
ES ≥ 3	0.642	0.667	0.857	0.374
ES ≥ 4	0.467	0.836	0.899	0.335

Abbreviations: BEDQ, Brief Esophageal Dysphagia Questionnaire; ES, Eckardt score; NMSS, no or mild symptom severity (PGIS < 2).

When calculating MID, results in this cohort were similar to the derivation one: Five points and 1 point for BEDQ and ES, respectively. Table 1 summarizes the three methods and depicts the derivation cohort for comparison.

3.2. Test–Retest Analysis

During the September–December 2025 period, 89 candidate patients were identified. Thirty patients could not be successfully re‐contacted, 4 declined participation and 1 started amitriptyline during the last days. Therefore, 54 patients constituted the final test–retest cohort. 30 patients (55.6%) were female, and the mean age was 45 ± 12 years old (Supplementary S1). The median interval between the test and retest was 4 days (full range: 2–5 days). Using Koo criteria [47], the test–retest reliability analysis revealed an excellent degree of temporal stability for the BEDQ scale, with an ICC of 0.97 (95% CI (0.94–0.99), p < 0.001). In contrast, the ES scale demonstrated good reliability, yielding an ICC of 0.85 (95% CI (0.71–0.92), p < 0.01). A comparison of the two instruments indicates that the confidence intervals do not overlap, suggesting that the BEDQ provides a significantly higher level of consistency compared to the ES in the current sample (Figure 4).

FIGURE 4.

ICC (95% CI) for ES and BEDQ. Reference line shows the accepted limit for good reliability (Koo et al. [47]). Confidence intervals do not overlap between BEDQ and ES, indicating significantly superior reliability of BEDQ.

3.3. Treatment Responsiveness Cohort

Eighty‐seven patients were included in this cohort from September to December 2025, with 48 (55.2%) female and a mean age of 50 ± 17 years old. Achalasia II (46, 52.9%) was the most prevalent pretreatment diagnosis. Seventy‐one patients (81.7%) were treated with POEM and 16 (18.3%) with Heller Myotomy (Supplementary S1). Seventy‐five patients (86.2%) described adequate relief. These patients showed significantly lower BEDQ total scores compared to those with inadequate relief (median 3.0 (IQR 0–9) vs. 24.5 (IQR 20–28); U = 28.5, z = −4.67, p < 0.001).

In 21 cases, pre and post‐treatment evaluations were available. The first evaluation occurred at a median of 2 (IQR 0–4) months before treatment. The second evaluation occurred at a median of 3 (IQR 1–6) months after treatment. Both BEDQ Total Score and Eckardt Score showed statistically significant reductions following treatment (Wilcoxon signed‐rank test: W = 0, Z = 4.02, p < 0.001 and W = 0, Z = 3.93, p < 0.001, respectively) (Figure 5). The standardized response mean (SRM) was 2.14 for the BEDQ and 1.98 for the Eckardt Score, both exceeding the threshold of 0.8 considered excellent sensitivity to change.

FIGURE 5.

Spaghetti plot illustrating intra‐patient changes in BEDQ between pre and post treatment. The dashed black line indicates the group mean. A statistically significant reduction was observed (p < 0.001). n = 21.

4. Discussion

In this large multicenter study, we have shown that a cutoff of 9 and 4 points is optimal to define NMSS, for BEDQ and ES, respectively. These thresholds have not been previously determined for clinical symptom severity, even considering their extensive use to define treatment success [15, 24, 50, 51].

In dysphagia there is a critical discordance between perception and objective findings in esophageal physiological tests [52, 53, 54]. Thus, description and validation of PROs are necessary. This study evaluated psychometric characteristics not previously addressed for BEDQ and ES.

The use of the self‐reported symptomatic burden questionnaire, generically called PGIS, has been repeatedly recommended for psychometric evaluation of PROs [55, 56, 57, 58]. PGIS has been previously used as an anchor for validation in IBD [59] and IBS‐related PROs [60]. Using this tool, we have recognized a group of patients with no or mild symptoms that we called NMSS. This distinction seems practical, as these patients are usually regarded as not requiring treatment escalation or eventually considered as treatment success. For example, using ES ≤ 3 as criterion, patients with mild achalasia symptoms have been excluded from treatment studies [61] and traditionally considered as treatment success [51]. Besides, we found no difference in BEDQ nor ES scores in patients with PGIS 0 and 1, supporting the adequacy of this classification.

