The American Foregut Society endoscopic classification outperforms Hill in phenotyping the esophago-gastric junction to guide gastroesophageal reflux disease management

Abstract

Gastroesophageal reflux disease (GERD) is driven by anti-reflux barrier (ARB) disruption, requiring precise endoscopic tools to guide diagnosis and management. The Hill classification, based solely on flap valve (FV), lacks precision. The American Foregut Society (AFS) classification, integrating hiatal hernia length (L), hiatal opening diameter (D), and FV (F), offers a comprehensive approach to phenotype esophago-gastric junction (EGJ) dysfunction. This study aims to assess the superiority of the AFS classification over the Hill and to weight the single AFS components contribution.

A retrospective analysis of adult patients evaluated with upper-GI endoscopy (EGD), high-resolution manometry, and reflux monitoring study for GERD symptoms at our Institution (2022–2025) was performed. GERD was defined by Lyon 2.0 criteria. EGJ was graded using AFS and Hill classifications. GERD prevalence was compared across AFS and Hill grades and by the number of disrupted AFS components. Logistic regression assessed individual AFS component contributions.

Of 249 patients (median age 52 years, 47% male, BMI 23.9 kg/m²), 127 had GERD. At least one AFS component was disrupted in 71.9% of the patients. GERD prevalence differed significantly across AFS grades 1–2 vs. 3 and 3 vs. 4, unlike Hill grades, where 2 and 3 overlapped. Patients with 1 and 2 impaired components were significantly different from those with 3 and 4. Pathologic D (OR = 2.537) and F (OR = 3.336) were independent GERD predictors. ROC analysis confirmed AFS superiority over Hill (AUC 0.750 vs. 0.653, P < 0.001).

The AFS classification enhances EGD diagnostic yield, outperforming Hill in EGJ phenotyping. The AFS improves patient stratification for pathophysiological testing and tailored therapies, offering a practical tool for GERD management.

INTRODUCTION

Gastroesophageal reflux disease (GERD) affects 4–14% of the general population, potentially reducing quality of life.¹ Its main pathophysiologic mechanism is the disruption of the anti-reflux barrier (ARB), whose main components are the lower esophageal sphincter (LES) and the crura diaphragm (CD).^2,3 Upper-GI endoscopy (EGD) is a key diagnostic tool, providing Lyon 2.0 criteria (Los Angeles grade B, C and D esophagitis, Barrett’s esophagus) and offering clear visualization of the esophago-gastric junction (EGJ).⁴

The Hill classification (1996) grades the EGJ, based on the retroflexed view of the flap valve (FV). This musculo-mucosal fold serves as a functional barrier, preventing the reflux of gastric content into the esophagus.⁵ The Hill classification categorizes the EGJ into four grades, from I (normal) to IV (disrupted), based on the degree of FV competence. However, it does not take into account other ARB components (i.e. hiatal hernia [HH] and CD separation), failing to precisely quantify hiatal disruption. Additionally, the lack of standardization, and the inconsistent results in predicting GERD severity or proton-pump inhibitor (PPI) response, limited its impact and use in clinical practice.^6–9

In 2022 the American Foregut Society (AFS) developed an endoscopic EGJ classification based on endoscopic retroflexed view that integrates HH axial length (component L), hiatal opening diameter (D) and presence of the FV (F).² Moreover, the AFS introduced a standardized terminology and protocol, grading the three components based on objective measurements. Preliminary studies suggest AFS superiority in predicting GERD and correlating with ARB disruption, assessed with high-resolution manometry (HRM) parameters, namely EGJ morphology, EGJ contractile integral (EGJ-CI) and straight leg raise (SLR) maneuver.^10,11 However, these studies involved a limited number of patients and could not fully validate the AFS classification or perform a comprehensive evaluation of its pathophysiologic features.

The primary aim of the study was to explore the ability of the AFS classification to accurately assess EGJ disruption and to stratify GERD severity. Secondary aims were to demonstrate its superiority over the Hill classification and to weight the single AFS component contribution to the overall pathophysiology of GERD.

MATERIAL AND METHODS

We retrospectively reviewed a prospective database of consecutive patients evaluated for GERD symptoms at IRCCS Policlinico San Donato (2022–2025). Inclusion criteria consisted of age 18–75 years and completion of EGD, HRM and a reflux monitoring study performed within 1 month. Exclusion criteria included prior foregut surgery, incomplete studies, or major esophageal disorders (e.g. achalasia, eosinophilic esophagitis).

GERD was defined by Lyon 2.0 criteria: Los Angeles grades B–D esophagitis, Barrett’s esophagus, peptic stricture, acid exposure time (AET) >6% on pH-impedance (MII-pH), or ≥ 2 days with AET > 6% on 96-hour wireless monitoring.⁴ Inconclusive AET (4–6%) with adjunctive criteria (reflux episodes >80/day, mean nocturnal baseline impedance [MNBI] <1500 Ω, or positive reflux-symptom association) qualified as GERD. Severe GERD was defined as AET > 12%, or DeMeester score (DMS) > 50, or grade C-D esophagitis or Barrett’s esophagus. The study protocol was approved by the internal review board (246_2025) and was conducted following the Declaration of Helsinki.

Clinical evaluation

Clinical and demographic data included age, BMI, symptom onset, PPI use and effectiveness. Symptoms were assessed using validated questionnaires, including GERD-Q,¹² GERD Health-Related Quality of Life (GERD-HRQL),¹³ Reflux Symptom Index (RSI),¹⁴ and Global Symptom Severity (GSS).¹⁵

Esophago-gastroduodenoscopy

EGD was performed under conscious sedation by expert endoscopists (>5 years of experience) and each of them met the caseload threshold recommended by the current guidelines.¹⁶

The EGJ was assessed in retroflexed view using the AFS classification protocol. Adequate visualization was obtained using endoscopic insufflation for 30–45 seconds or until the gastric rugal folds were flattened, followed by rotational maneuvers in retroflexed view along the lesser curvature to provoke potential sliding hiatal herniation and evaluate the hiatal opening in real time. Each component was graded according to the original paper (Fig. 1) and recorded in the endoscopic report. The Hill classification was also collected prospectively.^2,5

Fig. 1.

Visual representation of the L and D components of the AFS Classification.

High-resolution esophageal manometry

HRM was performed by expert physicians using the Medtronic system (Duluth, GA) per Chicago Classification 4.0 (CCv4.0) protocol.¹⁷

After overnight fasting, the catheter was inserted through the nostrils. Patients completed ten 5-mL swallows of room-temperature water in the primary position (upright or recumbent) after a 30-second swallow-free period, followed by five swallows in the secondary position. Metrics included LES length, basal pressure, EGJ-CI,¹⁸ and EGJ morphology. Ineffective esophageal motility (IEM) was defined as >70% ineffective or ≥ 50% failed swallows. Multiple rapid swallows (MRS), consisting of five 2-mL water swallows every 2–3 seconds were performed, with contraction reserve defined as an MRS-DCI to mean ten-swallows DCI ratio > 1.

After completion of the CCv4.0 protocol, patients performed the SLR, raising both legs to a 45° angle for 5 seconds. The SLR was considered effective if intra-abdominal pressure during the maneuver increased by 15.6 mmHg and positive if intra-esophageal pressure rose by 11 mmHg.^19,20 Finally, the Milan Score (MS), a comprehensive manometric tool that integrates EGJ morphology, IEM, EGJ-CI and SLR maneuver to quantify ARB disruption, was calculated.²¹

Reflux monitoring studies

Patients stopped acid-suppressive medications two weeks prior to reflux monitoring studies,^22,23 with alginate-based medications allowed as rescue therapy.²⁴ MII-pH studies employed catheters with eight impedance and two pH electrodes, with the esophageal pH electrode 5 cm above the LES. Data collected included total, upright, and recumbent AET, DMS, number of acid, weakly acid, and weakly alkaline reflux episodes, and reflux-symptom association, defined by positive symptom index (SI) or symptom association probability. All MII-pH studies were reviewed using Wingate consensus criteria.²⁵ Additionally, MNBI and post-reflux swallow-induced peristaltic wave index were also recorded.^26,27

The Bravo® system was used for prolonged catheter-free monitoring, with the capsule endoscopically placed six cm above the EGJ. Study duration was 96 hours, allowing normal daily activities while assessing reflux frequency, duration, and symptom correlation.²⁸

Outcomes and statistical analysis

Categorical variables were reported as frequencies, continuous as median and interquartile range (IQR). Comparisons used chi-squared, Fisher’s exact, t-tests, or Kruskal–Wallis tests. The population was stratified according to GERD diagnosis and compared. The EGJ phenotypes were defined by AFS components (L: 3 levels, D: 4, F: 2), (Fig. 1), yielding up to 24 possible combinations. Healthy patients were defined as those with no pathologic components (L1D1F+ or L1D2F+), while all other combinations reflected varying degrees of altered EGJ. The population was further stratified by the number of altered AFS components (0 to 3). GERD prevalence, DMS, MNBI, MS, and severe GERD prevalence were compared across Hill and AFS grades and disrupted components. Logistic regression evaluated AFS component contributions. Receiver operating characteristic (ROC) analysis compared AFS and Hill performance. Analyses used R 4.4.2 (The R Foundation, Austria), with P < 0.05 indicating significance.