Both PROs showed an acceptable discrimination capacity [62] for predicting NMSS, with an AUROC of approximately 0.7. This can be regarded as useful but also addresses the fact that symptomatic severity in these patients cannot be explained only by dysphagia severity. Other factors not addressed by these PROs, such as severity of other symptoms [63], psychological distress [64], the impact on daily living [63], and cultural aspects [8, 65], are likely contributors. A similar situation has been described for a classical PRO to grade severity in IBS (IBSSS), which has shown modest agreement with a patient self‐severity graduation (ƙ = 0.3) [63].

Figure 2 shows that no significant difference in BEDQ levels could be demonstrated among the PGIS 2–4 subgroup. This could be secondary to Type II error, as the sample size in those subgroups led to a low statistical power (20% for some comparisons). On the other hand, the association of BEDQ ≥ 9 with more severe symptoms (PGIS 3–4) and abnormal HRM, while directionally consistent, did not survive Bonferroni correction in these exploratory secondary analyses. Whether BEDQ > 15 has capacity to discriminate between moderate and severe symptoms will require a dedicated, adequately powered study.

For cutoff's analysis internal validation, we performed a randomized split‐sample analysis as previously recommended [29]. This approach does not only address overfitting, optimism, and miscalibration of the derivation cohort, but also supports the generalizability of the results. This analysis showed similar predictive capacity for both BEDQ and ES, supporting model reliability and applicability across different populations. The derivation and validation cohorts suggested slightly different cutoffs for BEDQ (9 and 8) and ES (4 and 2), respectively. As shown in Table 2, BEDQ ≥ 8 and BEDQ ≥ 9 are functionally equivalent, with almost no difference in any operating characteristic. In the case of ES, significant differences were found at different cutoffs. It has been suggested that the Youden Index can be more susceptible to bias in smaller samples, such as our validation cohort [66]. On the other hand, as dysphagia evaluating tools are used to decide treatment/retreatment, the use of low specificity cutoffs can derive in unnecessary treatments. Therefore, the use of ES ≥ 4 appears to be a safer approach. Finally, given the overlap of these cutoffs within the bootstrapping confidence intervals of the derivation cohort (BEDQ: 4–13; ES: 1–5), and the established prior use of BEDQ = 9 and ES = 4 in the literature, we maintained the derivation cohort cutoffs as the recommended thresholds.

The use of BEDQ < 9 is in line with the cutoff suggested by Taft et al. in the original description [18]. Nevertheless, they used a major motor dysfunction as anchor. In our study, the best cutoff to predict abnormal HRM was 15, coinciding with the cutoff to predict severe symptomatic burden (PGIS 3–4). This discrepancy can be explained by the use of CC3.0 as manometric criteria in the Taft study, which uses less strict criteria than CC4.0 for diagnosis [37].

Using three different methods, we have shown that BEDQ's MID was 5. From the statistical point of view, it means that in the case of a difference of < 5 a random effect cannot be ruled out. Similarly, we showed the same difference of 5 points allows change between the two clinical categories (NMSS and PGIS ≥ 2). Using the same approaches, MID for ES was 1 point. All these findings were consistent across the three methods, and also across both the derivation and validation cohorts (Table 1). This constitutes a strong and novel result of this study. Nevertheless, these recommendations for clinically relevant change need to be confirmed in future studies, using a prospective follow‐up assessment [10, 46].

Using a test–retest evaluation in a 4‐day interval, we showed that BEDQ has an excellent and ES a good stability over time. BEDQ's ICC appears significantly better than ES. We also confirmed previous data [18, 20, 67] showing a higher BEDQ's Cronbach's α. Therefore, BEDQ offers advantages over ES in two critical determinants of reliability. Another advantage found in this study is BEDQ's independence of GERD symptoms. ES, probably in the context of the regurgitation question, correlates with GERD, making it less specific for dysphagia.

We found that BEDQ differs significantly among those self‐perceiving adequate versus inadequate relief. In the group of patients prospectively evaluated pre‐ and post‐treatment, we found a strong standardized response mean (SRM) of 2.14, largely exceeding the threshold of 0.8 considered excellent sensitivity to change [49]. Therefore, our results show BEDQ's excellent responsiveness to treatment, a finding that has not been previously addressed.