RESULTS

Among the 249 included patients (median age 52 years, 47% male, BMI 23.9 kg/m²), 127 had pathologic GERD per Lyon 2.0. GERD patients were older (54 vs. 51 years, P = 0.038), more often male (61% vs. 31%, P < 0.001), and had higher BMI (25.6 vs. 22.5 kg/m², P < 0.001). GERD-Q and GERD-HRQL were significantly higher and RSI significantly lower in patients with GERD (Table 1). HRM showed higher EGJ disruption, HH, IEM, positive SLR, lower EGJ-CI, and higher MS in GERD patients. All MII-pH parameters differed significantly (Table 1). Of the 24 possible combinations, only 11 (45.8%) were observed in our population Figure 2A. An intact ARB (L1D1F+ or L1D2F+) occurred in in 70 patients (28.1%). 71.9% had ≥1 disrupted component (59.2% all three, 22.3% two, 18.4% one; Fig. 2B). GERD prevalence ranged from 15% in patients with an intact EGJ (L1D1F+) to 91% in those with a severely impaired EGJ (L4D4F-) (Table 2).

Table 1.

Demographic, clinical, manometric, and pH-impedence features of the study population. Continuous values are expressed as median [IQR]

	Total (n = 249)	No GERD (n = 122)	GERD (n = 127)	P-value
Male, n (%)	116 (47)	38 (31)	78 (61)	<0.001
Age (years)	52 [21]	51 [26]	54 [19]	0.038
BMI (Kg/m2)	23.9 [5.4]	22.5 [3.7]	25.6 [5.4]	<0.001
Waist circumference, (cm)	90 [19]	83 [18]	96 [14]	<0.001
Symptoms duration (months)	36 [108]	36 [108]	36 [108]	0.747
PPI use, n (%)	203 (86)	99 (86)	104 (86)	0.976
Response to PPI n (%)	134 (68)	60 (63)	74 (72)	0.192
Gerd-Q	8.0 [4.5]	7.0 [4.5]	9.0 [5.0]	0.003
GERD-HRQL	12 [16]	10 [14]	14 [17]	0.040
RSI	8 [13]	9 [14]	6 [10]	0.032
GSS	60 [40]	60 [38]	70 [40]	0.163
HRM findings
EGJ type				<0.001
1, n (%)	137 (55)	87 (71)	50 (39)
2, n (%)	82 (33)	32 (26)	50 (39)
3, n (%)	30 (12)	3 (2.5)	27 (21)
HH, n (%)	119 (48)	38 (31)	81 (64)	<0.001
HH size, (cm)	0.60 [1.50]	0.00 [0.70]	1.10 [1.70]	<0.001
LES total length (cm)	1.70 [0.60]	1.90 [0.70]	1.60 [0.40]	<0.001
EGJ-CI (mmHg*cm)	30 [37]	39 [47]	24 [23]	<0.001
Patients with IEM, n (%)	61 (24)	18 (15)	43 (34)	<0.001
Positive SLR, n (%)	101 (42)	12 (10)	89 (73)	<0.001
MS (points)	107 [123]	60 [34]	174 [56]	<0.001
pH-impedance findings
AET (%)	5 [9]	2 [3]	11 [10]	<0.001
Total reflux episodes	31 [29]	21 [24]	37 [28]	<0.001
Acid reflux episodes	22 [25]	13 [19]	32 [25]	<0.001
MNBI, (Ω), median [IQR]	2090 [2370]	3450 [1755]	1080 [940]	<0.001
MNBI <1500 Ω, n (%)	78 (39)	5 (5.1)	73 (72)	<0.001
Reflux-symptom association (%)	27 (23)	6 (12)	21 (32)	0.014
DMS	21 [37]	9 [7]	46 [38]	<0.001

Fig. 2.

(A) Distribution of AFS phenotypes considering the ‘L,’ ‘D,’ and ‘F’ variables among the study population. (B) Frequency and combinations of patients with increasing number of disrupted components.

Table 2.

Proportion of patients with pathologic GERD in the different AFS phenotypes

EGJ phenotypes, n (%)	Total (n = 249)	No GERD (n = 122)	GERD (n = 127)
L1D1F+	20	17 (85)	3 (15)
L1D2F+	50	37 (74)	13 (26)
L1D2F-	17	11 (65)	6 (35)
L3D2F+	16	14 (88)	2 (13)
L1D3F-	22	7 (32)	15 (68)
L3D2F-	12	6 (50)	6 (50)
L4D2F-	6	1 (17)	5 (83)
L3D3F-	32	12 (38)	20 (63)
L3D4F-	14	3 (21)	11 (79)
L4D3F-	28	11 (39)	17 (61)
L4D4F-	32	3 (9.4)	29 (91)

When pathologic GERD, DMS, MNBI, and MS were explored across the different grading systems, in the Hill classification grades 2 and 3 consistently showed similar values for all parameters, but both differed significantly from grades 1 and 4. In contrast, the AFS classification revealed significant differences in all parameters between grades 1–2 vs. grade 3, as well as between grades 3 and 4. Additionally, patients with one or two impaired components exhibited significant differences compared to those with three or four disrupted components (Fig. 3). Severe GERD was significantly different between patients with Hill grade 1 vs. 2–3-4 (P = 0.024) and with AFS 1–2 vs. 3–4 (P = 0.003) (Fig. 4).

Fig. 3.

Trend and comparison of (A) percentage of patients with pathologic GERD, (B) median values of DMS, (C) median values of MNBI, and (D) median values of MS between Hill and AFS classification and number of altered components.

Fig. 4.

Percentage of patients with severe GERD in Hill and AFS classification.

On the multivariate model, we found that pathologic D and F components were independent predictors of pathologic GERD (OR 2.537, CI [1.146–5.659], P = 0.021 and OR 3.336, CI [1.404–8.056], P = 0.006, respectively) (Table 3).

Table 3.

Logistic regression including AFS features associated with GERD

Variable	Odds Ratio	95% CI	P-value
Pathologic L	1.233	(0.616–2.401)	0.543
Pathologic D	2.537	(1.146–5.659)	0.021
Pathologic F	3.336	(1.404–8.056)	0.006

ROC analysis confirmed AFS superiority (AUC 0.750 vs. Hill 0.653, P < 0.001; Fig. 5).

Fig. 5.

ROC curve of the AFS and Hill classification in predicting GERD.

DISCUSSION

Our study provides a comprehensive evaluation of the AFS classification of the EGJ in patients with suspected GERD, demonstrating its superiority over the Hill classification in predicting objective diagnosis and stratifying disease severity. By assessing the three key components of the EGJ (HH and opening and FV), the AFS classification provides a comprehensive EGJ assessment potentially increasing the diagnostic yield of EGD.

EGD is a crucial test in the GERD workup, offering Lyon 2.0 criteria⁴ and exclude other conditions (i.e. motility disorders, malignancies, gastric diseases) presenting with similar symptoms.^29–31 Moreover, its ability to offer a detailed understanding of pathophysiology may lead to better disease severity stratification and even treatment outcome, as shown in other medical fields.³²

The pathophysiology of GERD is complex, primarily driven by the disruption of the ARB, which comprises the LES, the CD and the musculo-mucosal FV.^2,3 The Hill classification represented the first attempt to endoscopically assess EGJ disruption. In 2019, Inoue et al.³³ proposed a more comprehensive EGJ evaluation, by integrating sliding hernia (SH) and cardiac opening (CO) measurement, demonstrating a correlation with pathologic AET for a CO >3 cm and an SH >2 cm. Subsequently, in 2022 the AFS introduced a protocol with clear terminology and methodology, integrating two similar components (D and L) alongside with the FV assessment (F).

Our study explored the relationship between these components, assessing the ability of the classification to effectively phenotype the EGJ. Among the 24 possible AFS phenotypes, only 11 were observed in our population, suggesting that certain combinations may be more prevalent in GERD patients. Notably, 71.9% of patients had at least one disrupted component, with 59.2% exhibiting disruption of all three components, underscoring the multifactorial nature of EGJ dysfunction in GERD and the AFS ability to capture the heterogeneity of EGJ phenotypes. One disrupted component behaves similarly to healthy patients, suggesting a compensatory mechanism by the intact components. As shown in Figure 3, pH and HRM metrics (pathologic GERD, DMS, MNBI, and MS) significantly differed in patients with >1 compromised mechanism, where the degree of EGJ disruption was too advanced to protect from reflux.

When exploring the AFS classifications, we found that not only it effectively distinguished healthy individuals from those with GERD but also stratified GERD severity. Specifically, AFS grades 1 and 2 showed GERD rates of 12% and 25%, respectively, compared to 53% for grade 3 and 79% for grade 4 (P < 0.001). Furthermore, the rate of patients with severe GERD was significantly higher in AFS grade 3 and 4 (37% and 54%) than in AFS grade 1 and 2 (3.5% and 5.3%, respectively), highlighting its ability to discriminate both the presence and severity of GERD.