We recruited patients from those referred to HRM. These constitute a heterogeneous population, but are similar to real patients attending GI practice. Several initial patients were excluded. Nevertheless, most of the exclusions were due to previous surgery or incomplete filling of the questionnaires, which should not lead to a relevantly biased sample, making our results widely generalizable. The final cohort was significantly higher than the calculated required sample size for ROC analysis and included all HRM CC4.0 diagnoses. In fact, we could demonstrate statistical significance for all the analyses we performed. Nevertheless, the usefulness of our recommended cutoffs and evaluation of temporal stability and treatment responsiveness in longer follow‐up periods needs to be specifically addressed in future prospective studies.

BEDQ, ES, and PGIS are all self‐reported questionnaires that were administered at the same time. By design, they have a high shared‐method variance. Therefore, PGIS is not a fully independent anchor. Nevertheless, the finding of AUROCs of about 0.7 for both BEDQ and ES suggests a lack of complete construct overlap. As previously discussed, PGIS captured dimensions of severity not accounted in the other PROs. Accordingly, PGIS has been repeatedly suggested as an appropriate anchor for PROs evaluation despite this expected inherent limitation [55, 56, 57, 58]. A strict validation of these cutoffs will require a separate study using a fully independent construct as anchor.

Finally, our treatment responsiveness cohort was significantly biased, as 86% of patients reported adequate relief. This could lead to an overestimation of the standardized response mean. Nevertheless, it should not invalidate the conclusion of the Mann–Whitney analysis. This bias is fully explained by the excellent results of achalasia therapies like POEM [68].

In summary, using a multinational cohort, we studied several psychometric properties of BEDQ and ES, not previously evaluated. We demonstrated strong temporal stability and treatment responsiveness. We recommend the use of BEDQ < 9 and ES < 4 as a useful cutoff to describe dysphagia burden. We also suggest that a change in more than 5 points for BEDQ and 1 point for ES can be considered significant. These findings have direct clinical applicability, given the widespread use of these PROs. Finally, our data confirm superiority of BEDQ over ES, due to its superior test–retest reliability, higher internal consistency, and independence from GERD‐related symptoms, making it the preferred tool to grade dysphagia.

Author Contributions

Eugenia Morales: conceptualization; investigation; writing – review and editing. Esteban Glasinovic, José María Remes‐Troche, Karla García‐Zermeño, Juan Pablo García Marmolejo, Albis Hani, Andrés Ardila‐Hani, Andrés Ditaranto, Joaquin Agotegaray, Claudio Bilder, Hugo Monrroy, Paula Rey, José Ignacio Gran, Jordi Serra, Luis G. Alcalá‐González, and Claudia Barber‐Caselles: investigation; writing – review and editing. Dustin A. Carlson: writing – review and editing. Daniel Cisternas: conceptualization; formal analysis; investigation; writing – original draft; writing – review and editing.

Conflicts of Interest

Dustin Carlson: Medtronic (Speaking, Consulting, License); Diversatek (Consulting); Phathom Pharmaceuticals (Speaking, Advisory Board); Braintree (Consulting); Medpace (Consulting); Laborie (Consulting). Daniel Cisternas: Medtronic (Speaking). Jose Maria Remes‐Troche: Adium (Advisory Board); Carnot (Advisory Board, Speaking); PROMEDICS Praha a.s. (Advisory Board); Pisa (Advisory Board); Asofarma (Speaking); Abbott (Speaking); Chinoin (Speaking); Ferrer (Speaking); Johnson & Johnson (Speaking); Menarini Central America (Speaking); M8 (Speaking); Medix (Speaking); Medtronic (Speaking). KGZ: Carnot (Speaking); Ferrer (Speaking). The other authors declare no conflicts of interest.

Supporting information

Supplementary S1. Demographic and clinical characteristics of the three study cohorts.

Supplementary S2. Discriminative capacity of BEDQ > 9. (A) For severe symptoms (PGIS 3–4) (33.3% vs. 15.4%, p = 0.05). (B): For abnormal High‐resolution manometry (51.3% vs. 32.3%, p = 0.05).

Data Availability Statement

Data and study materials are available from the corresponding author (Dr. Daniel Cisternas, dcisterc@gmail.com) upon reasonable request.