In contrast, the Hill classification showed overlapping characteristics between grades 2 and 3, underlying its limited discriminatory power between GERD and no-GERD patients, as already shown in other studies.^34,35 When we compared the two classification systems, the AFS showed a significantly higher AUC at ROC analysis (0.750 vs. 0.653, P < 0.001) demonstrating a significantly higher ability to predict pathologic GERD. Also, GERD severity was not effectively stratified using the Hill classification. These results support preliminary findings demonstrating the AFS superiority over the Hill classification in predicting pathologic GERD.^10,11

The logistic regression analysis elucidated the independent contributions of the AFS components to GERD diagnosis. The D and F components were independent predictors of pathologic GERD, with odds ratios of 2.537 and 3.336, respectively, while the L component did not reach statistical significance in the multivariate model (P = 0.543). This finding suggests that the degree of hiatal opening and the integrity of the FV may play more critical roles in the pathophysiology of GERD than the axial length of the HH alone, as shown by Hansdotter et al.⁹ and by Seltman et al.³⁶ The wider hiatal opening (D) likely facilitates greater reflux volume by reducing the mechanical barrier provided by the CD, while the absence of the FV (F) compromises the functional valve mechanism that prevents gastric content backflow. These results confirm previous studies that underlined the importance of the CD, the FV and the synergy among the different components in ARB function.^3,36,37

The ability of the AFS classification to phenotype the EGJ by integrating different mechanisms of ARB disruption, enhances the diagnostic yield of EGD. Our study underlines the potential of this test to guide therapeutic decision-making, including the selection of patients for further pathophysiologic tests and for surgical intervention or optimization of medical therapy. Patients with severe EGJ disruption (e.g. AFS grades 3–4 or patients with all three components altered) may benefit from early referral for anti-reflux surgery, as their disease is likely to be more severe and less responsive to PPIs or Potassium-Competitive Acid Blockers (PCABs).³⁸ Conversely, those with AFS grades 1–2, with minimal EGJ disruption, may achieve adequate symptom control with PPI or PCAB therapy, reducing the need for invasive procedures. This stratification aligns with personalized approaches to GERD management, optimizing resource allocation and improving clinical outcomes. By integrating AFS with reflux monitoring and HRM metrics, clinicians can further refine patient selection for surgical referral earlier in the diagnostic pathway.

These findings are further supported by the correlation between AFS grades and HRM metrics of ARB competence, including EGJ-CI, SLR and the comprehensive MS. These parameters have previously shown strong correlations with GERD severity³⁹ and with clinical outcomes following medical and surgical treatments.⁴⁰ Thus, the AFS classification offers potential to integrate endoscopic findings with physiologic data, enhancing diagnostic accuracy.

Besides the strengths of the study, in particular the large cohort of patients that underwent a complete pathophysiology and endoscopic study with prospectively collected AFS and Hill grades, this study presents some limitations. First, the retrospective nature may introduce potential selection bias. Second, the monocentric nature limits the cohort of patients selected. The high rate of patients with at least one disrupted ARB mechanism may reflect the population of our surgical center, where many patients are referred for potential surgical treatment. Third, the relative novelty of the AFS classification in our clinical practice necessitated an adequate learning curve that may have impacted the early diagnosis. Finally, the study did not evaluate long-term outcomes, such as the response to surgical or medical interventions based on AFS phenotypes.

In conclusion, our study demonstrates that the AFS classification enhances the diagnostic yield of EGD, providing an accurate phenotypic classification of the EGJ and a quantification of ARB disruption. The extensive use of this grading system in the common clinical practice has the potential to impact decision-making to better select patients for pathophysiologic tests and for anti-reflux surgery.

Specific author contributions: Carlo Galdino Riva (Conceptualization, Data curation, Writing—original draft, Writing—review & editing), Stefano Siboni (Conceptualization, Data curation, Formal analysis, Writing—original draft, Writing—review & editing), Jiurgen Mema (Data curation, Writing—review & editing), Roberta de Maron (Conceptualization, Data curation, Writing—original draft, Writing—review & editing), Pamela Milito (Data curation, Visualization, Writing—review & editing), Andrea Lovece (Data curation, Investigation, Writing—review & editing), Daniele Bernardi (Visualization, Writing—original draft, Writing—review & editing), Marco Sozzi (Conceptualization, Data curation, Investigation, Methodology, Writing—original draft, Writing—review & editing), Emanuele Asti (Conceptualization, Supervision, Validation, Writing—original draft, Writing—review & editing).

Financial support: No financial support.

Potential competing interests: None of the authors have conflicts of interest or financial ties to disclose.